Wind design of solar panels for resilient and green communities: CFD with machine learning

Abstract

Climate change mitigation and adaptation in urban environments call for more reliance on clean energy sources. Large photovoltaic (PV) systems have been enjoying renewed interest in clean and renewable energy. However, designing resilient PV systems faces an increased risk due to windstorms. Whether wind loads on PV systems are well understood, properly accounted for, and the damage is mitigated are crucial questions. While computational fluid dynamics (CFD) is proven effective for quantifying wind loads on structures, accurate and affordable computations are challenging. In this paper, we employ CFD approaches and machine learning (ML) to obtain the design wind loads on solar panels. We validate the CFD simulations using experimental data and compare the results with the standard practice. Our findings suggest that experimentally validated CFD simulations can yield different results from the standard practice. Additionally, we recommend stowing solar panels at a -15° angle during wind events to reduce damage. CFD simulations are then employed to train an ML model to predict velocity and pressure distributions around a solar panel. The study demonstrates that integrating ML and CFD can significantly speed up simulations (up to 10,000 times faster) without sacrificing accuracy. Efficient designs can shape the future of PV systems and contribute to climate change adaptation and mitigation for improved disaster resilience and circular economy policies.