根据支持向量机(Support Vector Machine,SVM)的基本理论,并依据铅酸蓄电池荷电状态(State of Charge,SOC)随电压、电流而变化的特点,建立基于支持向量机SVM的铅酸蓄电池SOC估测模型。通过实验数据测试验证,比较不同核函数下模型的估测效果,利用扩展的粒子群(Extended Particle Swarm Optimization,EPSO)算法寻找最优参数,并观察在最优参数和最优核函数下支持向量机SVM估测模型的估测效果。结果表明,采用RBF作为支持向量机SVM的核函数,并用扩展的PSO算法优化的支持向量机SVM模型精度较高,适合在铅酸蓄电池的SOC估测上。
The basic theories of the SVM(Support Vector Machine,SVM) are introduced in this paper.According to the leadacid battery SOC affected by voltage and current,the lead-acid battery SOC estimation model based on SVM is built.And then three kinds of kernel are compared in mode and best parameters are searched by using extended PSO.