摘要
电力设备作为电力系统的基本要素,对其故障风险预测可以有效降低电网故障风险带来的损失。当前应用的高压电力设备故障预测模型忽略了对高压电力设备信号的盲源分离处理,无法去除虚假故障分量,导致故障预测结果不准确、耗时较长的问题。构建新的基于模糊神经网络的高压电力设备故障预测模型。将小波降噪方法引入到盲源分离中,对高压电力设备信号完成盲源分离和小波分解;通过互信息方法将分解结果中的虚假分量删除;利用插值形态滤波的方式提取故障特征,将其设定为模糊神经网络的输入变量,构建高压电力设备故障预测模型。实验结果验证了所构建的模型在30次实验迭代过程中的误差始终不超过2.5%,均方根误差低于3.4%,预测用时测试结果在14~23 ms之间。数据表明所构建模型的预测精度较高、预测速度更快,具有明显的应用优势。
As power equipment is the basic element of the power system,the prediction of its fault risk can effectively reduce the loss caused by grid fault risks.The currently applied fault prediction model of high-voltage power equipment ignores the blind source separation processing of high-voltage power equipment signals,and cannot remove false fault components,resulting in inaccurate fault prediction results and long timeconsuming problems.Therefore,a new fault prediction model of high voltage power equipment based on fuzzy neural network is constructed.The wavelet denoising method is introduced into BSS to complete BSS and wavelet decomposition of high voltage power equipment signals.The false components in the decomposition results are deleted by the mutual information method.The fault features are extracted by interpolation morphological filtering and set as the input variables of fuzzy neural network to construct the fault prediction model of high voltage power equipment.The experimental results show that the error of the model is always less than 2.5%and the root mean square error is less than 3.4%in the process of 30 experimental iterations.The prediction time test results are 14ms~23ms.The data suggests that the prediction accuracy of the model is higher,the prediction speed is faster,and it has obvious application advantages.
作者
谢国财
温锐
陈琛
XIE Guocai;WEN Rui;CHEN Chen(Guangdong Power Grid Co.,Ltd.,Guangzhou 510080,Guangdong,China;Guangdong Power Grid Energy Investment Co.,Ltd.,Guangzhou 510000,Guangdong,China;Qingyuan Power Supply Bureau,Guangdong Power Grid Co.,Ltd.,Qingyuan 511500,Guangdong,China)
出处
《电网与清洁能源》
北大核心
2022年第9期120-125,共6页
Power System and Clean Energy
基金
广东电网有限责任公司科技项目(C3762610K003)。
关键词
模糊神经网络
高压电力设备
故障预测模型
小波降噪
fuzzy neural network
high-voltage power equipment
fault prediction model
wavelet denoising
作者简介
谢国财(1985—),男,博士,教授级高级工程师,主要研究方向为电力设备状态监测与故障诊断、电力调度自动化、电力大数据分析;温锐(1972—),男,本科,高级会计师,主要研究方向为财务管理、资产管理、信息系统平台建设及运营;陈琛(1990—),男,本科,工程师,主要研究方向为电气工程、电力市场营销。
