Digital Twin Monitors and Controls District Heating Networks

Digital Twin and Modern Measurement Technology

The goal of the »DingFESt« project is to develop a high-resolution digital twin of complex district heating systems. This twin couples efficient dynamic flow simulation with robust consumption forecasting using artificial neural networks, model predictive control and modern measurement technology.

The sensors are installed and integrated according to the principle of minimality: How many and which sensors are minimally required and where exactly in the network topology in order to calibrate the digital twin in real time? The digital twin then proposes optimal control strategies that go well beyond the wealth of experience in previous control routines and consumption situations.

The aim is to enable utility companies to permanently ensure highly efficient network operation – even under increasingly complex and varying operating conditions – without jeopardizing their stability, resilience and security of supply. The use of the digital twin will then allow to decentralize producing infrastructures and significantly reduce operating costs as well as carbon dioxide emissions in the district heating supply. Disruptions in the heating network are thus detected at an early stage, such that maintenance cycles can be adjusted.

Key Aspect: Placement of Sensors

Most district heating networks are equipped with little to no sensor technology, so real measurement data from the network is rarely available to calibrate and validate a simulation – and to finally create a digital twin of the network. For this reason, an intelligent placement of sensors in the network topology is necessary, which

• provides a high information content for the identification of unknown model parameters for the simulation, e.g. heat transfer or roughness coefficients of the pipes,
  
• and requires as few measuring points as possible for economic reasons.

The sensor placement problem is based on sensitivities of thermal and hydraulic variables at network nodes with respect to unknown model parameters, which as derivatives can be easily determined by AD-Net Fernwärme. The sensor placement problem can be understood as an experimental design problem with measurements at one network node representing a single experiment. Thus, we can transform our problem into a continuous optimization problem, which as a convex problem then can be solved using well-known methods.

The developed procedure allows to take already existing measuring points at consumers or in the network topology into account and additionally yields a ranking of all proposed points according to their information content for the parameter identification.

When applied to the district heating network of TWL in Ludwigshafen, the algorithm provides comprehensible suggestions for the placement of measuring points in the network topology. The number of points is in an order of magnitude for which the installation of the corresponding sensors is feasible also from an economic point of view.

A comparison of the proposed measuring points with the occurrence of the various coefficients shows that the points are each located at the end of long section of pipes  in which one single coefficient occurs in isolation.

Our Project Partners

• GEF Ingenieur AG
• Technische Werke Ludwigshafen AG
   

Funding and Project Duration

The project is funded by the BMWK, is scheduled for three years and runs from 01.01.2021 to 31.12.2023.