Enhancing Spatial Load Forecasting Using Geographically Weighted Regression Based on Geographic Information Systems

Traditional electrical load forecasting often relies on macro approaches to analyzing historical data (time series) and energy consumption patterns (sectoral), which do not account for geographic variations and complicate the mapping of load centers. To address this, the study integrates Geographically Weighted Regression (GWR) model with Geographic Information Systems (GIS), providing a more nuanced analysis. GWR extends linear regression by incorporating geographic factors, resulting in location-specific regression coefficients that capture local variations in load density. GIS complements this by offering detailed spatial analysis and visualizing the distribution of electrical loads. The application of GWR revealed that land use for industry, land use for social purposes, GDP, and electrical load for business significantly affect load density in most subdistricts. The GWR model achieved a high R² of 99.82%, indicating that these variables explain almost all of the variance in load density, while GIS-based plots illustrated the spatial distribution and significance of these variables. The model has been validated as suitable for representing all subdistricts within the cluster, with an average MAPE of 3.36%. Therefore, the study demonstrates that the integration of GWR and GIS significantly improves the precision of load growth projections and the estimation of load centers. Additionally, it allows for a precise determination of the location and distribution of load centers, aligning forecasts with the underlying geographic structure and facilitating better-informed decision-making for energy management and infrastructure planning