Support vector machine regression based supercapacitor's dynamic characteristics model

Building an accurate model to describe supercapacitor's dynamic characteristics is the basis for energy management strategy and system simulation analysis. This paper presents the support vector machine regression (SVMR) based modeling approach which can describe the supercapacitor's nonlinear property and dynamic characteristics during the process of charging and discharging. The concrete modeling process has four steps. Firstly, the real-time synchronous acquisition data of current and terminal voltage from the charging and discharging experiments on supercapacitor were chosen and divided as the training and testing original data sets for the modeling process. Secondly, based on the principle of cross validation the optimized parameters of the SVMR model were acquired. Thirdly, the final SVMR prediction model was built through training. Finally, the accuracy of the model was analysed by means of comparing the error between the model's prediction output and the supercapacitor's real output. The results of experiments and simulations showed that the SVMR model built in the paper can describe the supercapacitor's dynamic characteristics accurately and the modeling method is feasible and valid.