Quantifying electric power consumption (EPC) at high spatial and temporal resolution (HSTR) is critical for evidence-based energy planning, especially in regions lacking detailed ground statistics. Nighttime light (NTL) data are consistent proxies for EPC; however, the relatively coarse spatial resolution of conventional NTL products derived from the Operational Linescan System (OLS) on board the Defense Meteorological Satellite Program (DMSP) satellites, as well as the Visible Infrared Imaging Radiometer Suite (VIIRS) payload on board the Suomi National Polar-Orbiting Partnership (SNPP) and Joint Polar Satellite System (JPSS) satellites, limits the accurate monitoring and analysis at sub-100-m scales. This study presents an ensemble machine learning framework for EPC estimation using fused NTL data generated through the Nighttime Light Spatiotemporal Fusion (NTLSTF) model, integrating VIIRS Black Marble monthly product (VNP46A3) with Sustainable Development Goals Satellite-1 (SDGSAT-1) panchromatic imagery to produce 10-m monthly and annual NTL composites. An ensemble machine learning model for EPC estimation was developed by leveraging three machine learning models, random forest (RF), artificial neural network (ANN), and simple linear regression (SLR), to enhance accuracy, capture nonlinear relationships, and mitigate overfitting across diverse spatial contexts. Punjab Province, a pivotal region along the China-Pakistan Economic Corridor, served as the case study. The 10-m highresolution NTL data substantially improved EPC estimation by reducing errors, preserving pixel-level variability, and aligning closely with observed values. The ensemble model achieved exceptional predictive accuracy [R-2 = 0.898, root-mean-square error (RMSE) = 1.132 GWh], representing a 35.6% improvement over the best individual model. Taylor diagram analysis further confirmed reliable preservation of spatial variance (SD ratio = 2.87) and correlation (0.889). The model's ability to resolve neighborhood-level dynamics was further validated through cityscale analysis in Lahore. These findings show that integrating fine-resolution NTL data with ensemble learning advances EPC estimation, offering a replicable framework for grid optimization, demand forecasting, and energy equity assessments in data-scarce regions.