Research On Key Technologies Of Data-driven Power Battery Fault Diagnosis And Life Prediction

With the rise and development of electric vehicles,lithium-ion batteries are widely used as the core energy storage device of electric vehicles.Due to the difficulty in extinguishing the spontaneous combustion of batteries and easy re-ignition,the safety of electric vehicle power batteries has always been an important direction of battery research.In this paper,based on the method of machine learning,combined with the physical equivalent circuit model of the battery,the research on the fault diagnosis and life estimation of the power battery is carried out.Firstly,this paper studies the fault classification model of BMS fault diagnosis based on the decision tree algorithm,and establishes a decision tree fault classification model for the undervoltage alarm,insulation alarm,and braking system alarm,and extracts the features with high correlation with the fault.Based on these characteristics,the laws and causes of BMS fault alarms including undervoltage,insulation and braking are analyzed.Secondly,this paper uses the equivalent circuit model and the long short-term memory network model,combined with the early fault-free data of the battery,to achieve an accurate estimation of the battery cell voltage.By comparing the accuracy of the two models,the equivalent circuit model with smaller error is selected as the voltage estimation model.Based on the estimation of cell voltage,a scheme to identify potential faulty cells is proposed,which can reasonably identify outlier cells.Finally,this paper studies the probability prediction model of battery cycle life based on the experimental data of battery charge-discharge cycles.The dominant features that characterize the cycle life are obtained,and the point estimation prediction method and the probability estimation prediction method of the battery cycle life are respectively studied by applying feature combinations with different complexity.Taking point estimation prediction as a reference and comparison,quantile regression is used to study the probability prediction model of battery cycle life under different confidence levels.By applying the experimental data,the original point estimation prediction method and the proposed probability prediction method are carried out,and the verification results of the two models are compared and discussed.