Life Prediction Of Power Battery Of Electric Vehicle Based On Degradation Mode Classification

With the shortage energy and serious polluted environment, electricalvehicles (EV) get more and more attentions, and come into everybody’s life. Asthe world’s energy becoming shortage and environmental pollution becomingincreasingly serious, electric vehicles are coming into people’s life. Theperformance of the EV is heavily affected by the life of the battery. The batterylife is an important factor that affects electric vehicle’s performance. This papermainly focuses on the life prediction of power battery of electric vehicle. A lifedegradation model of power battery based on degradation mode classification isestablished. The prediction on remaining useful life of power battery undervarying operating conditions is deeply researched.In the paper, the research on degradation mechanism of power battery isdeveloped by analyzing related literature and the data from the NASA’slithium-ion battery charge and discharge cycle experiment. The main externalfactors affecting battery life degradation, including discharge current, cut-offvoltage and ambient temperature, are confirmed. The specific research purpose isto predict remaining useful life of power battery through estimating the state ofhealth. A new characterization named declining point to measure batterydegradation is proposed. The effectiveness of the proposed characterization hasbeen verified.This paper establishes a life degradation model of power battery. After theconcept of the degradation mode is proposed, ART2neural network is used toclassify the degradation mode of power battery. Weighted Markov chain model isused to predict the sequence of degradation mode. Curve fitting method withtheoretical model is used to establish the degradation model of the single mode.Linear superposition method is used to complete the life prediction of battery.In order to verify the accuracy of life degradation model of power battery,the battery charge and discharge cycle experiment is designed, and theexperimental platform is built. The degradation data of power battery full lifecycle is collected. The life degradation model is established by these data, and itis used to estimate battery state of health and predict battery remaining useful life from different starting points. The predicted values based on degradationmode classification are compared with the predicted values of support vectormachines without classification and measured values. It is found that theabsolute error is within an acceptable range on the whole, and the accuracy ofprediction becomes better as the prediction starting point goes backward.This paper uses declining point as the characterization measure for batterydegradation, and establishes a new life degradation model of power battery basedon degradation mode classification. The model can realize the accurateprediction of remaining useful life of power battery under varying operatingconditions, and it is suitable for online detection. It has a good practicalsignificance for improving battery management system.