Earth Fault Protection of Transformer-less Power Conversion Systems

Dr Teng Long1, BEng PhD CEng MIET

Martin Butcher2, CEng MIET Dr Makhlouf Benatmane3, BSc(Hons) PhD CEng FIMarEST FIET

GE Power Conversion

This technical paper is prepared for the purposes of the 2016 International Naval Engineering Conference (INEC). It is based on the author’s opinion and information collected through various sources. Many variables may impact the technical considerations and data as well as the study results illustrated in the documents; GE’s intention is only to give examples of potential applications and related impact of certain enhanced technology and products and these technical documents are not meant to provide any guarantee relating to any conclusions and technical data contained herein. This technical paper shall not be reproduced nor the content used for different purposes.

SYNOPSIS

The Transformer-less, unearthed or high resistance grounded electric power system becomes popular for naval vessels. It is challenging to protect such systems from faults, because accurately isolating the faulty system part is difficult. Neither non-tripping nor incident tripping is acceptable in the interest of reliability. In this paper, the behaviour and characteristics of the earth fault in a Transformer-less low voltage high resistance grounding system are discussed and analysed. A novel earth fault detection method, which offers lower cost and simpler implementation, is introduced.

INTRODUCTION

Electric propulsion is preferred for many vessel types due to merits such as freedom of engine placement, reduction of weight and volume, and attenuation of acoustic noise. The power system of these vessels encompasses many power converters, which convert fixed voltage and frequency to variable voltage and frequency using power electronics devices and drive propulsion motors or other variable speed loads.

For vessels with an installed power capacity of up to 20MW, Transformer-less Active Front End (AFE) converters are increasingly popular due to their flexibility, lower installed footprint and weight. These converters connect directly to the main AC bus by using fully controllable power-electronic devices such as IGBTs, which offer bi-directional power flow, high power density and lower harmonics to the electric system, as shown in Figure 1.

Since no transformer galvanically isolates the converters from the main AC bus, all Transformer-less AFE converters are commonly coupled and their common mode voltages are associated. Hence, the entire system will 1 Teng Long received a BEng from the University of Birmingham and a PhD from the University of Cambridge in 2009 and 2013 respectively. He has been with GE Power Conversion since 2013. His research interests include Power Electronics, Drives and Power Systems. Teng has published 11 academic papers and filed 4 patents. Teng is a Chartered Engineer.2 Martin Butcher graduated from the University of Leicester with an MEng in Electrical and Electronic Engineering in September 2000 and joined GE Power Conversion (then ALSTOM Power Conversion) the same year. Mr Martin Butcher is currently a Principle Engineer. His expertise is in Power Systems Design and Development with a focus on Power System Stability and the integration of Power Electronic Converters.

3 Makhlouf Benatmane has extensive experience in electrical systems engineering, in industrial and marine applications. He has been a University lecturer in Power System Design, Electrical Power Station Design and Power Electronics. He holds a PhD in Electrical Engineering from The University of Nottingham and a BSc (Hons). He is a Chartered Engineer, Fellow of The Institute of Engineering and Technology, Fellow of The Institute of Marine Engineering, Science & Technology. He is currently the Fulfilment Leader Marine for GE Power Conversion.

experience the common mode voltage caused by an earth fault, no matter where this fault occurs [1]. This common mode fault voltage will stress the components in the system and the common mode fault current will bring thermal stress to components, in addition to resulting damage at the site of the earth fault [2].

This earth fault issue can be solved by increasing the rating of components in the system to withstand the enlarged common mode voltage and current. However, this solution will add more cost, does not mitigate the damage at the earth fault site and leaves the system vulnerable to the second earth fault propagating to a full short circuit fault [2]. The other solution is to detect the earth fault and trip the faulty drive. The challenge of this method is to discriminate between the faulty and healthy drives. A prudent detection method is required to ensure just the faulty drive trips because any nuisance tripping may cause instability or even a blackout of the electric propulsion system.

In this paper, characteristics of earth faults at various locations will be discussed and analysed. Subsequently, a novel detection method based on an equivalent circuit approach will be presented. This new method is able to identify the earth fault at any location and discriminate the faulty drive. The proposed method can be implemented in an easy way with minimal requirements on the hardware additional to the typical Transformer-less AFE and switchboard. Implementation and settings of the protection will be suggested and experimental results will be presented.

Figure 1 Typical LV system with transformer-less drives for offshore marine vessels

EARTH FAULT CHARACTERISTICS

A typical Transformer-less power conversion system is illustrated in Figure 1. Each AC bus feeds multiple Transformer-less drives and at least one High Resistor Grounding (HRG) per switchboard section. The drive has a network filter, including line reactors, PWM filter, and EMC filter (capacitor connected between the line and the earth), to attenuate harmonics at both line and phase voltages according to requirements. Each drive has a Network Bridge (NB) and a Machine Bridge (MB) to convert the fixed voltage, fixed frequency bus voltage to the variable voltage, variable frequency machine terminal [3].

Depending on the location of the earth fault, it has different characteristics in terms of common mode voltage and common mode current. There are three different locations to represent earth faults in the entire drive. In this paper, an earth fault at the network side is called Type 1; an earth fault at the DC link between the NB and MB is

called Type 2; and an earth fault at the machine side is called Type 3. The earth fault Type 1 and Type 3 can be caused by one phase short circuited to earth and Type 2 can be caused by one DC rail short circuited to earth.

The proposed earth fault detection methods utilise the different characteristics of the common mode current. During the fault, the common mode voltage generates the common mode current and energises the network impedance. In this section, characteristics of the common mode voltage and the common mode current during the earth fault are discussed as a function of locations.

Characteristics of Earth Fault Type 1

At the network side, the phase voltage is mainly the fundamental component at low input frequency (50 or 60 Hz) with the harmonics controlled to achieve the THD and EMC requirements. The common mode voltage is a third of the vector sum of all three phase voltages across the EMC filter capacitor at each phase, shown as Vcom_net in Figure 2. The line-to-earth impedance of the system consists of the EMC filter capacitor, the HRG and the line-to-earth parasitic capacitors. Although the parasitic capacitance is distributed throughout the system, only the parasitic capacitance at the generator and the machine are considered as the dominant parasitic capacitors, although this depends on installed cable length. If the system is healthy, all phase voltages are balanced and the common mode voltage is almost zero, as shown in Figure 3.

Figure 2 Common mode model of one LV drive with one HRG

Figure 3 Common mode voltage at faults

An earth fault at the network side causes a strong imbalance of phase voltages. As shown in Figure 3 and Figure4, the phase voltage of Phase C becomes zero and the phase voltages of Phase A and B become the line voltage. The common mode voltage then equals the phase voltage at the fundamental frequency. At the fundamental frequency, the HRG provides much lower impedance for the common mode current than the EMC filter capacitors and parasitic capacitors. Hence, most of the common mode fault current returns to the fault point via the HRG. Thus, switchboard HRG has become one of the industry standard earthing methods in marine electrical power systems. In summary, an earth fault Type 1 generates a common mode voltage at the fundamental frequency and the resulting current predominately returns via the HRG.

Figure 4 Modelled phase voltage of the fault occurrence at network side as fault Type 1

Characteristics of Earth Fault Type 2

In contrast to the network side, the DC link phase voltage (DC rail to earth) contains strong harmonics. These harmonics are generated by PWM switching at both the NB and MB and they are normally at high frequencies (switching frequency or integer multiples of the switching frequency). These harmonics result in non-zero common mode voltages on the DC link at the healthy state, causing the common mode current to flow through the EMC filter capacitors [3]. These voltages and currents are considered as the background common voltage and current, the latter of which is limited according to IEC 60533 [4]. The background common mode current can be reduced by limiting the parasitic capacitance and enlarging the common mode inductance in the line. In addition, the background common mode current should mostly return to the NB and MB via the EMC filter capacitor, so minimal background common mode voltage and current would be observed at the network side.

An earth fault at the DC link significantly enlarges common mode voltage. As shown in Figure 5, this common mode fault voltage becomes the constant DC rail to rail voltage, offsetting the common mode voltage at the AC side of the NB, shown as the Vcom_PWM in Figure 3. This Vcom_PWM contains much enlarged harmonics at high frequencies and becomes a common mode voltage source to generate the common mode fault current flowing through all the impedance between the lines to the earth at the high frequencies. The common mode DC voltage and common mode voltage at the AC side of the network bridge are shown in Figure 5. At high frequencies, the EMC filter capacitor provides the lowest impedance between the line to the earth compared with the HRG and parasitic capacitors so most of the common mode fault current caused by the V com_PWM returns to the fault point at the DC link via EMC filter capacitors and a smaller proportion of the fault current returns via the HRG.

In summary, an earth fault Type 2 significantly increases the common mode voltage, causing a much higher common mode current than the background current, both of which are at high frequencies. The major common

mode fault current path is from the DC link through the EMC filter capacitors and the NB and back to the DC link.

Figure 5 Modelled phase voltage of the fault occurrence at DC link as fault Type 2

Characteristics of Earth Fault Type 3

Most LV drives do not have an EMC filter between the machine and the MB, so the phase voltage at the machine side contains high frequency harmonics and this background common mode voltage generates background common mode current flowing through the EMC filter capacitor [3]. Similar to the principle of the Type 1 fault, the earth fault at the machine side causes a strong imbalance of phase voltages, so a strong common mode fault voltage is created. However, this imbalance is applied to all harmonics of the phase voltage at the MB, so the common mode fault voltage contains a large amount of high frequencies harmonics, shown in Figure6. Similar to the earth fault Type 2, the EMC filter capacitor is the preferable path for the common mode fault current, due to its low impedance at high frequencies. In summary, an earth fault Type 3 causes large common mode currents at high frequencies, flowing via EMC filter capacitors, NB and MB and back to the fault point.

Figure 6 Modelled phase voltage of the fault occurrence at machine side as fault Type 3

DISCRIMINATION METHOD

The principle of the proposed methods on discriminating the faulty drive is regardless of the number of drives in the system. Thus, two drives are used to represent multiple drives in the system to simplify the analysis. As illustrated in Figure 7, Drive 1 and Drive 2 are connected to the Point of Common Coupling (PCC). The PCC is also connected with one HRG.

The method utilises the path of the common mode fault current. As shown in Figure 8, two types of common mode voltage are considered in the equivalent circuit. The Low Frequency Common Mode Voltage (LFCMV), at the fundamental frequency is zero when there is no fault, but increases when an earth fault occurs at the network side (fault Type 1). The Common Mode Voltages generated at NB (COMV1) and MB (COMV2) always exist, but significantly increase when earth faults occur at the DC link and the machine side (fault Type 2 and 3). The common mode equivalent circuit model can represent the Transformer-less multi-drive system at the common mode and predict the common mode current with or without an earth fault in the system.

Discrimination Method for Earth Fault Type 1

The earth fault at the network side of one drive results in a common mode fault voltage at the fundamental frequency. Because all drives are directly connected to the PCC, this fault voltage will be observed by all drives at the same PCC, and can therefore be modelled as LFCMV as shown in Figure 9.

Because the line-to-earth parasitic impedances are very large at the fundamental frequency, they can be removed from the equivalent circuit for a Type 1 fault. As shown in Figure 9, both Drive 1 and Drive 2 are energised by the common mode fault voltage LFCMV, but most of the common mode fault current generated by this voltage returns to the fault point in Drive 1 via the HRG. This path indicates that an exclusive common mode fault current is enclosed in Drive 1 only, which can be used for discriminating the faulty drive. This path can be measured at the terminal of each drive, Point 1 in Figure 9. Although a proportion of this common-mode fault current flows into the healthy drives through their EMC filter capacitors, it should be negligible compared to the common-mode fault current in the faulty drive.

Figure 7 Simplified Transformer-less multi-drive system

Figure 8 Equivalent circuit of common-mode network for two drives system

Therefore, for the earth fault Type 1, the discrimination method is to use the common mode current sensed at the input of each drive (Point 1) in the system. The bandwidth of sensing at the Point 1 is suggested to be less than 100Hz to capture mainly the fundamental component. If a large value of the common mode current is sensed by one drive, this drive needs to trip due to its Type 1 earth fault. The threshold for tripping depends on the resistance of the HRG and the sensitivity requirement.

Figure 9 Common mode fault current loops of Type 1 earth fault

Discrimination Method for Earth Fault Type 2 and 3

The common mode voltages caused by earth faults of Type 2 and Type 3 contain harmonics at high frequencies. Common mode equivalent circuits for Type 2 and 3 faults are shown in Figure 10 and Figure 11. The equivalent circuit considers the harmonics only, so the fundamental component of the common mode fault voltage is removed. The High Frequency Common Mode Voltages (HFCMV) at Drive 1 and 2 are shown as the common mode voltage sources.

Figure 10 Common mode fault current loops of Type 2 earth fault

Figure 11 Common mode fault current loops of Type 3 earth fault

In contrast to fault Type 1, the EMC filter capacitors at every drive provide lower impedances than both the HRG and the parasitic capacitance at the high frequencies. The major common mode fault currents thus return to the fault points through all EMC filter capacitors in all drives, shown as Icom1 and Icom2 in Figure 10 and Figure 11. The sum of fault currents Icom1 and Icom2 can be measured at Point 2 of the faulty drive and no fault current is measured at that point of the healthy drive, as opposed to Point 1. Thus we propose to use the high frequency current measurement at Point 2 to prevent tripping of healthy drives.

We suggest using the switching frequency component for fault detection and discrimination, by applying a harmonics-selected band-pass filter at Point 2. The central frequency is set to be the switching frequency and the bandwidth is required to cover harmonics close to the switching frequency caused by sideband impacts. The process of the detection method for Type 2 and Type 3 fault is illustrated in Figure 12.

Figure 12 Implementation schematic of discrimination method for fault Type 2 and Type 3

TEST RESULTS

Test rig

Two drives are proposed to be used as depicted in Figure 13. Drive 1 is called 4QL drive and Drive 2 is called TG drive. Each drive is connected to a 1.5MW induction motor. The speed of each motor can be adjusted independently. The earth fault can be applied to any of these two drives at the DC link or the machine side for fault Type 2 and 3. A 10Ω resistor connects the fault location to the earth, limiting the earth fault current. Therefore, this arrangement represents a maximum fault detection sensitivity of 10Ω.

Figure 13 Test rig schematics

Concept Validation Test Results

An earth fault was applied for a short time and the tripping function was disabled in order to validate discrimination characteristics of the common mode fault at sensing Point 2 for Type 2 and 3 faults. The common mode current was measured at the NB and the sampling rate was set as 40kHz to ensure acquisition of harmonics at high frequencies. As shown in Figure 14, a strong common- mode current at high frequencies was sensed at the faulty drive (Drive 1), and the healthy drive (Drive 2) did not observe any change during the faults.

As shown in Figure 15, the components at the switching frequency area (considering the sideband) of the sensed common mode current increase significantly due to the fault. The healthy drive does not experience any change of the harmonics in the selected band. These results verify that the common mode fault current measured at Point 2 (between NB and EMC filter capacitor) discriminate fault Type 2 and 3 when using the harmonics around the switching frequency.

Figure 14 Test results of common mode current at Point 2 during an earth fault occurrence at Drive 1 machine side

Figure 15 Spectrum of common mode current at Point 2 from two drives at Pre-Fault and Fault

Influences from Operating Conditions

The proposed method uses common mode currents at the PWM switching frequency. The PWM patterns vary with the change of modulation, i.e. the change of the voltage and frequency of the fundamental. Generally, lower modulation leads to a higher common mode voltage and current at both healthy and fault conditions.

However, the lowest common mode fault current is still much larger than the lowest background common mode current at the same frequency and can therefore discriminate faults. As shown in Figure 16, for a Type 3 fault, the lowest common fault current was at 100% speed with maximum modulation and this fault current was 4 times larger than the highest background common mode current at very low speed with low modulations. These

phenomena mean there are enough margins to set a tripping threshold to be sensitive enough for the lowest fault current and robust enough to avoid any nuisance trip caused by background noise. According to [4], the maximal common mode current allowed for the drive is 5A, suggesting the threshold for the earth fault detection should be less than 5A. In order to set the threshold higher than the background common mode current, it should be lower, but close to, 5A. To a certain extent, the drive can have an even higher sensitivity and robustness when the threshold for tripping is adjusted dynamically with the modulation reference of the MB, which is available in the controller. This is an additional advantage of this software based earth fault detection method.

Figure 16 Comparison of processed common mode current of fault Type 2 and 3 at different speeds

Earth Fault Protection Test Results for Type 2 and Type 3 Faults

In total, 14 tests with different operating conditions have been conducted and all of these tests were witnessed and approved by the Classification Society DNV GL. The ‘PASS’ criteria require the faulty drive to trip and the healthy drive to continue with normal operation. Either faulty drive non-tripping or healthy drive tripping is marked as ‘FAIL’. All tests were regarded as ‘PASS’ and the summary of the test are shown in Figure 17. Furthermore, waveforms of three of these 14 tests are presented in Figure 18, Figure 19 and Figure 20.

Figure 17 Test results summary

4.3 4.35 4.4 4.45 4.5 4.55 4.6 4.65 4.7 4.75 4.8

[A]

0123456

Trip current (5%, 100%, D1/DC-)

D1D2

4.3 4.35 4.4 4.45 4.5 4.55 4.6 4.65 4.7 4.75 4.8

Boo

lean

0

0.5

1

1.5Trip signal (5%, 100%, D1/DC-)

D1D2

4.3 4.35 4.4 4.45 4.5 4.55 4.6 4.65 4.7 4.75 4.8

[p.u

]

0.998

0.999

1

1.001DC link voltage (5%, 100%, D1/DC-)

D1D2

4.3 4.35 4.4 4.45 4.5 4.55 4.6 4.65 4.7 4.75 4.8

[p.u

]

-0.1

0

0.1D1 NW phase current (5%, 100%, D1/DC-)

4.3 4.35 4.4 4.45 4.5 4.55 4.6 4.65 4.7 4.75 4.8

[p.u

]

-0.1

0

0.1D2 NW phase current (5%, 100%, D1/DC-)

4.3 4.35 4.4 4.45 4.5 4.55 4.6 4.65 4.7 4.75 4.8

[p.u

]

-0.2

0

0.2D1 MB phase current (5%, 100%, D1/DC-)

Time [s]4.3 4.35 4.4 4.45 4.5 4.55 4.6 4.65 4.7 4.75 4.8

[p.u

]

-0.2

0

0.2D2 MB phase current (5%, 100%, D1/DC-)

Figure 18 Earth fault discrimination tripping for Type 2 fault at Drive 1. Drive 1 runs at 5% speed and Drive 2 at 100%

5 5.05 5.1 5.15 5.2 5.25 5.3 5.35 5.4 5.45 5.5

[A]

0123456

Trip current (100%, 100%, D1/W)

D1D2

5 5.05 5.1 5.15 5.2 5.25 5.3 5.35 5.4 5.45 5.5

Boo

lean

0

0.5

1

1.5Trip signal (100%, 100%, D1/W)

D1D2

5 5.05 5.1 5.15 5.2 5.25 5.3 5.35 5.4 5.45 5.5

[p.u

]

0.998

1

1.002

1.004DC link voltage (100%, 100%, D1/W)

D1D2

5 5.05 5.1 5.15 5.2 5.25 5.3 5.35 5.4 5.45 5.5

[p.u

]

-0.1

0

0.1D1 NW phase current (100%, 100%, D1/W)

5 5.05 5.1 5.15 5.2 5.25 5.3 5.35 5.4 5.45 5.5

[p.u

]

-0.1

0

0.1D2 NW phase current (100%, 100%, D1/W)

5 5.05 5.1 5.15 5.2 5.25 5.3 5.35 5.4 5.45 5.5

[p.u

]

-0.2

0

0.2D1 MB phase current (100%, 100%, D1/W)

Time [s]5 5.05 5.1 5.15 5.2 5.25 5.3 5.35 5.4 5.45 5.5

[p.u

]

-0.2

0

0.2D2 MB phase current (100%, 100%, D1/W)

Figure 19 Earth fault discrimination tripping for Type 3 fault at Drive 1. Drive 1 and 2 run at 100% speed

4.1 4.15 4.2 4.25 4.3 4.35 4.4 4.45 4.5 4.55 4.6

[A]

0123456

Trip current (100%, 5%, D1/W)

D1D2

4.1 4.15 4.2 4.25 4.3 4.35 4.4 4.45 4.5 4.55 4.6

Boo

lean

0

0.5

1

1.5Trip signal (100%, 5%, D1/W)

D1D2

4.1 4.15 4.2 4.25 4.3 4.35 4.4 4.45 4.5 4.55 4.6

[p.u

]

0.998

1

1.002

1.004DC link voltage (100%, 5%, D1/W)

D1D2

4.1 4.15 4.2 4.25 4.3 4.35 4.4 4.45 4.5 4.55 4.6

[p.u

]

-0.1

0

0.1D1 NW phase current (100%, 5%, D1/W)

4.1 4.15 4.2 4.25 4.3 4.35 4.4 4.45 4.5 4.55 4.6

[p.u

]

-0.1

0

0.1D2 NW phase current (100%, 5%, D1/W)

4.1 4.15 4.2 4.25 4.3 4.35 4.4 4.45 4.5 4.55 4.6

[p.u

]

-0.2

0

0.2D1 MB phase current (100%, 5%, D1/W)

Time [s]4.1 4.15 4.2 4.25 4.3 4.35 4.4 4.45 4.5 4.55 4.6

[p.u

]

-0.2

0

0.2D2 MB phase current (100%, 5%, D1/W)

Figure 20 Earth fault discrimination tripping for Type 3 fault at Drive 1. Drive 1 runs at 100% speed and Drive 2 at 5%

CONCLUSIONS

The challenge of protecting unearthed or high resistance grounded systems with Transformer-less AFE drive from earth faults has been defined as the protective selectivity. In order to achieve high reliability, the faulty drive needs to be discriminated from the healthy drives and disconnected timely. According to the location of the earth fault, three types of faults have been defined and characteristics of each fault have been defined, discussed and analysed.

Novel detection methods for these different types of earth faults have been presented and an equivalent circuit approach has been introduced to derive the detection method. Furthermore, a low cost, software based and simple implementation method and the associated protection settings have been proposed. Modelling and practical testing have shown that the proposed methods can protect the system from earth faults effectively, regardless of the fault location. Finally, enhancements of these methods for different requirements have been presented and discussed.

REFERENCES

[1] Das, J.C.; Osman, R.H., "Grounding of AC and DC low-voltage and medium-voltage drive systems," in Industry Applications, IEEE Transactions on , vol.34, no.1, pp.205-216, Jan/Feb 1998

[2] Dunki-Jacobs, J.R., "The Effects of Arcing Ground Faults on Low-Voltage System Design," in Industry Applications, IEEE Transactions on , vol.IA-8, no.3, pp.223-230, May 1972

[3] Kastha, D.K.; Bose, B.K., "Investigation of fault modes of voltage-fed inverter system for induction motor drive," in Industry Applications, IEEE Transactions on , vol.30, no.4, pp.1028-1038, Jul/Aug 1994

[4] IEC 60533:2015, Electrical and electronic installations in ships – Electromagnetic compatibility (EMC) – Ships with a metallic hull, IEC Standard.

ACKNOWLEDGEMENTS

The authors are grateful for the permission of GE Power Conversion to publish this paper.