Study on the Impact of Low-Cost Sensor Alternatives for Photovoltaic Panel Modelling in Smart Grid Applications

Anabel Díaz-Labrador, Ángel Delgado, Héctor J. Pérez-Iglesias, Oscar Fontenla-Romero, José Luís Calvo-Rolle

Abstract: The growing use of photovoltaic systems in smart grids requires the proper monitoring and effective management of power generation. Accurate photovoltaic models typically use radiation sensors. However, their high cost limits their application in small or resource-constrained installations. This work investigates the use of low-cost alternatives, specifically ambient and panel temperature sensors, for photovoltaic performance modeling. Two experimental setups were evaluated: one using a traditional radiation sensor and another using only temperature-based sensors. Several regression models including linear, polynomial, ridge, and lasso regression were systematically compared. The results show that radiation sensors have better performance (\(R^2 \approx 0.93\), polynomial regression), but temperature sensors still give reasonable accuracy (\(R^2 \approx 0.86\)). Therefore, temperature sensors are a viable and much less expensive alternative. Future research should explore additional atmospheric variables and other types of machine learning techniques to further improve accuracy in a low-cost approach.

External IDs:dblp:conf/iwann/DiazLabradorDPFC25