Research On State Estimation Strategy Of Energy Storage Battery

With the continuous development of new energy power generation technology,energy storage battery’s technology also constantly develop.The use of energy storage batteries are monitored and managed by the battery management system,including two very important estimation values of State of Charge and State of Health.It is very important for the safety management、improving service life、decision and equilibrium control by improving these two estimation values’ accuracy.Because of the large proportion of lithium-ion battery in the energy storage system,the related methods to improve the estimation accuracy of the State of Charge and the State of Health of lithium-ion battery are studied.In order to solve the problem that the traditional lithium battery model has poor adaptive ability in SOC estimation,and the local estimation accuracy of the single SOC estimation algorithm is low,a weighted SOC online estimation method is proposed.The method using the recursive least squares algorithm to obtain direct identification parameters in real-time is used in the SOC estimation of open circuit voltage(OCV)method based on PI regulator.The value obtained by OCV and the value gained by ampere-hour method weight summary in order to improve the estimation accuracy and achieve the purpose of online estimation.There are different weights distribution for open circuit voltage method and ampere-hour method,which solved large estimation error of open circuit voltage method in platform area and the problem of being difficult to determine the initial value and error accumulation.In addition,the proposed method also solved the problem above two method that can not achieve online estimation.In view of small samples and nonlinear characteristics of lithium battery,a method for state of health estimation based on the feature vector of support vector regression model is proposed.The feature vector is selected by Mutual information.Taking into account that the representative of the input samples and the model parameter settings can affect the prediction results of support vector machine,the method of mutual information is used to select input samples.Finally,the average voltage and the maximum and minimum temperature difference in the constant current constant voltage charging process are chosen as the feature vector.Besides,the grid search algorithm is applied to optimize the model parameters.The proposed SOC estimation method is simulated and verified on MATLAB,and the comparison between the results of weighted SOC online estimation and the open circuit voltage estimation based on PI is given.Followed by the method of SOH,based on lithium battery experimental data from NASA,simulation experiment is carried out on the MATLAB platform.And the estimation results between the proposed method and BP neural network estimation results are compared and analyzed.