电力建设2026，Vol.47Issue(4)：63-81,19.DOI:10.12204/j.issn.1000-7229.2026.04.006
基于Informer-MATD3的风力发电商现货价格预测-日前竞价两阶段决策模型
Two-Stage Day-Ahead Bidding Decision Model for Wind Power Generator's Spot Price Forecasting Based on Informer and MATD3
摘要
Abstract
[Objective]To address the challenges of high electricity price volatility and forecasting difficulties caused by the high penetration of wind power in the new power system,this paper proposes a dynamic bidding strategy suitable for wind power producers.[Methods]The study first constructs a Market-Informer forecasting model that integrates multiple market factors.By introducing key variables such as carbon prices,green electricity certificate prices,and coal prices,it achieves day-ahead electricity price forecasting.Furthermore,the forecast information is embedded into a bidding decision framework designed based on the multi-agent twin delayed deep deterministic policy gradient(MATD3)algorithm.This framework is centrally trained in a market environment that includes hydropower,thermal power,and photovoltaic power producers,which finally contributes to the generation of optimal bidding strategies for wind power producers.[Results]Using data from a European electricity market in 2022 as a case study for bidding,the results show that the directional accuracy coefficient(DAC)of forecasting can reach 94.3%under a 10%error margin,representing an 18.6 percentage-point improvement over the traditional autoregressive integrated moving average(ARIMA)model.This strategy reduces the total system cost by 11.4%,increases the revenue of wind power producers by 9.8%,raises the bid-winning rate by 18.7%,and decreases revenue volatility by 22.3%.[Conclusions]The case study verifies the effectiveness of the"forecasting-decision"dynamic coupling mechanism in enhancing the bidding capacity of renewable energy and promoting low-carbon transition,providing an intelligent decision-making paradigm for power markets with a high penetration of renewable energy.关键词
电价预测/Informer模型/多智能体双延迟深度确定性策略梯度(MATD3)算法/市场竞价Key words
electricity price forecasting/Informer model/multi-agent twin delayed deep deterministic policy gradient(MATD3)/market bidding分类
信息技术与安全科学引用本文复制引用
张硕,王雨欣,李英姿,贺运政..基于Informer-MATD3的风力发电商现货价格预测-日前竞价两阶段决策模型[J].电力建设,2026,47(4):63-81,19.基金项目
国家重点研发计划项目(2021YFB2400704) （2021YFB2400704）
教育部人文社会科学研究规划基金(23YJA630133) （23YJA630133）
北京市自然科学基金项目(9232019) （9232019）
北京市社会科学基金项目(22GLB020) This work is supported by National Key Research and Development Program of China(No.2021YFB2400704),Humanities and Social Science Fund of Ministry of Education of China(No.23YJA630133),Beijing Natural Science Foundation(No.9232019)and Beijing Social Science Foundation(No.22GLB020). （22GLB020）
