Integrating Wavelet Reconstruction with Hybrid TCN-LSTM Model for Enhanced Fault Diagnosis in Distributed Photovoltaic System

As photovoltaic (PV) systems become more prevalent, effective fault detection and diagnosis in distributed PV systems (DPVS) are increasingly important. This study presents a novel approach combining Temporal Convolutional Networks (TCN) with Long Short-Term Memory (LSTM) networks for enhanced fault diagnosis in DPVS. By using wavelet transform-based signal denoising techniques, our method reduces noise and improves feature extraction. The TCN-LSTM hybrid model captures both long-term dependencies and sequential patterns, achieving a diagnostic accuracy of 99.33% on real-world PV fault datasets. This approach significantly enhances fault diagnosis, demonstrating strong performance and practical applicability in PV system maintenance.