Bio-inspired optimizer with deep learning model for energy management system in electric vehicles

Abstract

The rising popularity of electric vehicles (EVs) stems from their enhanced performance and environmental benefits. A critical challenge exists in optimizing the performance and extending the battery life of EVs, which depends on the accurate prediction of State of Charge (SOC) and State of Health (SOH). The Battery Management Systems (BMS) is essential for an EV’s Energy Management System (EMS). The current methodologies often fail to achieve the required precision, leading to suboptimal BMS that can compromise EV efficiency and reliability. To address these challenges, a merged SOC and SOH prediction approach is proposed. To maximize prediction accuracy, a hybrid Deep Learning (DL) model incorporating bio-inspired optimization algorithms such as Elephant Herding Optimization (EHO), Honey Badger Optimization (HBO), and Moth-Flame Optimization (MFO) is utilized. The architecture comprises two Convolutional Neural Networks (CNN) and an Autoencoder (AE), integrated with a Bidirectional Long Short-Term Memory (BLSTM) layer and a single Long Short-Term Memory (LSTM) layer for encoding and decoding tasks. The three optimized hybrid DL models were validated using standard benchmark datasets such as the Oxford Battery Aging Dataset, NASA, and CALCE. The prediction results of the merged SOC and SOH prediction from the three bio-inspired hybrid DL models were compared with those of the separate SOC prediction technique. The results of the merged SOC and SOH predictions were compared with traditional separate SOC prediction techniques, demonstrating superior performance. Notably, the HBO-Hybrid DL model achieved the highest R-squared (R2) values of 0.991 for SOC and 0.996 for SOH