Evaluation of solar radiation obtained from NASA and satellite imagery-based prediction models adjusted for microgrid sizing in Homer Pro.

Introduction: The optimization of renewable energy resources is transcendental to satisfy the world energy demand and to avoid the adverse effects produced by the burning of fossil fuels. Therefore, there are several studies that seek to estimate the capacity of renewable energy sources in a geographical location. Likewise, there are several software applications that seek a balance between the investment and the installed capacity of an electric power generating plant. Objective: This work uses the results of the Random Forest algorithm to predict solar radiation from satellite images. This technique achieved a performance in R2 of 0.82 and in RMSE of 107.05. The purpose of this study is to evaluate the results of 2 models of photovoltaic systems designed for 10 different locations in the Colombian territory. Model M1 uses solar radiation data from NASA. The M2 model uses solar radiation data generated by Random Forest. Methodology: The evaluation of solar radiation from NASA and the Random Forest algorithm is based on simulations provided by the energy resource optimization tool Homer Pro. Results: The simulations of both models in Homer Pro show a difference in the capacity of the system components of between 0.0% and 47.31%. The difference between electric power generation ranges from 0.0% to 11.99%. Similarly, the difference between system costs is between 1.34% and 23.64% respectively. Conclusions: The solar radiation data estimated by Random Forest is constituted as an alternative to the solar radiation data provided by NASA, given that the differences in the capacity of system components, electric power generation and total system costs are on average at around 27%. [ABSTRACT FROM AUTHOR]