AI-driven instrumentation for predictive maintenance of solar pv and thermal collectors in smart grids

Abstract

The growing deployment of solar photovoltaic (PV) and thermal collector systems requires intelligent maintenance frameworks to improve reliability and reduce operational costs. This study presents a hybrid Aidriven predictive maintenance system integrating Long Short-Term Memory (LSTM) networks for temporal forecasting and Random Forest (RF) classifiers for fault detection within an IoT-based multi-sensor framework. Using MQTT-enabled cloud-edge communication, the system achieved a forecasting RMSE of 0.98 °C and 27.9 W, an F1-score of 0.91, and a 37 % reduction in unscheduled downtime. The model improved energy yield by 5.6 % and shortened the return-on-investment period to 22 months, validating the economic and technical benefits of Aibased predictive maintenance.