最小二乘支持向量机在航空蓄电池剩余容量预测中的应用  

The estimation on the residual capacity of aeronautic battery based on LS-SVM

作　　者：刘勇智[1] 詹群[1] 盛增津[1] 陈杰[1] 

机构地区：[1]空军工程大学航空航天工程学院,陕西西安710038

出　　处：《蓄电池》2013年第3期118-120,144,共4页Chinese LABAT Man

摘　　要：本文提出了一种预测航空蓄电池剩余容量的新方法。介绍了电池容量常用的几种预测方法及缺点;阐述了用于航空蓄电池容量预测的最小二乘支持向量机(LS-SVM)回归算法原理,并说明了预测模型的建立步骤;提出了交叉验证法来优化惩罚因子和核函数参数,给出了两种误差标准;最后用实验数据验证了所建LS-SVM预测模型的准确性,并与BP神经网络进行比较,仿真结果证明,LS-SVM预测模型比BPNN模型的精度高,更适合用于航空蓄电池容量在线预测。A new estimation method of residual capacity of aeronautic battery is proposed in this text. A few common methods are introduced and their disadvantage are illuminated first, then the truth on the regression arithmetic of Least Square Support Vector Machine （LS-SVM） for the estimation of residual capacity is elaborated, and the steps of building the predicting model are illuminated. A new method of cross-validation is put forward to optimize the penalty factor and the parameter of kernel function, and two kinds of error standards are given. Finally, the accuracy of the LS-SVM predicting model is verified based on the experiment data, and the comparison is carried on with BP neural network. The simulating results show that the LS-SVM predicting model owns a higher accuracy than BPNN, which is more suitable for the on-line estimation on the residual capacity of aeronautic battery.

关 键 词：航空蓄电池 SOC LS—SVM 在线预测 

分 类 号：TM912.9[电气工程—电力电子与电力传动]