An adaptive physics-based reduced-order model of an aged lithium-ion cell, selected using an interacting multiple-model Kalman filter. (October 2018)

• Record Type: Journal Article
• Authors: Smiley, Adam; Plett, Gregory L.
• Is Part Of: Journal of energy storage. Volume 19(2018)
• Page Start: 120
• Page End: 134
• Publication Date: 2018-10
• Subjects: Lithium ion battery -- Electrochemical model -- Battery management systems -- Aged parameter estimation -- Interacting multiple model Kalman filter; Energy storage -- Periodicals; Energy storage -- Research -- Periodicals; 621.3126
• DOI: 10.1016/j.est.2018.07.004
• Languages: English
• ISSNs: 2352-152X

Abstract

Highlights: Physics-based reduced-order models enable battery management systems to limit aging. These models must adapt to capture present dynamics of cells over battery-pack life. We propose an approach to select model from precomputed set that matches measurements. Method uses nonlinear interacting multiple model Kalman filter to make selection. Results are guaranteed to be stable and physically meaningful. Abstract: Reduced-order physics-based models of lithium-ion cells provide the opportunity for a battery-management system to define battery-pack operational limits in terms of cell internal electrochemical processes in order to mitigate degradation and to avoid failure modes. For these physics-based models to be relevant over the lifetime of the battery pack, they must somehow adjust to describe the internal processes accurately at every stage of battery life. Two possible approaches to do so suggest themselves. First, an algorithm might somehow adapt the parameter values of the model during operation to match presently observed current–voltage behaviors; but, this must be done very carefully to avoid making the model unstable or physically nonmeaningful. Alternately, a set of models could be pre-computed at different feasible aging points and the model from this set that most closely predicts presently observed current–voltage dynamics could be selected from the set. This second approach guarantees stable and physically meaningful models since all models in the pre-computed set meet these criteria. We propose such an approach here. To do so, we first present a method for calculating a priori the changes to cell parameter values that will be produced by aging due to side reactions and/or material loss. These aged parameter values are utilized to produce reduced-order physics-based models at different stages of cell life. The reduced-order models are then used within a nonlinear interacting multiple-model Kalman filter to select the pre-computed model whose voltage predictions most resembles present measured voltage, so providing an estimate of the aged parameter values of a cell via the parameter values of this model. The selected model may then be used for state-of-charge estimation, state-of-power estimation, state-of-energy estimation, and other model-based battery-management tasks.