Improved State-of-health prediction based on auto-regressive integrated moving average with exogenous variables model in overcoming battery degradation-dependent internal parameter variation

Abstract

State-of-health (SOH) of lithium-ion battery is a health indicator that predicts the life of target application by estimating the internal state of battery. It may alarm the replacement of battery in the application and caution when it is abnormal. However, it is very complicated to accurately predict the battery lifetime because battery degradation depends on user`s driving paten, and external factor. Therefore, it is very important to predict SOH by selecting factors that are highly related to battery life. This study introduces the auto-regressive moving average (ARMA) time-series model and proposes SOH prediction with extended models. ARMA-extend models used in this paper are ARIMA, Seasonal ARIMA (SARIMA) and auto-regressive integrated moving average with exogenous variables (ARIMAX) models. To analyze the impact of additional external variables in the model on SOH prediction, SOH prediction performance is compared using ARIMA, Seasonal ARIMA, and ARIMAX models. Among them, ARIMA model was excluded since it did not take into account seasonality which is important attribute of battery degradation. SARIMA model has the advantage of being able to consider the seasonality of data and ARIMAX model has the advantage of being able to consider external factors which is highly related with battery SOH. The two models were compared using Means square error (MSE) technique, and it was proved that ARIMAX model had higher accuracy than SARIMA since ARIMAX can consider external factors affecting battery degradation.