Automated Statistical Methods for Fault Detection in District Heating Customer Installations

Abstract

In order to develop more sustainable district heating systems, the district heating sector is currently trying to increase the energy efficiency of these systems. One way of doing so is to identify customer installations in the systems that have poor cooling performance. This study aimed to develop an algorithm that was able to detect the poorly performing installations automatically using meter readings from the installations. The algorithm was developed using statistical methods and was tested on a data set consisting of data from 3000 installations located in a district heating system in Sweden. As many as 1273 installations were identified by the algorithm as having poor cooling performance. This clearly shows that it is of major interest to the district heating companies to identify the installations with poor cooling performance rapidly and automatically, in order to rectify them as soon as possible.