Predicting renewable power potentials is vital for generating clean sources of power and reducing the dependency on fossil fuels resulting in fewer greenhouse gas emissions. This paper illustrates a comparative analysis between two machine learning (ML) algorithms, the artificial neural networks (ANN) and the gene expression programming (GEP) models, in estimating the geospatial renewable wind and solar power potentials. The datasets, from which the ML models should learn, represent different locations obtained from Egypt’s map. Error Analysis including the mean square error (MSE) and the coefficient of determination was utilized to specify the most significant geospatial generated model regarding accuracy. The results indicate that the ANN has shown better results against the GEP by showing fewer errors. Furthermore, sensitivity and uncertainty analysis were implemented in the ANN model providing further accuracy measurements. The predictive key results were used as a main power component of an underground water pumping system used for irrigation to demonstrate the aspect and the replicability of the study. A geographic information system (GIS) management algorithm drives the pumping system by optimizing the performance of the underground water system to function sufficiently using the estimated renewable power potentials.