Fault Detection and Diagnosis Model for Photovoltaic Array Based on Multi-Scale Transformer

This paper presents an advanced automatic fault detection and diagnosis method for photovoltaic (PV) arrays aimed at enhancing the efficiency, reliability, and safety of PV systems. Traditional artificial intelligence methods, such as machine learning, although useful, are typically applied to structured, feature-rich two-dimensional data, thus lacking the capability to effectively classify samples represented as arrays of sampling points with only curve characteristics. This limitation may hinder the model's effectiveness and lead to local optimization issues. Existing deep learning techniques have limited feature extraction capabilities and often iterate repetitively over modules without extracting richer features from multiple levels and perspectives. To address these challenges, we propose a novel intelligent fault detection and diagnosis method utilizing a specifically designed multi-scale Transformer network. This network analyzes I-V curves through bilinear interpolation and incorporates irradiance and temperature data, ensuring a comprehensive input basis. We conducted thorough fault simulation experiments on the Simulink platform to evaluate the performance of our proposed method. The results highlight significant achievements in accuracy, generalization capability, reliability, and training efficiency.