[3F5-GS-10] AI application

[3F5-GS-10-02] Solar radiation estimation using Himawari-8 and LEO satellite data

〇Ryota Tobishima¹, Shoichi Urano¹ (1. Meiji University)

Keywords:Solar radiation estimation, Satellite data, Power system, Solar power, Neural network

When installing PV generation, it is necessary to consider the amount of solar radiation and judge whether sufficient power generation can be expected at that site. However, areas, where solar radiation data is continuously measured, are limited. Therefore, estimating solar radiation using satellite data is currently attracting attention. Satellites possess a large amount of observation data over a wide area. Hence, it is possible to obtain solar radiation data at any location on the earth by estimating solar radiation on the ground using satellite data as a characteristic quantity. However, the issue with satellite irradiance estimation is that the estimation accuracy is insufficient. Therefore, in this study, in addition to GEO satellite data, generally used for irradiance estimation, LEO satellite data is also treated as a feature. Although the data has a high resolution, it has not been used for irradiance estimation because of its insufficient temporal resolution. However, recent improvements in satellite constellation have made it possible to acquire LEO satellite data with high temporal resolution. The relationship between these two types of satellite data and ground-based data will be modeled by machine learning to evaluate the effectiveness and accuracy of satellite solar radiation estimation.