1. Multi-Agent Testbed for Emerging Power Systems Mark Stanovich, Sanjeev Srivastava, David A. Cartes, Troy Bevis

2. Outline Introduction Testbed description Example uses – Shipboard power system control – Characterization of data communication latencies on control Lessons learned Future work Conclusion

3. Emerging Power Systems More complex interactions Frequent changes to the electrical infrastructure Distributed generation (e.g., renewables) Decentralized and distributed Cyber-physical – Data communications facilities – Computational resources

4. Multi-Agent Testbed Testbeds – Validation and verification – Feasibility studies – Derisking Conventional power system testbeds – Focused on centralized paradigm – Limited tools to program complex control algorithms Large amounts of time prototyping Difficult to implement correctly

5. Requirements Driven Design Support a variety of software – Operating systems E.g., Linux, Windows, Vx Works – Applications and programming languages E.g, Matlab, C++, Java, JADE Data communications – E.g., TCP/IP Cost effective Portable Versalogic “Mamba” SBCs (x86 Core 2 Duo processor)

6. CAPS Multi-Agent Systems Testbed Power system simulation Computational platforms Data communication infrastructure – Inter-agent – Power system RT and non-RT simulation Controller hardware-in- the-loop

7. Example Use Cases Shipboard control Characterization of latency on controls

8. Distributed Control for Shipboard Power Systems Explore control capabilities for shipboard power systems Two-layer hierarchical control architecture – Intelligent agent modules interact to perform control functionalities Load shedding Human machine interface Enhance survivability by delivering power to needed loads (mission critical) (triage) in both normal and adverse conditions – Supervisory and zonal control – Demonstrates the loss and recovery of main generator

9. Shipboard Distributed Control *Work by Qunying Shen

10. Implementation on Testbed Each layer on separate Mamba board – Logical representation expressed physically – Characteristic of actual implementation Java Agent DEvelopment Framework (JADE) – Facilities to simplify data communications amongst agents – Easily configure agent to machine layout – Cross-platform (e.g., Windows, Linux) Electrical ship system on RTDS

11. Multi-agent System Layout System PCON Agent MAMBA 4 Zonal PCON Agent MAMBA 5 Load Agent MAMBA 1 Zonal ESM AGENT MAMBA 1 Main Gen 1 Agent MAMBA 6 Aux Gen 1 Agent MAMBA 6 Aux Gen 2 Agent MAMBA 2 Main Gen 2 Agent MAMBA 2 System ESM Agent MAMBA 6 Propulsion Agent 1 MAMBA 2 Propulsion Agent 2 MAMBA 2 Pulse Load Agent MAMBA 5 HMI Operator Status Control

12. Characterizing Latency Effects on Multi-Agent Control Use information theory to characterize the system Utilized case study to develop guidelines for applying information theory to energy systems Extracting timing requirements from characterization

13. Energy System Applications Previous Applications & Studies – PMU placement (state observation) – Multiagent communication (security) – Ship system reconfiguration (entropy rate) Possible Applications: – Extracting timing requirements from energy system – System control of large-scale renewable energy systems with many DER, such as a smart grid – Quantifying information loss in distributed intelligent control systems

14. Generator Sync. System Diagram Read n PMU msgs Find Δθ i for all G i Find Δθ av and θ new =Δθ i - Δθ av Send θ new and breaker = ON

15. Implementation on Testbed Implement on separate Mamba – Local generator controls – Supervisor control Intermediate Mamba to introduce communication variabilities RTDS – Electrical and low-level generator controls

16. Lessons Learned Need for automation – Artifacts from one experiment to the next – Less error prone – Time consuming to manually setup testbed for various configurations – “Wiring” for electrical connections (interface to power systems simulator) Benefits of real-time simulation

17. Lessons Learned Multiple computational facilities – General purpose High and low performance – FPGA – PLC Data communications simulation – Topologies – Communication technologies (e.g., wireless, fiber- optic)

18. Conclusion Testbed has been useful – Demonstrations – Study cyber-physical interactions – Quick prototyping – Simulate existing/planned systems Design decisions Composing systems Improvements – Increase diversity of agent hosts – More advanced automated setup – Data communication simulation

19. Acknowledgement This work was partially supported by the National Science Foundation (NSF) under Award Number EEC-0812121 and the Office of Naval Research Contract #N00014-09-C-0144.