Health Assessment Of Lithium Ion Batteries Based On Deep Gaussian Process Regression

In order to implement the carbon peak and carbon neutral decision deployment,it is crucial to improve the efficiency of energy resources utilization,promote the use of renewable energy and reduce carbon emissions in buildings.Lithium-ion batteries,as an essential storage part of renewable energy,have been widely used in building energy efficiency.Therefore,with the continuous development of the global new energy industry,lithium-ion batteries are attracting much attention as the core of new energy and intelligent manufacturing.However,the performance of lithium-ion battery systems deteriorates during long-term use.Therefore,accurately assessing the state of health(SOH)can effectively avoid unnecessary losses due to unexpected failures of lithium-ion batteries.To this end,a deep Gaussian process regression(DGPR)lithium-ion battery SOH evaluation method based on Gaussian process regression and deep neural network is established in this paper.Firstly,the heterogeneous features reflecting the SOH of Li-ion battery are extracted from the charge/discharge time,cumulative capacity curve(ICC)peak,internal resistance and energy;then,significant features are introduced into the DGPR model through gray correlation analysis(GRA)to establish the Li-ion battery SOH estimation method.Finally,the data sets provided by CALCE and NASA are used as experimental objects to compare with different data-driven models,and the results show that the MAE,MAPE,RMSE,and R~2indexes of the DGPR model are higher than those of other models,which verifies the accuracy,reliability,and applicability of the proposed model.The experimental results are also analyzed from a quantitative perspective,and the method has a high accuracy of SOH estimation for each Li-ion battery,with RMSE less than 0.7%and R~2greater than 98.2%.This thesis uses the continuous ranking probability score(CRPS)to evaluate the probabilistic model of SOH for Li-ion batteries.The DGPR model obtains a narrower 95%confidence bandwidth than the widely used GPR model.Its CPRS is only 50%of that of the GPR model,which indicates that the DGPR model is better than the GPR model in predicting the uncertainty of SOH for Li-ion batteries.Thus the method can provide a reliable basis for the health management of lithium-ion batteries.Figure[14]Table[5]Reference[56]...