Enter your username and password here in order to log in on the website
IEA-DHC Logo INTERNATIONAL ENERGY AGENCY TECHNOLOGY COLLABORATION PROGRAMME ON
District Heating and Cooling

Annex XIII Project 03

The Research / IEA DHC Annexes / 2020-2023 / Annex XIII / Annex XIII Project 03

Artificial Intelligence for Failure Detection and Forecasting of Heat Production and Heat demand in District Heating Networks (AI for Failure Detection and Heat Demand and Production Forecast).

Downloads

Register for free and log in in the upper right corner of the website to access the final report, an open data repository and two more presentations.

Videos

Project summary

AI is a rapidly growing field of research and application in many areas of daily life. It is probably one of the fastest growing businesses in the world. However, in heating applications the use of AI is very rare up to date. AI, especially “self-learning” approaches like Artificial Neural Networks (ANN) could contribute to a large extent to improve energy efficiency in district heating networks.

The big advantage is that such an approach allows forecasting without tedious modelling and simulation. The project will develop AI methods for forecasting heat demand and heat production and evaluate algorithms for detecting faults which can be used by interested stakeholders (operators, suppliers of DHC components and manufactures of control devices).

Regarding forecasting, Fraunhofer ISE has already developed an ANN-based control algorithm for single family house heating systems including renewable energy which showed energy savings of up to 12%  [9]. The ANN approach provides a “self-learning” capability allowing to create automatically black-box models just from measurement data. This low-effort/low-cost forecast approach provides the basis of a simple upgrade of current District Heating Network (DHN) control and is particularly important when fluctuating renewable energies are integrated into DHN enabling maximizing renewable heat yields, minimizing flow temperatures, optimizing demand side, minimizing faults and reducing cost.

Regarding fault detection, CEA LITEN has been investigating traditional methods based on modeling and simulation for several years. However, detecting faults in DH networks and components remains a challenge, and best performance can only be achieved when combining different approaches. While AI methods alone face the difficulty of a lack of accurate fault data for most systems, this project will investigate a relevant combination of simulation and experimental data for “self-learning” fault detection algorithms. The methods will be developed and evaluated based on data from a real District Heating Network at Stiftung Liebenau Meckenbeuren, Germany with multiple fossil and renewable sources (CHP, solarthermal, natural gas, oil, waste combustion) and a variety of heat sinks (residential, laundry, green house, workshop, medical station).

Some data from the experimental DHC network at CEA LITEN in Le-Bourget-du-Lac, France will be used for evaluating fault detection methods. AI allows for a combination of low cost implementation, significant reduction of CO2 emission and economic benefits. Its application is neither limited to any technology nor to any local conditions. Implementation can be done within a short period without changing infrastructure.

Thus, AI provides a powerful approach to an economically viable and fast pathway for DHN towards a transition to a carbon-neutral, sustainable energy system. The project team is managed by an experienced manager, all necessary scientific and technical skills including AI, computer science, DHN application and skilled craftsmanship are represented by qualified professionals. Additional scientific and economic expertise is provided by acknowledged experts. An effective communication plan will secure the transfer of the outcome of the project to the different stakeholders.

Different scientific and public dissemination channels will be used to explain how to apply the AI methods to real district heating network applications and show possible benefits in energy saving the stakeholders. Intense interaction between the project team and the international research community as well as final users are maintained throughout the whole project by being present on several international conferences, different IEA TCPs, technical papers, public webinars, newsletters, press releases and twitter. The project partners are willing to assist future product development of industrial partners.

Target groups

  • Suppliers of components and control systems for District Heating Networks
  • Energy Service Companies

Deliverables / Outcomes

  • AI algorithms to optimize energy efficiency and reduce faulty operation and set-up of DHN.
  • A report documenting the developed AI algorithm

Project lead

Fraunhofer Institute for Solar Energy Systems
Heidenhofstrasse 2
79110 Freiburg
Germany

Dr. Wolfgang Kramer,
Phone number: +49 172 821 58 94,
E-Mail: wolfgang.kramer@ise.fraunhofer.de

Project partners

  • CAE LITEN, France
  • Stiftung Liebenau, Germany