摘要
当前,新型电力系统中的负荷预测正面临多重挑战,传统基于历史负荷规律统计与机器学习的负荷预测方法难以有效应对数据可用性瓶颈、多模态因素解耦、极端场景下的预测建模难题。面对现有负荷预测技术的不足,文章回顾了时序大模型、语义大模型、多模态大模型、检索增强生成技术、智能体技术、思维链技术的研究现状,并分析其在数据异常检测及修复、负荷影响因素分析、小样本及零样本场景下预测等方面的成效与应用前景。展望了大模型及相关技术的发展前景,未来随着技术创新与制度供给,构建智能预测体系将推动电力电量预测向精准、高效、智能化发展。
Currently,load forecasting in a new type of power system is facing multiple challenges.Traditional load forecasting methods based on historical load pattern statistics and machine learning are difficult to effectively address data availability bottlenecks,multi-modal factor decoupling,and modeling difficulties in extreme scenarios.In the face of the shortcomings of existing load forecasting technologies,this paper reviews the research status of time series large models,semantic large models,multimodal large models,retrieval enhancement technologies,intelligent agent technologies,and thought chain technologies,and analyzes their effectiveness and application prospects in data anomaly detection and repair,load influencing factor analysis,small sample and zero sample scenario prediction,etc.Looking ahead to the development prospects of large models and related technologies,in the future,with technological innovation and institutional supply,the construction of an intelligent forecasting system will promote the development of accurate,efficient,and intelligent power and electricity forecasting.
作者
韦于思
金璐
高艳玲
邱敏
徐靖
赵伟博
周颖
陈宋宋
WEI Yusi;JIN Lu;GAO Yanling;QIU Min;XU Jing;ZHAO Weibo;ZHOU Ying;CHEN Songsong(China Electric Power Research Institute,Haidian District,Beijing 100192,China;National Center of Standards Evaluation,SAMR,Xicheng District,Beijing 100032,China)
出处
《电力信息与通信技术》
2025年第12期89-100,共12页
Electric Power Information and Communication Technology
基金
智能电网重大专项(2030)资助(2025ZD0804800)。
关键词
新型电力系统
用电需求分析预测
时序大模型
语义大模型
a new type of power system
electricity demand analysis and forecasting
time series large model
semantic large model
