浙江电力2025，Vol.44Issue(12)：12-21,10.DOI:10.19585/j.zjdl.202512002
基于M-BigST的电力系统频率预测方法
A power system frequency prediction method based on M-BigST
摘要
Abstract
To solve the problems of high computational complexity,modeling difficulty,and the challenge of balanc-ing accuracy with efficiency in conventional power system frequency analysis methods,this paper proposes a fre-quency prediction method based on M-BigST(modified BigST).Firstly,a block-level dynamic graph learning mod-ule and a linear spatial convolutional layer are employed to extract spatial correlation features embedded in the power grid topology,capturing local dependencies among nodes and generating high-dimensional spatial semantic information.Then,sliding convolution kernels are used to accurately capture the local temporal dependencies and short-term dynamic characteristics of system frequency,enabling a frequency prediction model that jointly considers temporal and spatial features.Finally,actual grid operation data from a certain region are used for validation.The re-sults show that,compared with other methods,the proposed method offers significant advantages in prediction accu-racy and robustness.关键词
电力系统频率/改进的BigST/时序预测/网络拓扑/系统频率预测Key words
power system frequency/M-BigST/temporal prediction/network topology/system frequency prediction引用本文复制引用
汪旸,董向明,张越,陈钟钟,乔咏田,杨丘帆,姜涛..基于M-BigST的电力系统频率预测方法[J].浙江电力,2025,44(12):12-21,10.基金项目
国家电网有限公司总部科技项目(5100-202404010A-1-1-ZN) （5100-202404010A-1-1-ZN）
