Leakage Localization in District Heating Networks Based on Real Network and Measurement Data

Abstract

Fast localization of leakages in district heating networks is a demanding yet important task for network operation. Every leakage results in a loss of medium thus of energy. In case of small leakages, the medium lost can be replenished to maintain network operation. However, if the loss of medium is too large, it may be inevitable to shut down network operation. This can be prevented by separating the damaged network part using exclusion areas. For this, the leakage must be assigned to the correct exclusion area as fast and accurately as possible. This is a major challenge for current district heating networks without existing or with leakage localization systems delivering results too slow, so that other methods of leakage detection and localization are needed here. In a joint research project, three model- and data-based approaches are developed to localize leakages in real time. These are based on operating data which are typically measured for operational needs such as pressure or flow rate. The first approach evaluates the pressure wave which traverses the network at speed of sound when a leakage occurs. The second, purely data-driven approach uses machine learning models to localize leakages. The training phase of this approach is based on a large number of simulated leakages using a network model. The last approach uses the same network model to numerically compute the network state and localize leakages depending on available measurement data. In previous publications these approaches are evaluated using simulated measurement data with encouraging results. In this paper the work is continued and carried out to measurement data of a real district heating network. It is shown that each method is individually applicable to localize leakages under real conditions and provides plausible results. This encourages further development for using synergies which may deliver even better results.

Keywords

• District Heating Network
• Leakage Localization
• Real Measurement Data
• Pressure Wave Evaluation
• Machine Learning
• Model-based Data Estimation