基于改进K-means算法的电力短期负荷预测方法研究  

Research on short-term power load forecasting method based on improved K-means algorithm

作　　者：荀超 陈伯建 吴翔宇 项康利 林可尧 肖芬 易杨 XUN Chao;CHEN Bojian;WU Xiangyu;XIANG Kangli;LIN Keyao;XIAO Fen;YI Yang(State Grid Fujian Electric Power Co., Ltd.,Fuzhou 350000, China;Electric Power Science Research Institute,State Grid Fujian Electric Power Co., Ltd.,Fuzhou 350000, China;Economic and Technological Research Institute,State Grid Fujian Electric Power Co., Ltd., Fuzhou 350000, China;College of Electrical Engineering and Automation, Fuzhou University,Fuzhou 350108, China)

机构地区：[1]国网福建省电力有限公司,福建福州350000 [2]国网福建省电力有限公司电力科学研究院,福建福州350000 [3]国网福建省电力有限公司经济技术研究院,福建福州350000 [4]福州大学电气工程与自动化学院,福建福州350108

出　　处：《电力科学与技术学报》2022年第1期90-95,共6页Journal of Electric Power Science And Technology

基　　金：国家自然科学基金(51177107);国家电网有限公司科技项目(52022319003P)。

摘　　要：现有方法预测电力短期负荷时忽略了对其进行聚类优化处理,导致预测耗时较长,短期负荷预测精度偏低。为此,提出一种基于改进K-means算法的电力短期负荷预测方法。该方法利用改进后的K-means算法聚类处理电力负荷大数据,使用聚类后获得的训练样本构建循环神经网络RNN拓扑结构,然后通过对RNN神经网络模型设置最优权值,实现电力负荷的短期预测。实验结果表明,所提方法具有高预测效率和高短期负荷预测精准度。The existing methods ignore the clustering optimization process when predicting the short-term load of electricity,which leads to a long prediction time and a low accuracy of short-term load prediction.Therefore,a short-term load forecasting method based on improved K-means algorithm is proposed.This method uses the improved K-means algorithm to cluster the big data of power load,uses the training samples obtained after clustering to construct the RNN topology structure of the recurrent neural network.Then the optimal weights are set for the RNN neural network model to realize short-term forecast of the power load.The experimental results show that the proposed method has high forecasting efficiency and high short-term load forecasting accuracy.

关 键 词：K-MEANS算法 数据聚类 RNN神经网络模型 电力负荷大数据 预测方法 

分 类 号：TM715[电气工程—电力系统及自动化]