District Heating Meets Smart Heat Meter Data: Can We Diagnose Buildings and Substations?

The rapid deployment of smart heat meters in Denmark and EU has catalysed significant digitalisation in district heating (DH). While smart heat meter (SHM) data is increasingly used for operational optimisation and fault detection, challenges persist in translating fault identification into actionable diagnostics. Key barriers include limited training data, lack of standardised metadata, and low-resolution measurements. This paper addresses these gaps by proposing a three-tiered data collection framework to enhance fault detection and diagnosis (FDD) in domestic DH substations: (1) standard hourly SHM data; (2) high-resolution data (4-second intervals, 1 W precision) via optical reading; and (3) supplementary sensor data on indoor climate and secondary-side of heating system. This approach aims to support the development of robust, data-driven diagnostic methods and accelerate digitalisation and energy optimisation in DH systems.

Keywords: fault detection and diagnosis, smart heat meter data, district heating, substation, labelled data

The widespread rollout of smart meters in Denmark over the past decade has driven numerous digitalisation initiatives using heat meter data. Following the 2021–22 energy crisis, 70% of Danish buildings are now connected to district heating (DH), with more expected as oil boilers are phased out [1]. This trend is expanding globally. The IEA’s Net Zero by 2050 report estimates DH will supply over 20% of global space heating demand—and up to 50% in the EU—by 2050. With EU regulations requiring all DH-connected customers to have smart meters by 2027, the potential for energy optimisation using heat data is significant [2].

Denmark and other Nordic countries lead in DH digitalisation, using data beyond billing purposes [3]. A 2022 Danish District Heat Association survey found that utilities increasingly use hourly heat meter data to optimise operations, lower costs, manage assets, and forecast demand. Digital service providers analyse smart heat meter (SHM) data to pinpoint poorly performing installations, prompting outreach to affected customers to inspect and improve their systems by lowering return temperatures.

Researchers primarily in Denmark, Sweden and China have explored fault detection in domestic heating systems using SHM data [3–7]. Techniques include outlier detection, probabilistic models, clustering, and machine learning. Analyses focus on consumption time series as well as derived hourly averages of supply/return temperatures and flow. Studies reveal a high share of DH substations with inefficient performance, with 43–74% showing faults [5].

Despite progress in fault detection using smart meter data, a major challenge remains in moving from fault identification to diagnosis. A key issue is the lack of standardised metadata and labelled, high-quality training data linking heat consumption patterns to specific faults. Another limitation is the low metering resolution—hourly readings with a minimum of 1 kWh, which hinders fault detection for small consumers with usage below this threshold. Although meters can record down to 1 Wh, only 1 kWh values are transmitted due to cost, battery life, and billing-focused use. AAU BUILD has developed a method to infer decimal values from recorded energy data [6], but it has yet to be adopted by utilities or companies.

Danish utilities have shown strong interest in using SHM data beyond billing. Integrating robust, data-driven diagnosis methods into commercial products—and providing anonymised training datasets for utility implementation—would significantly support ongoing digitalisation and energy optimization efforts.

How to facilitate the fault detection and diagnosis in domestic DH substations?

This paper advances the state of the art by addressing two key challenges in fault detection and diagnosis (FDD): (1) insufficient model training data and (2) the need for measurements beyond standard heat meter data. To tackle these, three levels of data collection are proposed:

·         Level 1 – Standard SHM Data: Hourly, 1 kWh-rounded data typically used for billing by DH utilities.

·         Level 2 – High-Granularity Data: Using an optical eye on SHMs, data is captured every 4 seconds with 1 W precision, offering much finer resolution.

·         Level 3 – Supplementary Sensor Data: Includes sensors on the secondary side (post heat exchanger) and indoor climate sensors, capturing variables such as indoor temperature, thermostat settings, and supply/return water temperatures, enabling deeper system insights for advanced FDD.

Figure 1. Data levels (1-3) with graphical indication of they belonging in the building.

Initiative 1 - Standardized digital fault inspection reporting.

Additionally, to the time series from SHM, a key component in proceeding with the FDD in domestic DH substations is the labels of the faults. The Danish D4Heat project [8] addresses this gap by developing a standardised digital survey for fault inspections. It is currently distributed to five district heating utilities participating in the project. Here the objective is to establish a consistent (standardized) method for collecting information on faults in building heating systems connected to district heating networks.

To facilitate easy access, a QR code was provided that links directly to the digital questionnaire. All submitted responses are stored on the AAU OneDrive server. The survey is intended to be completed by utility technicians conducting fault inspections in customer buildings. It gathers the following information:

·         Customer ID – in order to connect report with time series from SHM

·         Smart heat meter ID – in order to connect report with time series form SHM

·         Building address – in order to collect building characteristics gathered in the Danish building register (BBR) database [9]

·         Customer’s name and surname

·         Date of inspection – to be able to identify the timestamp of the fault start

The survey also notes whether the customer was contacted remotely (e.g., by phone) or through an on-site visit. Technicians are asked to provide key details about the building’s heating system, including:

·         Type of DH connection: direct or indirect (via heat exchanger)

·         Space heating system: radiators, floor heating, or both

·         DHW production type: heat exchanger or storage tank

·         Presence of a ventilation unit with a DH-connected heating coil

After entering this information, technicians select up to three observed faults each for the space heating and DHW systems. The full list of faults is shown in Table 1.

Table 1. Fault types and names in space heating and domestic hot water system

To ensure consistency and support technicians during the inspection process, the survey is supplemented with simple graphics that illustrate where each fault is located within the building and heating system, see Figure 2 for faults in space heating system.  These visuals help clarify the fault categories and improve the accuracy of fault reporting. Additionally, technicians are provided with a protocol that includes detailed descriptions of each fault type along with their typical symptoms. This support material is intended to enhance consistency and understanding during inspections. All efforts are aimed at collecting high-quality, well-labelled fault data to support further analysis and reliable FDD model development.

Figure 2. Graphical indication of the faults in space heating system

Initiative 2 – imposed faults, high resolution and high precision data logging.

This activity departs from the earlier work reported in [7, 10]. It focused on the classification and analysis of the most common faults registered by DH utility technicians. The analysis of reports led to the distribution of labelled faults. The objective in Initiative 2 is to impose in a controlled manner the most common faults while allowing for high resolution (4 seconds) and precision (down to 1 W) data logging by using optical eye device that is mounted on smart heat meter, see Figure 3. This activity is conducted in 10 single family buildings. There can be imposed only one fault at the time per building followed by filling in the protocol as indicated in the previous section of this article. Moreover, more than one fault can be imposed per building but in consecutive manner. Faults related to domestic hot water will be imposed during spring and summer, while those for space heating will be imposed during the heating season. This activity concentrates on faults originating from incorrect system settings.

Figure 3.  Optical eye (left), smart heat meter (right).

Table 2. Table with faults to be imposed in space heating and domestic hot water with indication of expected symptoms to be validated.

Initiative 3 – Smart heat meter data and additional sensors at the secondary side

Equipping DH substations with three additional temperature sensors on the supply and return pipes for space heating and the return pipe from the DHW unit (heat exchanger or water tank), as shown in Figure 4 adds costs for equipment and installation. This setup is more relevant for larger, more complex buildings requiring detailed fault detection. With these extra readings, SHM data can differentiate between space heating and hot water consumption, enabling precise fault identification for each category and prompt detection year-round.

Figure 4. Level 3 monitoring with heat meter data and additional sensors on the secondary side

A significant portion of the faults identified in [7, 10] are due to user behaviour, which cannot be fixed by replacing components or adjusting settings. Therefore, the activity under Initiative 3 focuses on combining heat meter data with indoor climate sensors to encourage more efficient district heating use. Two office buildings in Denmark will be equipped with IEQ sensors and serve as showcases for the D4Heat project [8].

Conclusions

We are on the verge of improving fault identification in building heating systems. The DH industry recognizes the potential of SHM data, but further development is needed before DH utilities can advise customers on installation faults. A 1°C reduction in return temperature can save utilities 0.5–2 million DKK/year, highlighting the significant market potential for fault detection and diagnostics. Users can avoid penalties for high return temperatures and secure better heat prices. With 50-60% of installations faulty or misadjusted, achieving 20% of the savings potential could generate 200 million DKK/year in Denmark, out of an estimated total potential of 1 billion DKK/year. In the EU, the market potential is 100 times larger.

References

[1]     https://klimaraadet.dk/sites/default/files/imorted-file/fra_gas_til_groen_varme.pdf.

[2]     European Parliament, Directive (EU) 2023/1791 on energy efficiency and amending Regulation (EU) 2023/955 (recast), Article 16, subsection 2, Official Journal of the European Union (2023). URL: https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=OJ%3AJOL_2023_231_R_0001&qid=1695186598766.

[3]     Lygnerud K., Nyberg T., Nilsson A., Fabre A., Stabat P., Duchayne C., Gavan V., A study on how efficient measures for secondary district heating system performance can be encouraged by motivational tariffs, (2023) Energy, Sustainability and Society, 13 (1), art. no. 38, DOI: 10.1186/s13705-023-00417-0.

[4]     Leiria D., Johra H., Anoruo J., Praulins I., Piscitelli M.S., Capozzoli A., Marszal-Pomianowska A., Pomianowski M.Z., Is it returning too hot? Time series segmentation and feature clustering of end-user substation faults in district heating systems, (2025) Applied Energy, 381, art. no. 125122, DOI: 10.1016/j.apenergy.2024.125122.

[5]     Månsson S., Johansson Kallioniemi P.-O., Thern M., Van Oevelen T., Sernhed K., Faults in district heating customer installations and ways to approach them: Experiences from Swedish utilities, (2019) Energy, 180, pp. 163 – 174, DOI: 10.1016/j.energy.2019.04.220.

[6]     Sun W., Chen D., Peng W., Anomaly detection analysis for district heating apartments, (2018) Journal of Applied Science and Engineering, 21 (1), pp. 33 – 44, DOI: 10.6180/jase.201803_21(1).0005.

[5]     Henrik Gadd, and Sven Werner, Applied Energy 157 (2015): 51-59 / Sara Månsson, et al., Energies 12.1 (2018).

[6]     Schaffer M., Leiria D., Vera-Valdés J.E., Marszal-Pomianowska A., Increasing the accuracy of low-resolution commercial smart heat meter data analysing its error, (2023) Proceedings of the European Conference on Computing in Construction, DOI: 10.35490/EC3.2023.208.

[7]     Leiria D, et al.: Towards automated fault detection and diagnosis in district heating customers: generation and analysis of a labelled dataset with ground truth. In: 18th International Proceedings of IBPSA conference, pp. 3620-3628, Shanghai, China (2023).

[8]     https://eudp.dk/en/node/17117.

[9]     https://bbr.dk/forside.

[10]   Marszal-Pomianowska A., Leiria D., Johra H., Pomianowski M., Praulins I., Anoruo J.C.A., Fault Detection in District Heating Substations: Overview of Real-Life Faults in Residential Heating Installations, (2025), Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) , 15271 LNCS, pp. 357 – 364, DOI: 10.1007/978-3-031-74738-0_23.