Attention-Enhanced Mlp Model for Robust Fault Diagnosis in Photovoltaic Systems Under Data Loss and Noise Interference

Photovoltaic (PV) fault detection and diagnosis is an important research topic in the field of renewable energy. In this study, we propose a novel attention-enhanced multilayer perceptron (MLP) network, referred to as MLP-AP, for the detection and diagnosis of photovoltaic (PV) faults. The network employs a dual-path feature extraction architecture that simultaneously processes both the sequential points and key parameters of the full I-V curves, leveraging attention mechanisms to enhance point-wise feature extraction. The performance of the proposed MLP-AP network is evaluated using simulated datasets, taking into account potential disturbances and data loss that can occur under realistic conditions. Specifically, we consider two types of challenges: (1) environmental noise and random measurement errors, and (2) data loss resulting from transmission packet loss and sampling failures. The results demonstrate that classification accuracy of the proposed model can achieve an overall accuracy greater than 98.5 %, exhibiting good reliability and generalization ability compared to other baseline neural networks.