Research On New Energy Vehicle Battery Failure Prediction System Based On Big Data

In recent years,with people's increasing attention to the environment and energy issues,new energy vehicles,as a low-carbon car,has become a trend.According to Annual Report On Big Data of New Energy Vehicle in China(2018),battery system failures account for 45% of vehicle failures,among which the problems of poor consistency of single units and low SOC account for 60%.This paper takes the battery,as the research core,combined with the real-time driving data of 10,000 new energy vehicles,and conducts in-depth research on the key technologies of the major battery faults and the big data monitoring platform system of new energy vehicles that integrates fault prediction.The main content is as follows:?.Fault prediction and diagnosis of battery cell consistency failure based on cosine similarity.Poor cell consistency will deteriorate battery performance and greatly affect battery life.In this paper,through the analysis of a large number of new energy vehicle driving data,it is determined that the unit consistency evaluation method used in the data is the unit poor range at the first start-up time after the end of charging,and the threshold is around 0.253.The abnormality of the internal parameters of the battery cell will be reflected in the cell range curve of the charging process to a certain extent.Therefore,a similarity algorithm is proposed to diagnose the cell consistency fault in the cell range curve of the charging process,which is based on the cosine similarity The algorithm has excellent diagnostic capabilities.When the cosine similarity threshold is 0.99946,the true positive rate is 70.2% and the false positive rate is 9.8%,which is at least 5 minutes earlier than the original method to obtain the fault diagnosis result.?.Fault prediction and diagnosis of low SOC based on CNN-LSTM.When the battery is in deep discharge,the current and voltage will be unstable,and frequent deep discharge will greatly affect the battery life.This paper presents a univariate low SOC prediction based on ARIMA(autoregressive moving average model)and a multivariate low SOC prediction based on CNN-LSTM combined with weather data information.Although the former method has 100% true positive ate,it has 13.5% false positive rate,while the latter has 83.5% true positive rate,but the false positive rate is 0%.?.Design and development of fault prediction system for new energy vehicles.Most of the existing vehicle factories or provincial or national monitoring systems for new energy vehicles only have the functions of vehicle traffic data display,fault display and statistics,as well as operation and maintenance.This section combines the fault prediction algorithm in the first two sections to complete the technical selection and development of the fault prediction system based on big data for key components of new energy vehicles This system consists of two main parts.I.Web system developed by back-end framework of Egg.js and front-end framework of Ant Design.II.Big data computing platform developed by Spark Streaming.Finally,the failure prediction and data display of key components of new energy vehicles are realized.