ALTERNATIVE MEASUREMENT APPROACH USING INVERSE SCATTERING...
Transcript of ALTERNATIVE MEASUREMENT APPROACH USING INVERSE SCATTERING...
Florida State University Libraries
2015
Alternative Measurement Approach UsingInverse Scattering Theory to ImproveModeling of Rotating Machines inUngrounded Shipboard Power SystemsPatrick Ryan Breslend
Follow this and additional works at the FSU Digital Library. For more information, please contact [email protected]
FLORIDA STATE UNIVERSITY
COLLEGE OF ENGINEERING
ALTERNATIVE MEASUREMENT APPROACH USING
INVERSE SCATTERING THEORY TO IMPROVE MODELING OF
ROTATING MACHINES IN UNGROUNDED SHIPBOARD POWER SYSTEMS
By
PATRICK RYAN BRESLEND
A Thesis submitted to theDepartment of Electrical & Computer Engineering
in partial fulfillment of therequirements for the degree of
Master of Science
2015
Copyright c© 2015 Patrick Ryan Breslend. All Rights Reserved.
Patrick Ryan Breslend defended this thesis on August 3, 2015.The members of the supervisory committee were:
Chris S. Edrington
Professor Directing Thesis
Lukas Graber
Committee Member
Mischa Steurer
Committee Member
The Graduate School has verified and approved the above-named committee members, and certifiesthat the thesis has been approved in accordance with university requirements.
ii
I dedicate this dissertation to my loving fiance Alicia and my entire family. Without theirunderstanding, moral support, inspiration and most of all love, the completion of this dissertationwould not be possible. I must express the fact that I could not visit my family as much as I wouldhave liked to during the course of my masters and they wholeheartedly accepted it. I owe themquality time, which they missed during my studies.
iii
ACKNOWLEDGMENTS
I express my deep gratitude to my major professor Dr. Chris Edrington and advisor Dr. Mischa
Steurer for providing a significant opportunity to work on this project at the Center for Advanced
Power Systems funded under the Office of Naval Research. I thank them for their guidance,
encouragement, understanding, and patience throughout the duration of my graduate studies at
FSU. Without their research ideas, and financial support, this thesis would not be possible.
Special thanks to my friend and colleague Behshad Mohebali who worked with me on this
project. He provided generous help and time to help solve the problems I faced while working on
the research. His suggestions and ideas will always be cherished.
I would like to express my sincere thanks to my master’s thesis committee members Dr. Edring-
ton, Dr. Graber and Dr. Steurer for providing valuable suggestions as well as serving on the
committee and providing assistance where needed.
Thanks are due to all my friends and family who have helped me in various capacities in the
completion of this thesis. If it were not for them, I would have honestly struggled to finish.
iv
TABLE OF CONTENTS
List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vi
Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viii
1 Introduction 11.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.2 High Frequency Transients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41.3 Significance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61.4 Organization of Master Thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2 State of the Art 102.1 High Frequency Electric Machine Modeling Approaches . . . . . . . . . . . . . . . . 11
2.1.1 Lumped Element Equivalent Circuit Models . . . . . . . . . . . . . . . . . . . 122.1.2 Distributed High Frequency Models . . . . . . . . . . . . . . . . . . . . . . . 142.1.3 Common-mode and Differential-mode Equivalent Circuit Models . . . . . . . 16
2.2 Scattering Methodology for Measuring Electric Machines . . . . . . . . . . . . . . . . 182.2.1 Linear Time-Invariant Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . 192.2.2 Instrumentation for Characterizing Machine Coupling to Ground . . . . . . . 202.2.3 Port Determination and Measurement Process . . . . . . . . . . . . . . . . . 23
3 Problem Statement 29
4 Inverse Scattering Theory Applied to Three-Phase Electric Machine Modeling 314.1 Validity of Linear Time-Invariant Networks Applied to Rotating Machines . . . . . . 324.2 Parameter Extraction for Lumped Circuit Model . . . . . . . . . . . . . . . . . . . . 364.3 Virtual Measurement Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
5 Results and Comparison to Conventional Techniques 435.1 Resonance and Anti-Resonance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 435.2 Short Circuit Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 465.3 Virtual Measurement Comparison to Impedance Analyzer . . . . . . . . . . . . . . . 51
6 Conclusions and Future Considerations 576.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 576.2 Future Considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 586.3 A Note to the Reader . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
Appendix
A Motor Models 62
Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
Biographical Sketch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
v
LIST OF FIGURES
1.1 All-electric ship proposed by the U.S. Navy [1] . . . . . . . . . . . . . . . . . . . . . . 2
2.1 Equivalent circuit models for capacitors and inductors [2] . . . . . . . . . . . . . . . . 13
2.2 Equivalent lumped model plots for capacitors and inductors [2] . . . . . . . . . . . . . 13
2.3 Distributed high frequency circuit model with 17 turns per slot [3] . . . . . . . . . . . 15
2.4 Parasitic coupling of the ship MVDC power system and ship hull ground [4] . . . . . . 16
2.5 This is a visual representation of scattering parameters for a two–port network . . . . 22
2.6 Port representation for electric machine scattering measurements where T is for termi-nal and P is for port. T2P2 is terminal 2 and also port 2 of the measurement matrix.Each port of the matrix is measured by referencing it to a common ground. . . . . . . 24
2.7 Differential-mode and common-mode capacitive coupling [5] . . . . . . . . . . . . . . . 25
2.8 CM and DM impedance measurement connections [6] . . . . . . . . . . . . . . . . . . 27
2.9 Voltage waveforms corresponding to the 4-port PMSM network with Z0 = 50Ω forboth VNA cables and both single ended terminations . . . . . . . . . . . . . . . . . . 28
4.1 N-connector shield setup for scattering parameter measurements on electric machines(Left) using a vector network analyzer (Right) . . . . . . . . . . . . . . . . . . . . . . 33
4.2 Variation of port 1 reflection (s11) of the brushless DC motor or synchronous machinewith cogged rotor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
4.3 Variation of port 1 reflection (s11) of the induction or asynchronous machine withsmooth rotor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
4.4 Per-phase representation of the induction motor including the high frequency modeland the dynamic dq model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.5 Common-mode impedance of a lumped parameter model compared to impedanceanalyzer measurements and scattering parameter model representation . . . . . . . . . 38
4.6 Single–port scattering parameter measurement to obtain common–mode impedanceinformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.7 Single–port scattering parameter measurement to obtain differential–mode impedanceinformation like T1 + T2 to T3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
vi
4.8 Single–port scattering parameter measurement to obtain phase–neutral impedanceinformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
4.9 Virtual measurements of impedance for phase-to-neutral, common-mode, and differential-mode are compared to measurements made on the physical machine . . . . . . . . . . 41
5.1 Induction motor impedances measured using IA and VNA measurements . . . . . . . 44
5.2 Common-mode impedance of a PMSM equivalent T-circuit . . . . . . . . . . . . . . . 47
5.3 PMSM fault current simulation using a measured fault to trigger a controlled voltagesource coupling terminal T1 to ground and phase T2 and T3 unloaded. . . . . . . . . 48
5.4 Measurement setup for a single phase to ground fault of a three-phase PMSM . . . . 49
5.5 Single phase to ground fault of a three-phase PMSM . . . . . . . . . . . . . . . . . . 50
5.6 T-circuit simulated fault current . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
5.7 PMSM measured impedance compared to simulated T-circuit equivalent . . . . . . . . 51
5.8 Virtual measurement of common–mode impedance . . . . . . . . . . . . . . . . . . . . 52
5.9 Phase–neutral mode impedance data was measured virtually . . . . . . . . . . . . . . 53
5.10 Differential–mode impedance data was measured virtually . . . . . . . . . . . . . . . . 55
5.11 Impedance data was virtually measured and compared to known measured data withhigh accuracy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
vii
ABSTRACT
The Navy has proposed to use a shipboard power system operating at medium voltage direct cur-
rent to distribute power for their all-electric ship. The power is generated by electric machines as
alternating current and requires power electronic rectifiers to output direct current. Power elec-
tronics converters are needed to convert the direct current to alternating current for ship propulsion
and service loads. An increase in the use of fast switching power electronics is expected in future
ships. The increased voltage rise time on switches is known to produce unwanted high frequencies
with corresponding wavelengths of the same order of magnitude as the length of the ship hull.
These high frequency transients can cause the ship system to couple with the surrounding ship hull
causing adverse effects. The amount of high frequency content and the impact it has on the ship
system performance is difficult to calculate with current models. Increased voltage and performance
requirements for power electronics has led to advancements in switching frequencies into the 10s
to 100s of kilohertz and increased voltage edge rates. The faster switching corresponds to higher
frequency responses from the shipboard power system.
Research has shown that high frequency content in electrical power systems is responsible for
parasitic coupling and ultimately damage to the equipment. Electric machines, for instance, have
increased winding and iron losses, overvoltages at the terminals, and even bearing currents via
shaft voltages. The Navy is interested in simulating ship systems to test their electromagnetic
compatibility before implementing or committing to a specific design.
There are numerous techniques used to acquire machine parameters that have been proven to
be useful in modeling electric machine behavior. The approaches were considered by the amount of
proprietary information needed to acquire accurate results, the complexity of the modeling methods,
and the overall time it takes for implementation. A majority of system simulations gravitate towards
simple solutions for machine behavior which require assumptions to be made that deviate from
the actual machine behavior. Exact inner dimensions, winding layouts, end winding dimensions,
insulation thickness, and other information are proprietary and often not accurate representations
of the physical machine once built. It is time consuming to obtain an accurate working model when
assumptions are made or when detailed computer aided design models are needed to calculate
machine response quantities.
viii
The research modeling approach put forth in this paper is not aimed at capturing the steady-
state behavior of the machine. It is shown that a detailed understanding of the motor may not
be necessary to accurately model the high frequency effects. It is the transient behavior at non-
operating frequencies that need to be modeled correctly to develop new models of shipboard power
systems for grounding research. The frequency dependent information is most useful to determine
frequencies of interest that other modeling techniques are less likely to capture and point out.
Previously suggested measurement techniques have been considered useful in determining pa-
rameters of machines but are not always accurately implemented without in-depth knowledge of
the motor that may be proprietary. Lumped-parameter models are based on extracting informa-
tion at transitional frequencies or looking at the slope of a variable over a frequency range. These
models tend to be over simplified representations of the component by averaging the parameters
for given ranges. In reality a machine’s impedance varies with all frequencies. Lumped parameter
based models typically over simplify the grounding behavior of the machine by not varying the
impedance as a function of frequency.
The technique used in this research is based on scattering parameters, a way of determining
the terminal behavior of the machine without the knowledge of the actual inner workings of the
machine. The inverse scattering technique uses steady-state stimuli to calculate reflection and
transmission coefficients of system components allowing the device to be considered as a black box.
This can be understood as electrical snapshots of how the machine would respond when subjected
to a range of spectral content. The approach could have a significant impact on the modeling of
ground interactions with machines. The machine can now be measured and characterized with no
prior knowledge of the machine. The measurements are placed in simulation software in the typical
measurement configurations used in other approaches to extract parametric data. It was discovered
that these different configuration setups could now be measured in software without the need to
physically reconfigure the machine’s wiring for each measurement. This modeling approach was
coined ’virtual measurement modeling.’
To the best of the author’s knowledge there are not any known techniques for fast model
prototyping of electric machines which cover a broad range of frequencies with high accuracy. This
thesis will present a possible solution for consideration in future models developed for grounding
studies. This approach outlines a promising technique that can be easily implemented with high
ix
accuracy and reproducibility. The technique was derived from inverse scattering theory and was
implemented on electric machines for characterizing high frequency behaviors.
x
CHAPTER 1
INTRODUCTION
The information disclosed in this manuscript will cover research on alternative measurement tech-
niques for characterizing high frequency electric machines and build grounding models. The re-
searcher has demonstrated in previous work how new measurement techniques can be applied to
modeling of various components in a shipboard power system for grounding studies. Although
research has been carried out on multiple devices, the focus in this manuscript will be limited to
the development of three-phase rotating electric machine techniques. The following chapter will
cover the motivation for the work, a brief description of high frequency transients, the significance
of the proposed approach with regards to modeling, and the outline of the remaining chapters of
the manuscript.
1.1 Motivation
The all-electric ship (AES) has been an ongoing research topic for the Office of Naval Research
(ONR) Electric Ship Research Development Consortium (ESRDC) [7, 8]. One of the main focus
of our group’s research is on the proposed shipboard electrical power system and the need for
full detailed models of electrical components as well as the environment in which the system will
operate. The all-electric ship proposed by the U.S. Navy can be seen in Figure 1.1. The classical
methods of system modeling cannot be assumed to directly apply to ship electrical power systems,
considering the physical differences between the domains of terrestrial and shipboard applications
[9]. The current approach is to operate the shipboard power system (SPS) ungrounded from the
ship hull. This will have an effect on the attenuation of high frequency content so it is pertinent that
research be conducted to understand these events and model their effect on the overall performance
of the entire ship system.
The SPS will mainly be comprised of power electronics, cabling, and machines. The SPS
will undoubtedly be subject to electromagnetic interference (EMI) and therefore electromagnetic
compatibility (EMC) will need to be explored further [10, 11, 12, 13]. Models which capture the
1
Figure 1.1: All-electric ship proposed by the U.S. Navy [1]
electromagnetic interaction between the SPS and ship hull are important for grounding considera-
tions such as fault location algorithms based on common mode (CM) and differential mode (DM)
noise [12, 14]. The use of SPS modeling could also mitigate issues prior to building the ship and
reduce equipment sizing through optimization and proper placement of components to lower the
destructive high frequency content in the system.
The SPS preferred configuration will be a Direct current (DC) distribution network. An un-
grounded DC distribution network is said to be the most suitable option because it has the ability
to operate under certain fault conditions and it removes the need for redundant conversions found
in AC distribution [15]. The latest technology in power electronics need highly reliable intelli-
gent systems that can reconfigure quickly to maintain survivability. Power electronics continue
to increase in efficiency while being designed for higher voltages and currents that come with the
increased power density requirements for future ship design. The rectification and inversion of cur-
rent by means of power electronic switching is known to produce EMI due to steep rise times [16].
The study of EMC has produced substantial research describing transient behavior. Section 1.2
will discuss the effects of transients such as overvoltages, insulation break down, unwanted bearing
currents, and ultimately the coupling of the SPS with the surrounding ship hull through parasitic
2
paths. The frequency spectrum of the EMI in the system is related to the transient rise times and
can be accurately captured with wide-band frequency models [17, 18].
To properly study the effects of different grounding strategies [19], models must be developed
that capture the parasitic coupling effects of components and the ship hull. The current design
for the SPS is said to be likely an ungrounded system topology, also known as non-intentionally
grounded because it is impossible to have a power system completely isolated from ground due
to inherent parasitic elements [9, 14]. The parasitic coupling of the SPS and ship hull provides
a possible low impedance path back to the source. It is for this reason that studies need to be
done in order to capture high frequency responses of the SPS components. Our work focuses on
developing higher order grounding models of the SPS that properly characterize each component
at the frequencies determined to be of interest. The grounding schemes will have a large impact
on the system performance. Models which typically capture steady state time scales will need to
incorporate smaller time steps in order to properly describe the high frequency response.
High frequency models of power cables, power electronic devices, and electric machines were
designed in [20], [21], and [22] respectively. The common methods or techniques used to develop the
current high frequency models can be highly complex and specialized to each type of device in order
to capture wider frequency responses [16]. Techniques used to extract model parameters of various
components have been done primarily through on-line testing, lumped parameter approximations
from standstill frequency response , and finite element analysis [23], which are all further outlined
in Chapter 2 Section 2.1 where high frequency electric machine modeling approaches are discussed.
The main focus of this work is to introduce a new measurement method to this field of research
that provides information over a wide range of frequencies needed for modeling an electric machine.
The measurement of frequency response characteristics is already widely used by obtaining common-
mode (CM) and differential-mode (DM) impedances [24]. These impedance quantities can be used
for many different modeling approaches. In contrast using the inverse scattering theory to measure
a machine is a new approach, where input and output ratios for each port of the machine are
recorded. The scattering parameters can be transformed into familiar configurations such as CM
and DM impedances. Scattering parameters are most useful when considering the machine as a
black box were the internals of the machine are not necessarily known or readily available. Thus the
3
motivation for this research is to explain the significance of quality high frequency measurements
that can accurately be modeled with little to no knowledge about the machine.
1.2 High Frequency Transients
Transients in electrical systems can be defined as disturbances in voltage or current that re-
sult in a deviation from steady-state conditions for a short period of time. Transient responses
typically oscillate about a steady-state quantity. The oscillating behavior brought on by the tran-
sient will eventually attenuate and return to steady-state based on the circuit parameters and the
corresponding system configuration. The transients can operate at frequencies considered high for
electric ships due to the physical length of the ship and the corresponding cable lengths of the SPS.
Switching events in electrical power systems are known for producing high frequency transients.
The transient response of the various power system components can affect the behavior of the
system if not properly mitigated. A component’s parasitic elements produce varying electrical
responses at these higher frequencies. This behavior should be included in simulation models or
the system controls may misinterpret and calculate the incorrect response causing a cascade of
issues. The transient current radiates electromagnetic energy from the circuit that can couple with
neighboring conductors, causing unwanted interference [25]. This potentially unwanted coupling
between circuits and their connected ground paths can be highly complicated and difficult to predict
the full effect of the energy distribution if not modeled correctly [11]. The transient response of
the system can lower the impedance of unwanted paths through parasitic elements inherent to the
system’s construction and connected circuitry.
Research of transients and their high frequency effects on electric machines is largely due to the
damage it causes to insulation in windings and unwanted leakage currents to the frame through these
various parasitic paths [15]. The inductive coils in an electric machine try to impede the change
in current by producing a complementary voltage in the opposite direction known as electromotive
force (e.m.f.). High frequency transients are significant and can cause e.m.f. voltages to peak across
the windings. The e.m.f. can be much higher than the operating voltages due to the impedance
ratios of the windings. The higher the frequency the more irregularities in voltage distribution
across windings. Energy is thus stored in the magnetic field and its flux is proportional to the
rate of change in current flow. The magnetic field is a response produced to oppose the transient.
4
The collapse of the field in the magnetized coil can have faster fall times then the e.m.f. rise time
and produce even higher frequencies. These high frequency behaviors are what make transient
responses so destructive to electric machines [26].
A periodic switching transient can also produce constructive and destructive wave behaviors
in an electrical power system. When these waves constructively combine, the voltage can rise
more than twice the operating voltage in some cases [24, 27]. Each component has an inherent
impedance when seen from the source called its characteristic impedance. This impedance will
either be capacitive or inductive in nature with some real resistive behavior. Surges in energy from
the source traveling on a direct path through conductive material like cables and windings will be
met by the impedance of the machine called surge impedance.
The various impedance magnitudes will rise and fall at certain frequencies in the machine.
These specific frequencies are known as resonant and anti-resonant frequencies. Machine transient
responses can be mitigated by controlling or filtering the frequencies, which lower impedances
through unwanted parasitic paths [11, 25]. The impedance changes as a function of frequency and
it ultimately shapes the transient response of the system. To understand which frequencies will
have negative effects, the system’s characteristics and surge impedances are analyzed across the
frequency spectrum to determine frequencies of interest. Resonant frequencies with low impedance
can cause a ringing that decays slowly if there is nothing to attenuate the waveform. The transient
system behaviors need to be studied, not only at the operating frequency, but also at the frequencies
of the system’s response, which lower the impedance of parasitic paths.
High switching frequencies of inverters will have steep voltage edge rates between stator wind-
ings and motor frame, i.e. transients that cause electromagnetic interference in the motor. The
common–mode voltage (CMV) arises when the terminal voltage changes at a fast rate. Surge
propagation in stator windings can cause voltages to develop in the end winding, influence of in-
terturn voltage distribution, and voltage across the end winding. These voltages are not uniformly
distributed along windings due to smaller wavelengths corresponding to the higher frequencies
found in transients. The CMV on the motor windings produces a common-mode current that flows
through the distributed stray capacitance between the windings and motor frame. The frame of the
rotor is generally connected to the ground circuit giving CM current a low impedance path through
5
the cables capacitive coupling to ground. This has been well-documented in literature such as [20]
and explored through many different full spectrum models of electric machines.
Access to accurate models will help system performance by balancing line voltages and miti-
gating the negative sequence currents in the machine [25]. Reduction of the unwanted fields which
rotate in a direction counter to the normal field can decrease current losses due to heating and in-
creased torque, efficiency, and power. Typically models of power systems do not account for these
effects due to the complexity of the modeling methods needed to describe the high frequency be-
havior. Most of the models are simplified and designed to work well for terrestrial power systems.
Terrestrial systems are less concerned by these issues of ringing frequencies due to the attenua-
tion available with various circuit grounding configurations [15]. Designing an ungrounded circuit
that can attenuate the high frequency noise created from switching events requires larger model
bandwidth.
Our research involves the modeling of machines for use in all-electric ships, which will potentially
have an ungrounded power system with a lower attenuation characteristics. If there is nothing in
place to stop these frequencies or at least recognize their occurrences during certain electrical events
then the system will continue to oscillate and potentially cause degradation of the system. This
oscillation can be prevented by changing the capacitive or inductive characteristics of the impedance
at these specific frequencies causing issues. Reducing these effects can be achieved by changing the
switching frequencies to cause different responses that are less harmful or contain less energy at
the anti-resonant frequencies.
1.3 Significance
The significance of this work can be applied to many fields, but this research was conducted
for Naval applications. In terrestrial systems, higher frequencies are generally attenuated by the
earth ground. The hull of the ship provides a low impedance path for current to freely travel across
during parasitic coupling of the the ship power system and the ship hull. The large size of the
motors and generators used in electric ships are a large contributor to the capacitance present in
the system [28]. Understanding how the system will operate under various frequencies is of high
concern because of the adverse effects it could have on the rest of the system; especially unwanted
ground currents flowing in the ship hull [26, 8].
6
The research was first aimed at exploring state of the art literature on high frequency machine
modeling and to provide insight into new techniques that could simulate the ground coupling
and other high frequency effects. High frequency electric machine models presented in previous
literature were developed through many different techniques. Each model presented had a different
approach and results would vary in resolution. Whether computer aided design (CAD) or equation
based approaches were used, detailed information about the physical dimensions and design of the
machine were required in order to design an accurate model. Most of these models also required
short, open, and load based measurements to determine their parametric values. This can cause
damage to the motor if not done correctly. Much of the literature written on high frequency ship
power system models was focused on the inverter drive CM voltage and how to reduce this effect
using filters [15, 21, 29]. These system models were based on various approaches and normally used
simplified motor models such as π, RLC, or distributed ladder networks. Filtering can mitigate
these effects but models still need to accurately capture the frequencies of interest to build such
filters. Reducing the effects of high frequency through component design and physical location
considerations will also be a significant part of ship building. Accurate models will allow filters to
be designed specifically for the inherent response of the system. Also, the physical size of the filters
could be reduced and the placement of them in the ship system would be designed to lower the
impact of unwanted frequencies. After a thorough literature review on this topic, it was clear that
there were no universal methods capable of easy implementation for all types of electric machines.
The need for more accurate system models directed the research towards network based theories.
The field of communication and antenna theory relied heavily on a technique which standardized
their approach to obtaining measurements on various network components. This idea was proposed
to power system components and research in this area has been growing steadily [9, 30]. The
thought of using inverse scattering theory to determine or identify characteristics of power devices
has become a new field of exploratory research. This method has been applied to cables and even
three-phase EMI filters [30] which are typically considered non-linear devices. Early research in this
area first focused on simple geometries such as power cables [4, 31]. This provided a simple path
to determine viability for further implementation of other components such as electrical machines.
The standard approach to measuring devices based on the number of interface ports has proved
7
to be a powerful approach to simplifying the process needed to generate accurate and repeatable
models.
The originality of this research was to focus on applying inverse scattering theory measurements
to high frequency machine modeling. Until now, to our knowledge, no one has recorded or used
scattering parameters to model an electric machine. This required a large effort to be put forth
into understanding previous machine modeling approaches for inspiration. The measurements
were fairly easy to obtain in hindsight, but involved a substantial learning curve to understand
and prepare them for use in models. The combining of two separate fields of research (machine
modeling and scattering parameters) slowed the initial progress of model development due to the
lack of overlapping literature between the two topics. In the end, the work suggests to be a
significant addition to high frequency machine modeling approaches presented to date.
A simple methodology to develop highly accurate models could be used to better understand and
develop the ship system prior to installation [7, 4, 31, 32, 8]. The time saved and the knowledge
gained by the use of the proposed approach would translate into dollars saved in damages and
wasted time spent on other more intense modeling approaches. Previous modeling techniques
don’t offer the simplicity in model development and the ease of connection with other components
for full system modeling. The scattering parameters can also serve as a snapshot of the machine’s
parameters. These measurements can be used to monitor the current state of the machine’s health.
This could be used to study how aging and pollution on the windings of the machine can change
the machines characteristics.
1.4 Organization of Master Thesis
The manuscript begins with an introductory Section 1.1 where the motivation for high frequency
machine models which accurately address grounding for shipboard power systems was discussed.
In Section 1.2 an explanation of transient behaviors in shipboard power systems was covered. The
following Section 1.3 details the significance that the proposed work will have on machine modeling
techniques for future all-electric ship design. The main focus of this work is to fully characterize the
model of an electric machine using a scattering measurement technique previously used on other
components of the SPS with great success.
8
In Chapter 2 state-of-the-art research is discussed, which covers machine modeling approaches
to date and how scattering parameter methodology could be applied to model the entire SPS.
Section 2.1 can be further broken into the most common modeling approaches; lumped element
models in Subsection 2.1.1, distributed models in Subsection 2.1.2, and common–mode/differential–
mode equivalent models in Subsection 2.1.3. Section 2.2 will cover the scattering methodology
the linear time-invariant requirements for use in Subsection 2.2.1, the instrumentation needed
Subsection 2.2.2, and the different port determinations for setting up the measurement process
Subsection 2.2.3.
In Chapter 3 the scope of work was presented and the plan for conducting the necessary exper-
iments were organized. This chapter will also cover how the work will be accomplished, and how
it pertains to the motivation and significance of high frequency grounding models.
In Chapter 4 the proposed scattering parameter theory will be applied to three-phase electric
machines. This chapter is divided into three sections starting with linear time-invariant testing,
Section 4.1, then parameter extraction techniques Section 4.2, and finally virtual measurement
models Section 4.3.
In Chapter 5 the results from experiments are discussed and compared to conventional tech-
niques. This chapter will cover the importance of resonances and anti-resonances in Section 5.1,
short circuit high frequency response of an electric machine in Section 5.2, and Section 5.3 will
cover the results of virtual measurements and how they compare to impedance measurements.
The manuscript closes with Chapter 6, which discusses conclusions to this work and future
considerations are given to help expand the research further.
9
CHAPTER 2
STATE OF THE ART
This chapter will discuss the literature review that was conducted. The researcher chose to focus his
search to literature on high frequency electric machine models and scattering parameter based mea-
surements. The majority of high frequency electric machine models found to be useful for grounding
research are discussed in Section 2.1. The approaches found most useful include lumped element,
distributed, and Common-/Differential–mode models. A majority of the models found required
detailed information on the exact design of the machine followed by a series of measurements and
calculations. Some of the most accurate models focused heavily on parameter extraction, in some
instances, one section of the motor or captured one high frequency phenomenon in great detail. In
other high frequency modeling approaches, the literature proposed different testing procedure for
isolating specific parameters, which would be used to make a full model. The search found very few
approaches that could be used as a standard modeling procedure across multiple machine types.
The models were also developed with great insight coming from the researcher’s experience with a
specific machine. This was determined to be an issue because it made these approaches difficult to
implement with ease on other machines with differing construction characteristics.
Research and previous experience with scattering parameter based models was influential in
determining alternate approaches to obtaining high frequency models of electric machines. The re-
search goal of our group was to create grounding models based solely on one measurement method-
ology that could be applied to all components in the ship’s power system. Essentially the group’s
research on scattering parameter network methodology was determined to be plausible and worth
exploring for devices such as the electric machine. In Section 2.2 the scattering methodology pro-
posed is applied to electric machines. The theory behind linear time-invariant networks is discussed
first. This is followed by instrumentation needed for measuring the terminals of large power equip-
ment. Lastly the ports of the machine were determined and the proposed measurement process is
outlined.
10
2.1 High Frequency Electric Machine Modeling Approaches
This section will explore the current state-of-the-art research involving high frequency charac-
terization of electric machines. In addition this chapter will cover a wide range of machine research
including models that only focus on the characterization of machine windings based on equivalent
circuits all the way to complete physics based models. The models vary in accuracy as well as
modeling capabilities and resulting resolutions. The state-of-the-art research explores numerous
machine models and the varied techniques used to obtain the modeling parameters as well as the
frequency ranges of interest. It is pertinent that all forms of machine modeling be explored before
proposing a new methodology. The coverage of such a wide field of science allows for some con-
clusions to be made about the current state-of-the-art, and points to why the statement of work,
discussed in Chapter 3, is desired.
The need for accurate high frequency machine models has been expressed in Chapter 1 and is
agreeable with the current research thrusts by the Navy. Developing high frequency machine models
has been approached many different ways through prior art seen in Appendix A. Not only are the
models different by design but the methods to determine each parameter differ across this field of
research. This may have happened due to each approach attempting to solve a different research
topic or developed from research groups with differing backgrounds. In the next few subsections an
overview of the most common approaches to modeling and parameter extraction will be covered.
High frequency electric machine models have been developed and tested in prior work. Cur-
rently there are many machine designs that incorporate parasitic and high frequency effects, but
these models have their limitations. In some cases, the modeling efforts were developed based
on previous models in order to closely approximate the electromagnetic effects. These models
were limited because they were unique to a specific machine or focused on one test condition and
were not considered to be useful models for the current SPS modeling requirements. The work
discussed in this section has been reduced to focus only on the high frequency electric machine
modeling approaches found in literature. High frequency characterization of machines has been
accomplished typically in either a lumped element circuits found in Subsection 2.1.1, distributed
circuits in Subsection 2.1.2, or mode type circuits for separating out CM and DM behavior found
in Subsection 2.1.3.
11
The author has chosen to not further study any finite element based models because they are
dependent on a physical understanding of the inner dimensions, connections, and full details on
material properties of the machine, which are typically proprietary and unavailable to the engineer.
These models are obviously useful for machine design because they include all the physics to simulate
but this is the most time consuming approach. Typically finite element methods (FEM) are one-off
solutions, which require heavy resources to solve even simple geometries. It’s highly unlikely that a
data sheet on a machine’s design will accurately match each machines actual parameter values let
alone be available to the end user. It was determined not to be feasible for rapid model generation.
The aim of this work was to investigate and suggest an approach that can be universally followed
for any machine. Also, most FEM approaches eventually extract parameters to be used in a more
simplified circuit representation to reduce the computational burden [33]. These simplified models
are not guaranteed or expected to provide the same level of accuracy seen with the FEM. Therefore,
this approach has been summarized and excluded from further discussion.
2.1.1 Lumped Element Equivalent Circuit Models
The process of extracting parameter values from frequency data for equivalent circuit modeling
is a common approach to high frequency modeling for overvoltage studies [34]. The equivalent
circuit model approach typically starts with three per phase circuits that are electromagnetically
coupled. Each circuit is equivalent, assuming symmetrical machine construction. The parameters
are normally found via analytical calculations [35, 36, 37], slopes and resonances in impedance data
[13, 34, 38], or CAD software [23, 39].
Previously suggested measurement techniques have been considered useful in determining pa-
rameters of machines but are not always accurately implemented without in-depth knowledge of
the motor. Looking at Figure 2.1a one can see the equivalent capacitance for a standard capacitor
with lead inductances and leakage current included. The general model for the inductor can be
seen in Figure 2.1b where the series resistance and skin resistance are included in the model [2].
Lumped-parameter models are based on extracting information at transitional frequencies or
looking at the slope of the impedance as a function of frequency. These models tend to be simplified
representations of the component by generalizing the parameters based on the specific impedance
found at a few frequency points. It is common for these parameters to be extracted in frequency
domain and converted to a single lumped circuit component in the time domain. In practice the
12
(a) Generic equivalent circuit for capacitor [2] (b) Generic equivalent circuit for inductor [2]
Figure 2.1: Equivalent circuit models for capacitors and inductors [2]
machine’s impedance varies over all frequencies as should the parameter values to most accurately
model the machine. In Figure 2.2 the impedance plot is shown for both an inductor and a capacitor.
The dotted line indicates the impedance of an ideal component. Any deviation from the dotted
line represents the real impedance due to the parasitic effects of the physical components.
(a) Actual versus ideal capacitor [2] (b) Actual versus ideal inductors [2]
Figure 2.2: Equivalent lumped model plots for capacitors and inductors [2]
Lumped parameter based models can over simplify the grounding behavior of the machine.
Models have been proposed that neglect the second resonance while others build more elaborate
circuits to account for these missing behaviors [24]. Capacitive currents are an issue in ungrounded
shipboard power system behavior and should not be neglected in the modeling approach. For this
reason it is not recommended to use single–phase equivalent circuit models to describe three–phase
circuits when capacitive currents are strong [26]. Any variations in the potential of each phase could
13
impose coupling and ultimately ground current. This effect could further increase high frequency
power loss due to the harmonics and imbalances in the electric machine phases.
Through extensive research on the topic, models were also discovered that included the bearing
capacitances and their high frequency contributions to CM currents [24, 40, 28, 41]. This was very
interesting to find because many papers mention the effects bearings have on the capacitance mea-
sured but choose to neglect these effects in the final model. It was first discovered when literature
was found were standstill frequency response (SSFR) techniques warned that measuring a machine
at rest would undoubtedly change the capacitance once the machine was moving. Furthermore it
was noted that the grease formed an insulation layer between ball and race [42, 43, 28] when in
rotation. This time-variant behavior was considered when researching new measurement methods
for machines.
There are numerous approaches to measuring a machine that lead to parameter estimation
and implementation into various high frequency models. Many authors started building tables to
organize the different models found on these topics of high frequency electric machine modes. An
example of these tables can be found in Appendix A on page 62. The extensive list of models are
all valid approaches to modeling but lack the ability to be applied to other machine types without
significant effort.
2.1.2 Distributed High Frequency Models
Some distributed HF parameter models of electric machines have been proposed by other au-
thors, and they are briefly introduced here. The concept of distributed high frequency (DHF)
models is typically based on ladder circuits to describe the machines stator windings and slot area
[3, 36, 44]. These models can be indexed continuously in the frequency domain to represent the
resistance, inductance, and capacitance for modeling the effects of incident waves on the motor
windings. Each winding circuit can be made up many latter circuits to model the distributed effect
of voltage across the windings during fast pulses. These models tend to have higher resolution when
compared to lumped element models which typically catch only the self-resonance frequency. The
tradeoff comes with a more complicated model that can have upwards of 500 elements per winding
and the need for numerical models to calculate the computationally heavy model. Examples of
distributed models are explained in literature such as [36, 45, 46, 47, 20, 3] or seen in Figure 2.3
found in [3].
14
Figure 2.3: Distributed high frequency circuit model with 17 turns per slot [3]
In [36] the values of the distributed winding model are calculated via a matrix of equations
representing the various positions of each winding and where it falls in the slot of the stator. It
was common to find contradicting information among these models. For instance in [3] the author
breaks their work into simplified high frequency (lumped equivalent circuit) and distributed high
frequency models. The first circuit was said to be used if measurements were available and the
second set of models was used for predicting high frequency effects. The contradiction is found
in [45] where the distributed equivalent circuit model is called a simplified high frequency model
similar to the lumped models name in [3]. The one thing that seems to hold true across all the
models is a notion that the values can be calculated as a per unit length value. Interesting work
was conducted in [46] where electromagnetic interference was studied with and without shielding
to show the ground plane interactions seen in Figure 2.4. This work used a distributed and lumped
model using the same parameters as input to the model reducing the complexity of simulating high
frequency behavior.
In [47] the machine was measured over a wide frequency range and was broken up into non
iterative stages. Each stage was made of linear RLC circuits bringing a new meaning to distributed
modeling. This model distributed the model over various frequencies while other models were
distributed based on per unit length positions within the machine. Interesting enough more work
into network theory has started to surface [47, 3] where the former utilizes line-impedance stabilizing
networks and the latter uses scattering parameters for higher frequency measurements.
15
Figure 2.4: Parasitic coupling of the ship MVDC power system and ship hull ground [4]
Among the DHF models mentioned, it was clear that these models took time to develop and
required special fitting techniques with measured data for better parameter estimations. The
results from these models were much closer to their measured values that seen in lumped element
approaches. An issue was observed where the effort needed for parameter extraction from measured
results in each DHF algorithm increased with complexity the higher the frequency range of interest.
2.1.3 Common-mode and Differential-mode Equivalent Circuit Models
When describing a machine using high frequency parametric data there are three types of circuits
that seem to be most prevalent. The first is the lumped equivalent model which approximates values
for equivalent circuit representations of three-phase machines. Equivalent circuit components are
added to the measurement circuit with their own tuned frequency response. In these circuits the
researcher begins with a basic equivalent circuit model and expands the circuit to capture effects
seen through measurement or calculation. The second is when direct quadrature zero (DQ0) theory
is used to describe the machine at low frequency and a complimentary coupling circuit is used to
adjust for high frequency content [38, 34]. The third is the most commonly found high frequency
modeling method which uses common-mode and differential-mode circuits to simplify the behavior
of the machine as a function of frequency. In these circuits the effects of high frequency on the
machine terminals would develop either a common or differential current to form in the circuit. This
circuit representation is quite useful for studies because each mode can isolate specific problems
such as parasitics with the ground, which is relevant for our own studies.
16
To clarify, common and differential measurements are one method of measuring which can be
used to determine all other model parameters. For example, a machine can be both measured
and modeled in common-mode and differential-mode circuits. All measurement techniques can
be converted to CM and DM modeling equivalents and all modeling parameters can be derived
from CM and DM measurements. The main methods for obtaining these measurements will be
discussed further in Section 2.2.3 on Page 23 or seen in Figure 2.8. As mentioned before, some
of the distributed models used CM and DM circuits to isolate certain high frequency behaviors.
Their contributing currents and values were measured in the frequency domain in order to build
the model found in [47].
In [40] the most basic CM and DM models are built using lumped circuit equivalent models.
The simulation to measured data was not as accurate as other methods but showed how simple it
was to get a general model built that was able to capture most of the key resonances and seemed to
fit for extracting slope values of the impedance as a function of frequency. This is accomplished by
updating the values of capacitance and inductance to match the measured data similar to [3, 13]
where the model was fit more closely to measured results.
Literature found discussing CM and DM circuits did not always identify high frequency behav-
iors in machines. For instance in [48] the modeling of the machine was accomplished with a simple
RL motor model. The point to be made is that not all CM and DM models are high frequency
models. This paper describes a design for removing CM currents without effecting DM currents
but does not mention the accuracy of the motor model just techniques on how to mitigate the
unwanted currents. There were papers in the scattering parameter field that also used the terms
common and differential to describe circuits being developed [49]. In the definitions of differential
and common mode signals apply to the port of the network and are used to redefine the scattering
matrix into separate matrices so the device model has common ports and differential ports. This
method is used to develop a mixed mode model based on scattering theory. This is worth further
investigation to determine its merits when applied to motor models.
In [50], and more recently expanded in [51], the use of CM and DM terminology differs from
other high frequency motor models available in literature. A new improved DM cable model in
conjunction with a simple lumped element motor model was proposed in [40]. The motor model
parameters are extracted from CM and DM impedance characteristics. The equations used in this
17
model for calculating parameters from CM plots was useful in defining our own circuit models for
comparison and are detailed further in Chapter 4.
2.2 Scattering Methodology for Measuring Electric Machines
The use of scattering parameters are well documented in any radio frequency (RF) or microwave
textbook but can be briefly defined as the amplitude of incident, transmitted, and reflected waves
for steady-state stimuli from matched sources of sinusoidal electrical signals of a certain frequency
[52]. The parameters are complex because they illustrate how the magnitude and phase of input
signals are changed by the device’s network. Ports, or electrical terminals, are typically identified
as the inputs or outputs of the machine, where voltage and current can be delivered or transmitted.
The machine contains a finite number of ports that are subject to incoming waves as well as
interactions internally from other ports transferring energy. The common device used to measure
scattering parameters and the preferred instrument used in this research is the Vector Network
Analyzer (VNA). The VNA connects to the various ports using measurement cables and special
high frequency termination equipment defined further in Section 2.2.2 on page 20.
Scattering parameters contain a given range of frequency data for each port-to-port combination.
The number of ports, N , are defined by N2 coefficients representing a possible input or output path.
There are N rows and N columns of scattering coefficients in the scattering matrix. The diagonal
parameters are the reflection coefficients referring to single port reflections e.g. s11, s22. The off
diagonal parameters are the transmission coefficients referring to the interactions between differing
ports e.g. s12, s21. A complete scattering matrix is essentially a tool used to determine voltage out
compared to voltage in for the various port combinations on a machine.
The following sections will cover linear time-invariant network theory, the instrumentation
needed for conducting scattering measurements, and the determination of port configurations for the
measurement process. Port selection for an electric machine will be discussed using two approaches;
single–port and multi–port scattering parameter measurements. Single–port measurements are
closely related to commonly used machine measurements done using CM and DM circuits. The
multi–port measurement is the most complete measurement matrix, which contains all reflection
and transmission coefficients defined between each port and the reference ground. The benefits and
shortcomings of each method will also be discussed further in this chapter.
18
2.2.1 Linear Time-Invariant Theory
Scattering parameters are only relevant when the network being measured is considered linear
and time-invariant for the frequency range considered. This can be defined by the Equation 2.1.
The output signal F (x) follows the rules of superposition where its inputs can be broken into smaller
pieces and applied separately to receive the same output.
F (ax1 + bx2) = aF (x1) + bF (x2) (2.1)
Linearity is also true if the network’s parametric values don’t change as a function of the voltage
and current inputs. This is accomplished when the network contains only resistors, inductors, and
capacitors and is not operating in saturation where hysteresis, and eddy current are present [53].
For the purposes of machine modeling we will consider electric machines to be linear for the range
of frequencies being applied. Each frequency is fired independently in the VNA allowing each signal
to be measured independent of the next measurement point.
The second stipulation is whether or not the machine can be considered time-invariant. Time-
invariance can be defined as any input signal producing a consistent response; regardless of a shift
in time (considering the machine is linear for small time steps and is not operating in saturation).
Similarly, in [54] the electric machine was considered as a discrete-time linear time-invariant system
because the measurements were taken at stand allowing these theories to hold true.
A test to determine the accuracy of these assumptions was designed to capture machine mea-
surements during specific rotor positioning. The time varying aspect of a machine comes from the
interactions between stator and rotor. The machine would, in theory, see the largest effects of the
stator to rotor interactions during the passing of the poles of the rotor against the poles of the
stator. This test is further described in Section 4.1 found on Page 32 where multiple positions of
the rotor where measured to observe the overall impact in measurements. Even though a machine
is not truly time-varying, it is possible to use for small time steps and therefore scattering param-
eter method is considered a viable measurement method for the discrete-time standstill frequency
response measurements conducted in this research.
19
2.2.2 Instrumentation for Characterizing Machine Coupling to Ground
There are many parameters that can be used to define an electric machine and most instrument
methods require an open or short circuit condition to properly characterize these parameters.
The vector network analyzer (VNA) uses a matched load instead of a short or open termination.
The matched load has a wider bandwidth allowing for higher frequency measurements without
energizing the machine. A VNA was used to extract scattering parameters because high frequency
measurements are of interest when characterizing the machine’s coupling to ground. Scattering
parameters for high frequency modeling were used in previous research of other power components
such as power cables [31] and three–phase EMI filters [30] with good results. The basic equipment
used to perform scattering measurements included a two–port VNA, two phase–stable leads, a
calibration test kit, and as many termination clamps as there are ports present on the equipment.
In the proposed approach a two–port VNA is used to measure scattering parameters of a
three–phase machine following the techniques used in [19]. A VNA manufactured with two signal
generators and two receivers is used to measure amplitude and phase properties of transmitted and
reflected power waves. The VNA uses an internal direct digital synthesizer to produce a variable
frequency continuous wave source signal. Each signal will pass through the device with a swept
frequency. The device will have a unique response corresponding to each test frequency. During
each measurement step a portion of the incident wave will return to the sending port as a reflected
wave while the remaining incident wave will transmit through to other ports. This scattering
measurement records how the traveling waves interact with the various structures and electrical
paths within the machine. The measured response to these changes in scattered energy throughout
the device, allows for full models to be developed that describe the machine across a wide range of
frequencies. Similar to the way light reflects at certain frequencies giving our surroundings color,
the inverse scattering reflects at certain frequencies that can be used to define the device in an
electrically visual representation.
When performing inverse scattering theory the measurement plane of reference is preferably as
close to the device as possible to ensure only the device is being measured. The VNA measurement
plane has to be calibrated to ensure the measurements are reflections and transmission due to the
electric machine and not the instrument or the measurement leads. Cables reaching from the VNA
to the electric machine are subject to calibration tests using a known calibration kit, purchased
20
separate from the VNA. The calibration kit consists of high frequency rated short, open, and load
terminations to be used in a series of calibration steps. Following the steps given in the VNA
calibration guide, the measurement plane can be properly set to the ends of the measurement
cables or the terminals of the electric machine. Calibration is recommended to be performed each
time the VNA is used and more importantly every time the measurement settings are changed such
as number of ports, frequency range, frequency points, power level, and bandwidth. Calibration is
not difficult to administer and is comparable in setup time to other measurement setups.
The electric machine is connected to the measurement cables at two of its defined ports (port
determination is covered in Section 2.2.3). The remaining ports are connected to the ground
plane through a high precision wide bandwidth load resistor equal to the VNA system impedance,
typically 50Ω. Using known terminations, measurements of each two–port combination were later
assembled into one complete scattering matrix. The complete scattering parameter measurement
matrix is stored in the industry standard Touchstone [55] or CITIfile format [56]. The scattering
parameters are each denoted by snm and form to make an n×m square matrix where the n is the
measurement port and m is the sending port.
Each of the two–port measurement combination creates four parameters which are s11, s12,
s21, and s22 seen in Figure 2.5 on page 22. Where a1 and a2 are the incident wave measurements
and b1 and b2 are reflected wave measurements. The off diagonal parameters are the transmission
coefficients and the diagonal parameters are the reflection coefficients. If the electric machine is
directional in nature then it can be described as having forward port and reverse port signals and
more importantly snm 6= smn. For modeling, the electric machine it is not necessary to classify
them as having directional ports. It was common to set snm = smn to simplify the calculations by
assuming symmetry between the phases. When transforming scattering data from the frequency
domain to the time domain it is better to have symmetry in the off-diagonal positions. Extra
confidence in measurements can be obtained by measuring all reverse measurements and averaging
the off-diagonal matrix parameters [19].
Measuring a machine with four–ports will produce 16 scattering parameter measurements for
each swept frequency. This is based on the 16 possible port-to-port interactions between each
machine terminal when neutral is accessible or 12 when neutral is inaccessible. Two machine ter-
minals, at a time, are measured by the two test ports of the vector network analyzer (VNA). While
21
Figure 2.5: This is a visual representation of scattering parameters for a two–port network
the other two ports of the electric machine are terminated with a 50Ω single-ended termination.
The matched termination load acts as an infinite transmission line and absorbs the wave without
reflection. The ports of the VNA are also matched similarly to allow the measurements to describe
the machine exactly without reflections from the VNA ports. This procedure is repeated by alter-
nating the measurement cable leads through all possible port combinations on the machine. The
magnitude and angle of each signal is stored under the respective test frequency being sent.
Scattering measurements will paint a frequency response picture of the entire machine. The two–
port measurement data are stored and assembled into multi-port representations, building the full
scattering matrix of the electric machine. In Section 4.3, this scattering matrix serves as a virtual
measurement model. The model can be connected to simulated probes allowing for measurements
to be taken as if the researcher were measuring the electric machine directly. The scattering
vector measurements contain magnitude and phase data that can later be used for complete time
domain characterization, device modeling, and vector-error correction. The resulting parameter
measurements are assembled and can be used to determine reflected voltages using the matrix
Equation 2.2
V −
1
V −
2
V −
3
V −
N
=
S11 S12 S13 S1N
S21 S22 S23 S2N
S31 S32 S33 S3N
SN1 SN2 SN3 SNN
V +
1
V +
2
V +
3
V +
N
(2.2)
22
where the positive voltage sign denotes an incident voltage and the negative sign represents
reflected voltages. Further explanation on the scattering parameter procedures can be found in
[19]. The act of measuring an N–port device using scattering measurements is well documented
[4, 31] for simple geometries, but to the author’s knowledge this is the first time it has been used
to measure machines for ground modeling.
When measuring at frequencies in the megahertz range and higher, the series resistance is much
smaller. In-turn the phase angle is more susceptible to large fluctuations and the overall impedance
of the machine will vary accordingly. Meaning, when obtaining measurements, the equipment must
be calibrated and the recommended procedures should be followed closely.
Special equipment was designed to interface between measurement cables and the machine due
to the physically large terminals of the machine. VNAs use high frequency coaxial connectors of
Type–N for interfacing with typically smaller devices. This is not the only connector available but
it was the connector of choice in this research for its simple construction and availability. The new
equipment was designed, built, and its frequency limitations were characterized by measurements.
Moving the measurement leads caused errors well above the desired maximum frequency of 10MHz.
The losses and distortion in data seen above 10MHz were discarded similarly to other motor models.
Effects above 10MHz are considered unnecessary for consideration because their effects are minimal
in SPS.
2.2.3 Port Determination and Measurement Process
There are many ways to measure and describe a machine depending on the variables of interest
and the level of detail that is desired. The most common way to determine the ports of a machine are
to look at how it interacts with its surroundings. The machines electrical connections to the power
system and any nearby capacitive interactions can be observed as ports. The ports are typically
three–phase connections and one neutral connection, seen in Figure 2.6. The measurements of all
the ports must reference a common ground to be valid for use in the scattering matrix.
Typically a three–phase electrical machine is measured with different configurations to help
isolate specific parameters. The same concept can be applied when measuring scattering param-
eters. There can be anywhere between 2-16 scattering measurements needed to characterize the
machine. Some useful machine models are constructed using single-port impedance data extracted
23
Figure 2.6: Port representation for electric machine scattering measurements where T is for terminaland P is for port. T2P2 is terminal 2 and also port 2 of the measurement matrix. Each port of thematrix is measured by referencing it to a common ground.
from scattering data when measured in a common-mode (CM), differential-mode (DM), and phase-
neutral configuration (phase-neutral is considered DM impedance with neutral terminal accessible
[6]). These single-port measurements will be covered in Section 2.2.3 where each measurement pro-
vides information that can be used for modeling and has been verified using alternative equipment
(impedance analyzer).
The machine can also be measured as a multi-port model arrangement. Where each electrical
terminations, on the machine is seen as a separate port and a full set of scattering measurements
are performed and assembled into one complete scattering matrix. The full port representation of
the machine is detailed in Section 2.2.3 where the machine can be modeled directly from scattering
parameters taken at each port.
Deciding whether a single–port or multi–port method is better for a specific application, was
not determined in this work because both are valid and are often useful in different modeling
approaches. For the purposes of laying the foundation for this research, both methods were used
to form a complete measurement set. The measurement matrices can then be used as needed to
help build the machine model that best represents the machine configuration and the frequencies
24
of interest. The measurement setup for the PMSM machine seen in Figure 2.9 details the possible
voltage waveforms to measure at each port of the four-port network.
Single-Port Scattering Measurements. There are two types of port configurations often
used to measure impedance spectrum of an electric machine for parametric modeling and can
be seen in Figure 2.8. The CM and DM configurations are quite useful for modeling because it
allows noise to be isolated from common and differential sources. In Figure 2.7 the source voltages
are applied to the corresponding differential-mode and common-mode circuit [5]. The capacitive
coupling between phase-to-phase and phase-to-ground sections allow for a return path to be created
and the current produced is measured as the disturbing CM or DM current. CM and DM noise
produce unwanted conduction paths for current to flow. Modeling the electric machine this way
makes it practical for technicians to locate problems and mitigate component fatigue. Also this
measurement configuration is widely accepted in high frequency modeling literature as a key step
for parameter extraction.
Figure 2.7: Differential-mode and common-mode capacitive coupling [5]
The CM configuration is setup by connecting the three phases of the machine in parallel to
make a single measurement node. This node is then measured with respect to the machine case
or grounding point seen in Figure 2.8a. This measurement setup provides a frequency response
25
representation of the parasitic coupling between the machines windings and the case ground. If the
neutral terminal is available then another CM configuration can be seen in Figure 2.8b.
DM is similar to a single-port CM measurement but the terminal setup consists of two parallel
phases connected in series with the remaining phase. This setup will provide information about
the interaction between phases.
Phase–neutral measurement is achieved only if the neutral point of the machine is accessible.
The three phases of the machine are connected in parallel, like the CM setup, to create one node
and then measured with respect to the neutral point. This measurement is not necessarily needed
for parameterizing lumped models because it is a redundant DM measurement but it could be
useful in increasing accuracy of calculated parameters found using multiple CM and DM methods
[34].
Describing the machine in the three proposed single–port representations is considered a com-
mon approach to high frequency motor modeling [22, 34]. This approach allows for standstill
measurements to be made that help determine lumped parameter quantities of common machine
circuit representations. Using a VNA for this approach offers the same results found when using an
impedance analyzer but with higher resolution due to the ease of frequency sweeps with no delay
between measurements. In most cases the measurement process is less involved when using scatter-
ing parameter methods because it is standardized. This work presents an alternative measurement
approach that has the potential to be applied universally to other components in the system such
as the cables and power electronic devices. Having a similar measurement process and data post
processing method for all devices shows promise for future research.
Multi-Port Scattering Measurements. Subsection 2.2.3 discussed how to use widely ac-
cepted machine measurement protocols using a VNA instead of an impedance analyzer. In this
section the machine is measured as having at least three main nodes corresponding to the three–
phase connections and an optional fourth node for the neutral point. This would be the standard
approach to following the scattering methodology, where the device under test is viewed by the
ports in which energy in the system is transferred through to the device. The problem with this
approach is that it has not been verified for three–phase machines and no power systems modeling
software packages are using S-parameters for motor modeling are known.
26
(a) CM neutral inaccessible (b) CM neutral accessible
(c) DM neutral inaccessible (d) DM neutral accessible
Figure 2.8: CM and DM impedance measurement connections [6]
Scattering parameter software has recently been used in new ways to describe other power
equipment [31, 32]. It would be safe to assume that a release of new software packages in the
future will meet the demand for more complex networks like electric machines. This would allow
for full measurement sets to be used directly as the machine model. The model would be capable
of connecting with other components using the same measurement approach. In the single port
approach the machine was only looked at for the CM and DM operating modes. The single port
approach also lost accuracy when applied to a lumped model approximate circuit discussed in
Chapter 4 Section 4.2.
The multi–port measurement approach will need further research to determine its accuracy
as a complete model. However, it was found that this technique could be used to develop a
standalone model for simulating the single–port measurements termed ‘Virtual Measurements’.
This is described further in the Section 5.3 and Section 4.3. This method allowed for the full matrix
model to be probed as though it was being physically measured in all the possible configurations.
Measured scattering data converted to the time-domain from the frequency-domain using Fourier
transforms were not extensively considered because the measurements were still being analyzed for
their accuracy to capture the machine at frequencies above the operating point. In this research
27
it was found that transforms must operate on uniformly sampled data that extends across the
entire frequency range; 5 Hz to 10 MHz. Scattering parameter data can be captured using either
a linear or logarithmic frequency sweep. Sampling is valid only if the amplitude and phase charac-
teristics are adequately sampled to capture all salient frequencies. Logarithmically swept data can
sometimes be used, but linearly swept data is always preferred.
Figure 2.9: Voltage waveforms corresponding to the 4-port PMSM network with Z0 = 50Ω forboth VNA cables and both single ended terminations
The proposed scattering parameter measurement approach can be found in the companion paper
[32] explaining how measurements are to be conducted on shipboard power system components.
The network parameters defining reflection and transmission coefficients for measured frequency
vectors of n-port devices is defined in [57]. The measurement setup for the PMSM machine seen
in Figure 2.9 which details the voltage waveforms measured at each port of the four-port network.
There are four reflected signals, one for each port depicted in Figure 2.9. The other 12 signals needed
to assemble the scattering matrix describe the four-port internal network interactions between each
port. It is not certain if any of the measurements can be discarded by assuming the matrix to be
symmetrical because each machine type is constructed differently. If the machine is symmetrically
designed then it is possible to average redundant measurements to make the matrix symmetrical
and take advantage of mathematical transforms.
28
CHAPTER 3
PROBLEM STATEMENT
The current process to design models using current available methods are time consuming and
require special circuit representations and detailed information about the equipment, which is typ-
ically considered proprietary. The models are commonly found to be unique to one problem or can
only give insight into a specific phenomenon while neglecting others or deeming them insignificant.
Characterization of different type electric machines usually requires multiple test measurements to
isolate each parameter and are not always applicable with other machines. The process to develop
such models can exhaust research hours and provide no standard approach that can be easily im-
plemented on all machines. The models to be developed are not guaranteed to be accurate due to
assumptions made with each step of the available modeling approaches. Even the more accurate
models that describe the physics and material properties of the system are not ideal because they
are difficult to solve with current software. Also, in most cases the motor information is proprietary
or the information given lacks enough detail to fully depict the machine’s construction geometry.
These models are also prone to solver issues, and can be more computationally demanding versus
a scattering type model.
One of the main objectives in the scope of work is to research the prior art which is outlined
in Chapter 2 Page 10. The goal of this work is in determining the validity of proposing a new
measurement approach to capture the high frequency behavior of an electric machine. A secondary
objective is to use techniques found in scattering research and apply them to electric machines
found in our facility. This will require the machines to be validated under the network theory
principles which govern scattering theory i.e. linear and time-invariant, which will be discussed in
4.1.
Another objective is to compare the results from scattering measurements to other known
measurement techniques. This requires managing the data, converting, and comparing it to known
characterization approaches seen in 4.2. The measurements are expressed in impedance parameters
29
instead of the native scattering parameters to utilize prior art techniques of extraction and allow
for more definitive comparison between the various measurement approaches.
The concept and process of conducting virtual measurements will be formulated. The virtual
measurement model is created from a physical electric machine measurement using inverse scatter-
ing methods. The data is assembled into an equivalent multi-port scattering matrix, which defines
the electric machine as having ports for each terminal with respect to a common ground. Each port
in the simulated model becomes a measurement point or load to be used for various experiments.
Scattering parameters can be converted to many different network parameters with impedance
being the most useful to this body of research. This unique approach to measuring machines is
further discussed in Section 4.3. The proposed statement of work is captured in Chapters 2 and 4
and results from select experiments are discussed in Chapter 5.
30
CHAPTER 4
INVERSE SCATTERING THEORY APPLIED TO
THREE-PHASE ELECTRIC MACHINE MODELING
Primarily due to the complex geometry, non-linearity, and time variance of machines; state of
the art research has yet to explore inverse scattering theory (IST) for machine characterization.
The work summarized in this chapter will serve as the initial foundation for researchers to follow
when modeling electric machines using IST to capture the frequency response. The extent of this
research was confined by available equipment, time, and current state of the full SPS model. Typical
approaches to measuring machines are based on voltage and current measurements, which are not
suitable at high frequency. The scattering methodology applied in previous research of power cables
can be utilized to set a foundation for this work [31].
Scattering parameter methodology is desired because the universal measurement process for
developing models is well documented in [19] and provides the researcher with a structured approach
to modeling all the components in the system. Currently, there is a disconnect between what
researchers can perform in a lab and what can be performed in the field. Equipment specifications
are typically limited to what is found on the nameplate. This limits the options for high frequency
modeling approaches. IST can be used to capture information about the internal interactions of the
machine. If the machine is repaired or replaced on a ship the typical control models in place would
no longer accurately model the new changes. Using the scattering parameter approach allows for
any device to be quickly described as an N–port device and can be connected to other devices.
This modeling method is simple enough that a technician could generate a model directly from
measurement. The measurements are transferable in the Touchstone format to any simulation
software package. Once the model is set to accept scattering parameter measurements, it can be
maintained by updating the measurement data when the system configurations change. To grasp
this concept, imagine each component described by an electrical image that captures multiple
snapshots of the frequency responses of the machine. Each machine model is based on a frequency
sweep of snapshots that can be retaken at any time to maintain the most accurate models. Also
31
the model can be used in other forms such as CM, DM, or even full three–phase configurations for
virtual measurements.
This chapter will cover how scattering parameters could be adopted and how this approach
compares to other measurement techniques when it comes to developing accurate high frequency
electric machine models. The Instrumentation and measurement setup is critical to receiving valid
data and was discussed in detail in Section 2.2. An experiment was conducted to validate the
assumptions that a motor can be considered to be a linear and time-invariant (LTI) network, a
prerequisite for applying LTI network theory. The stator to rotor interaction, which tests LTI
compliance, is further detailed in Section 4.1. Parameter extraction for a lumped circuit model is
discussed in Section 4.2. It was also found that measurements could be used to build a full model
of the machine. The simulated model, detailed in Section 4.3, can be virtually measured as though
the machine were physically present.
4.1 Validity of Linear Time-Invariant Networks Applied to
Rotating Machines
Scattering parameters by definition apply to linear time-invariant (LTI) networks covered in
Section 2.2.1. Early research was set out to verify scattering theory as a valid approach to measure
electric machine’s high frequency behavior. An electric machine is typically described as having a
time varying field due to the interactions between the stator and rotor. An electric machine is also
considered non-linear due to the hysteresis found in the back iron [53].
Electric machines are technically not considered LTI, but the inside of the machine is made
up of easily measurable impedances such as resistors, inductors, and capacitors. Literature has
shown that non-linear devices have been modeled using scattering measurements in [39]. Also the
LTI network theory was proven to be capable of accurately measuring the impulse response of an
induction motor at standstill for [54].
Time variance was expected to be caused by the interactions between stator and rotor. An
experiment was setup with two similarly rated power machines, one with a smooth rotor and one
with a cogged rotor. Differences were noted in the physical resistance to rotating the synchronous
cogged motor shaft when compared to the asynchronous machine’s smooth rotation and lack of
cogging. It was hypothesized that measuring the two machines would generate a non significant
32
error between cogging positions. If any error would be seen it would likely come from the cogged
rotor and not from the motor with no cogging resistance. The complexity of mapping the scattered
incident wave as it reflects and transmits through the machine made the mathematical approach
unreasonably difficult. It was determined that the time-variant conditions could be tested by
analyzing the scattering measurements as a function of rotor position. The measurement setup
can be seen in Figure 4.1 where the VNA is using N–type connectors and measurement leads to
interface with the machine.
Figure 4.1: N-connector shield setup for scattering parameter measurements on electric machines(Left) using a vector network analyzer (Right)
In [16] it was concluded that impedance response measurements had substantial error due to
rotor position effects below the first resonance frequency. The test was setup so a comparison
could be made when varying the rotor position during stator measurements. Three rotor positions
were chosen, 0, 270, and 345, corresponding to three different cogging positions of the machine.
Scattering parameters were measured at each cogging position on a 250W brushless DC motor,
which has a similar design as a permanent magnet synchronous machine. Figure 4.2 is the resulting
deviation in scattering results due to rotor position. The maximum had only a 0.4 dB difference
found in the lower frequency ranges given by Equation 4.1. The measurements from initial testing
were scrutinized for errors and determined to be accurate representations of the machine with the
deviations less than 3 dB, thus the scattering measurements are considered valid measurements.
max(S110, S11270
, S11345)−min(S110
, S11270, S11345
) ≤ 0.4dB (4.1)
Measurements were also carried out on an asynchronous machine and the deviations were also
found to be insignificant with a maximum value less than 0.025 dB seen in Figure 4.3. Therefor, the
33
rotor position does not have a significant effect on the measured results. However, measurements
made with the absence of the rotor did affect the frequency response below the first resonance of
the common–mode impedance plot Zm and it had less of an effect on the phase–neutral impedance
plot Z0n confirmed in [24].
Figure 4.2: Variation of port 1 reflection (s11) of the brushless DC motor or synchronous machinewith cogged rotor
When calibrating the measurement equipment, it is impossible to make a perfect short circuit
required, as there will always be some inductance measured in the short. It is also impossible
to make a perfect open circuit, as there will always be some fringing capacitance. Most error
in these forms can be neglected to some degree if the majority of the impacted frequencies are
further removed through proper calibration prior to making measurements. The model number
of the calibration kit is entered into the VNA to ensure proper settings when using that specific
calibration set. The calibration was not able to remove signal noise from the power mains found
at 60Hz. Suggestions for error reduction are given in Chapter 6 conclusions. In comparing our
34
Figure 4.3: Variation of port 1 reflection (s11) of the induction or asynchronous machine withsmooth rotor
results to other published work we found that our data was erroneous at points well below the
first resonance. Depending on the port configuration being measured, the machine was considered
to be either heavily dominated by the parallel capacitance to ground or dominated by the series
inductance of the machine windings . This is where lumped circuit models of machines approximate
the lower frequencies to a specific reactance value. The reactance can be found using the slope of
the impedance in the frequency domain and solving for the corresponding capacitance or inductance
in Equation 4.2.
C =1
ωImZ ; L =ImZ
ωwithω = 2πf (4.2)
Even though the VNA available for research was capable of capturing frequencies as low as
5Hz it was found that the measurements were corrupted by the noise floor on the VNA below a
few 100Hz. Later it was determined that this problem originates in low signal power and could
have been avoided if known. It was good to stumble into this problem because we researched
ways to correct the missing data and learned how the simulation software handled noisy data.
Our data was considered corrupted in the frequency ranges where our measurements were effected.
35
Two techniques are typically used in the proposed method for extrapolating to the DC point or
0Hz. Constant extrapolation uses the closest known point to determine the next point or linear
extrapolation which uses two points as a reference for calculation.
It was concluded that for our purposes, the lower frequency measurements were not critical to
our models performance. Since the research was focused on modeling the interactions between the
machine and ground, our lower frequency limit was moved to 1 kHz and our upper bounds were set
at 10MHz. These frequency ranges were considered acceptable for developing the grounding models
of three–phase electric machines [50, 13, 40, 58]. For frequencies well above the power frequency,
the motor can be considered linear and the time variance at lower frequencies was mitigated by
raising the lower frequency limit to an acceptable frequency where the effects were negligible [59].
Using the slope of the impedance at 1 kHz allowed for extrapolation to approximate the DC point.
The measurement range chosen will capture the leakage currents [60] or CM currents [58] found
oscillating between 100 kHz and 10MHz.
4.2 Parameter Extraction for Lumped Circuit Model
The proposed approach to measure a three phase machine will produce either single-port or
multi-port measurements across a given frequency range. The multi-port measurement is the
preferred method because it contains all the data necessary for multi-port modeling capabilities like
virtual measurements discussed in Section 4.3. The process of extracting parameter values from
frequency data for equivalent circuit modeling is common approach to high frequency modeling
for overvoltage studies [34, 35], etc. The equivalent circuit model approach typically starts with
three per-phase circuits that are electromagnetically coupled. Each per-phase circuit is equivalent
assuming a symmetrical machine construction (Figure 4.4). The parameters are normally found via
analytical calculations, slopes and resonances in impedance data, or FEA software. In some cases
the winding and slot geometry were used separately to calculate more accurate representations of
the windings [35].
The process to determine each parameter and the overall accuracy of the model is highly de-
pendent on the circuit chosen to represent the device. This approach insist that there is knowledge
about the internals of the machine and specific details not typically released by the manufactures
of the machine. In research it is common to see work done based on an abundance of information
36
Figure 4.4: Per-phase representation of the induction motor including the high frequency modeland the dynamic dq model
about the machine construction [44]. One can argue that a high frequency model could be general-
ized based on a much simpler machine type, which would loosely fit measured data like the phase
model in [38]. The problem with this thought is that small errors in the frequency domain can
result in large errors in the time domain.
Using scattering parameter data is preferred by the author because it is based specifically on
the device being measured. In the IST approach the researcher develops a circuit model based
on the data extracted instead of measuring the device to fill in parameters of a general machine
model. It is safe to assume that a machine’s construction and parameters are unique to the
designer. The modeling approach should accommodate this uniqueness by first measuring the
device, and second determining circuit representation that best fits the data. This approach is the
least favorable because one must use transforms to breakdown the frequency data into separate
pieces with equivalent capacitance, inductance or resistance. This approach was found to be fairly
accurate but still considered a tedious approach.
The equivalent per-phase circuit in [34] and seen in Figure 4.4 was designed. Special equations
were made to help with parameter extraction. These equations are unique to this model and are
not universal to other machine types. The extraction Equations 5.15-4.9 will complete the high
frequency portion of the model, while the low frequency is controlled via the commonly used dq
model.
37
Figure 4.5: Common-mode impedance of a lumped parameter model compared to impedance ana-lyzer measurements and scattering parameter model representation
Cg ≈ 1
2
(
1
3
)
1
(2πflow)‖Zpg‖flow(4.3)
Rg ≈ 3xReZpgfhigh (4.4)
Ld ≈ 2
Cg
(
1
2πfpole−Zpn
)2
(4.5)
Re ≈ 3x‖Zpn‖fpole−Zpn(4.6)
Ct ≈Cg
10(4.7)
Lt ≈1
Ct
(
1
2πfzero−Zpn
)2
(4.8)
Rt ≈ 3xReZpnfzero−Zpn(4.9)
The equivalent model common-mode impedance was plotted excluding the dynamic dq branch
and compared to impedance measurements and scattering parameter model results. The equivalent
model clearly misses the higher order behavior, while the scattering parameter model has high
fidelity and tracks the actual measurements with little error (Figure 5.11). The most detailed
modeling approach found [44] covers just the inductive portion of the machine from 20 kHz to
20MHz.
38
4.3 Virtual Measurement Model
Electric machine models based on multi-port scattering data are capable of virtual measure-
ments. Scattering measurements from the machine were assembled into one complete measurement
matrix and brought into the Agilent’s Design Software (ADS). The scattering data was constructed
in a four-port model matrix and placed in the software for simulation. The probes from the simula-
tion software were placed on the terminals (now referred to as ports) of the electric machine model
to mimic typical measurements made on physical machines. The probes can act as an impedance
analyzer and return impedance information drawn from the measurement model or the model can
be wired to match specific designs and measured for the resulting characteristics. Previously dis-
cussed measurements of phase-neutral, common-mode, and differential-mode are virtually measured
and the model connections can be seen in Figures 4.6-4.8.
Figure 4.6: Single–port scattering parameter measurement to obtain common–mode impedanceinformation
Figure 4.7: Single–port scattering parameter measurement to obtain differential–mode impedanceinformation like T1 + T2 to T3
The results were found to be identical with the actual impedance analyzer measurements taken
on the machine. The impedance plots in Figure 4.9 capture more resonances and anti-resonances
than found by using any equivalent circuit, finite element analysis (FEA), frequency response
39
Figure 4.8: Single–port scattering parameter measurement to obtain phase–neutral impedanceinformation
function (FRF), or stand still frequency response (SSFR) techniques given in previous literature
[50, 23, 39, 42]. This is a significant step in the direction to develop accurate models for future
all-electric ships, considering the accuracy and the effort needed to develop such models. The
computation method to combine scattering measurements into one matrix model, discussed further
in [19], can be left in the default state and only need to be considered if any of the reference
impedances entered are complex. Complications with making accurate scattering measurements
for building models of ship power system components are discussed in [61] and used in conjunction
with the work in [62].
Virtual measurements of CM and DM impedances were significant parts of my research The
important concept realized is how much information was captured by scattering measurements and
how it can be extracted by different modeling techniques. This approach has significant advantages
considering that all devices in a shipboard power system can be connected to a VNA and measured
in the same protocol with differences only in the amount of ports per device. The component
connections can be configured in different combinations and probed for measurement information.
One application relevant to our work is determining the grounding of the shipboard power system
to the ship hull. The simulations could be configured in the various grounding arrangements of
interest. Then each simulation would be scrutinized and ultimately suggestions would be passed
to the final design.
Another application that is significant but most likely overlooked is the sharing of component
models. The physical machine does not need to be present in the lab for researchers to administer
measurements and configuration testing. A machine could be measured by the manufacturer in
40
(a) 2.5MW Squirrel-cage induction machine with neutral accessible
(b) 17kW Permanent magnet synchronous machine with neutral accessible
Figure 4.9: Virtual measurements of impedance for phase-to-neutral, common-mode, anddifferential-mode are compared to measurements made on the physical machine
41
standard scattering format and distributed out to various researchers, all working on different
simulation tests. The machine’s performance can be tested without ever leaving the manufacturers
factory. Ship power system components are large and shipping cost can be unreasonable when
shipped to labs individually. The model can be distributed and used as a simulated device in
hardware-in-the-loop testing.
When a specific impedance profile is needed, the scattering parameters could be chosen first
and the motor could be designed to fit the desired parameters. These design requirements can be
passed back to the manufacturer where the machine could be reconfigured to fit approved scattering
profiles. This concept can be seen in commercial lighting for plants, where a specific frequency is
desired for stronger yields. The researchers would determine the appropriate wavelengths of interest
and the manufacturer would design accordingly.
42
CHAPTER 5
RESULTS AND COMPARISON TO
CONVENTIONAL TECHNIQUES
The results discussed within this chapter will be added to the foundational prior work towards
alternative grounding models using IST. There were three main research experiments that were
conducted to explore this work further and they will be discussed in the next three sections.
The first important discovery was knowing that the VNA measurements were giving valid IST
measurements. The VNA captured the unique behavior of multiple electric machine types using a
universal measurement procedure. The resonant and anti-resonant behaviors were captured with
high fidelity and are further discussed in Section 5.1. The next insight came while measuring
the permanent magnet synchronous machine (PMSM) in the lab and devising a measurable event
to capture the electric machine’s high frequency response to transients. A short circuit test was
designed and prepared for testing the response of the motor when a terminal voltage collapses
to zero due to a short circuit. The short circuit test was used to capture a measurable event to
compare simulation approaches. The results from the short circuit model and measurement model
are discussed in Section 5.2. In Section 5.3, a proposed modeling approach opens up the possibility
for virtual measurements. Virtual measurement models are capable of being expressed in terms
of scattering data. The virtual measurements are useful in capturing parameter information when
setting up simulations. The benefits and results from these models are discussed towards the end
of Section 5.3.
5.1 Resonance and Anti-Resonance
It was convenient to convert scattering parameters to impedance parameters in order to compare
measurements conducted by other approaches. Typically these impedance plots are inspected
manually to determine inductance and capacitance. The calculated magnitude of common-mode
impedance can produce estimated values for inductance and capacitance by measuring the slope
before and after the anti-resonance point. This can be visually spotted because the plot follows
43
a ‘V’ shape where the local minimum marks the anti-resonance frequency. This shape holds true
for most models found in prior art. The more accurate and detailed the model is made the better
chance the model will capture all the effects seen by resonances and anti-resonances. It was found
that the lowest order models can still capture the global minimum impedance, which is at the
lowest frequency point in the plot where the impedance slope caused by the capacitance found in
the lower frequencies and the slope in impedance from the inductance in higher frequencies cross,
causing resonant behaviors of interest.
Figure 5.1: Induction motor impedances measured using IA and VNA measurements
Looking at Figure 5.1 the magnitude of CM data is noisy in the lower frequencies. As the fre-
quency increases, the measurement fidelity was much more accurate between the two anti-resonant
points at 35 kHz and 1.0MHz and alternated between capacitive and inductive behaviors.
The machine is mainly capacitive in nature at frequencies below the first resonance. The first
resonance point can be used to calculate the CM inductance of the machine. The coil is much higher
44
and reduces the effect of parallel capacitances found closer to the neutral point of the machine.The
CM impedance also contains a global minimum considered the anti-resonance point, where the
lowest impedance is found. The anti-resonance also is a frequency at which the resistance can be
calculated. This frequency point is to be noted when developing shipboard power system layouts,
because CM noise passing through the machine at this frequency has a low impedance path to
ground.
The common-mode impedance is usually modeled by two capacitances and in some cases only
one. It is clear though from Figure 5.1 that a second order model is not accurate enough to
capture all the parasitic effects. The impedance generally starts between 1-10 kΩ and decreases
with frequency until the first resonance based on the capacitance to ground. The inductance due
to skin effect will increase the impedance until the second resonance and then continues decreasing
as a function of the combined ground and turn-to-turn capacitances. This is a generalization of the
impedance plots for typical machines. The actual number of resonances and anti-resonances can
vary based on the machine type. Having a universal measurement protocol and model development
procedure would allow for fast model building and tends to have higher accuracy based on this
work.
The anti-resonance point will generally happen in the higher frequencies where the machine is
more strongly coupled to the ground. This coupling adds an inductive path between the phases
and ground at an anti-resonance point, where the lowest local impedance path is present. If a
frequency in this range were to perturb the machine it would have the least impedance path and
could ultimately cause damage. Setting boundaries with minimum impedances for each machine
or avoiding specific frequencies should be considered when designing the SPS. If we remove all low
impedance paths to ground then we can avoid or maintain the ground coupling issues. Adding
resistance can also raise the bottom line on impedance but this can produce unwanted losses. A
low impedance path can cause partial discharge or degradation of the insulation material. In severe
cases this can manifest into a short circuit path. Looking at the graphical representation of these
measurements will help designers to choose specific operating frequencies and reduce the risk of
problematic resonances with voltage escalations.
A unique finding was brought to our attention in the results around 1.0MHz. It seems as
though all three measurement types have a minimum impedance near this frequency. This needs
45
to be explored further to determine the cause and whether this can be used to build better models.
The DM measurements approach their resonant impedances at frequencies lying to the left and
right of the CM’s anti-resonant frequency.
5.2 Short Circuit Test
In a MVDC ship board power system it is more likely that a fault will happen on the DC
network and not the AC due to the percentage of the system that is DC versus AC. In the actual
testing of the effects of faults, the short circuit path should be faulting the DC side of the inverter
not the phase winding of the AC motor or generator. The purpose of the test described herein
was to capture the parasitic characteristics of the motor itself without the added coupling of the
inverter and cables. In this test the motor is isolated from all other effects and can be modeled
as such. This would create a better representation of the motor’s true response to high frequency
effects. The focus is not on system modeling but rather machine modeling for this test.
The electric machine model should characterize a wide frequency spectrum in order to pre-
dict accurately how the machine will act in the shipboard power system. The model is initially
developed using the conventional method of an equivalent circuit (EC), based on parameterized
lumped elements, to help benchmark the accuracy of the test [25]. The equivalent T-circuit has
three phases consisting of two resistive-inductive (RL) components split by a capacitance to ground
on each phase and a wye connected neutral. The RL values are to represent the resistance and
inductance of the machine windings while the capacitor represents the equivalent parasitic capac-
itance to ground. The common mode impedance is measured from the equivalent T-circuit where
the terminals are connected in parallel and measured with respect to ground (Figure 5.2).
The electric machine model was subjected to a test designed to characterize the high frequency
response without connecting an inverter or long cables which could misrepresent the actual response
of the modeled PMSM. The simulation was set up such that the neutral was ungrounded and
the machine was left unloaded with one phase equipped to fault to ground through a switch.
The unloaded generator will produce alternating phase voltages that are functions of the rotor
revolutions and the number of pole pairs. The parasitic capacitance between the windings and
frame will provide a low impedance path to ground at higher frequencies. If a phase is shorted
to ground then the reflected voltage wave back into the machine will contain a wide spectrum of
46
Figure 5.2: Common-mode impedance of a PMSM equivalent T-circuit
frequencies corresponding to the fall time of the short. The phase voltage short to ground will
produce frequencies that correspond to the inverse of the fall time tshort (Equation 5.1). At these
higher frequencies the low impedance path will cause a transient current into the ground path.
fshort = 1/tshort (5.1)
For simulation the capacitors and inductors are given an initial voltage and current correspond-
ing to the phase voltages and currents of an unloaded system prior to fault time tfault. The short
is characterized to match the experimental fault waveform by using a controlled voltage source.
Sending the experimental waveform into the simulation will better match the switch fall time and
provide a more accurate response from the model seen in Figure 5.3.
The model parameters for the T-circuit approximation are given in Equations 5.2-5.15. The
capacitance was calculated at 1 kHz from the phase to ground impedance measurement. The induc-
tance and resistance was calculated at 10MHz from the phase to neutral impedance measurement.
Each phase is modeled by a T-circuit with the phase inductance and resistance split by a capaci-
tance to ground. The neutral is wye connected without any source voltage because the capacitors
are initially charged.
Vrms = 480V (5.2)
47
Figure 5.3: PMSM fault current simulation using a measured fault to trigger a controlled voltagesource coupling terminal T1 to ground and phase T2 and T3 unloaded.
Vpk = Vrms ×√2√3= 391.92V (5.3)
Vc1 = Vpk = 391.92V (5.4)
Vc2 = Vc3 = Vpk ×−1
2= −195.96V (5.5)
tfault = 2× 10−9 s (5.6)
tsim = 1× 10−6 s (5.7)
∆ttol = 1× 10−9 s (5.8)
tstep,max = 1× 10−8 s (5.9)
C =−1
2× π × f × ‖Zpg‖= 1.16× 10−8 F (5.10)
L =‖Zpg‖
2× π × f= 1.95× 10−6H (5.11)
R = ReZpn = 24.66Ω (5.12)
Rs =3
2×R = 36.99Ω (5.13)
Ls =3
2× L = 2.93× 10−6H (5.14)
Cg =C
3= 3.86× 10−9 F (5.15)
48
The physical experiment involved an unloaded 8 pole PMSM rotating at nominal speed and
setup for a terminal fault to ground via solid state relay (SSR). A virtual push button in the Real
Time Digital Power System (RTDS) controls the initiation of fault logic when pressed and stays
on for greater than one full cycle i.e. 1/60Hz = 16.67 s. The terminal voltages are measured with
a differential probe that has CM rejection capability for an isolated measurement. The voltage
of terminal T1 is relayed through to the RTDS program called RSCAD. A push button in RTDS
controls the initiation of the fault logic. Fault logic checks for next positive edge zero crossing,
calculates a 90 phase shift, and sets the terminal fault signal initiation process at the peak voltage
Vpk of terminal T1. RSCAD then initiates the fault signal to output an analog 10VDC signal to
activate the SSR and trigger an oscilloscope to capture the transient response. The full setup of
the physical test performed can be seen in Figure 5.4.
Figure 5.4: Measurement setup for a single phase to ground fault of a three-phase PMSM
The results can be seen in Figure 5.5, where the top plot is the fault current and the bottom
plot shows terminal voltages T1, T2, and T3. The transient fault current peaked at 1.4A for 1 s.
Terminals T2 and T3 rise to 3 times their voltage before the fault and settle out at twice their
voltage 80 s after the fault.
49
Figure 5.5: Single phase to ground fault of a three-phase PMSM
Figure 5.6: T-circuit simulated fault current
50
The modeled fault was found to be inaccurate and displayed a peak current more then 5.5A
and lasted less than 0.2 s, seen in Figure 5.6. The impedance plots in Figure 5.7 shows where the
equivalent circuit does not capture the higher frequencies that the VNA captures.
Figure 5.7: PMSM measured impedance compared to simulated T-circuit equivalent
5.3 Virtual Measurement Comparison to Impedance Analyzer
The virtual measurement model was not a goal for this research but it was found to be a
useful tool for delivering model data in a form that allowed quick model information. This concept
was found when data was converted from scattering data to impedance data. The data captured
across all frequency measurements was found to be highly accurate to the true impedance at those
corresponding frequencies. This meant a motor could be measured using data in the scattering
form. An interesting thing was observed when the terminals were tied together in the CM and
DM configurations leading to a new measurement set later used for virtual measurement models.
These models are capable of extracting the impedances with high accuracy. In ADS the model was
considered a four–port network. The results can be seen in Figures 5.8, 5.9, and 5.10 where the 3
port configurations were discussed in Subsection 4.3.
51
Figure 5.8: Virtual measurement of common–mode impedance
52
The common-mode impedance Z0g was measured from a full four-port scattering matrix. The
results were analyzed at the lower frequencies and found to match the capacitance measured by
the impedance analyzer. There are distinct frequencies in the data where three points of interest
stand out, one at 100kHz, and the other two at 3 MHz and 7 MHz. These frequencies of interest
are useful for modeling and should be measured in detail. This was difficult to accomplish with the
impedance analyzer (IA) which manually stepped each frequency. This is why there are less data
points seen in Figure 5.11. The scattering results were at much higher resolutions due to the ease
in linearly sweeping the test frequencies for more than ten thousand data points. When compared
to the IA measurements which were difficult to transition and manually record each point. It was
found that 10 - 20 IA measurements per logarithmically spaced frequency sections were enough
to infer that both were valid approaches. It was clear after running multiple experiments that
scattering parameters were obtained faster with higher resolutions and less effort.
Figure 5.9: Phase–neutral mode impedance data was measured virtually
53
In Figure 5.9 the typical inductive behavior is captured before the first resonance which is
found at approximately 200 kHz. The typical phase pattern was also captured in this figure, where
early impedances are not effected by the added inductance. Skin effect takes place and brings
the phase towards the highest inductive behavior before transitioning to capacitive behavior due
to the capacitance between turns in the windings of each phase. In the megahertz and higher
regions, the machine transitions multiple times from capacitive to inductive. Capturing all these
transitional frequencies in detail allow for accurate models when using scattering measurements.
This measurement is considered another form of DM impedance measurements and is comparable to
the more accepted form found in Figure 5.10. The interesting effect is seen when you compare both
of these DM methods. In the Zm plot the same phase characteristics were found at lower frequencies
starting inductive and then drop to a lower inductance when the series resistance was introduced.
The plots have similar visual features such as the shape of the resonances and anti-resonances.
The difference is that they occur at different frequencies due to the terminal connections used to
measure signals passing from one terminal phase to the opposing parallel phase combination.
Taking a further look at the results in Figure 5.11 it is clear that the VNA was more dominant
and covered more frequency points with higher accuracy than the IA. This does not mean that
IA measurements are incorrect but it was found to take much longer to measure each individual
frequency. Newer models of IA may have corrected this but the current model that was used lacked
a quick measurement operation.
54
Figure 5.10: Differential–mode impedance data was measured virtually
55
Figure 5.11: Impedance data was virtually measured and compared to known measured data withhigh accuracy
56
CHAPTER 6
CONCLUSIONS AND FUTURE CONSIDERATIONS
6.1 Conclusions
Similar to the way light reflects at certain frequencies giving our surroundings color, the incident
wave reflects at certain frequencies that can be used to characterize the device in an electrical
representation. Scattering is when a known incident wave is sent into a device/network and the
reflections and transmission ratios of input voltage to output voltage of each port are recorded.
A full scattering parameter matrix is very useful for modeling the higher order behaviors inherent
to electric machines. The matrix can be used as a stand alone model, in conjunction with other
scattering models, or for extraction of other parameters for use in existing models. Use of the inverse
scattering theory (IST) method has been explored and analyzed for accuracy. In conclusion, it is
recommended that the IST methodology be considered for use in future shipboard power system
analysis when concerned with ground interactions with ship-hull ground. These ground interactions
are typically best understood as common-mode coupling. The causes of common-mode currents
or stator ground currents have been covered extensively in [58, 60, 63, 64]. Different methods to
suppress these currents have been proposed. However, little has been done to model the parasitic
effects, electromagnetic interference (EMI), or circulating bearing currents that are part of common–
mode (CM) currents in the detail found when using scattering data.
According to [13] it was found that plotting various motors for size and power ratings did not
produce any trends that were consistent. Some parameters would follow a trend while others would
not track well. It is too difficult to derive any trends of parameters based on power ratings and size
with the given scope of this research. The inductances for example are strongly related to turns
and slot characteristics. Each machine is different due to how it was physically built. The same
machine sizes were measured and produced varying results. Low frequency circuits will not see
these differences. However, current high frequency models are needed and each machine can have
its own scattering parameter fingerprint essentially.
57
Understanding what causes high frequency phenomenon in shipboard power systems (SPS) and
using proper machine models capable of capturing these effects are critical to future ship building.
Using techniques such as changing the switching frequencies or reconfiguring cable lengths are
possible solutions to minimize unwanted effects. Research proposed in [65] also suggested criteria
to follow when designing a CM filter. These types of studies are all realizable with the data obtained
using IST.
Describing the machine in the three proposed single–port representations is considered a com-
mon approach to HF motor modeling [16, 22, 34]. This approach allows for standstill measurements
to be made that help determine lumped parameter quantities of common machine circuit represen-
tations. Using a vector network analyzer (VNA) for this approach does not offer differing results
from results found when using an impedance analyzer. In most cases, the measurement process was
simplified to a set of standardized measurement sets, when using scattering parameter methods.
This work presents an alternative measurement approach that has the potential to be applied uni-
versally to other components in the system such as the cables and power electronic devices. Having
a similar measurement process and data post processing method for all devices makes this method
promising for future research.
Preliminary high frequency models were proposed and analyzed for their merits. It was found
that IST methods for measurement were valid in accurately measuring ground interactions between
the motor and the surrounding system. A systematic approach was followed to extract high fre-
quency characteristics from machines using universally accepted scattering methods for devices not
typically considered linear and time–invariant (LTI). Passive models for motor applications were
completed and the introduction of virtual measurements was discussed.
6.2 Future Considerations
Leading the efforts in machine modeling using IST methods has opened up many possibilities
that should be considered for future work. The work covered within was nearly scratching the
surface with possible scenarios and improvements that can be made. Further work into the relative
error is an obvious thrust that must be considered if these measurements are to standardize mod-
eling efforts. There was literature found that could help reduce error in measurements and clean
parametric data [66, 67, 57]. Even the use of other types of network analyzers is an area that could
58
be explored further. The use of microwave transition analyzers (MTA) or large signal network
analyzers (LSNA), which measure both amplitude and phase of the fundamental and harmonics.
It was barely touched on in this work but equipment health snapshots could be a critical tool to
monitor the behavior of devices after certain events and can be used to predict possible failure when
series of data are taken and trends are established. On a Navy ship, large amounts of energy are
being transfered back and forth throughout the ship sources and loads. These events can degrade
the equipment and snapshots may help with preventative maintenance and serve as a living model
that is always held current. Measurements used to time stamp the current state of the machines
health considering aging and pollution of windings in machines is based on an existing method but
this particular approach is new and needs further work to define all the procedures.
Finite element analysis (FEA) could be used to extract scattering parameters for design and
implementation without building the machine. Aging can be predicted and modeling can be done to
see the effect of aging on the machine. Adding this known progression to update models over time
of use can be a simple approach to predict current state of equipment or predicting the lifetime of
equipment. This can allow the Navy to be proactive in maintenance schedules and new snapshots
can be taken to track the change in system parameters. Models can be easily updated to properly
represent current state of the ship.
Speaking of validation it is pertinent these models be further scrutinized at higher frequencies
using faster switches. Validation methods using a fast switching MOSFET in [31] helped to set a
standard for IST measurements on power equipment. The MOSFET used has a steep rise time in
the nanoseconds range which corresponds to higher frequencies, subsequently found with available
measurement leads used in this research. Measurement responses of the incident pulse were captured
using S-parameter models with reasonable accuracy and should be considered to advance the upper
frequency bounds set in this work.
For future modeling considerations, it is important to develop active models. The models
developed in this research were all passive loads acting as variable impedances based on the input
frequencies into the model. These models are well suited as motors but do not contain internal
sources needed when discussing electric machines as power generators. Work will need to be done
to incorporate dependent voltage sources that respond based on scattering data provided in typical
IST measurement protocols.
59
It was also found that the termination resistance was not highly accurate in the higher fre-
quencies which caused deviations in similar measurements. There was also slight variations in
repeatability when the resistance was lower then the complex impedance. It was determined to
be most likely caused by improper terminations into non-ideal 50Ω impedances. Another issue
with accuracy was found when machines did not have an accessible neutral terminal. In [57] a new
method for working around this problem seems possible. The use of CM and DM circuits with the
neutral terminal inaccessible was previously discussed but should be brought up again and explored
in further detail.
Engineers have continued to have difficulty modeling machine bearing contact impedances be-
tween stator and rotor when measuring at standstill i.e. SSFR. Future work could determine the
frequency that this interaction will effect and remove this from the scattering matrix to better rep-
resent its response quantities when the motor is rotating and the bearings are no longer in direct
contact. SSFR research could be further looked at with scattering theory in mind in order to help
further the usability of scattering measurements.
Model validation on other machine types and sizes must be considered in the future. Also, full
system models with various grounding schemes have not been explored and would likely bring up
new information on the usability of scattering parameters on all components of the power system.
The first use suggested from this work would be to identify certain frequencies of interest that may
cause low impedance to ground paths. This would require further developments on test conditions
that can be reproduced in the lab for validation and verification of scattering models. The last
and probably more crucial step for future work would be the implementation of scattering models
into various simulation platforms. Researchers all have their own ”go to” software packages for
simulation. Using Touchstone format, models data could be easily shared via email and then
dropped into the modeling software of the researchers choice.
6.3 A Note to the Reader
Throughout this manuscript and body of work, extensive research of the state of the art was
conducted. The references that did not get mention in this document were all inspiration to the work
in some capacity but were not drawn on specifically when writing. Each reference has been left in
folders for others to draw from when extending this work further. Discussed in this document were
60
high frequency transients and frequency dependent measurement methods which capture transient
effects.
61
APPENDIX A
MOTOR MODELS
Summary
of m
odels
Litera
ture
Zhongand
Lipo(1995)
Consoliet
al.
(1996)
Grandiet
al.
(1997b,a,
2004
)
Ranet
al.
(1998a,b)
Bogliettiet
al.(1999,
2001
,20
07)
Moreira
etal.(2002)
Llaquet
etal.
(2002)
Gubıa-
Villabona
etal.(2002)
Web
eret
al.
(2004)
Maki-Ontto
andLuomi
(2005)
Proposed
model
Mach
inety
pe
Induction
motor
Induction
motor
Induction
motor
Induction
motor
Induction
motor
dual-vo
ltage
acinduction
motor
Induction
motor
acmotor
Induction
motor
Induction
motor
wound-rotor
Modelty
pe
Distributed
Lumped
Lumped
Distributed
Lumped
Lumped
Distributed
Lumped
Lumped
Distributed
Lumped
Use
ofmodel
EMI
EMI
EMI
EMI
EMI
EMI
EMI
EMI
EMI
Bearing
current
Dynamic
simulation
Fre
quency
range(H
z)
1K–500K
10K–2M
10K–1M
1K–3M
1K–1M
1K–2M
10K–1M
N/A
1K–100M
10K–10M
0.01–51.2K
Typeofdata
DM
CM/DM
CM/DM
CM/DM
CM/DM
CM/DM
CM
CM
CM/DM
CM,machine
dim
ensions
Admittance
matrix
Fitting
meth
od
N/A
Analytical
Analytical
Analytical
Least
squares
Analytical
Analytical
Analytical
Analytical
Analytical
GA
Variables
abc
qd0
abc
abc
abc
abc
abc
abc
abc
abc
abc/qd
DC
resistance
YN
NN
YN
NY
NY
Y
Eddycurrent
loss
YY
YY
YY
YY
YN
Y
Sta
torleakage
inducta
nce
YY
YY
YY
YY
YY
Y
Phase-to-
phase
inductive
coupling
YN
YN
NN
NN
YY
Y
Phase-to-
phase
capacitive
coupling
YN
NN
NN
YN
YN
N
Turn
-to-turn
capacitance
YY
YN
YY
YY
YN
N
Phase-to-
gro
und
capacitance
YY
YY
YY
YY
YY
Y
Iron
resistance
YN
YY
YY
YN
YY
Y
Main-fl
ux
satu
ration
NN
NN
NN
NN
NN
Y
Low-fre
quency
behavior
NN
NY
YY
NN
NN
Y
Roto
r
components
NN
NN
NN
NN
NN
Y
62
Summary
of m
odels (C
ont.)
Litera
ture
Schinkel
etal.(2006)
Mirafzal
etal.(2007,
2009)
Henze
etal.
(2008)
Idir
etal.
(2009)
Magdunand
Binder
(2012)
Wanget
al.
(2010)
Deganoet
al.
(2010)
Deganoet
al.
(2012)
Boucenna
etal.(2012)
Proposed
model
Mach
inety
pe
Induction
motor
Induction
motor
Induction
motor
Induction
motor
squirrel-cage
squirrel-cage
induction
motor
Induction
motor
Induction
motor
wound-rotor
Modelty
pe
Lumped
Lumped
Distributed
Lumped
Lumped
Lumped
Lumped
Lumped
Distributed
Lumped
Use
ofmodel
EMI
EMIand
bearing
current
EMI
EMI
EMI
EMI
EMI
EMI
Bearing
current
Dynamic
simulation
Fre
quency
range(H
z)
10K–30M
10–10M
100K–10M
100K–40M
1K–100M
100–10M
10K–30M
10K–30M
10K–200K
0.01–51.2K
Typeofdata
CM/DM
CM/DM,
machine
dim
ensions
Machine
dim
ensions
CM/DM
CM/DM,
machine
dim
ensions
CM/DM
CM/DM
CM/DM,
machine
dim
ensions
Machine
dim
ensions
Admittance
matrix
Fittingmeth
od
Analytical
Analytical
FEM
Analytical
Analytical
Analytical
GA
GA
FEM
GA
Variables
abc
abc
N/A
N/A
abc
abc
abc
abc
N/A
abc/qd
DC
resistance
YY
YN/A
NN
NN
N/A
Y
Eddycurrent
loss
YY
YN/A
YY
YY
N/A
Y
Sta
torleakage
inducta
nce
YY
YN/A
YY
YY
N/A
Y
Phase-to-p
hase
inductive
coupling
YY
NN/A
NY
YY
N/A
Y
Phase-to-p
hase
capacitive
coupling
NN
NN/A
NN
NN
N/A
N
Turn
-to-turn
capacitance
NY
YN/A
NY
YY
N/A
N
Phase-to-
gro
und
capacitance
YY
YN/A
YY
YY
N/A
Y
Iron
resistance
YY
NN/A
YY
YY
N/A
Y
Main-fl
ux
satu
ration
NN
NN
NN
NN
NY
Low-fre
quency
behavior
YY
NN
NN
NN
NY
Roto
r
components
NY
NN
NN
NN
NY
63
Results for 3-phase Induction motor (AsynM)
103
104
105
106
107
108
109
-8
-6
-4
-2
0
S-Parameter Magnitude S11
Frequency [Hz]
Magnitude [
dB
]
103
104
105
106
107
108
109
-600
-400
-200
0
200
S-Parameter Angle S11
Frequency [Hz]
Angle
[]
Rotor Position 0° Rotor Position 270° Rotor Position 345°Rotor Position 0° Rotor Position 270° Rotor Position 345°
103
104
105
106
107
108
109
-50
-40
-30
-20
-10
0
S-Parameter Magnitude S12
Frequency [Hz]
Magnitude [
dB
]
103
104
105
106
107
108
109
-600
-400
-200
0
200
S-Parameter Angle S12
Frequency [Hz]
Angle
[]
Rotor Position 0° Rotor Position 270° Rotor Position 345°Rotor Position 0° Rotor Position 270° Rotor Position 345°
64
103
104
105
106
107
108
109
-40
-30
-20
-10
0
S-Parameter Magnitude S13
Frequency [Hz]
Magnitude [
dB
]
103
104
105
106
107
108
109
-200
-100
0
100
S-Parameter Angle S13
Frequency [Hz]
Angle
[]
Rotor Position 0° Rotor Position 270° Rotor Position 345°Rotor Position 0° Rotor Position 270° Rotor Position 345°
103
104
105
106
107
108
109
-60
-40
-20
0
S-Parameter Magnitude S21
Frequency [Hz]
Magnitude [
dB
]
103
104
105
106
107
108
109
-600
-400
-200
0
200
S-Parameter Angle S21
Frequency [Hz]
Angle
[]
Rotor Position 0° Rotor Position 270° Rotor Position 345°Rotor Position 0° Rotor Position 270° Rotor Position 345°
65
103
104
105
106
107
108
109
-8
-6
-4
-2
0
S-Parameter Magnitude S22
Frequency [Hz]
Magnitude [
dB
]
103
104
105
106
107
108
109
-600
-400
-200
0
200
S-Parameter Angle S22
Frequency [Hz]
Angle
[]
Rotor Position 0° Rotor Position 270° Rotor Position 345°Rotor Position 0° Rotor Position 270° Rotor Position 345°
103
104
105
106
107
108
109
-40
-30
-20
-10
0
S-Parameter Magnitude S23
Frequency [Hz]
Magnitude [
dB
]
103
104
105
106
107
108
109
-1000
-500
0
S-Parameter Angle S23
Frequency [Hz]
Angle
[]
Rotor Position 0° Rotor Position 270° Rotor Position 345°Rotor Position 0° Rotor Position 270° Rotor Position 345°
66
103
104
105
106
107
108
109
-40
-30
-20
-10
0
S-Parameter Magnitude S31
Frequency [Hz]
Magnitude [
dB
]
103
104
105
106
107
108
109
-200
-100
0
100
S-Parameter Angle S31
Frequency [Hz]
Angle
[]
Rotor Position 0° Rotor Position 270° Rotor Position 345°Rotor Position 0° Rotor Position 270° Rotor Position 345°
103
104
105
106
107
108
109
-40
-30
-20
-10
0
S-Parameter Magnitude S32
Frequency [Hz]
Magnitude [
dB
]
103
104
105
106
107
108
109
-1000
-800
-600
-400
-200
0
S-Parameter Angle S32
Frequency [Hz]
Angle
[]
Rotor Position 0° Rotor Position 270° Rotor Position 345°Rotor Position 0° Rotor Position 270° Rotor Position 345°
67
103
104
105
106
107
108
109
-8
-6
-4
-2
0
S-Parameter Magnitude S33
Frequency [Hz]
Magnitude [
dB
]
103
104
105
106
107
108
109
-600
-400
-200
0
200
S-Parameter Angle S33
Frequency [Hz]
Angle
[]
Rotor Position 0° Rotor Position 270° Rotor Position 345°Rotor Position 0° Rotor Position 270° Rotor Position 345°
68
Results for 3-phase brushless DC motor (SM)
103
104
105
106
107
108
109
-10
-5
0
S-Parameter Magnitude S11
Frequency [Hz]
Magnitude [
dB
]
103
104
105
106
107
108
109
-600
-400
-200
0
200
S-Parameter Angle S11
Frequency [Hz]
Angle
[]
Rotor Position 0° Rotor Position 270° Rotor Position 345°Rotor Position 0° Rotor Position 270° Rotor Position 345°
103
104
105
106
107
108
109
-60
-40
-20
0
S-Parameter Magnitude S12
Frequency [Hz]
Magnitude [
dB
]
103
104
105
106
107
108
109
-1500
-1000
-500
0
S-Parameter Angle S12
Frequency [Hz]
Angle
[]
Rotor Position 0° Rotor Position 270° Rotor Position 345°Rotor Position 0° Rotor Position 270° Rotor Position 345°
69
103
104
105
106
107
108
109
-60
-40
-20
0
S-Parameter Magnitude S13
Frequency [Hz]
Magnitude [
dB
]
103
104
105
106
107
108
109
-1000
-500
0
S-Parameter Angle S13
Frequency [Hz]
Angle
[]
Rotor Position 0° Rotor Position 270° Rotor Position 345°Rotor Position 0° Rotor Position 270° Rotor Position 345°
103
104
105
106
107
108
109
-60
-40
-20
0
S-Parameter Magnitude S21
Frequency [Hz]
Magnitude [
dB
]
103
104
105
106
107
108
109
-1500
-1000
-500
0
S-Parameter Angle S21
Frequency [Hz]
Angle
[]
Rotor Position 0° Rotor Position 270° Rotor Position 345°Rotor Position 0° Rotor Position 270° Rotor Position 345°
70
103
104
105
106
107
108
109
-15
-10
-5
0
S-Parameter Magnitude S22
Frequency [Hz]
Magnitude [
dB
]
103
104
105
106
107
108
109
-500
0
500
S-Parameter Angle S22
Frequency [Hz]
Angle
[]
Rotor Position 0° Rotor Position 270° Rotor Position 345°Rotor Position 0° Rotor Position 270° Rotor Position 345°
103
104
105
106
107
108
109
-100
-50
0
S-Parameter Magnitude S23
Frequency [Hz]
Magnitude [
dB
]
103
104
105
106
107
108
109
-1000
-500
0
S-Parameter Angle S23
Frequency [Hz]
Angle
[]
Rotor Position 0° Rotor Position 270° Rotor Position 345°Rotor Position 0° Rotor Position 270° Rotor Position 345°
71
103
104
105
106
107
108
109
-60
-40
-20
0
S-Parameter Magnitude S31
Frequency [Hz]
Magnitude [
dB
]
103
104
105
106
107
108
109
-1000
-500
0
S-Parameter Angle S31
Frequency [Hz]
Angle
[]
Rotor Position 0° Rotor Position 270° Rotor Position 345°Rotor Position 0° Rotor Position 270° Rotor Position 345°
103
104
105
106
107
108
109
-100
-50
0
S-Parameter Magnitude S32
Frequency [Hz]
Magnitude [
dB
]
103
104
105
106
107
108
109
-1000
-500
0
S-Parameter Angle S32
Frequency [Hz]
Angle
[]
Rotor Position 0° Rotor Position 270° Rotor Position 345°Rotor Position 0° Rotor Position 270° Rotor Position 345°
72
103
104
105
106
107
108
109
-10
-5
0
S-Parameter Magnitude S33
Frequency [Hz]
Magnitude [
dB
]
103
104
105
106
107
108
109
-500
0
500
S-Parameter Angle S33
Frequency [Hz]
Angle
[]
Rotor Position 0° Rotor Position 270° Rotor Position 345°Rotor Position 0° Rotor Position 270° Rotor Position 345°
73
BIBLIOGRAPHY
[1] S. Littlefield and A. Nickens. (2005, feb) Roadmap for the all-electric warship.Power-JF05-R&D 2.gif. [Online]. Available: http://www.powermag.com/roadmap-for-the-all-electric-warship
[2] F. F. Jack ou. (2012, September) S2p, parameter extraction. [Online]. Available:www.sonoma.edu/Fusers/o/ouj/classes/ES590/lectures/Lecture 1 Components ResonantCircuits Fall 2012.pptx
[3] O. Magdun, A. Binder, C. Purcarea, and A. Rocks, “High-frequency induction machine mod-els for calculation and prediction of common mode stator ground currents in electric drivesystems,” in Power Electronics and Applications, 2009. EPE ’09. 13th European Conferenceon, Sept 2009, pp. 1–8.
[4] J. Ciezki, L. Graber, S. Pekarek, and M. Steurer, “Evaluation of rail harmonics in alternativemvdc grounding strategies,” in ASNE Electric Machines Technologies Symposium (EMTS),Philadelphia, PA, May 2010, pp. 19–20.
[5] C. Cassiolato, “Tips on shielding and grounding in industrial automation,” SMAR IndustrialAutomation, Brazil, Tech. Rep., Nov 2011.
[6] Y. Xu, “Advanced wound-rotor machine model with saturation and high-frequency effects,”M.S. Electrical Engineering. Thesis, Iowa State University, Ames, Iowa, August 2014.
[7] N. Schulz, R. Hebner, S. Dale, R. Dougal, S. Sudhoff, E. Zivi, and C. Chryssostomidis, “Theu.s. esrdc advances power system research for shipboard systems,” in Universities Power En-gineering Conference, 2008. UPEC 2008. 43rd International, Sept 2008, pp. 1–4.
[8] L. Graber, S. Pekarek, and M. Mazzola, “Grounding of shipboard power systems - results fromresearch and preliminary guidelines for the shipbuilding industry,” Perdue and FSU and MSU,FL, Tech. Rep. N0014-08-1-0080, Feb. 2014.
[9] L. Graber, M. Steurer, J. Kvitkovic, M. Kofler, S. Pekarek, R. Howard, A. Taher, M. Maz-zola, and A. Card, “Time and frequency domain methods to evaluate grounding strategiesfor medium voltage dc shipboard power systems,” in Electric Ship Technologies Symposium(ESTS), 2013 IEEE, April 2013, pp. 43–48.
[10] B. Revol, J. Roudet, J.-L. Schanen, and P. Loizelet, “Emi study of three-phase inverter-fedmotor drives,” Industry Applications, IEEE Transactions on, vol. 47, no. 1, pp. 223–231, Jan2011.
74
[11] L. Arnedo and K. Venkatesan, “High frequency modeling of induction motor drives foremi and overvoltage mitigation studies,” in Electric Machines and Drives Conference, 2003.IEMDC’03. IEEE International, vol. 1, June 2003, pp. 468–474 vol.1.
[12] Y. Pan, M. Steurer, and T. Baldwin, “Ground fault location testing of a noise-pattern-basedapproach on an ungrounded dc system,” Industry Applications, IEEE Transactions on, vol. 47,no. 2, pp. 996–1002, March 2011.
[13] S.-P. Weber, E. Hoene, S. Guttowski, W. John, and H. Reichl, “Modeling induction machinesfor emc-analysis,” in Power Electronics Specialists Conference, 2004. PESC 04. 2004 IEEE35th Annual, vol. 1, June 2004, pp. 94–98 Vol.1.
[14] Y. Pan, “Feasibility study of noise pattern analysis based ground fault locating method for un-grounded dc shipboard power distribution systems,” in Electric Ship Technologies Symposium,2009. ESTS 2009. IEEE, April 2009, pp. 18–22.
[15] H. Hamilton and N. Schulz, “Impact of grounding on overcurrents for a naval ac/dc shipboardpower system,” in Power Symposium, 2007. NAPS ’07. 39th North American, Sept 2007, pp.170–176.
[16] E. Zhong, T. Lipo, and S. Rossiter, “Transient modeling and analysis of motor terminal voltageon pwm inverter-fed ac motor drives,” in Industry Applications Conference, 1998. Thirty-ThirdIAS Annual Meeting. The 1998 IEEE, vol. 1, Oct 1998, pp. 773–780 vol.1.
[17] R. Kerkman, D. Leggate, and G. Skibinski, “Interaction of drive modulation and cable param-eters on ac motor transients,” Industry Applications, IEEE Transactions on, vol. 33, no. 3, pp.722–731, May 1997.
[18] I. Metwally, “Simulation of the impulse response of electrical machines,” Energy Conversion,IEEE Transactions on, vol. 14, no. 4, pp. 861–867, Dec 1999.
[19] M. Kofler, “Scattering parameters for transient analysis of shipboard power systems,” M.S.thesis, address = Wels, Austria, month = February, year = 2013,, University of AppliedSciences Upper Austria School of Engineering and Environmental Sciences.
[20] H. De Paula, D. de Andrade, M. Chaves, J. Domingos, and M. de Freitas, “Methodology forcable modeling and simulation for high-frequency phenomena studies in pwm motor drives,”Power Electronics, IEEE Transactions on, vol. 23, no. 2, pp. 744–752, March 2008.
[21] W. Shen, F. Wang, and D. Boroyevich, “Conducted emi characteristic and its implicationsto filter design in 3-phase diode front-end converters,” in Industry Applications Conference,2004. 39th IAS Annual Meeting. Conference Record of the 2004 IEEE, vol. 3, Oct 2004, pp.1840–1846 vol.3.
75
[22] A. Moreira, T. Lipo, G. Venkataramanan, and S. Bernet, “High frequency modeling for cableand induction motor overvoltage studies in long cable drives,” in Industry Applications Con-ference, 2001. Thirty-Sixth IAS Annual Meeting. Conference Record of the 2001 IEEE, vol. 3,Sept 2001, pp. 1787–1794 vol.3.
[23] R. Escarela-Perez, T. Niewierowicz, and E. Campero-Littlewood, “Synchronous machine pa-rameters from frequency-response finite-element simulations and genetic algorithms,” EnergyConversion, IEEE Transactions on, vol. 16, no. 2, pp. 198–203, Jun 2001.
[24] B. Mirafzal, G. Skibinski, R. Tallam, D. Schlegel, and R. Lukaszewski, “Universal induc-tion motor model with low-to-high frequency-response characteristics,” Industry Applications,IEEE Transactions on, vol. 43, no. 5, pp. 1233–1246, Sept 2007.
[25] S. Ogasawara, H. Ayano, and H. Akagi, “Measurement and reduction of emi radiated bya pwm inverter-fed ac motor drive system,” Industry Applications, IEEE Transactions on,vol. 33, no. 4, pp. 1019–1026, Jul 1997.
[26] D. Maly, D. Novotny, and C. Thompson, “The influence of winding capacitance on high fre-quency time harmonic losses in induction motors,” in Industry Applications Society AnnualMeeting, 1992., Conference Record of the 1992 IEEE, Oct 1992, pp. 33–39 vol.1.
[27] A. Bonnett, “Analysis of the impact of pulse-width modulated inverter voltage waveformson ac induction motors,” Industry Applications, IEEE Transactions on, vol. 32, no. 2, pp.386–392, Mar 1996.
[28] A. Muetze and A. Binder, “Calculation of motor capacitances for prediction of the voltageacross the bearings in machines of inverter-based drive systems,” Industry Applications, IEEETransactions on, vol. 43, no. 3, pp. 665–672, May 2007.
[29] W. Shen, F. Wang, D. Boroyevich, and Y. Liu, “Definition and acquisition of cm and dmemi noise for general-purpose adjustable speed motor drives,” in Power Electronics SpecialistsConference, 2004. PESC 04. 2004 IEEE 35th Annual, vol. 2, June 2004, pp. 1028–1033 Vol.2.
[30] J. Wei, D. Gerling, and M. Galek, “S-parameters characterization and sequence model of three-phase emi filter,” in Industrial Electronics Society, IECON 2013 - 39th Annual Conference ofthe IEEE, Nov 2013, pp. 1254–1259.
[31] B. Mohebali, P. Breslend, L. Graber, and M. Steurer, “Validation of a scattering parameterbased model of a power cable for shipboard grounding studies,” in ASNE Electric MachinesSymposium (EMTS), May 2014, pp. 28–29.
[32] B. Behshad, P. Breslend, L. Graber, and M. Steurer, “How to measure scattering parametersfor modeling the components of shipboard power systems,” in Submitted in ASNE NavalEngineers Journal (NEJ), Dec. 2014.
76
[33] H. De Gersem and A. Muetze, “Finite-element supported transmission-line models for calcu-lating high-frequency effects in machine windings,” Magnetics, IEEE Transactions on, vol. 48,no. 2, pp. pp. 787–790, Feb 2012.
[34] A. Moreira, T. Lipo, G. Venkataramanan, and S. Bernet, “High-frequency modeling for cableand induction motor overvoltage studies in long cable drives,” Industry Applications, IEEETransactions on, vol. 38, no. 5, pp. 1297–1306, Sep 2002.
[35] Z. Xian, J. Jiang, L. He, D. Liu, and H. Cao, “Measurement of large ac electrical machinestator windings parameters at high frequency,” in Electric Machines and Drives Conference,2003. IEMDC’03. IEEE International, vol. 2, June 2003, pp. 1331–1336 vol.2.
[36] S. Luff, S. Garvey, and W. Norris, “Predicting high frequency characteristics of the windingsof large electrical machines-a transmission line analysis approach,” in Electrical Machines andDrives, 1999. Ninth International Conference on (Conf. Publ. No. 468), 1999, pp. 218–222.
[37] W. Liu, E. Schaeffer, D. Averty, and L. Loron, “A new approach for form wound machinegoundwall insulation diagnosis by means of high frequency model parameter monitoring,” inIndustrial Electronics, 2007. ISIE 2007. IEEE International Symposium on, June 2007, pp.1270–1275.
[38] A. Boglietti and E. Carpaneto, “Induction motor high frequency model,” in Industry Appli-cations Conference, 1999. Thirty-Fourth IAS Annual Meeting. Conference Record of the 1999IEEE, vol. 3, 1999, pp. 1551–1558 vol.3.
[39] A. Repo and A. Arkkio, “Numerical impulse response test to estimate circuit-model parametersfor induction machines,” IEEE Proceedings Electric Power Applications, vol. 153, no. 6, pp.883–890, Nov 2006.
[40] M. Schinkel, S. Weber, S. Guttowski, W. John, and H. Reichl, “Efficient hf modeling andmodel parameterization of induction machines for time and frequency domain simulations,” inAPEC ’06. Twenty-First Annual IEEE. Applied Power Electronics Conference and Exposition,2006., March 2006, p. 6.
[41] S. Guttowski, S. Weber, M. Schinkel, W. John, and H. Reichl, “Troubleshooting and fixing ofinverter driven induction motor bearing currents in existing plants of large size - an evaluationof possible mitigation techniques in practical applications,” Fraunhofer Institute for Reliabilityand Microintegration, Germany, Tech. Rep., nov 2006.
[42] A. Keyhani and H. Tsai, “Identification of high-order synchronous generator models from ssfrtest data,” Energy Conversion, IEEE Transactions on, vol. 9, no. 3, pp. 593–603, Sep 1994.
[43] R. Schiferl, M. Melfi, and J. Wang, “Inverter driven induction motor bearing current solutions,”in Petroleum and Chemical Industry Conference, 2002. Industry Applications Society 49thAnnual, 2002, pp. 67–75.
77
[44] P. Tavner and R. Jackson, “Coupling of discharge currents between conductors of electricalmachines owing to laminated steel core,” Electric Power Applications, IEE Proceedings B, vol.135, no. 6, pp. 295–307, Nov 1988.
[45] S. Mahdavi and K. Hameyer, “High frequency equivalent circuit model of the stator windingin electrical machines,” in Electrical Machines (ICEM), 2012 XXth International Conferenceon, Sept 2012, pp. 1706–1711.
[46] I. Dolezel, V. Valouch, and J. Skramlik, “High frequency models of transistor voltage inverter-fed induction motor drives,” in Industrial Technology 2000. Proceedings of IEEE InternationalConference on, vol. 2, Jan 2000, pp. 32–37 vol.1.
[47] J. Sun and L. Xing, “Parameterization of three-phase electric machine models for emi simula-tion,” Power Electronics, IEEE Transactions on, vol. 29, no. 1, pp. 36–41, Jan 2014.
[48] C. Mei, J. Balda, W. Waite, and K. Carr, “Minimization and cancellation of common-modecurrents, shaft voltages and bearing currents for induction motor drives,” in Power ElectronicsSpecialist Conference, 2003. PESC ’03. 2003 IEEE 34th Annual, vol. 3, June 2003, pp. 1127–1132 vol.3.
[49] D. Bockelman and W. Eisenstadt, “Combined differential and common-mode scattering pa-rameters: theory and simulation,” Microwave Theory and Techniques, IEEE Transactions on,vol. 43, no. 7, pp. 1530–1539, Jul 1995.
[50] L. Wang, C.-m. Ho, F. Canales, and J. Jatskevich, “High-frequency cable and motor mod-eling of long-cable-fed induction motor drive systems,” in Energy Conversion Congress andExposition (ECCE), 2010 IEEE, Sept 2010, pp. 846–852.
[51] L. Wang, C. Ho, F. Canales, and J. Jatskevich, “High-frequency modeling of the long-cable-fedinduction motor drive system using tlm approach for predicting overvoltage transients,” PowerElectronics, IEEE Transactions on, vol. 25, no. 10, pp. 2653–2664, Oct 2010.
[52] L. Graber, D. Infante, M. Steurer, and W. Brey, “Validation of cable models for simula-tion of transients in shipboard power systems,” in High Voltage Engineering and Application(ICHVE), 2010 International Conference on, Oct 2010, pp. 77–80.
[53] I. Boldea and S. Nasar, “Super-high frequency models and behavior of ims,” in The InductionMachines Design Handbook, Second Edition, December 2009, pp. 639–658.
[54] W. Dezheng and S. Pekarek, “A multirate field construction technique for efficient modelingof the fields and forces within inverter-fed induction machines,” Energy Conversion, IEEETransactions on, vol. 25, no. 1, pp. 217–227, March 2010.
78
[55] I. Agilent Technologies. (2009, April) Touchstone file for-mat specification. touchstone ver2 0.pdf. [Online]. Available:http://www.eda.org/ibis/touchstone/ ver2.0/touchstone ver2 0.pdf
[56] A. Technologies, Using Circuit Simulators, Working with Data Files. Palo Alto, CA: OxfordUniv. Press, 2005, ch. 4.
[57] I. Rolfes and B. Schiek, “Multiport method for the measurement of the scattering parametersof n-ports,” Microwave Theory and Techniques, IEEE Transactions on, vol. 53, no. 6, pp.1990–1996, June 2005.
[58] A. Muetze and A. Binder, “Generation of high-frequency common mode currents in machinesof inverter-based drive systems,” in Power Electronics and Applications, 2005 European Con-ference on, Sept 2005, pp. 10 pp.–P.10.
[59] W. Hribernik, L. Graber, and J. Brunke, “Inherent transient recovery voltage of powertransformers-a model-based determination procedure,” Power Delivery, IEEE Transactionson, vol. 21, no. 1, pp. 129–134, Jan 2006.
[60] S. Ogasawara and H. Akagi, “Modeling and damping of high-frequency leakage currents in pwminverter-fed ac motor drive systems,” Industry Applications, IEEE Transactions on, vol. 32,no. 5, pp. 1105–1114, Sep 1996.
[61] B. Mohebali, P. Breslend, L. Graber, and M. Steurer, “Challenges of frequency domain mea-surement for modeling the components of shipboard power systems,” Naval Engineers Journal(NEJ), vol. 126-4, pp. 123–128, dec 2014.
[62] P. Breslend, B. Mohebali, L. Graber, and M. Steurer, “High frequency models for rotatingmachines in ungrounded shipboard power systems,” Naval Engineers Journal (NEJ), vol. 126-4, pp. 36–42, dec 2014.
[63] D. Busse, J. Erdman, R. Kerkman, D. Schlegel, and G. Skibinski, “System electrical parametersand their effects on bearing currents,” Industry Applications, IEEE Transactions on, vol. 33,no. 2, pp. 577–584, Mar 1997.
[64] E. Zhong and T. Lipo, “Improvements in emc performance of inverter-fed motor drives,”Industry Applications, IEEE Transactions on, vol. 31, no. 6, pp. 1247–1256, Nov 1995.
[65] D. Graovac, T. Hoffmann, and A. Haltmair, “A transfer function approach to a common modefilter (cmmf) optimization in the pwm inverter supplied motor drives,” Energy Conversion,IEEE Transactions on, vol. 26, no. 1, pp. 93–101, March 2011.
[66] J. Tippet and R. A. Speciale, “A rigorous technique for measuring the scattering matrix of amultiport device with a 2-port network analyzer,” Microwave Theory and Techniques, IEEETransactions on, vol. 30, no. 5, pp. 661–666, May 1982.
79
[67] J. C. Rautio, “Techniques for correcting scattering parameter data of an imperfectly termi-nated multiport when measured with a two-port network analyzer,” Microwave Theory andTechniques, IEEE Transactions on, vol. 31, no. 5, pp. 407–412, May 1983.
80
BIOGRAPHICAL SKETCH
Patrick R. Breslend was a research Assistant at the Center for Advanced Power Systems, Florida State Uni-
versity in Tallahassee, Florida from the years 2012–2015. In this position, he was responsible for
the development of research models for use in the Electric Ship Research Development Consor-
tium (ESRDC). His work has led to the development of rapid model generation tutorials for use on
MathWorks website. Currently working on M.S. in Electrical Engineering at Florida State Univer-
sity, where he has focused his efforts in the areas of electrical machine grounding techniques. His
approach to modeling the electric machine incorporates scattering parameters in order to analyze
ground currents due to higher switching frequencies present in future electric ships.
Prior to these positions, he was cofounder and President of Sustainable Engineered Solu-
tions (SES) a student organization at Florida State University from 2011-2012. Through his direct
leadership the group completed multiple projects in the area of sustainability. The successes of
SES lead to a conference trip to Association of Energy Engineers World Energy Engineering Con-
ference. The trip was fully funded due to the hard work of Patrick and his supporting staff. He
also contributed to the group by acquiring the management responsibilities for the FAMU/FSU
Solar Car in the hope that the group could spread awareness and support for sustainability in our
community.
Also, in his masters program he found time to Co-found and hold the position of Vice-President
for the Society of Engineering Entrepreneurs (SEE). His business ideas and breadth of experience
was instrumental in the growth of this organization during its infancy. He has lead future planning
and put together the first entrepreneur group specifically aimed at young engineers looking to start
businesses.
Mr. Breslend has held various jobs since the age of eleven where he started his own lawn
mowing company Leprechaun Lawn Service. It was at this job where the seed for his first patent
was planted. It was years later in college, working for FedEx , he decided to patent his idea
for use in the shipping industry. He is currently in the process of starting up his own business
Optimal Bagging and its first product the Quick-Pack. He looks forward to turning his patent into
a commercial product seen in industries such as shipping, hotels, restaurants, etc. He plans to soon
file for a new patent and start the next business with his entrepreneurial spirit in hand.
81
If starting a business was not enough, Mr. Breslend has now begun new research into high
voltage pulsed power plasma reaction engineering for production of various chemical species needed
in the growth of plants. His research is currently funded by the Chemical & Biomedical Engineering
department at Florida State University under the direction of Dr. Bruce Locke. His experience in
entrepreneurship and background in high voltage and high frequency applications has moved this
research in great strides towards commercialization.
82