基于聚类分析和GRNN神经网络的短期负荷预测  

Short Term Load Forecasting Based on Clustering Analysis and GRNN Neural Network

作　　者：李春涛 LI Chun-tao(State Grid Yongzhou Power Supply Company,Yongzhou 425000,China)

机构地区：[1]国网湖南省电力有限公司永州供电分公司,湖南永州425000

出　　处：《电气开关》2025年第1期51-54,共4页Electric Switchgear

摘　　要：为改善神经网络模型短期负荷预测精度,提出了一种基于K均值聚类分析和改进广义回归神经网络(general regression neural network,GRNN)的短期地区电网负荷预测方法。首先对某地区原始电网负荷数据进行预处理,剔除坏数据,选取相似日负荷序列,利用K均值聚类方法将地区电网相似日负荷序列进行归类处理,然后通过改进的GRNN神经网络进行地区电网负荷预测。该数据处理方法改善了单一GRNN神经网络模型的预测精度,并通过预测实例进行有效验证。To improve the accuracy of short-term load forecasting using neural network models,a short-term regional power grid load forecasting method based on K-means clustering analysis and improved general regression neural network(GRNN)is proposed.Firstly,preprocess the original power grid load data of a certain region,eliminate bad data,select similar daily load sequences,use K-means clustering method to classify and process similar daily load sequences of the regional power grid,and then use an improved GRNN neural network for regional power grid load forecasting.This data processing method improves the prediction accuracy of a single GRNN neural network model and is effectively validated through prediction examples.

关 键 词：聚类分析 神经网络 短期负荷预测 

分 类 号：TM71[电气工程—电力系统及自动化]