ANN-based prediction of photosynthetically active radiation (PAR) in an agrivoltaic greenhouse system

Agrivoltaics represents an innovative approach that mitigates land-use conflicts between the energy and agricultural sectors. The integration of semi-transparent photovoltaic modules into agriculture is a valid strategy for increasing the availability of solar radiation incident on crops. These modules modify the spectral distribution of solar radiation and the amount of Photosynthetically Active Radiation (PAR) available for photosynthesis. However, direct PAR measurement is limited worldwide, making it necessary to use empirical models for its prediction. The aim of this paper is to fill this gap by presenting a Multi-Layer Perceptron (MLP) Artificial Neural Network (ANN) model developed to predict PAR in semi-transparent agrivoltaic greenhouses. The network was trained using experimental data collected at the University of Ja´en on two agrivoltaic greenhouses: a control unit with transmissivity of 92% and a semi-transparent configuration with transmissivity of 20%. Specifically, the input variables considered for this study are global radiation on the array plane (G poa ), angle of incidence (AOI), air mass (AM) and transmittance coefficient (CT) of the modules. The model shows excellent convergence between the predicted and target values for both greenhouse configurations, with an average correlation coefficient (R) of 0.99 and a Normalized Mean Squared Error (nMSE) equal to 0.0135.