基于广义回归模型的充电桩运行异常预测方法

Prediction method for abnormal operation of chargingpile based on generalized regression model

针对现有充电桩运行异常预测方法存在预测误差大、置信度低的问题,提出基于广义回归模型的充电桩运行异常预测方法。划分充电桩的异常类型以及异常区域,设置异常类型特征作为判断设备运行异常的标准。根据充电桩的工作原理,收集其动态多维运行数据。分别从人为操作以及环境天气两个方面确定充电桩运行影响因素。构建充电桩的供应商评价模型,确定充电桩的初始质量。在考虑充电桩初始质量和影响因素的前提下,采用收集的运行数据,利用广义回归模型估算电力负荷、充电量等运行参数。最终经过与设置标准特征的比对,得出充电桩的运行异常预测结果。通过性能测试实验得出结论:在两种实验环境下,设计方法的电力负荷预测误差均低于0.5 kW,置信度始终高于85%,且识别时间为24 s。

Aiming at the problems of large prediction errors and low confidence in the existing abnormal operation prediction methods of charging piles,a generalized regression model-based method for predicting abnormal operation of charging piles is proposed.Dividing the abnormal type and abnormal area of the charging pile,and setting the abnormal type characteristics are used as the standard for judging the abnormal operation of the equipment.According to the working principle of the charging pile,its dynamic multi-dimensional operating data are collected.The factors affecting the operation of the charging pile are determined from two aspects of human operation and environmental weather.A supplier evaluation model is constructed for charging piles to determine the initial quality of charging piles.Under the premise of considering the initial quality and influencing factors of the charging pile,the collected operating data and the generalized regression model are used to estimate the operating parameters such as power load and charging capacity.After comparing with the set standard features,the abnormal operation prediction result of the charging pile is obtained.Through the performance test experiment,it is concluded that in the two experimental environments,the power load forecast error of the designed method is less than 0.5 kW,the confidence is always higher than 85%,and the recognition time is 24 s.