Leakages in the district heating networks is a current and growing problem. To find the leakages today many district heating companies uses manual techniques that are both time consuming and insecure, the methods can leave a lot of leakages in older pipes undiscovered for a very long time.
These undetected leakages costs the district heating companies a lot of money and can even be fatal. It is therefor of great importance that the leakages is found in time, thus the methods for leak detection needs to be improved.
The main purpose of this thesis was to investigate the ability to use thermal images to automatically search for leakages in district heating systems. To investigate this aerial thermal images from 2013 were collected. Image analysis was performed using ArcGIS and ENVI. This included, among other things, image preprocessing such as to define the projection of the images and unsupervised isodata classification to find potential leakages in the thermal images.
This automatic analysis resulted in many false alarms. One example were false alarms caused by vegetation, since vegetation absorbs heat during the day it appears warmer than the surroundings at night. To deal with this problem an unsupervised classification algorithm, isodata, was used again to classify the vegetational areas and the non-vegetational. This algorithm decreased the number of false alarms drastic and thereby increased the usability of the algorithm.
Other false alarm that has not been automatically rejected in this thesis were for example false alarms caused by heat leaking from buildings. One way to map such false alarms could be to analyse the shape and the linearisation of the potential leakages close to buildings. This would hugely increase the accuracy of the used algorithm.
The provided thermal images used in this thesis consisted of several confirmed leakages. All these confirmed leakages was found by the used algorithm. Although, the accuracy of the used algorithm could be discussed since many false alarms were generated. Nevertheless, to reject false alarms are much less time consuming than manual leak detection for an entire city.
Thereby the conclusion that an automatic leak detection in district heating networks is possible, furthermore a leak detection tool like this would be usable for the district heating companies. The evaluations from several different district heating experts who are using Digpro’s district heating application, dpHeating, today shows that a leak detection tool using thermal images would be a useful addition in dpHeating.
Source: KTH
Author: Ekroth, Natalie