Sensorless estimation of irradiance and temperature for renewable energy applications: An experimental examination

Abstract

This paper prescribes and examines a sensorless Neural Network (NN) model for the precise estimation of essential climatic resources-irradiance and temperature-integral to optimizing renewable energy systems. Reliable data on these variables is crucial across multiple disciplines, especially in renewable energy, where it drives numerous technical and economic objectives. However, achieving exact, real-time estimation remains complex, hindered by the dynamic and variable nature of these variables. This work proposes an NN approach that estimates irradiance and temperature using only the maximum power point (MPP) outputs from a modern photovoltaic (PV) system, eliminating the need for direct sensor measurements. This approach not only offers high adaptability but also integrates seamlessly into existing PV infrastructure, enabling real-time, cost-less implementation. To rigorously validate the model, extensive experimental evaluations were conducted across multiple days, demonstrating its accuracy and resilience. The model achieved a Mean Absolute Error (MAE) of 0.87 and 2.728 for irradiance and temperature, respectively; and a Root Mean Square Error (RMSE) of 2.1127 and 9.1008. These metrics highlight the model's precision and reliability, establishing it as a powerful tool for enhancing the efficiency and intelligence of renewable energy systems. The findings offer significant contributions to renewable energy development, providing a robust, sensorless solution for real-time climatic resource estimation with broad interdisciplinary applications, ultimately empowering smarter and more sustainable energy systems.