Demand Estimation and Optimization of Domestic Hot Water Supply in District Heating Networks

For new as well as existing buildings, the provision of domestic hot water is increasingly taking on a more important role in the optimization of district heating networks. One reason for this is that in new buildings the relative share of domestic hot water of the total heat demand is increasing. This must be taken into account in the controls for the substations. On the other hand, hot water generation with storage systems is often responsible for high return temperatures, which should be avoided in the operation of district heating networks. In the first issue of our Research Newsletter in 2023, we would like to summarize the main findings of two articles that focus more strongly on the preparation of domestic hot water in district heating systems.

The article A methodology to estimate space heating and domestic hot water energy demand profile in residential buildings from low-resolution heat meter data by D. Leiria et al. presents a new methodology to disaggregate domestic hot water and space heating use from a single heat meter of the district heating supply. The unique feature of this new methodology is that no in-depth knowledge of the buildings, the occupants, or their behavior patterns need to be known. Different algorithms for disaggregation are applied and tested on 28 Danish apartments. For these apartments, the demands are available separately, so that this data set is very well suited to validate the methodology. It can be shown that the methodology used so far in Denmark to estimate the annual domestic hot water demand can be significantly improved with the new methodology.

In the article Staged control of domestic hot water storage tanks to support district heating efficiency A. Tahiri et al. propose two new control strategies for loading domestic hot water storage tanks in substations for district heating networks. The motivation is that inefficient and unoptimized storage tank loading often results in high return temperatures and thus inefficient district heating network operation. The first control strategy uses a smart meter signal, while the second does not require advanced metering technology, only an additional sensor in the storage tank. The second control strategy is first validated and optimized using a Modelica model, and then applied in a real building. In the end, the return temperature was reduced by an average of 7 K and the volume flow by as much as approx. 24 %.

Disaggregation of Smart Meter Data into Domestic Hot Water and Heating Demand Profiles.

The aim of D. Leiria et al. is to develop new disaggregation methods which only require hourly measurement data (e.g. from smart meters) and the outdoor temperature at the location as input data. Measurement data of outdoor temperatures are publicly available in Germany, e.g. from the DWD. For the development and validation of the methodology, detailed data from 28 homes in Denmark are available to the researchers. Outdoor temperatures at the site are also available through the meteorological service in Denmark. Each apartment is equipped with two heat meters, one measuring the heating-only demand and the other measuring the total heat demand. From this measurement data, the domestic hot water demand can be calculated and used for validation.

Four different approaches are tested to calculate the two types of demands from a measurement time series. The first approach uses the hypothesis that the heating demand changes only little over the day, because the changes in the outdoor temperature over a day are not very distinct. Thus, all peak demands can be assigned to domestic hot water demand. The 7 highest demands are identified as domestic hot water demands. For disaggregation, the heating demand is calculated using the remaining 17 data points for the day. Subtraction of these two values then yields the domestic hot water demand. The second method works similar, but assumes that the highest domestic hot water demand occurs at certain times of the day (e.g., morning, noon, evening). These times are identified as domestic hot water use and the same procedure is followed as in the first approach. Additionally, different days (e.g., weekdays and weekends) can be considered. The third method uses the difference in demand between winter and summer periods. It is estimated that the domestic hot water demand has little variation over the year and thus the domestic hot water demand in summer is transferred to the profile in winter. In the fourth approach, all three previously used approaches are combined to identify the times with domestic hot water demand.

The result of all previously presented methods are time series where points of time with primary domestic hot water demand have been identified. In the next step, these data points are replaced with "NA values" and the heating demand for the missing data is determined. Five different methods are tested for this purpose. These range from a simple linear interpolation to a moving average, to more complex machine learning methods such as a Kalman filter, support vector regressions and the combination of both approaches.

Application to the above described data shows that the best results are obtained when the 7 highest daily demands are identified as domestic hot water demands. In this case, about 80 % of the times with domestic hot water demand are correctly identified. The demand estimation of the heating demand is best achieved with the combination of the two machine learning methods. The method shows promising results. It could be shown that for the 28 apartments the error for the determination of the heating demand is in the range of +/- 10 % and that of the domestic hot water in the range of +/- 15 %.

Optimization of Domestic Hot Water Storage in District Heating Networks

A. Tahiri et al. develop in the paper a new optimized control strategy for the loading of domestic hot water storage tanks in district heating substations. Normally, a thermostat controls the domestic hot water flow through the heating coil of the storage tank, often with a proportional-integral controller, to maintain a set point temperature at the top of the sensor in the storage tank. The two new strategies are aimed at preventing overcharging of the storage tank, i.e., minimizing situations with excessive heat input from the primary side. The first strategy uses high-resolution measurement data from the building automation system for this purpose, especially the heat and domestic hot water demand. The second methodology requires only one additional temperature sensor in the domestic hot water storage tank compared to conventional controls. Both strategies are modeled, simulated and evaluated in Modelica. The second control strategy is applied and validated in a multi-family house in Denmark.

The first strategy uses detailed measurement of domestic hot water demand to ensure demand-based loading of the storage tank. Four different states have been identified for this purpose. Two different conditions are defined, one is a low temperature difference, where only the power necessary to raise the temperature from 50 °C to 55 °C is required, and the other is a condition where the temperature difference between the cold water temperature and the useful temperature is necessary. The transition between the states is described by a threshold value.

• periods without domestic hot water demand: During these times, only the energy consumption of the circulation losses must be covered.
• times with low domestic hot water demand: In addition to the circulation losses, the domestic hot water demand must now also be compensated. Since the demand is low, only the power with low temperature difference must be provided.
• times when the domestic hot water demand corresponds to the limit value: in contrast to the second condition, the power must be provided at a higher temperature difference. However, it still corresponds exactly to the energy for domestic hot water and circulation losses.
• times when there is a high domestic hot water demand: When there is a high demand for domestic hot water, the loading capacity is limited. In this way, the entire storage volume is utilized, peak outputs are reduced and an increase in return temperatures at high outputs is avoided.

The second strategy uses less measurement technology. The exact domestic hot water demand does not have to be known at all times. For this purpose, an additional sensor is installed in the upper third of the storage tank, according to which the loading (i.e. valve position) of the storage tank is controlled proportionally. The slope of the proportional controller is divided into four ranges to meet the different demand situations (see above). The lower the temperature of the sensor, the higher the power with which the storage is loaded.

In a simulation study, the two control strategies are tested and further optimized. Finally, the second strategy is implemented in an apartment building in Denmark. The results of a 45-day test phase, in which the conventional control was alternated with the new more innovative method, show that the new control strategy reduces the return temperature of the district heating networks by 7 K on average. The volume flow is reduced by about 24 %. Return temperatures below 30 °C could be achieved with the new methodology, and on days with a lot of domestic hot water demand (and thus a lower proportion of circulation), temperatures below 25 °C could be realized. A. Tahiri et al. emphasize the great added value of the simulation study, which simplifies both the implementation and the determination of the optimal parameters.

Further information

In both studies, high-resolution data from smart meters in the district heating sector are used. For the first paper, hourly measured values are used, which make it possible to determine the demand for demand simulations, for example. For the optimization and validation of control strategies in the second paper, even higher resolution time series are used. This shows that for a better understanding and optimization of district heating networks, higher temporal resolution measurement data is needed. We are very happy that our heatbeat Digital Twin will be used for exactly this purpose this year. To this end, we are working together with Stadtwerke Aachen, regioIT and RWTH Aachen University in the FunkSTA research project to digitalize all substations in the Aachen district heating network.

Both articles explore the complex topic of substation optimization and the necessary input data (demand profiles) in great detail, so we recommend reading both articles in full: https://doi.org/10.1016/j.energy.2022.125493 and https://doi.org/10.1016/j.energy.2022.125705 respectively.

The next issue of our newsletter will be published on February 1, 2023. Until then, feel free to follow us on LinkedIn where we share smaller application examples and information.