GMM clustering for heating load patterns in-depth identification and prediction model accuracy improvement of district heating system

Abstract

The identification of heating load patterns, also known as load profiles, is of vital importance to effective management and operation of district heating system (DHS). Clustering algorithms have been successfully applied in identifying heating load patterns. In this paper, we propose that the heating load patterns should be analyzed more specifically, and Gaussian Mixture Model (GMM) clustering is selected to extract sub-patterns. The novelty of this paper is that a new GMM clustering is applied to identify temperature related sub-pattern and people behavior related sub-pattern, and the clustering result is further utilized to improve the accuracy of prediction models. An energy station in Tianjin is used as case studies and four typical operation patterns are found with their characteristic of occur time and energy signature, which are defined as working pattern, on-duty pattern, daytime–nighttime pattern and nighttime–daytime pattern. The results reveal that this proposed method can make an in-depth identification of heating load patterns and also proves that the prediction accuracies of regression and artificial intelligence model are significantly improved by utilizing the GMM clustering results.