Bayesian Hyperparameter Optimization of Stacked Long Short-Term Memory Neural Network for the State of Charge Estimation

Abstract

Electrification and reduction of fossil fuel usage placed the battery industry in a prominent position. The state of charge (SOC) of the battery is a crucial parameter for the optimal operation of the battery management system (BMS). In the last years, with the development of electronic materials and the ability to store a big amount of data, data-driven methods for the SOC of the battery can to capture the dynamics of the battery. A big problem of the data-driven methods is the selection of the hyperparameters of the model. Heuristic methods have been used mostly with manual tuning or through exhaustive methods, such as random search and grid search. In this paper, a Bayesian Hyperparameter Optimization with a Gaussian process is proposed and applied to a stacked Long Short-Term Memory (LSTM) neural network. The proposed method is validated using a public dataset and compared with other state-of-the-art methods, achieving low errors.