【摘要】 依据最小二乘支持向量机(LS_SVM)的基本理论,针对蓄电池荷电状态(state of charge,SOC)随温度、电压、电流而变化的特点,建立基于LS-SVM支持向量机的蓄电池SOC估测模型。通过数据验证,比较不同核函数下的效果,利用网格搜索寻找最优参数。观察在最优参数和最优核函数下LS_SVM支持向量机的预测效果。结果表明,与其他算法相比,采用RBF核函数,并用网格搜索优化的LS_SVM模型精度较高,适合用在蓄电池的SOC估测上。更多还原

【Abstract】 The basic theories of the LS_SVM(Least Square Support Vector Machine) were introduced. According to the battery SOC affected by temperature, voltage and current, the battery SOC estimate model based on LS_SVM was built. After training the LS_SVM model by experiment data, three kinds of kernel were compared in the mode and the best parameters were searched by using the grid search. The results show that RBF kernel and the grid search are the best for LS_SVM, and LS_SVM is very suitable for the prediction of SOC.更多还原