基于联邦学习的能源聚合服务商负荷预测  

Federated Learning-Based Load Forecasting for Energy Aggregation Service Providers

作　　者：黄一川 宋瑜辉 荆朝霞[1] HUANG Yichuan;SONG Yuhui;JING Zhaoxia(School of Electric Power Engineering,South China University of Technology,Guangzhou 510641,China)

机构地区：[1]华南理工大学电力学院,广州市510641

出　　处：《电力建设》2025年第1期37-47,共11页Electric Power Construction

基　　金：国家自然科学基金项目(52107106)。

摘　　要：能源聚合服务商有效的负荷预测能力是其参与电力市场的基础,并且是其最终获取最大化市场利润的重要因素。然而,其面临的一大挑战是,预测所需的数据大都分散在各个能源个体手中,由于数据隐私和安全的考虑,各能源个体通常不愿意进行数据共享。这一难题不仅制约了能源聚合服务商运营效率的提升,还对相关能源供应商运营模式的研究与分析形成了一定障碍。为此,提出一种基于联邦学习的能源聚合服务商负荷预测方法,该方法允许在不接触真实数据集的情况下共享它们的信息,从而获得准确有效的负荷预测能力。首先,根据任务需求建立多维环境特征选择的人工神经网络;然后,利用加权平均策略联合训练人工神经网络,在各个能源个体和环境特征之间架起桥梁;最后,将其各自训练过的模型特征发送到能源聚合服务商的服务端进行拟合处理。通过对中国南方某市大型能源聚合服务商的部分数据进行算例分析,结果表明,该方法在提高整体运营效率和平均预测精度方面均取得良好成效。Energy aggregation service providers in the electricity market are highly dependent on the load-forecasting accuracy to maximize their market interests.However,the required forecasting data are typically scattered among different energy suppliers,and owing to data security and privacy concerns,these suppliers are often unwilling to share their data.Inaccurate load forecasting reduces the operational efficiency of energy aggregation service providers and limits the in-depth understanding of supplier operating models.To address this issue,this study proposes a load-forecasting model based on federated learning,which achieves an effective improvement in load-forecasting accuracy while ensuring data security and privacy.First,an artificial neural network with multidimensional environmental feature selection is established according to the task requirements.Subsequently,a weighted-average strategy was used to jointly train the artificial neural network,bridging the gap between individual energy entities and environmental features.Finally,the model features trained separately are sent to the server of the energy-aggregation service provider for the fitting process.The results of a case analysis using actual data from a large energy aggregation service provider in a southern city in China demonstrate that the proposed model significantly improves operational efficiency and forecasting accuracy.

关 键 词：能源聚合服务商 负荷预测 联邦学习 数据隐私 神经网络 

分 类 号：TM73[电气工程—电力系统及自动化]