Identification heat user behavior for improving the accuracy of heating load prediction model based on wireless on-off control system. (15th May 2020)

• Record Type: Journal Article
• Authors: Yuan, Jianjuan; Zhou, Zhihua; Tang, Huajie; Wang, Chendong; Lu, Shilei; Han, Zhao; Zhang, Ji; Sheng, Ying
• Is Part Of: Energy. Volume 199(2020)
• Publication Date: 2020-05-15
• Subjects: Wireless on-off control system -- Heating load prediction -- Open valve number -- Heat user behavior -- K-means algorithms; Power resources -- Periodicals; Power (Mechanics) -- Periodicals; Energy consumption -- Periodicals; 333.7905
• DOI: 10.1016/j.energy.2020.117454
• Languages: English
• ISSNs: 0360-5442

Abstract

Abstract: In-depth knowledge of the customers and a better understanding of their heat use is the basis of accurate heating load prediction model, it is also the cornerstone for realizing the intelligent and multi-energy development of the heating system. Most of existing heating load prediction models were established mainly based on outdoor meteorological parameter, while few of them took indoor temperature and user behavior into account. This paper took the wireless on-off control system as an example, and the heat user behavior was analyzed by the open valve number. Different heat users have different heat behaviors in different time periods, which are reflected in the change of open valve number. The number of open valve was first time introduced into the heating load prediction model (Model-2) in this paper. The results showed that the prediction accuracy of Model-2 was significantly improved compared with the traditional model (Model-1) including multiple linear regression (MLR) and support vector machine (SVM). K-means algorithm was used to identify and cluster the heat user behavior. Comparing the heating load prediction accuracy of MLR and SVM before and after clustering (Model-2 and Model-3), the results showed that Model-3 had a much higher accuracy. The rank of the models was Model-3, Model-2 and Model-1 in turn, in terms of the performance coefficients for model training and testing. The Model-3 proposed in this paper provides a reliable heating load prediction method for wireless on-off control system. Highlights: A new heating load prediction model is introduced. The heat user behavior is introduced into the prediction model. K-means clustering algorithm is used for identifying the heat user behavior. The prediction accuracy of new model is significantly improved after clustering.