“22nd PSCC 2022 papers submission and review platform

Full Program »

Learning to Solve DCOPF: A Duality Approach

View File
PDF
3.2MB

The optimal power flow (OPF) problem is a fundamental tool in power system operation and control. Because of the increase in uncertain renewable resources, solving OPF problems fast and accurately provides significant values because of a large number of load and generation scenarios need to be accounted for. Recent works have focused on using neural networks to replace iterative solvers to speed up the computation of OPF problems. A critical challenge is to ensure solutions satisfy the hard constraints, which is difficult to do in end-to-end machine learning. In this work, by leveraging the rich theory of duality and physical interpretations of OPF, we design a learning approach that guarantees constraint satisfaction. This approach is an order of magnitude faster than standard solvers and outperforms other learning methods in terms of feasibility and optimality.

Author(s):

Yize Chen    
University of Washington
United States

Ling Zhang    
University of Washington
United States

Baosen Zhang    
University of Washington
United States

 

Powered by OpenConf®
Copyright ©2002-2014 Zakon Group LLC