Speaker
Description
Photovoltaic-thermal (PVT) collectors are hybrid solar systems that efficiently produce both thermal and electrical energy in a single device. However, the interplay between their control parameters is complex, necessitating accurate yet computationally efficient modeling. Recently, artificial neural networks (ANNs) have proven effective in addressing this challenge due to their data-driven learning capabilities. This study evaluates the performance of a PVT system—considering electrical output, thermal energy, and exergy—under the climatic conditions of Chichaoua, Morocco. The analysis is based on a 2D energy balance model, which results cross-checked against literature data. Subsequently, the study compares the performance of two ANN models: a standard Multilayer Perceptron (MLP) and a Particle Swarm Optimization (PSO)-enhanced MLP, in predicting the exergy and energy outputs of the PVT system. The models utilize input parameters such as time, ambient temperature, and solar radiation intensity. Results demonstrate that ANNs can reliably predict the performance of the hybrid collector under real-world Moroccan climate conditions, with PSO optimization significantly improving prediction accuracy.
