Development of an Integrated DBN-ELM, CNN-SVM and CNN-BiGRU Photovoltaic Array Fault Diagnosis Model Based on Weighted Probability Averaging

Photovoltaic arrays are continuously exposed to complex environmental conditions over long periods, making them susceptible to seven types of single and multiple failures such as shadow blocking and module aging. Fault diagnosis of photovoltaic arrays is essential to prevent failures that may lead to reduced power generation efficiency and potential safety hazards. This paper proposes a photovoltaic array fault diagnosis model based on weighted probability averaging, integrating DBN-ELM, CNN-SVM, and CNN-BiGRU methods. The model is calculated and experimentally verified. The results demonstrate that the integrated model achieves an overall accuracy, recall, precision, and F1-score of 99.0% across the four evaluation metrics, indicating a highly effective fault recognition capability.