Optimized model of 1-min PV power vs experimental data for an Agrivoltaic tracking system with bifacial modules

Nowadays, Photovoltaic (PV) energy has become one of the most affordable and reliable carbon-free energy sources, playing a key-role in the energy transition scenario. In recent years, Agrivoltaic systems have emerged as a promising solution for integrating solar energy production with agricultural activities, enhancing the sustainability of these systems. This integration introduces new complexities for the maintenance and operation of such systems: in this context, efficient PV solutions, like bifacial modules and solar tracking systems, are adopted to maximize the energy production. The goal of this thesis is to develop an optimized power model, based on a 1-min timescale, in order to estimate the energy production of a utility scale Agrivoltaic plant located in Mazara del Vallo (Italy), and managed by the Italian company ENGIE ENERGIES ITALIA S.r.l. This PV system was built in 2022, has a rated power of 66 MW and occupies an area of 115 hectares. The plant includes more than 120’000 monocrystalline silicon PV modules with bifacial technology and efficiency of 21%. A mono-axial solar tracking system permits to change the inclination of the modules during each day to maximize their energy production. Moreover, the plant is divided into 10 sections, each one identified by an Inverter Transformer Station (ITS), and a variable number of string boxes (between 15 and 32) is connected to each ITS, with a total number of 272 string boxes. The power model is based on an irradiance-proportionality law and relies on the utilization of optimized parameters like the bifaciality factor and the temperature coefficient for maximum power of the PV modules, as well as the DC/AC conversion efficiency of the inverters. Such quantities are numerically optimized on a string box level after applying a set of filters aiming to exclude measurements affected by low irradiance, clipping phenomenon, unstable weather conditions. The quantities are optimized considering the dataset acquired by the Supervisory Control and Data Acquisition (SCADA) system of the plant in the month of April 2023. This month is considered to exclude any effect due to early-stage degradation. The effectiveness of the optimization is evaluated by calculating the Normalized Root Mean Square Errors (NRMSEs) for each string box under test. The last part of the analysis includes the energy estimation for the month of April by the model and the comparison with the actual performance of the plant.