Research On Joint Estimation Of SOC And Capacity For Ternary Lithium Ion Battery Based On Gaussian Process Regression

Non-renewable energy is becoming more and more harmful to the environment,and the world energy situation continues to develop toward cleanliness and low carbonization.Therefore,new energy vehicles have received more attention and research due to the trend of energy development.As the power source of new energy vehicles,the power battery for vehicles is regarded as the "fuel gauge" of new energy vehicles,which can intuitively remind the remaining mileage of electric vehicles.Accurate SOC estimation can ensure the safety of power batteries and improve battery capacity utilization.In order to reduce the estimation error of power battery SOC,this paper takes ternary lithium ion battery as the research object,fully considers the influence of various factors on SOC,and proposes a method for joint estimation of battery capacity and SOC based on Gaussian process regression.as follows:According to the performance parameters of ternary lithium-ion batteries and the factors affecting performance,the battery characteristics test platform was designed in conjunction with the research purpose,and the battery aging life cycle experiment was carried out,and the experiments of different working conditions were interspersed during the battery aging process to obtain ternary The battery aging data of lithium ion battery in 180 cycles and the experimental data of different working conditions,the SOC reference curve is obtained by the ampere-hour integration method,which provides a training data set for the establishment of the GPR model.The modeling and prediction process of GPR is systematically deduced.The key to GPR estimation performance lies in the selection of appropriate kernel functions and accurate hyperparameter values.The learning ability of a single kernel function for nonlinear characteristics of training data is limited.Add the combined kernel function,and use the artificial bee colony algorithm(ABC)with strong global optimization performance to search for the optimal value of the hyperparameters in the combined kernel function,so as to improve the estimation performance of the GPR model.The battery capacity increment was studied.The peak value of the capacity increment curve under different cycle times was used to accurately estimate the battery capacity.The relationship between the battery discharge voltage,discharge capacity and SOC under different aging conditions was analyzed.The battery current,voltage and estimated battery capacity data are used as one of the input parameters of the GPR model to realize the estimation of the SOC of the battery under different aging conditions.The experimental results prove that the joint estimation method has the lowest estimation error and the estimation accuracy remains unchanged under different cycle times compared to the GPR algorithm and the extended Kalman filter method that do not consider the battery capacity change.The combined battery capacity and SOC estimation method based on GPR model optimized by ABC algorithm has the characteristics of accurately estimating the battery capacity and SOC under different aging conditions,which can provide a reliable basis for the battery management system to implement the energy distribution strategy,thus has valuable significance improving the battery application level.