Grammatical Evolution for Temperature Prediction Models in Different Photovoltaic Technologies

Abstract

This work presents a comparative analysis of thermal models for photovoltaic modules using Grammatical Evolution (GE) and Differential Evolution (DE) across four photovoltaic (PV) technologies: crystalline silicon (c-Si), amorphous silicon (a-Si), cadmium telluride (CdTe), and organic (OPV), under three sky conditions: sunny, cloudy, and diffuse. Temperature data were collected through a monitoring system in a photovoltaic cube, measuring temperatures on the horizontal face (top PV module) as well as environmental parameters (irradiance, ambient temperature, wind speed and direction, relative humidity). Three empirical models from the literature (Sandia, Faiman, and Obiwulu) were compared with 10 models generated by GE+DE using a Global Performance Index (GPI) to evaluate the accuracy of the models, considering five statistical metrics. The results show that in 11 out of 12 scenarios, the generated models outperform the empirical models, highlighting the importance of relative humidity in model accuracy. This work extends previous research, providing more accurate predictive models for the operating temperature of photovoltaic modules.

Lucía Serrano-Luján

Associate Professor

Lucía Serrano-Luján is an Associate Professor in the Department of Computer Science. Her research field is multidisciplinary. She developed a Life Cycle Assessment methodology to assess renewable energies and applied AI to their data. Her main goal is to impact energy-related materials production and find a more sustainable way to develop them. She has applied LCA to reduce graphene oxide and perovskites solar cells, build integrated photovoltaics, etc.

J. Manuel Colmenar

Full Professor

My research interests are focused on metaheuristics applied to optimization problems. I have worked on different combinatorial optimization problems applying trajectorial algorithms such us GRASP or VNS. Besides, I am very interested in applications of Grammatical Evolution, specifically in model and prediction domain, as alternative to machine learning approaches.