A Comprehensive Framework for Network System Fault Diagnosis using CNN–LSTM and Optimization Algorithms

As reliance on battery systems increases in electric vehicles and renewable energy applications, accurate fault diagnosis of current sensors is essential for maintaining performance and safety. This research addresses challenges in detecting current sensor faults, particularly noise, data variability and algorithm inefficiencies that can compromise battery health monitoring. We propose a comprehensive framework utilizing MATLAB R2023a, starting with constructing a five-phase permanent magnet synchronous motor (PMSM) simulink model, followed by an extensive data collection process. The collected data are preprocessed using an adaptive smooth variable structure filter with a time-varying boundary layer (ASVSF-VBL) to mitigate noise interference. To detect current sensor failures, we implement a Kalman filter and Cuckoo algorithm (KFCA) as an equivalent circuit model complemented by deep learning called CNN–LSTM architecture for advanced fault detection. Additionally, we employ a pulse width modulation voltage source inverter (PWM VSI) to monitor transistor open-circuit failures. The self-adaptive Bonobo optimizer with least mean squares (SABO-LMS) optimizes battery condition assessments, enabling dynamic adjustments to varying operational conditions. By plotting these metrics, we aim to demonstrate the effectiveness of our approach in enhancing the operational integrity of battery systems, ultimately contributing to safer and more efficient energy storage solutions. Integrating MATLAB R2023a in our simulations facilitates robust analysis and validation of the proposed methodology.