A novel nonlinear decreasing step-bacterial foraging optimization algorithm and simulated annealing-back propagation model for long-term battery state of health estimation.

With the rapid development of electric energy storage, more and more attention has been paid to the accurate construction of energy storage lithium-ion battery (LIB) model and the efficient monitoring of battery states. Based on this requirement, a simulated annealing-back propagation (SA-BP) model is proposed, and the long-term state of health (SOH) of LIBs can be estimated online by combining with the battery single particle (SP) model. Among them, simulated annealing (SA) algorithm is used to optimize the initial parameters of back propagation (BP) network. In order to improve the identification efficiency and avoid the local optimization, the nonlinear decreasing step-bacterial foraging optimization (NDS-BFO) algorithm is introduced into the parameter identification process. On the basis of adopting the SOH sequence as the output of the SA-BP model, two electrochemical parameter sequences are used as the input of the model for training and testing. In addition, in this paper, the contributions in terms of the SOH estimation task mainly include two aspects. Firstly, the SOH estimation results can provide suggestions for the timely replacement of batteries in actual energy storage power stations. Secondly, the electrochemical parameters identified before SOH estimation are strongly related to the quality of the LIB. Therefore, they can provide references for the economy of LIBs. At 25 °C, the accuracy of the SP model is verified under three different working conditions. Degradation experiments are carried out under a constant current condition and a self-designed energy storage condition. The experimental results show that, under the 0.5 rate constant current condition, the root mean square error (RMSE), mean absolute error (MAE) and mean absolute percentage error (MAPE) of the long-term SOH estimation result are 0.42 %, 0.34 % and 0.38, respectively. And under the self-designed energy storage condition, the RMSE, MAE and MAPE of the result are 0.33 %, 0.26 % and 0.29, respectively. Under the same working condition, the SOH estimation results have a significant improvement in various performance evaluation indicators. The improved algorithm provides theoretical and experimental basis for the reliability of energy storage battery monitoring