报告题目：Feasible Region Analysis and Hierarchical Optimization Algorithm for AC Power Flow
报告人：Prof. Changhong Zhao
Dr. Changhong Zhao is an Assistant Professor with the Department. Of Information Engineering at the Chinese University of Hong Kong (CUHK). He received BE in Automation from Tsinghua University and PhD in Electrical Engineering from Caltech. He spent years on research and development at the US National Renewable Energy Laboratory (NREL). He received the Demetriades Prize for best thesis in renewable energy, the Charles Wilts Prize for outstanding research in EE at Caltech, the Early Career Award from Hong Kong Research Grants Council, and the IEEE Power and Energy Society Prize Paper Award.
The growth of distributed and renewable energy sources in power system is calling for more scalable and responsive operation schemes than before, particularly new methods to analyze and optimize power flows in distribution networks. I will introduce two pieces of work on this topic. In the first work, we develop an optimization framework to approximate the feasible region of renewable power injections under a nonlinear AC power flow model. We first formulate a power injection feasibility problem, which is then relaxed to a convex second-order cone program (SOCP). We utilize the strong duality of SOCP to characterize the SOCP-relaxed feasible region as a convex polytope. Finally, we apply a heuristic method to remove power injections that make the SOCP relaxation inexact and obtain a more accurate approximation of the feasible renewable power injection region. The second work is a hierarchical distributed algorithm to solve a large-scale optimal power flow (OPF) problem. The proposed algorithm exploits the tree structure of a distribution network to significantly reduce the computational burden of the primal-dual gradient method to solve OPF. To reduce the risk of voltage violation caused by the modeling inaccuracy due to power flow linearization, we propose an improved gradient evaluation method that is more accurate with mild impact on computational efficiency.