关键词: SOC, 高斯过程回归, 锂离子电池

Abstract:

Battery state estimation (state of charge, SOC) is particularly important in a battery management system (BMS). It is difficult to ensure its accuracy because SOC estimation is vulnerable to temperature, load, charging and discharging efficiency, and other external factors. Currently, many scholars use machine learning algorithms to estimate the SOC. However, the estimation accuracy of a neural network (NN) is dependent on the number of samples. The support vector machine (SVM) falls into a local optimum in parameter optimization. An online estimation method is proposed based on Gaussian process regression (GPR) for lithium-ion batteries to improve the estimation accuracy of the SOC. Based on the battery measurement parameters, including current, voltage, and temperature, as input to the GPR model and the SOC as an output of the model, the model is trained, and the parameters are optimized using the gradient descent method. The validity of the model is verified by simulation and data collected from the constant current charging and discharging experiments. Compared with the SVM, LSSVM, and NN, the validity and feasibility of the model are verified.

Key words: SOC, Gauss process regression, lithium-ion battery