Short - term Load Forecasting Method for Distinguishing Building Types in the Heating System

First, the heat usage characteristics of different building types in the heating system were analyzed. Then, a correlation analysis was conducted on quantifiable factors that may affect the demand-side load, and the top five factors with the highest correlation coefficients were selected as input features for the load forecasting model. A Support Vector Machine (SVM) algorithm, optimized using Grid Search (GS) and Cross-Validation (CV), was used to construct the forecasting models. This resulted in separate load forecasting models for public buildings, residential buildings, and a mixed model. These models were then used to predict the hourly load demand for the next 12 hours at the heat stations. The results show that constructing separate forecasting models for different building types improves the short-term load forecasting capability and enhances the prediction accuracy. This provides reliable load forecasts for predictive control in the heating system, reducing energy waste due to over-supply and improving user comfort by avoiding under-supply