Micro Hydro Power Site Characterization in Indonesia: Variable Optimization for Site Selection Using GeoDetector and RFE-Random Forest

Abstract

Micro hydropower (MHP) systems are a promising alternative renewable and sustainable energy source to conventional fossil fuels, particularly in regions with abundant water resources like Indonesia. The success of MHP initiatives is contingent upon identifying suitable sites and remains challenging related to influencing parameters in site selection for the regional/national scale. Therefore, this study aimed to determine the essential influencing variables for MHP site selection by evaluating multiple variables related to the existing MHPs. The method used for analysis was the GeoDetector and Recursive Feature Elimination-Random Forest (RFE-RF) approach in the Geographic Information System (GIS) framework. Combining GeoDetector and RFE-RF models proves to be a potent tool for essential influencing variables screening in MHP site selection. The eight essential variables were obtained, down from nineteen original variables, with a better performance statistically. This hybrid approach considers spatial patterns in data for variable selection, ensuring alignment with the chosen machine learning method. This study result is expected to assist decision-makers in the preliminary evaluation stage of MHP site exploration and promote Indonesia’s transition to a cleaner, more renewable energy future and participatory forest conservation.