Akademik Veri Yönetim Sistemi
Arabian Journal for Science and Engineering, 2024 (SCI-Expanded)
Accurate state-of-charge (SOC) estimation is a cornerstone of reliable battery management systems (BMS) in electric vehicles (EVs), directly impacting vehicle performance and battery longevity. Traditional SOC estimation models struggle with the computational complexity versus prediction accuracy trade-off. This study introduces a new “Deep Neural Network (DNN)” architecture, providing precise SOC estimations with minimal computational requirements. The model utilizes Dense and Concatenate layers, alongside ReLU and linear activation functions, to efficiently process data while maintaining accuracy.Trained on both the LG 18650HG2 and Panasonic NCR18650PF li-ion battery datasets, which encompass diverse ambient temperatures (-20, -10, 0, 10, and 25∘C), our proposed DNN model operates with a reduced parameter set, requiring only a fraction of the trainable parameters. For the LG 18650HG2 dataset, the model achieves mean absolute errors (MAEs) ranging from 0.69 to 1.17% when tested on separate temperature datasets,while for the Panasonic NCR18650PF dataset, the MAE ranges from 0.36 to 2.53%, showcasing its robust effectiveness across varying conditions. Remarkably, the model’s training time is substantially reduced compared to other models. Additionally, we employ explainable AI (XAI) to interpret the model’s predictions, enhancing transparency by highlighting feature contributions to SOC estimation. Integrating the neural network into the BMS ensures accurate SOC estimation, optimizing battery performance and extending power electronics’ lifespan through improved control strategies. This balance of speed and accuracy makes the model suitable for real-time BMS applications and the use of XAI improves reliability, and interpretability, which are crucial for the safety, adoption, and widespread use of EVs.