基于LSSVR-AR模型的电解电容状态预测方法

摘要

针对传统的状态预测方法预测精度不高的问题,提出了一种基于最小二乘支持向量回归机(LSS-VR)和AR模型相结合的非平稳时间序列建模的方法(LSSVR-AR),并应用于Buck电路的电解电容等效串联的状态预测中.对非平稳时间序列进行最小二乘支持向量回归,得到非平稳时间序列的趋势项及剔除趋势项后的随机项;对随机项建立AR模型并与趋势项的LSSVR模型组合,得到非平稳时间序列模型;用组合模型对电解电容的等效串联电阻进行状态预测.用本文所提出的方法对其预测的平均绝对百分比误差为6.57%,低于单一的LSSVR模型.实例证明:本文所提出的模型能对电解电容的状态进行准确预测.%To improve prediction accuracy of the traditional prediction methods,a LSSVR-AR (least square support vector regression and Autoregressive Model)based non-stationary time series are put for-ward and applied in the electrolytic capacitor equivalent Buck circuit on the state prediction.Firstly, least squares support vector regression is used to non-stationary time series to abstract the tendency. The stochastic components are obtained after the tendency is eliminated.Secondly,to give non-stationa-ry time series model,the AR model of stochastic models in combination with LSSVR model need to be established.Finally,the equivalent of electrolytic capacitor is predicted by the combined model of series resistance.By using this mean absolute percentage error of the forecast is 6.57%,which is lower than the single LSSVR model.Examples show that the model presented in this paper for electrolytic capacitor accurate prediction of the state.