Identifying the most significant input parameters for predicting district heating load using an association rule algorithm

High-energy consumption of a district heating system is mainly caused by low-efficiency operation. Real-time regulation and control of the system in the operation stage has great energy-savings potential, and accurate load prediction is the basis of system control and on-demand supply. The key for obtaining an accurate hourly heating load forecast value is to select reasonable input parameters. Therefore, this paper summarizes the methods of predicting heating loads and selecting input parameters, proposes the Eclat algorithm of association rules to obtain the best combination of input parameters, and introduces the Spearman parameter selection method for comparison; the Eclat-Support Vector Regression (E-SVR) and Spearman-Support Vector Regression (S-SVR) prediction models are established. The heating load data of a district in Xi'an, China is taken as an example, and the results show that the water supply temperature of the historical primary network was the most significant factor that affects the district ultrashort-term heating load prediction. The E-SVR prediction model has better performance than S-SVR with an R-2 value of 0.92, an accuracy improvement of 8.2%, and a root mean square error (RMSE) reduction of 28.1%. Single factor analysis of different prediction models showed that with an increase in the length of the prediction cycle, the influences of the historical water supply temperature and flow rate on the heating load decrease gradually, and the influence of outdoor temperature increases gradually. (C) 2020 Elsevier Ltd. All rights reserved.