基于深度学习的电动汽车充电桩故障自动预测方法  

Automatic Fault Prediction Method of EV Charging Pile Based on Deep Learning

作　　者：赵宇琦 ZHAO Yuqi(China Railway Fifth Survey and Design Institute Group Co.,Ltd.Tianjin Institute,Tianjin 300199,China)

机构地区：[1]中铁第五勘察设计院集团有限公司天津院,天津300199

出　　处：《自动化应用》2024年第15期153-155,共3页Automation Application

摘　　要：常规的电动汽车充电桩故障自动预测节点形式一般为定点定向,预测的效率受限制,导致最终得出的平均故障间隔预测次数减少。为此,提出对基于深度学习的电动汽车充电桩故障自动预测方法的设计与验证分析。首先,根据当前的自动预测需求,采用交叉的方式测定,打破预测效率受到的限制,实现自动预测节点交叉部署及故障特征提取。然后,以此为基础,基于深度学习中的LSTM神经网络完成电动汽车充电桩故障自动预测模型的构建,并采用自适应校验处理的方式实现自动预测。最后,针对选定的5个充电桩小组,按照顺序导入3组虚拟故障指令,发现平均故障间隔预测次数均可以达到15次以上。结果表明,该自动预测方法在复杂环境下可实现对电动汽车充电桩故障的自动预测识别,具有实际应用意义。The conventional automatic prediction node form of EV charging pile fault is generally fixed-point orientation,and the prediction efficiency is limited,resulting in the final average fault interval prediction times are reduced.The design and verification analysis of the automatic fault prediction method for EV charging pile based on deep learning are proposed.Firstly,based on the current demand for automatic prediction,cross testing is adopted to break the limitations of prediction efficiency and achieve automatic prediction node cross deployment and fault feature extraction.Then,based on this,the LSTM neural network in deep learning is used to construct an automatic prediction model for EV charging pile faults,and adaptive verification processing is adopted to achieve automatic prediction.Finally,for the selected 5 charging pile groups,3 sets of virtual fault instructions were imported in order,and the average number of predicted fault intervals reached 15 or more.The results indicate that this automatic prediction method can achieve automatic prediction and recognition of EV charging pile faults in complex environments,and has practical application significance.

关 键 词：深度学习 电动汽车 充电桩 故障监测 自动预测方法 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置]