中国电机工程学报2025，Vol.45Issue(19)：7456-7467,中插6-中插16,23.DOI:10.13334/j.0258-8013.pcsee.240616
基于STD-ST-Former的现货电价长步时空预测
Long Step Spatial-temporal Prediction of Spot Electricity Price Based on STD-ST-Former
摘要
Abstract
Accurate prediction of spot price can provide beneficial guidance for market participants in trading strategies.In order to improve the long step spatial-temporal prediction accuracy of regional spot electricity price,a prediction framework using dual channel spatial-temporal transformer(ST-Former)based on seasonal trend decomposition is proposed.This prediction framework adapts the idea of decomposition,and obtain more predictable results including a stable trend part and a fluctuating seasonal part by untangling the entangled time patterns in regional spot electricity prices.On this basis,a ST-Former model based on the combination of variable Patch temporal attention and spatial self-attention is proposed,to explore the long step spatial-temporal features of regional spot electricity prices.Then,the ST-Former model is used to model the trend part and seasonal part,forming a dual channel model framework.Finally,the output results of the dual channel model framework are aggregated to obtain the final prediction of regional spot electricity prices.Taking the spot electricity price data of the Australian electricity market as an example,the effectiveness and superiority of the proposed prediction framework are verified by comparing with twelve prediction methods.关键词
电力市场/区域现货电价预测/季节趋势分解/注意力机制/时空相关性/长步时空预测Key words
electricity market/prediction of regional spot electricity price/seasonal trend decomposition/attention mechanism/spatial-temporal correlation/long step spatial-temporal prediction分类
信息技术与安全科学引用本文复制引用
张鹏飞,胡博,胡展硕,罗桓桓,刘桁宇,邢作霞..基于STD-ST-Former的现货电价长步时空预测[J].中国电机工程学报,2025,45(19):7456-7467,中插6-中插16,23.基金项目
国家自然科学基金项目(U22B20115).Project Supported by National Natural Science Foundation of China(U22B20115). （U22B20115）
