基于机器学习方法的现货电价预测研究综述  

Review of Spot Electricity Price Prediction Studies Based on Machine Learning Methods

作　　者：加鹤萍 郭宇辰 马乾鑫 杨争林 郑亚先 曾丹 刘敦楠 JIA Heping;GUO Yuchen;MA Qianxin;YANG Zhenglin;ZHENG Yaxian;ZENG Dan;LIU Dunnan(Beijing Key Laboratory of New Energy and Low-Carbon Development,Beijing 102206,China;School of Economics&Management,North China Electric Power University,Beijing 102206,China;China Electric Power Research Institute,Beijing 100192,China)

机构地区：[1]新能源电力与低碳发展研究北京市重点实验室,北京市102206 [2]华北电力大学经济与管理学院,北京市102206 [3]中国电力科学研究院有限公司,北京市100192

出　　处：《电力建设》2025年第2期160-179,共20页Electric Power Construction

基　　金：国家重点研发计划项目(2022YFB2403000);北京市社会科学基金决策咨询项目(24JCC101);北京市自然科学基金面上项目(9242015)。

摘　　要：全国统一电力市场建设背景下,现货市场建设有助于推动电力资源在更大范围共享互济和优化配置。现货电价作为市场参与者的重要决策信息,对现货市场的辅助决策、市场运行监测及风险管理等至关重要。机器学习方法的快速发展为电价预测提供了可行途径。首先,分析全国统一电力市场下现货电价的特点及其影响因素,并基于现有电价研究的预测机制阐述预测模型的类别及现货电价预测面临的挑战。其次,基于数据标签、特征提取和数据流动控制等特点总结各类机器学习预测模型的研究现状,分析不同预测模型的特点及适用性。然后,分析基于机器学习的现货电价预测模型评价指标,并总结相关预测方法的模型参数训练要求及实际应用情况。最后,针对机器学习方法在电价预测研究中的挑战,对未来技术研究方向进行展望,以期对全国统一电力市场建设下的现货市场发展提供具有建设性意义的参考。In the context of developing a unified national electricity market,the development of a spot market helps promote the sharing and optimal allocation of electricity resources on a larger scale.As important decision-making information for market participants,spot electricity prices are crucial for auxiliary decision-making in the spot market,market operation monitoring,and risk management.The rapid development of machine learning methods provided a feasible approach for electricity price prediction.This study first analyzed the characteristics of spot electricity prices and their influence on the unified national electricity market.The types of prediction models and challenges faced by spot electricity price prediction can be elaborated based on existing research on electricity price prediction mechanisms.In addition,based on the characteristics of data labeling,feature extraction and data flow control,the research status of various machine learning prediction models was summarized,and the characteristics and applicability of different prediction models were analyzed.This study then analyzed the evaluation criteria for spot electricity price prediction models based on machine learning,and summarized the model hyperparameter training requirements and the practical application of relevant prediction methods.Finally,in view of the challenges of machine learning methods in electricity price prediction research,this study outlined future research directions to provide constructive references for the development of the spot market under the construction of a unified national electricity market.

关 键 词：全国统一电力市场 现货市场 电价预测 机器学习方法 

分 类 号：TM76[电气工程—电力系统及自动化]