ANN–IoT Enabled Predictive Framework for Performance Assessment and Maintenance Scheduling in Floating Solar Photovoltaic (FSPV) Systems

Abstract

With the rise of solar Photo-Voltaic (PV) energy, the need for optimized land usage and enhanced energy yield has drawn significant attention. Although the solar industry is highly developed and extensively researched, conventional PV installations require large land areas. To address this challenge, Floating Solar Photo-Voltaic (FSPV) systems have emerged in recent years as a promising solution, as they substantially reduce land requirements. While FSPV systems are installed on water surfaces and are often assumed to experience lower dust exposure, airborne particulates resulting from high pollution levels, nearby agricultural activity, and regional atmospheric conditions still lead to measurable soiling losses. The present research utilizes an Internet of Things (IoT) and Artificial Neural Network (ANN)-based predictive architecture, originally developed for land-based PV modules, and successfully generalizes it for FSPV systems. The real time data from ACME Solar power plant database of PAVAGADA AC-50MW and DC-67.46MW with latitude-14.260099^oN/longitude-77.471534^oE of site are taken in account to create the prediction model. was conducted to adapt and validate the model. The results demonstrate notable improvements, with annual energy yield increasing by 8–15% and cleaning costs reduced by 30–40%.