基于数据分解的多区域个性化联邦负荷预测方法  

Personalized Federated Multi-region Load Forecasting Method Based on Data Decomposition

作　　者：焦润海[1] 褚佳杰 李俊良 张炜杰 JIAO Runhai;CHU Jiajie;LI Junliang;ZHANG Weijie(School of Control and Computer Engineering,North China Electric Power University,Changping District,Beijing 102206,China;College of Intelligence and Computing,Tianjin University,Jinnan District,Tianjin 300072,China)

机构地区：[1]华北电力大学控制与计算机工程学院,北京市昌平区102206 [2]天津大学智能与计算学部,天津市津南区300072

出　　处：《中国电机工程学报》2025年第5期1691-1703,I0005,共14页PROCEEDINGS OF THE CHINESE SOCIETY FOR ELECTRICAL ENGINEERING

基　　金：中央高校基本科研业务费专项项目(2022JG004)。

摘　　要：开放电力市场中的小规模主体由于缺乏数据导致负荷预测准确度低,联邦学习在保证数据隐私前提下利用多方数据训练得到考虑多方共性的全局模型,但该模型由于忽略了个性特征无法保证在每个参与方都达到最优预测效果。为此,提出一种基于数据分解的多区域个性化联邦负荷预测方法(personalized federated multi-region load forecasting method based on data decomposition,pFedD)。首先,对原始负荷数据序列分解得到包含不同数据特征的本征模态函数(intrinsic mode functions,IMF);其次,中央服务器根据信号过零率将所有IMF分为高频、低频和趋势分量;最后,根据分量相关性分析,客户端将高频和趋势分量作为个性化分量进行本地模型训练,将低频分量作为联邦分量参与全局模型训练。在中国北方10个地区的真实负荷数据上进行实验,结果表明,pFedD的平均绝对百分比误差(mean absolute percentage error,MAPE)为3.09%,比经典的联邦平均(federated averaging,FedAvg)方法降低了1.67%。Some entities of power market lack sufficient data,leading to low accuracy in load forecasting.The traditional federated learning forecasting method can fully train a global model with multi-client data while preserving each client’s data privacy.However,the global model may not perform optimally on each client due to the neglect of personalized characteristics.Therefore,this paper proposes a personalized federated multi-region load forecasting method based on data decomposition(pFedD).First,the original load data sequence is decomposed to obtain intrinsic mode functions(IMF)containing different data features.Then,using the zero-crossing rate,all IMFs are divided into high-frequency components,low-frequency components,and trend components.Finally,through component correlations analysis,the high-frequency components and trend components are retained in the client as personalized components for local model training,and the low-frequency components participate as federated components of the central server for global model training.Experiments conducted on the load data of 10 areas in northern China demonstrate that the proposed method achieves an average mean absolute percentage error(MAPE)of 3.09%,representing a 1.67%reduction compared to the traditional federated averaging(FedAvg).

关 键 词：负荷预测 联邦学习 个性化 数据分解 分量选择 

分 类 号：TM71[电气工程—电力系统及自动化]