Research On Data-Driven Fault Diagnosis And Early Warning Of Power Batteries For New Energy Vehicles

In light of the ongoing energy crisis and environmental pollution,our country is actively promoting energy transformation and has deemed new energy vehicles a national strategy.As a crucial component of new energy vehicles,the power battery is one of the primary sources of failure under complex operating conditions,and its performance directly affects the safe operation of new energy vehicles.Based on real-world operational data of new energy vehicles,this paper focuses on lithium-ion power batteries and employs a data-driven approach to research the diagnosis and warning of battery failures.The main research work is as follows:Data preprocessing and dimensionality reduction.The raw data contains duplicate values,missing values,and anomalies.Clustering,outlier detection,and box plot methods are used for data cleaning.In the fault diagnosis and warning process,dimensionality reduction is performed on high-dimensional data.For data that only contains alarm information,dimensionality reduction is carried out using correlation analysis combined with principal component analysis.For data that contains both alarm and normal information,the stepwise regression method is used to select feature parameters with positive contributions to the target value,and then principal component analysis is used to reduce the selected data.The results show that the proposed dimensionality reduction method can effectively remove useless data feature parameters,reduce interference to model training,and reduce data dimensionality while retaining useful information.Fault diagnosis based on convolutional neural network.To avoid the challenges brought by imbalanced data sets,the SMOTE(Synthetic Minority Over-sampling Technique)algorithm is used to expand data and ensure that the amount of all types of data reaches the same level.A CNN(Convolutional Neural Network)model is constructed to diagnose faults in power batteries,and the results show that the CNN model can diagnose multiple faults.Considering the high level of harm caused by the highest-level alarm,the logistic regression model is used to diagnose faults in the highest-level alarm during diagnosis.The results show that the recall rate of the logistic regression model reaches 100%,and there is no missed diagnosis of faults in the highest-level alarm.Fault warning based on LSTM(Long Short-Term Memory Recurrent Neural Network,LSTM RNN).For data that contains both alarm and normal information,an LSTM model is constructed to predict important characterization parameters such as the voltage and temperature of the power battery pack,single-cell voltage,and SOC,and to detect power battery faults through abnormal detection of these parameters.The results show that the LSTM model can accurately predict the selected characterization parameters and achieve a60-second advance warning of power battery faults based on abnormality detection of these parameters.