Volume 49

Enhancing Multimodal Meteorological Data Resolution via Diffusion Model for Accurate PV Potential Estimation Jiaze Li, Zhiling Guo, Huan Zhao, Hongjun Tan, Qing Yu, Rui Zhang, Jian Xu, Jinyue Yan

Abstract

High-resolution meteorological data is crucial for accurately assessing photovoltaic potential at the microclimatic level. However, due to the scarcity of highresolution data, modeling microclimates often becomes severely inaccurate, highlighting the need for technologies that improve the resolution of solar data. Additionally, current super-resolution models often fall short in accurately processing multimodal solar energy data, which is essential for improving the precision of PV potential estimation. This paper proposes a novel downscaling framework for meteorological data that leverages deep generative models. By integrating multimodal meteorological data from the National Solar Radiation Database (NSRDB), including Direct Normal Irradiance (DNI), Diffuse Horizontal Irradiance (DHI) and other relevant variables, with a diffusion model, the framework produces meteorological data at a high spatial resolution. Experimental results demonstrate that our superresolution approach not only outperforms baseline methods but also significantly enhances the accuracy of PV potential estimation when compared to using coarse resolution data. Consequently, the diffusion-based super-resolution framework shows great promise for widespread adoption in the field of photovoltaic energy.

Keywords Renewable energy, super resolution, diffusion model, multimodal learning, high resolution meteorological data, PV potential estimation