APyV: Designing Agrivoltaic Facilities Considering Crop Needs

The integration of agricultural and photovoltaic (PV) systems, known as agrivoltaics, offers a promising solution to simultaneously produce food and generate electricity, optimizing land use especially in regions where arable land is scarce. This work presents a novel Python-based tool designed to optimize the geometrical dimensions of agrivoltaic installation for any PV setup at any location. Techniques from bifacial_radiance (Deline, 2020) are used with improvements to simulation speed with parallelization, additional analysis and reporting functionality. It also employs dedicated methods for analyzing agrivoltaic systems and deriving characteristic key performance indicators easing the informed design of such systems. As shown in Figure 1, the tool leverages advanced ray tracing techniques to accurately simulate solar radiation distribution and its impact on both photovoltaic panels and crops beneath them. By the flexibility of integrating any suitable crop model, the tool adapts to various types of vegetation, assessing how different configurations of PV panels affect crop growth and in return the land equivalent ratio (LER) (Trommsdorff, 2021). Finally, it is designed to fit well into optimization algorithms supporting timestep classification and the automized spawn of new system configurations. In this case study, the tool is used in a Bayesian optimization framework to fine-tune system parameters. This optimization method systematically improves the design by focusing on key performance indicators such as electricity output and optimized sunlight exposure for crops resulting in maximized yields. Additionally, it accommodates the nonlinear behaviors of the LER function, influenced for example by regulatory frameworks imposing minimal crop yield levels or in some cases crop quality. The Bayesian framework iteratively then updates the system parameters based on prior performance metrics, leading to an optimal balance between energy production and crop yield. The user-friendly interface of the tool allows practitioners and researchers to easily modify input parameters, making it adaptable to a any agrivoltaic scenario across any geographical locations. Figure 2 depicts the results for a 1P tracked system in Magdeburg, Germany with Alfalfa as crop. Under the given boundary conditions, the optimum system design is characterized by a row-to-row distance of 5.7 m, leading to the maximum value of LER equal to 152.4 %. These results indicate that our tool can significantly enhance the productivity of agrivoltaic systems by tailoring designs to specific environmental conditions and crop types. In conclusion, this tool not only facilitates the precise design of agrivoltaic systems but also contributes to the sustainable co-development of agriculture and renewable energy sectors. Future developments will aim to incorporate real-time data feeds from monitored facilities and other machine learning techniques to further enhance the predictive accuracy, speed and functionality of the tool.