Global horizontal and direct normal solar irradiance modeling by the machine learning methods XGBoost and deep neural networks with CNN-LSTM layers: a case study using the GOES-16 satellite imagery

Restrictive legislations on the use of fossil fuels encourage the research and development of clean and renewable energies. Renewable energy is characterized by random behavior, which hampers its integration into the current energy base system. Thus, estimating solar irradiation is important for the adoption of renewable energies into the current energy matrix. In this paper, two machine learning estimation models for global horizontal (GHI) and direct normal solar irradiance (DNI) are proposed: the first uses XGBoost and the second employs a convolutional neural network (CNN) combined with a long short-term memory (LSTM) network, forming the hybrid CNN-LSTM model. The case studies apply both models to process images from the GOES-16 satellite, taken from the city of Petrolina, Pernambuco, Brazil. Their results are compared against the reference Copernicus Atmosphere Monitoring Service, Solcast and the Physical Solar Model (PSM) provided by the National Solar Radiation Database. For the GHI estimation, the PSM model achieved the lowest RMSE, 147.23 W/m 2 , while for DNI estimation, the CNN-LSTM model performed best, with an RMSE equal to 238.22 W/m 2 . In this case, the proposed models achieved lower RMSE for DNI estimation when compared against the benchmark models, improving by 2.89% and 1.70% for the CNN-LSTM and XGBoost models, respectively.