Lithium-ion Battery Health Prediction Based On Gaussian Process Regression

With the development of social economics and energy storage technology,lithium-ion battery is widely used in various fields of the whole society because of its high stable voltage,longevity,light weight,environment-protecting and other advantages.But the performance of lithium-ion battery will gradually degrade under continuous charging and discharging,sometimes unexpected failure may occur and it will lead to lithium battery failure and serious consequences.Therefore,predicting the health of lithium battery based on its historical monitoring data is of great significance to the operation and maintenance of lithium batteries.In this thesis,Gaussian Process Regression(GPR),one of the data-driven approaches,is applied to the health predicting field of lithium battery,and the State of Health(SOH)and Remaining Useful Life(RUL)of the lithium battery are predicted.The thesis emphasized on several aspects as follows:(1)Since GPR doesn’t have unified theoretical support when selecting kernel functions,the influence of kernel functions and the hyperparameters on the predictive distribution is analyzed in this thesis.Based on this,the composite kernel function is derived.The composite kernel functions were applied to predict the State of Health and the Remaining Useful Life of lithium battery.Compared with the prediction result based on a single kernel function,the Gaussian process regression based on the composite kernel function improves the generalization ability while guaranteeing the local learning ability and has higher prediction accuracy.(2)Since GPR cannot feasibly be applied to big datasets,the Sparse Pseudo-input Gaussian Process and Recursive Gaussian Process are selected to simulate the model.The results show that the Recursive Gaussian Process regression method with higher accuracy than Sparse Pseudo-input Gaussian Process.And it can effectively reduce the prediction time under the premise of ensuring accuracy compared with the general GPR.(3)Since GPR cannot update the prediction model in real time according to the newly acquired data,the Recursive Gaussian Process is applied to predict the State of Health of lithium-ion battery.The experimental results compared with the general GPR show that the Recursive Gaussian Process can update the prediction model faster,and it does not need to repeat historical data training.