电网技术2025，Vol.49Issue(7)：2680-2690,中插5-中插7,14.DOI:10.13335/j.1000-3673.pst.2024.0692
基于物理引导神经网络的静态电压稳定预防控制策略在线计算
Online Computation of Static Voltage Stability Preventive Control Strategy Based on Physics-guided Neural Network
摘要
Abstract
The preventive control of static voltage stability in power systems is a typical non-convex and non-linear problem.Traditional model-driven methods have difficulties balancing decision-making accuracy and efficiency.While data-driven methods improve the above issues,their generalization ability and constraint satisfaction are questionable.This paper proposes an online computational method for static voltage stability preventive control based on a physics-guided neural network(PGNN)to address this.Firstly,based on the input convex neural network(ICNN),the non-convex static voltage stability margin constraint is transformed into a convex proxy model that retains nonlinearity.Then,based on PGNN,an end-to-end parameterized preventive control model of ICNN proxy is constructed.Combined with the linearized DC power flow equation considering voltage and reactive power,a convex physical information loss function conducive to the convergence of PGNN training is designed to enforce PGNN to satisfy physical laws.The test results of the IEEE 118-bus system and the n×118-bus system show that compared with traditional model-driven algorithms and purely data-driven methods,the proposed method can better balance the accuracy and efficiency of preventive control while strictly satisfying safety constraints.关键词
静态电压稳定性/预防控制/物理引导神经网络Key words
static voltage stability/preventive control/physics-guided neural network分类
信息技术与安全科学引用本文复制引用
明巧,刘友波,邱高,刘季昂,刘俊勇..基于物理引导神经网络的静态电压稳定预防控制策略在线计算[J].电网技术,2025,49(7):2680-2690,中插5-中插7,14.基金项目
国家自然科学基金项目(52307124).Project Supported by the National Natural Science Foundation of China(52307124). （52307124）
