基于机器学习与预测分析的直流光测量系统远程模块的寿命预测方法

Lifetime Prediction Method for Remote Module of DC Optical Measurement System Based on Machine Learning and Predictive Analysis

针对高压直流换流站中直流光测量系统的远程模块,基于机器学习方法,采用预测分析理论,建立了其剩余寿命的预测模型。搭建了远程模块特征参数采集系统,以采集远程模块的特征参数,并确定了其工作状态阈值;接着采用离群点检测算法检查远程模块运行情况;最后基于多元线性回归算法采用训练数据样本集搭建远程模块剩余寿命的预测模型,实现对远程模块剩余寿命的预测。进一步,通过测试数据样本集来评价了寿命预测模型的精确程度。结果显示,该模型寿命预测结果的均方误差和均方根误差分别为0.009072和0.0952,从而说明该寿命预测模型的预测结果较好,精确度较高。

A prediction model for the residual lifetime of the remote module of the DC optical measurement system is established by using machine learning method and predictive analysis theory.A parameter acquisition system is built to collect the characteristic parameters of the remote module,and determine the operating status threshold of the remote module.Then,the LOF-based outlier detection algorithm is designed to diagnose the operating status of the remote module.Finally,the prediction model for the residual lifetime of the remote module is constructed based on the multiple linear regression algorithm with the training sample dataset,to realize the prediction of the residual lifetime of the remote module.Furthermore,the accuracy of the lifetime prediction model is evaluated on the test sample dataset.The evaluation results show that the mean square error and root mean square error of the predicted residual lifetime are 0.009072 and 0.0952 respectively,which indicates that the prediction model is good and accurate.