Applied AI and Machine Learning for Next-Generation Power Distribution
SCADS combines realistic testbeds with learning-based control. CPSyNet generates full cyber-physical distribution feeders, and an open-source Gym wrapper lets reinforcement-learning agents interact with OpenDSS in real time. Using this platform, we (i) pre-train a soft-actor-critic agent with OPF solutions for near-optimal battery-storage dispatch, (ii) run a cluster-level MPC that coordinates PV inverters and EV chargers for voltage control, and (iii) deploy an adaptive model that disaggregates behind-the-meter PV from net smart-meter data. Together these tools advance reliable, data-driven operation in DER-rich grids.
Papers
- L. Wang, J. Halvorsen, S. Pannala, A. Srivastava, A. H. Gebremedhin, N. Schulz.
CPSyNet: A Tool for Generating Customized Cyber-Power Synthetic Network for Distribution Systems with Distributed Energy Resources, IET Smart Grid 5 (6): 463–477 (2022).
Abstract | Paper - L. Wang, A. Dubey, A. H. Gebremedhin, A. Srivastava, N. Schulz.
MPC-Based Decentralized Voltage Control in Power Distribution Systems with EV and PV Coordination, IEEE Transactions on Smart Grid 13 (4): 2908–2919 (2022).
Abstract | Paper - R. Saeedi, K. S. Sajan, K. Davies, A. Srivastava, A. H. Gebremedhin.
An Adaptive Machine Learning Framework for Behind-the-Meter Load/PV Disaggregation, IEEE Transactions on Industrial Informatics 17 (10): 7060–7069 (2021).
Abstract | Paper - G. Krishnamoorthy, A. Dubey, A. H. Gebremedhin.
An Open-source Environment for Reinforcement Learning in Power Distribution Systems, IEEE Power & Energy Society General Meeting (PESGM 2022).
Abstract | Paper - G. Krishnamoorthy, A. Dubey, A. H. Gebremedhin.
Reinforcement Learning for Battery Energy Storage Dispatch Augmented with Model-based Optimizer, IEEE SmartGridComm 2021.
Abstract | Paper