基于纵向联邦回归的隐私保护电力负荷预测模型  

A Privacy-Protected Electric Load Forecasting Model Based on Vertical Federated Regression

作　　者：王翀 高莉莎[2] 许梦晗 WANG Chong;GAO Lisha;XU Menghan(Information&Communication Branch,State Grid Jiangsu Electric Power Co.,Ltd.,Nanjing Jiangsu 210024,China;State Grid Nanjing Power Supply Company,Nanjing Jiangsu 210019,China)

机构地区：[1]国网江苏省电力有限公司信息通信分公司,江苏南京210024 [2]国网南京供电公司,江苏南京210019

出　　处：《电子器件》2025年第1期187-193,共7页Chinese Journal of Electron Devices

基　　金：国网江苏省电力有限公司科技项目(J2022057)。

摘　　要：在智能电网广泛应用的背景下,电力部门需要和政府、气象、金融和其他企业进行合作。如何在隐私保护的前提下实现多方的数据交互、共享和计算是亟待解决的问题。提出了隐私保护电力负荷预测模型并阐述了对应的模型参数提取算法,该模型可以将各合作方数据进行变换,并利用纵向联邦回归和同态加密技术实现数据的隐私共享。最后,利用某市的电力负荷、气温和限电政策数据对本文所提出的模型和算法进行了验证。可为电力负荷的协同预测提供了基本理论和框架。With popularization of the smart grids applications,the power sector needs to cooperate with government,meteorological,fi-nancial,and other enterprises.How to realize multi-party data interaction,sharing and computing under the premise of privacy protection is an urgent problem to be solved.A privacy-protected electric load forecasting model is proposed and the corresponding model parame-ter extraction algorithm is described.The proposed model can transform the data of each partner,and use vertical federated regression and homomorphic encryption technology to realize data privacy sharing.Finally,the proposed model and algorithm are verified by using the realistic electric load,temperature,and power rationing policy data of a city.This can provide the basic theory and framework for the collaborative prediction of electric load.

关 键 词：电力负荷预测 纵向联邦回归 隐私保护 同态加密 多方安全计算 

分 类 号：TM743[电气工程—电力系统及自动化]