SOH Estimation Method for Lithium-ion Battery Based on Discharge Characteristics

Abstract

In order to accurately estimate the state of health (SOH) of lithium batteries under the condition of constant current charging and discharging, a multi-scale health indicator (HI) selection strategy based on discharge characteristics is proposed in this paper, which can estimate SOH when the battery is not fully charged and discharged. Firstly, the initial voltage drop in the discharge process and the time interval of an equal discharging voltage difference were treated as the indirect HIs of battery capacity estimation. Secondly, the optimal discharge voltage segment was selected using the genetic algorithm based on elitist model (e-GA), the selection criterion of time interval of an equal discharging voltage difference was determined, and its correctness was verified. Finally, a lithium battery degradation model was established by the long and short term memory (LSTM) neural network according to the HI selection strategy as mentioned above, and the experimental results were compared with those as obtained using other selection strategies. As suggested by the experimental results, the aforementioned HI selection strategy can reduce the mean absolute percentage error (MAPE) by 0.41 % on average. In addition, the MAPE of the prediction results fell below 2.22 % in twelve experiments conducted on four batteries at different allocation ratios of training set and test set.