Coluccia, Simone (2026) Crop performance analysis and AI-driven optimisation in different agrivoltaic systems under variable radiative conditions. [Tesi di dottorato]

Documento PDF PhD_thesis_Coluccia_Simone_Final_Version.pdf Download (18MB)

Collection description

This doctoral thesis investigates the behaviour of agrivoltaic systems through an integrated approach that combines agronomic experimentation, ecophysiological analysis, and advanced artificial-intelligence-based modelling tools. The primary objective is to assess the impact of different agrivoltaic configurations on crop performance and to develop methodologies for the optimised management of such systems. Experimental activities were conducted under real field conditions in Sweden, where vertical agrivoltaic systems were compared with conventional ground-mounted photovoltaic configurations using oat crops (Avena sativa L.). Monitoring included biophysical and physiological parameters, such as leaf area index (LAI), photosynthetic rate, and biomass production, as well as qualitative yield indicators. Statistical analysis revealed pronounced spatial variability in crop performance within the agrivoltaic systems, primarily driven by radiation distribution and shading patterns. Notably, certain portions of the agrivoltaic system exhibited yields comparable to or exceeding those observed under full-field conditions. In contrast, ground-mounted photovoltaic systems generally resulted in a significant reduction in productivity. In parallel, data obtained from a meteorological station installed within the vineyard hosting the experimental agrivoltaic system were utilised to support the development and validation of predictive models of crop behaviour in an agrivoltaic environment. Within this framework, a management module based on machine learning and reinforcement learning techniques was developed, capable of integrating environmental, energy, and agronomic variables to dynamically optimise the configuration of photovoltaic panels. The problem was formulated as a multi-objective optimisation task to balance energy production and crop performance. The results demonstrate the effectiveness of a data-driven approach to managing agrivoltaic systems and highlight the importance of integrating experimental data with computational tools to enhance system design and operation. This work contributes to the development of advanced strategies for sustainable land use and for the integration of agricultural and energy production.

Downloads

Actions (login required)

View Item