Three-Phase Inverter Fault Diagnosis Based on an Improved Deep Residual Network

This study addresses the challenges of limited fault samples, noise interference, and low accuracy in existing diagnosis methods for three-phase inverters under real acquisition conditions. To increase number Wavelet Packet Decomposition (WPD) denoising a Conditional Variational Auto-Encoder (CVAE) are used sample enhancement based on faulty samples. The resulting dataset is then normalized, pre-processed, to train an improved deep residual network (SE-ResNet18) model with channel attention mechanism. Results show that augmented samples improve compared original Furthermore, SE-ResNet18 achieves higher fewer iterations faster convergence, indicating its effectiveness accurately diagnosing inverter open-circuit faults across various situations.