Real‐Time Data‐Driven Fault Diagnosis of Photovoltaic Arrays Using an Edge‐Server Machine‐Learning Framework

With the rapid global expansion of photovoltaic (PV) systems, reliable fault detection and classification have become crucial to ensuring system efficiency, safety, and longevity. This paper presents a comprehensive approach to diagnosing common PV array faults, such as line‐to‐line faults, open‐circuit faults, short‐circuit faults, and partial shading using machine‐learning (ML) techniques. A 3 × 2 PV array is modeled in MATLAB/Simulink to simulate various fault scenarios, and corresponding data sets are generated. The Raspberry Pi Zero 2W is employed for real‐time acquisition of voltage, current, and temperature signals. Raspberry Pi 4 serves as an edge intelligence node that performs online fault classification and notification. Sensor data are transmitted wirelessly using a lightweight communication protocol, enabling low‐latency decision‐making and real‐time fault alerts. Five supervised ML classifiers, such as decision tree (DT), k ‐nearest neighbors, support vector machine, artificial neural network, and bagged DT, are systematically evaluated based on their accuracy, robustness, and generalization capabilities under varying operating conditions. Furthermore, an ensemble model is implemented to enhance classification performance by aggregating multiple learners. The results from both simulation and hardware experiments demonstrate that the proposed ML‐based approach can detect and classify PV array faults with high precision and low computational overhead. This integrated data‐driven framework offers a scalable, adaptive, and cost‐effective solution for intelligent PV monitoring systems.