A Novel Adaptive Gradient Descent Based Parameter Estimation Algorithm for the Single Particle Model of a Lithium-Ion Battery

Lithium-ion batteries (LiBs) have seen wide adoption for several applications, and in turn, the implementation of accurate battery models in battery management systems (BMSs) has seen growing research interest. The Single Particle Model (SPM) is a high-fidelity battery model, that while robust and accurate, is challenging to parameterise for BMS platforms. Hence, in this work, we propose a novel application of the adaptive gradient descent (Adagrad) optimisation algorithm for parameter estimation (PE) of the Single Particle Model. The proposed algorithm is the first implementation of Adagrad optimisation for a physics model of a LiB and requires only a single voltage discharge curve as experimental data. Using a novel multi-objective optimisation criterion, the PE algorithm generates parameter sets of reasonable accuracy with fast estimation speeds, enabling online parameter estimation on BMS platforms.