Performance Assessment of ANN-Based FDIRM in PV Boost Converters

Renewable energy systems, especially those in aerospace, automotive, industrial, and medical fields, require high reliability and performance to ensure continuous and reliable operation. However, photovoltaic (PV) boost converters are vulnerable to power switch faults, such as open and short circuits (OCF and SCF), which can significantly affect system reliability and lead to false alarms. This study introduces a fault detection, identification, and reconfiguration method (FDIRM) that utilises artificial neural networks (ANN) for detecting, identifying, and compensating the switches faults. The MPPT method used incorporates an ANN-driven sliding-mode maximum power point tracking strategy (ANN-SM MPPT) for enhanced PV performance. The investigated FDIRM method uses real-time data on power, voltage, current, and duty cycle of the PV module, allowing the ANN to rapidly detect and compensate for faults using only current and voltage sensors, simplifying system complexity and reducing costs. A redundant leg is also combined into the boost converter topology to maintain PV module functionality following fault reconfiguration. In this study, the diagnostic and reconfiguration approaches of the PV boost converter are designed and implemented using hardware-in-the-loop (HIL) control managed by the dSPACE 1104 board programmed in MATLAB/SIMULINK. The simulation results confirm the high performance of FDIRM, demonstrating effective fault detection, isolation, and compensation to support autonomous and robust operation.