DEEP LEARNING–BASED STATE OF CHARGE ESTIMATION FOR LITHIUM-ION BATTERIES IN ELECTRIC VEHICLE FOR LOCAL MOBILITY GOVERNANCE

The rise of using electric vehicles (EVs) as a sustainable means of transportation has augmented the necessity of precise monitoring and regulation of lithium-ion batteries as part of the Battery Management Systems (BMS). To avoid overcharging, enhance energy consumption, minimize the cost of the operation of lithium-ion batteries, and improve battery safety during the actual working conditions, reliable estimation of the State of Charge (SOC) of the lithium-ion battery is necessary. Nonlinear electrochemical characteristics of lithium-ion batteries and dynamically changing loads, however, become major challenges to the traditional approach to SOC estimation. This study proposes a deep learning-based system of predicting SOC in lithium-ion batteries in EV applications. The proposed model, as opposed to recurrent and convolutional neural network methods, is based on an Attention-assisted Temporal Convolutional Network with a Deep Autoencoder to provide an effective feature extraction and temporal learning that is stable. The autoencoder learns compact and noise-resistant feature representations of the battery voltage, current, temperature, and operating profiles, and the attention-enhanced temporal model learns long-range dependencies of lithium-ion battery behavior.,To achieve the implementation of the SOC estimation model in real-time, an edge computing architecture compatible with low-cost-BMS hardware is deployed. The experimental assessment shows that the technique is better than ampere-hour counting and baseline deep learning techniques in terms of prediction accuracy, reduced computational time, and higher robustness in testing different conditions of temperature and loads. The suggested framework facilitates effective management of lithium-ion batteries and also facilitates the effective operation of the EVs in a smart mobility environment that is deployed locally.