Download - Large scale simulation ship power system hebner-herbst-gatozzi - july 2010

Transcript
Page 1: Large scale simulation ship power system  hebner-herbst-gatozzi - july 2010

Large Scale Simulation of a Ship Power System with Energy Storage and Multiple Directed Energy Loads

R. E. Hebner, J. D. Herbst, A. L. GattozziCenter for ElectromechanicsUniversity of Texas, Austin

July 13, 2010

Page 2: Large scale simulation ship power system  hebner-herbst-gatozzi - july 2010

Presentation Overview

• Ship Design Challenges & Power System Studies• CEM Ship Power System Model• Modeling Issues & Simulation Alternatives• Path Forward

Page 3: Large scale simulation ship power system  hebner-herbst-gatozzi - july 2010

Challenges to Naval Power System Designers

• Wide variety of loads on board – Continuous duty to pulsed– Loads with different requirements

(low freq. ac, high freq. ac, dc)

• Generated power capacity minimally larger than averagetotal load and smaller than peak load– Reliance on energy storage to supply intermittent loads

• Increased use of power electronics• Need to maintain power quality and stability margins• Need flexible architecture suitable for fault

management and reconfiguration

Page 4: Large scale simulation ship power system  hebner-herbst-gatozzi - july 2010

Ship Power System Studies

• Modeling and Simulation play a crucial role– Experimental data not easy to generate– Difficult to reproduce the complexity of system

interactions in subscale physical models– Scarcity of accumulated experience with non-

traditional energy resources and loads• Models provide a common base to evaluate

alternatives and study component interactions• Key concerns:

– Scale of the model– Modeling technique– Software and Hardware PlatformsNeed a flexible tool for the ship designer, not application specific codes

Page 5: Large scale simulation ship power system  hebner-herbst-gatozzi - july 2010

CEM’s Notional Ship Power System

• Two turbo-generators with cross-connect option• Two flywheel energy storage systems• System designed around common 6 kVdc bus• Subsystems modeled :

– Propulsion– Hotel Loads– Free Electron Laser (FEL)– AN/SQQ-90 Sonar System– Electromagnetic Rail Gun– Active Denial System– Advanced Radar– Laser Weapon System– Electromagnetic Aircraft Launch System

Page 6: Large scale simulation ship power system  hebner-herbst-gatozzi - july 2010

Conceptual Diagram of CEM’s Ship Model

Page 7: Large scale simulation ship power system  hebner-herbst-gatozzi - july 2010

Interactive ModelOperator can change in real time,as the simulation is running,the status of allanalog gains,pushbuttons,switches.

Page 8: Large scale simulation ship power system  hebner-herbst-gatozzi - july 2010

FEL Operational Modes

Pier Side FEL in minimal power state; necessary maintenance performed (425 kW)

Underway Ship is crossing open waters; no imminent threat (625 kW)

Hot Standby Ship is in combat theater; threat could appear at any time (1 MW)

Engagement FEL is firing upon incoming threat (17 MW)

Page 9: Large scale simulation ship power system  hebner-herbst-gatozzi - july 2010

Collaboration with NAVSEA, NPS, Stanford, AES, JLab

FEL Component Overview

Linac10 kW filaments

16 MW RF

Page 10: Large scale simulation ship power system  hebner-herbst-gatozzi - july 2010

Current

Power:Active – yellowReactive-pink

Voltage

ACTIVE DENIAL EM GUN

Voltage

Current

Heat

ControlSignals

Page 11: Large scale simulation ship power system  hebner-herbst-gatozzi - july 2010

Modeling Issues

• Modeling some very unusual loads, many still experimental or in the R&D stage

• Complex model results in long simulation times

• Typical values are σ ≈ 100,000 (real time is σ = 1)– 6 seconds simulated time = 1 week running time on a

64-bit, 3.16 GHz, 3.93 GB, dual core desktop.

,

simulation time

simulated time

Page 12: Large scale simulation ship power system  hebner-herbst-gatozzi - july 2010

• Segmentation of the simulation model – Run one section at a time– Creates interface issues similar to parallel processing– Makes interpretation of results more difficult

• Compression of operating scenarios– Not reflective of realistic operating scenarios– Affects component interactions– Prevents real time operator engagement

• Eliminated GUI due to impact on simulation times

– Makes interpretation of results more difficult

Simulation Alternatives & Issues

Page 13: Large scale simulation ship power system  hebner-herbst-gatozzi - july 2010

Multi-Rate / Multi-Core Options• Expanded use of multi-rate techniques

– Models run now on dual rates– Expansion to further levels is possible

• Multi-core calculations:– MATLAB/Simulink version for parallel computations (Parallel Computing

Toolbox) run on quad core computer resulting in speed gains of ~2-3:1– Third-party supported parallel MATLAB not being pursued now but

remains an option• MATLAB/Simulink run on computer cluster (Distributed

Computing Server) in cooperation with the Texas Advanced Computing Center (TACC)– Work is ongoing– Goal is 10:1 speed gain

• Develop custom code to fully exploit parallel processing– Kept as a last option due to cost and specificity of resulting code

Page 14: Large scale simulation ship power system  hebner-herbst-gatozzi - july 2010

Heterogeneous ComputationFPGA assisted processing:

FPGAs outperform CPUs by >1 order of magnitude in speed

Potential solution should: 1. Retain broad utility of programs developed 2. High degree of generality

Pursue COTS suppliers e.g.National Instruments (NI),Xilinx, etc.

Page 15: Large scale simulation ship power system  hebner-herbst-gatozzi - july 2010

Heterogeneous Computation

• NI offers a hybrid architecture that can be exploited by their LabView software

• LabView is an intrinsically parallel language – Simulink is sequential

• Our model can be transferred to LabView and executed on NI’s RT-HPC system

The University of Texas has a Long Term Working Relationship with National Instruments

Page 16: Large scale simulation ship power system  hebner-herbst-gatozzi - july 2010

Path Forward• Heterogeneous computation appears to be a

promising path to significant reductions in simulation times– Capabilities of FPGA’s and GPU’s are increasing

• UT-CEM is seeking opportunities to apply these techniques to simulation of Naval power systems– Exploring collaborations with National Instruments

and Xilinx

Questions?