An improved association rule mining-based method for discovering abnormal operation patterns of HVAC systems. (February 2019)

• Record Type: Journal Article
• Authors: Zhang, Chaobo; Zhao, Yang; Zhang, Xuejun
• Is Part Of: Energy procedia. Volume 158(2019)
• Page Start: 2701
• Page End: 2706
• Publication Date: 2019-02
• Subjects: Association rule mining -- Data transformation -- Post mining -- HVAC systems; Power resources -- Congresses; Power resources -- Periodicals; Power resources; Conference proceedings; Periodicals; 333.7905
• DOI: 10.1016/j.egypro.2019.02.025
• Languages: English
• ISSNs: 1876-6102

Abstract

Abstract: Association rule mining has been widely employed to discover energy-saving knowledge from massive operation data of HVAC systems. However, it is still challenging to utilize the method in practice. One of the major barriers is that mined rules are always highly redundant and numerous. And, there is a lack of effective data transformation approaches specified for HVAC system to transform numerical data into categorical data adaptively. In this study, a kernel density estimation-based data transformation approach is proposed to transform data adaptively. Numerical data are grouped into several categories according to an estimated density function. Measurements in the same category is transformed into a unified text form. A rule clustering-based post mining approach is proposed to filter the mined rules. Only rules which are significantly different from others in the same cluster are extracted. Evaluations are made on one-year historical data obtained from chiller plants of the HVAC system in a commercial center in Shenzhen, China. Results show that the proposed data transformation approach is effective in outlier identification and data transformation. The proposed post mining approach filters out 43.9% of the mined rules successfully.