Robust state of charge and state of health estimation for batteries using a novel multi model approach

关键词:  Adaptive estimation multiple models · Connected embedded systems · Extended Kalman filter · Nonlinear observability · State-of-charge · State-of-health · State and parameter estimation

Key words:  Adaptive estimation multiple models · Connected embedded systems · Extended Kalman filter · Nonlinear observability · State-of-charge · State-of-health · State and parameter estimation

摘要:

Estimation of state-of-charge and state-of-health for batteries is one of the most important feature for modern battery management system (BMS). Robust or adaptive methods are the most investigated because a more intelligent BMS could lead to sensible cost reduction of the entire battery system. We propose a new robust method, called ERMES (extendible range multi-model estimator), for determining an estimated state-of-charge (SoC), an estimated state-of-health (SoH) and a prediction of uncertainty of the estimates (state-of-uncertainty—SoU), thanks to which it is possible to monitor the validity of the estimates and adjust it, extending the robustness against a wider range of uncertainty, if necessary. Specifically, a finite number of models in state-space form are considered starting from a modified Thevenin battery model. Each model is characterized by a hypothesis of SoH value. An iterated extended Kalman filter (EKF) is then applied to each model in parallel, estimating for each one the SoC state variable. Residual errors are then considered to fuse both the estimated SoC and SoH from the bank of EKF, yielding the overall SoC and SoH estimates, respectively. In addition, a figure of uncertainty of such estimates is also provided.

Abstract:

Estimation of state-of-charge and state-of-health for batteries is one of the most important feature for modern battery management system (BMS). Robust or adaptive methods are the most investigated because a more intelligent BMS could lead to sensible cost reduction of the entire battery system. We propose a new robust method, called ERMES (extendible range multi-model estimator), for determining an estimated state-of-charge (SoC), an estimated state-of-health (SoH) and a prediction of uncertainty of the estimates (state-of-uncertainty—SoU), thanks to which it is possible to monitor the validity of the estimates and adjust it, extending the robustness against a wider range of uncertainty, if necessary. Specifically, a finite number of models in state-space form are considered starting from a modified Thevenin battery model. Each model is characterized by a hypothesis of SoH value. An iterated extended Kalman filter (EKF) is then applied to each model in parallel, estimating for each one the SoC state variable. Residual errors are then considered to fuse both the estimated SoC and SoH from the bank of EKF, yielding the overall SoC and SoH estimates, respectively. In addition, a figure of uncertainty of such estimates is also provided.