Enhancing reliability and safety of uncertain grid-connected photovoltaic systems based on intelligent transient regime analysis

Ensuring the uninterrupted operation of Grid-Connected Photovoltaic (GCPV) systems is crucial, as these systems are highly susceptible to faults and downtime caused by various factors, which can lead to significant system damage. To address these challenges, fault detection and diagnosis (FDD) methods are essential to maintain the reliability and safety of GCPV systems. This paper presents a transient regime based on the FDD approach of uncertain GCPV systems, employing deep learning techniques to detect and classify faults effectively. Furthermore, the proposed method takes advantage of the transition phase between healthy and faulty states in renewable energy systems to enable early fault detection by identifying anomalies in performance signals. By combining transient regime analysis with deep learning techniques, the approach facilitates rapid and accurate fault detection, thereby enhancing the reliability and extending the lifespan of photovoltaic systems. To handle uncertainties in the measured data, an interval-valued data representation is utilized, ensuring robust fault analysis under varying conditions. However, the hyperparameters of the proposed techniques are optimized using the Genetic Algorithm, improving their adaptability to diverse operating scenarios. The robustness of the methodology is further validated by introducing varying levels of noise into the data, simulating real-world perturbations and dynamic variations. The processed outputs are used to train deep learning classifiers to distinguish between various operating modes in GCPV systems. Experimental validation with real-world data demonstrates the efficacy and robustness of the proposed approach, enabling immediate decision-making and preventing fault propagation. The results highlight the strategy’s high accuracy and computational efficiency, contributing to improved reliability and safety of GCPV systems.