Prediction and Factors Determination of District Heating Load Based on Random Forest Algorithm

High energy consumption of district heating system can be improved by control strategy. Accurate prediction of heat load is very important for optimizing system control. Selecting reasonable input parameters is also the key to accurate prediction. Therefore, this paper establishes a short-term heat load forecasting model based on random forest regression (RFR), forecasts the heating load of a district in Xi’an, analyzes the most influential parameters in different month, and compares the forecasting results with the support vector regression (SVR). The results show that the performance of RFR model is better than that of SVR model by 10.2%. The load factors in different heating stages are not identical, indicating that energy operation mode has changed. Therefore, in different heating periods, the change of influencing parameters can be considered appropriately, and the prediction model can be adjusted to help the reasonable operation of the heating system and improve the energy efficiency.