Lithium-ion Battery Data Processing Based On Gaussian Process Regression Model

Lithium-ion battery(referred as lithium battery)is a green high-energy and rechargeable battery.It has a lot of advantages such as high stable voltage,wide temperature limit,long lasting,small volume,light quality and it is harmless to the environment.Therefore,lithium batteries are widely used in electronic products,electric vehicles,aerospace and other fields.But the lithium battery in the course of the performance will gradually decay,sometimes may also occur unexpected failure and it will lead to lithium battery failure and serious consequences.Therefore,the lithium battery health monitoring and residual cycle life prediction is essential,this research to guide the operation and maintenance of lithium batteries is of great significance.This thesis emphasize on several aspects as follows:(1)This thesis chooses the data-driven method to establish the Gaussian process regression model,and the data of lithium battery voltage and battery capacity are processed.The health status and remaining service life of lithium battery are predicted and analyzed.At the same time,the prediction results are compared with the artificial neural network method,and the superiority of Gaussian process regression is analyzed.(2)The problem of kernel function selection in Gaussian process regression model is studied.The kernel function is divided into local kernel function and global kernel function.The optimal kernel function or combined kernel function is selected for the characteristics of battery data,which improves the accuracy of the prediction result by analyzing,comparing and combining different kernel functions.(3)Propose to apply the sparse Gaussian process to the lithium battery data processing.The sparse pseudo-input method is used to model the model,and the computational complexity of the model is effectively reduced under the premise of ensuring the accuracy.The real-time performance of the battery data is improved by the Gaussian process regression model.