Capacity prediction for Lithium-ion batteries using optimized long short-term memory networks

Lithium-ion Batteries (LIBs) play an important role in many applications, such as electric vehicles. The prediction of the capacity of LIB is crucial in maintaining the performance of the batteries. Since data-driven methods outperform model-based methods in various ways, time series machine learning models have been used to predict the capacity. In this paper, we evaluate Long Short-Term Memory (LSTM) networks to predict the future capacity of two NASA LIB datasets. We compare two stochastic optimization algorithms, i.e., Particle Swarm Optimization (PSO) and Genetic Algorithm (GA), to optimize the hyperparameters of the LSTM. We also observe the effects of different train-test data split ratios, i.e, 50%-50% and 80%-20%. The results indicate that the 80%-20% ratio yields superior outcomes, with GA-LSTM outperforming PSO-LSTM, achieving a mean RMSE of 0.0130 for dataset B0005 and 0.0102 for dataset B0007.