Tracking Optimization in Agrivoltaic Systems. A Comparative Study for Apple Orchards.

Agrivoltaic (APV) systems, based on the co-location of solar panels and crops, are an innovative solution to land-use conflicts that often arise between agriculture and energy production. Their optimal functioning starts with efficient management and sharing of light between solar panels and underlying plants. This is where tracking systems come into play, as they offer the flexibility needed to strike a balance between energy production and crop growth. This thesis presents several tracking optimization techniques that focus on the availability and distribution of light. To simulate and analyze the performance of these strategies, a simulation model was created, with reference to a Fraunhofer ISE research project in Bavendorf, Germany where semi-transparent solar panels are installed over an apple orchard. The chosen developmental environment was Simtool, a Fraunhofer Python package based on the ray-tracing tool Radiance. Considering the computational cost of the simulation, a Bayesian black-box optimization algorithm was leveraged to relieve the latter from such a computational burden. For the first scenario, the goal was to maximise the radiation reaching solar panels. The algorithm developed, Diffuse-Track Optimization, proved particularly effective during overcast days, allowing daily energy gains of up to 9%. Plants were prioritized in the second scenario, Apples-Track Optimization with the goal of minimising their shading rates, which were seen to fall below 10% despite the presence of the tracking system. Lastly, a compromise between the two objectives was achieved in the final scenario through an overall optimization approach, called APV-Track Optimization. By assigning equal importance to the irradiation reaching trees and that which reaches photovoltaic panels, shading rates of less than 40% can be guaranteed throughout the year, with a reduction of the electrical yield by only 8% compared to astronomical solar tracking conditions. The study showcased the potential of the proposed methodology, representing a good starting point to develop holistic optimisations methods that are still lacking in the literature. Future developments will reduce runtime costs, integrate weather forecasts and validate results by means of accurate field measurements.