Optimised model for the energy estimation of an Agri-PV plant with high-efficiency bifacial modules

Photovoltaic energy has emerged as one of the cheapest and most reliable carbon-free energy sources. Recently, the combination of solar energy production with agricultural activities in a new paradigm called Agri-PV has enhanced the sustainability of these systems. However, these new solutions have increased their overall complexity with new technologies, such as bifacial modules and solar trackers. The goal of this thesis, developed in collaboration with the international energy company ENGIE, is to elaborate a model able to predict the ideal performance of an Agri-PV plant from ambient parameters, specifically, solar irradiance and air temperature. The facility is located in Mazara del Vallo (TP), Sicily, and is managed by the Italian unit of ENGIE. At its inauguration in 2023, the Mazara solar farm represents the Italian Agri-PV plant with the largest installed capacity at 66 MWp, implementing more than 120,000 bifacial modules with a rated efficiency of 21.5%. It is important to underline the power limit applied to the plant by the Italian Transmission System Operator (TERNA), indeed, the system cannot inject more than 50 MW into the national grid. This thesis work focuses on the numerical identification of the optimised coefficients for the model. First, a proper dataset is selected by applying several filters to the data acquired by the Supervisory Control And Data Acquisition (SCADA) system for a period of four months, from May 2023 to August 2023. Indeed, the first filter permits to exclude night-time data; the second filter selects measurements not affected by the TERNA power injection limitation; the third filter eliminates data of days with inverter availability different from 100%; the fourth filter keeps out irregular values of DC-AC efficiency; the last filter excludes data with unstable environmental conditions by imposing a maximum data variation. Starting from the obtained dataset, the coefficients of a proper energy model, based on an irradiance-proportionality law, have been numerically optimised. The optimisation process has been carried out in Matlab environment selecting the proper algorithm to solve non-linear constrained optimisation problems. To validate the process, the daily and monthly Normalized Root Mean Square Errors (NRMSEs) have been calculated with respect to the SCADA power. In the final analysis, unfiltered data have been employed to evaluate the effectiveness of the optimised model.