Medium- and Long-Term Load Forecasting for Distribution Network Transformers with Highproportion Coal-to-Electricity Conversion Based on Trend-Regression Fusion Method

Against the backdrop of the “Dual Carbon” goals and energy transition, large-scale coal-to-electricity projects have led to stepped growth, strong seasonal fluctuations, and spatial heterogeneity in distribution network transformer loads, which traditional forecasting methods struggle to address accurately. This paper proposes a novel load forecasting approach for highproportion coal-to-electricity distribution networks: it first analyzes spatio-temporal load characteristics and screens strongly correlated factors via Spearman correlation analysis, then constructs a trend-regression fusion model integrating a time-seasonal trend analysis method (with seasonal differencing) and a spatial heterogeneity regression method, and finally employs dynamic weight fusion for medium- and long-term forecasting. Case studies show the model keeps prediction errors within 15 % for both medium- and long-term forecasts, accurately captures coal-to-electricity load characteristics, and provides a scientific basis for grid expansion, effectively supporting large-scale coal-to-electricity projects and stable operation of new power systems.