Real Time Environmental Fault Detection and Diagnosis in Photovoltaic Systems

Renewable resources are adopted nowadays to promote sustainable development, enhance energy security, and build resilient communities. The increase in demand for renewable resources has led to the use of photovoltaic systems. These systems must be efficiently monitored for fault and maintained properly for maximum productivity. Faults like cracks, dust, bird droppings and shadows may arise in the solar panel over a period, reducing the overall performance of photovoltaic systems. These faults can significantly reduce the lifespan of the solar panel. In this paper, voltage, current and surface temperature are measured using sensors from the photovoltaic panel. Real-time data from the solar cell via sensors are collected under no-fault, dust-induced, and partial shading-induced fault conditions. A fault detection methodology for photovoltaic (PV) systems is proposed using a combination of real-time power prediction and classification algorithms. Linear regression is employed to predict the fault. The fault is then classified using a random forest algorithm.