ElectroPredictNet: Accurate estimation of battery state of charge and state of health using advanced deep learning models

Recently, the usage of Electric Vehicles (EVs) has tremendously increased due to the development of green cars in the automotive sector. One of the most important components in addressing monitoring and safety issues in an EV is the battery. Specifically, assessing the state of charge (SoC) and State of Health (SoH) is crucial in battery-only EVs. Lithium-ion batteries are now the most popular energy source for electric cars. The safety and dependability of EV operation are ensured by monitoring the battery life and maintaining it using a Battery Management System (BMS). To calculate an EV battery's SoC and SoH, a novel ElectroPredictNet is developed to estimate the battery life by assessing the BMS parameters (SoC and SoH). For this process, pre-processing and feature smoothing phases are added, and five models are explored, including Convolutional Neural Network (CNN), Recurrent Neural Network (RNN), Long Short-Term Memory (LSTM), Deep Neural Network (DNN), and R-squared (R2) regression with a self-attention mechanism. These models are applied to accurately estimate battery performance by analyzing charge‒discharge cycles and predicting degradation patterns. Each model is evaluated for accuracy and error metrics to determine which model forecasts SoC and SoH best. It also provides valuable insights for enhancing BMS in EVs, achieving a Mean Squared Error (MSE) of 0.002 and a Mean Absolute Error (MAE) of 0.0096. Lastly, a comparative study is conducted with the State-of-The-Art (SOTA) models to validate the effectiveness of the developed ElectroPredictNet in predicting SoC and SoH.