Assessment of Solar Irradiation Data Sources and Prediction Models for Rural Villages in the Colombian Amazon Region

Abstract

Despite global efforts to adopt renewable energy, many remote regions still lack reliable electrical services. Addressing this requires a thorough analysis of solar resource data to identify viable solutions for these underserved areas. We evaluate the error in solar radiation data from a satellite image-based Random Forest (satellite RF) model by using data from IDEAM meteorological stations and NASA sources. By rigorously comparing these datasets, we aim to assess the reliability of predictive sources of solar radiation in the Amazon region. The results help establish confidence in various data sources, essential for utilizing estimated solar energy data in renewable energy research. We compared the data using the Relative Root Mean Squared Error (Relative RMSE). On the one hand, the relative RMSE between NASA and IDEAM ranges from 6.86% to 20.93%. On the other hand, the error between satellite RF model and IDEAM fluctuates between 6.56% and 12.33%. Similarly, the error between satellite RF model and NASA ranges from 4.80% to 15.27%. The findings indicate that the error in NASA data is higher compared to the error in satellite RF model data when benchmarked against IDEAM. Despite the limited number of meteorological stations and a maximum error of 20.93% between the two predictive data sources compared to ground-based observed data, we consider it reliable to use estimated solar radiation data for developing effective renewable energy solutions in remote locations.