贝叶斯神经网络下的配电网异常负荷数据修正  

Correction of abnormal load data in distribution networks using Bayesian neural network

作　　者：马麒 刘宝林 卢耕儒 MA Qi;LIU Baolin;LU Gengru(Hainan Power Supply Company,State Grid Qinghai Electric Power Company,Hainan 813099,China)

机构地区：[1]国网青海省电力公司海南供电公司,青海海南813099

出　　处：《电子设计工程》2025年第8期188-191,196,共5页Electronic Design Engineering

基　　金：国网青海省电力公司科技项目(522803240001)。

摘　　要：在处理配电网异常负荷数据时,小波神经网络常用于获取负荷预测值,但小波神经网络模型参数固定,动态场景下修正值误差大。因此,提出贝叶斯神经网络下的配电网异常负荷数据修正方法。采用小波阈值收缩法分解和重构负荷信号,得到去噪后的负荷数据。通过模糊C均值聚类实现负荷曲线的分类,并建立特征曲线,以此为参考辨识出异常负荷数据。构建基于贝叶斯神经网络的预测模型,通过学习正常负荷数据的变化特征预测未来负荷,利用预测数据置换异常数据,从而完成异常负荷数据修正。实验结果表明,该方法给出的修正值相对误差保持在1%以内,满足异常负荷数据处理要求。When dealing with abnormal load data in distribution networks,wavelet neural networks are commonly used to obtain load prediction values,but the parameters of the wavelet neural network model are fixed,and the correction value error is large in dynamic scenarios.Therefore,a method for correcting abnormal load data in distribution networks using Bayesian neural network is proposed.Using wavelet threshold shrinkage method to decompose and reconstruct load signals,obtain denoised load data.Classify load curves through fuzzy C-means clustering and establish characteristic curves to identify abnormal load data as a reference.Build a prediction model based on Bayesian neural network,predict future loads by learning the changing characteristics of normal load data,and replace abnormal data with predicted data to complete the correction of abnormal load data.The experimental results show that the relative error of the correction value provided by this method remains below 1%,meeting the requirements for processing abnormal load data.

关 键 词：配电网 异常负荷数据 贝叶斯神经网络 辨识 修正 

分 类 号：TN82[电子电信—信息与通信工程]