基于PBES-LS-SVM的锂离子电池组SOC预测

SOC prediction of Li-ion battery pack based on PBES-LS-SVM

针对锂离子电池组荷电状态(state of charge,SOC)难以预测的问题,提出应用主成分分析(principal component analysis,PCA)选取影响因素和秃鹰算法(bald eagle search,BES)优化最小二乘支持向量机(least squares support vector machine,LS-SVM)的SOC预测模型。首先,采用PCA筛选出主成分特征值较大的因素后,将其组成多输入样本集合;其次,利用秃鹰搜索算法的全局搜索能力不断优化最小二乘支持向量机的惩罚系数C、核函数g,建立PBESLS-SVM预测模型;最后,应用某储能设备历史数据,采用GA-BP、BES-SVM和PBES-LS-SVM等模型分别对锂离子电池组的完整放电过程数据集与部分放电过程数据集进行仿真研究。结果表明,提出的模型SOC预测均方根误差、平均绝对误差、平均绝对百分比误差分别减小至1.79%、1.30%和3.39%。与其余预测模型相比,PBES-LS-SVM模型预测精度高、预测时间短,具备良好的收敛性、泛化性。

To address the problem that the state of charge(SOC)of lithium-ion battery pack was difficult to predict,principal component analysis(PCA)was applied to select the influencing factors and bald eagle search(BES)to optimize the least squares support vector machine(LS-SVM)for SOC prediction model.Firstly,the factors with large principal component eigenvalues were screened out using PCA and then composed into a multi-input sample set;secondly,the global search capability of the bald eagle search algorithm was used to continuously optimize the penalty coefficient C and kernel function g of the least squares support vector machine to establish the PBES-LS-SVM prediction model;finally,the historical data of an energy storage equipment was applied and the GA-BP,BES-SVM and PBES-LS-SVM models were applied to simulate the complete discharge process data set and partial discharge process data set of a lithium-ion battery pack,respectively.The results show that the root mean square error,mean absolute error,and mean absolute percentage error of SOC prediction of the models proposed are reduced to 1.79%,1.30%and 3.39%,respectively.Compared with the rest of the prediction models,the PBES-LS-SVM model has high prediction accuracy,short prediction time,and good convergence and generalization.