AI-Assisted Optimization of Dynamic Operating Envelopes for Flexible Integration of Distributed Energy Resources

The increasing penetration of distributed energy resources (DERs) brings opportunities and challenges to the operation of power distribution systems. Conventionally, the DER hosting capacity of a local electric power system is commonly represented as a static value that is determined by evaluating DER’s impacts on grid reliability considering “worst-case” conditions, resulting in a conservative DER integration and  underutilization of grid infrastructure. However, hosting capacity is not a static value as it is dependent on the coincidence of DER generation and load demand that vary over time. Dynamic Operating Envelopes (DOEs) establish upper and lower operational bounds for DERs over a given time period, within which they can operate without adverse grid impacts, thus unleashing the full potential of grid infrastructure and maximizing the DER integration. This talk will introduce an industry-accepted model to optimize DOEs, which has been integrated into major distribution system management software. Due to the non-convex nature of power flow model, the optimization of DOEs faces a challenge of balancing solution accuracy and computational efficiency. We propose an AI-assisted constraint embedding method that convexifies the problem by replacing power flow equations with trained input convex neural networks (ICNNs). We compare its performance with two conventional optimization methods, i.e., gradient descents and linearization, under various operating scenarios.

Speaker: Zhaoyu Yang, Iowa State University

Register at weblink.  Attend in person or online