Search Survive Optimization Based Deep Incorporated Model for Electric Vehicle Battery Fault Detection

Abstract

With the progressive switching from a conventional transportation system to an intelligent transportation system (ITS), the eco-friendly alternative is made possible in metro cities. Moreover, electric vehicles (EVs) gained more attention due to their low charging costs, low energy consumption, and reduced greenhouse gas emissions. However, a single failure or malfunction in an EV's intrinsic components due to poor charging infrastructure can bring about a high tendency of fault occurrence that needs to be diagnosed earlier for efficient safety management. In addition, ensuring the safety and reliability of these EV batteries remains a critical challenge that underscores the importance of an efficient battery fault detection system, pivotal in enhancing battery safety and lifespan. Hence, the research centers on developing a well-structured battery fault detection model leveraging a Search- Survive optimization (SSO) based deep incorporated model. This incorporated model combines Deep Convolutional Neural Network (Deep CNN), Deep Bidirectional Long-Short Term Memory (Deep BiLSTM), and Deep Belief Network (DBN) that assists in extracting the hierarchical representations and the spatial–temporal features associated with the various EV faults. The deep incorporated model is optimized with SSO that aids the model to perform enhanced battery fault detection of EVs. Performance assessment relies on key parameters like accuracy, sensitivity, and specificity, based on the NASA battery dataset. Impressively, the SSO-based Deep Incorporated model attains an accuracy of 96.00%, sensitivity of 96.29%, and specificity of 95.72 for 80% of training. With k-fold 10 validation, the proposed model attained the metric values of 96.31%, 97.29%, and 95.32% respectively using the NASA dataset and surpassed other existing techniques.