Fault detection and operation optimization in district heating substations based on data mining techniques

Abstract

The present generation of district heating (DH) technologies will have to be further developed into the 4th generation to fulfil the important role in future smart energy systems. At present, automatic meter reading systems have been installed in DH systems. These systems make hourly or even minutely meter readings available at low cost. However, the sheer quantity and complex of the data poses a challenge at various levels for traditional data analysis approaches. Data mining is a promising technology and is used to automatically extract valuable knowledge hidden in large amounts of data. To investigate the potential application of descriptive data mining techniques in DH systems, this study proposes a method based on descriptive data mining to improve the energy performance of DH substations. The proposed method consists of five steps: data cleaning, data transformation, cluster analysis, association analysis, and interpretation/evaluation. Data cleaning and transformation are implemented to improve data quality and transform data into forms that are appropriate for mining. Cluster analysis is performed to identify distinct operating patterns of substations. Based on each pattern, association analysis is then adopted to discover the unsuspected knowledge in the form of rules. Interpretation/evaluation is performed to select and interpret potentially useful rules. To demonstrate its applicability, the proposed method is used to analyze the datasets obtained from an automatic meter reading system at two substations in the DH system in Changchun, China. This application reveals that the method can effectively extract potentially useful knowledge and thereby provide essential guidance for the fault detection and operation optimization of DH substations.