A Scalable Random Forest-Based Scheme to Detect and Locate Partial Shading in Photovoltaic Systems

Abstract

Photovoltaic (PV) systems are prone to partial shading (PS) due to the environmental factors that they function in such as vegetation, nearby structures, and clouds. All types of PS scenarios can lead to power loss and hot spots in the PV system due to module mismatch and heating of shaded cells. To mitigate the power loss that occurs due to PS, it is imperative to detect PS and its characteristics, such as the number of shaded modules and the associated shading factor (SF), in a reliable manner. This paper proposes a three-step framework to detect and locate PS, the number of shaded modules, and the SF in the PV system using a random forest (RF)-based approach. The proposed approach utilizes independent string current and voltage measurements to distinguish different PS scenarios. This approach allows for a scalable data acquisition through an uncoupled modeling scheme. PS, the number of shaded modules and the SF are deduced with accuracies of 99.5%, 92.3%, 90.2%, respectively. Further, the proposed approach is validated through two testing tiers, and its ability to detect multiple PS scenarios in a PV system has been highlighted. The results observed through different PS scenarios confirm the high reliability and demonstrate the effectiveness and scalability of the proposed RF-based approach.