Optimal Temperature Control for 5th Generation DHC

while it's clear that low network temperatures are the key to efficient 5th Generation District Heating and Cooling networks, it's less clear at which exact temperatures a given network will achieve optimal efficiency. This is because the network temperature influences the network efficiency in several ways, including the network heat losses or gains, the efficiency of building heat pumps using the network as their heat source, and the feasibility of using low-temperature heat sources to feed into the network. And as such systems get even more complex when heating and cooling demands are met simultaneously and flow directions can reverse dynamically, both the potential and the difficulty for optimizations increases. Therefore, we need good tools to optimize the network operation.

One promising option for optimizing the network temperatures in such networks is offered in the paper Temperature control in 5th generation district heating and cooling networks: An MILP-based operation optimization by Marco Wirtz et al. at the Institute for Energy Efficient Buildings and Indoor Climate at RWTH Aachen University. At this point we would like to offer a short disclaimer: We're a bit biased here because heatbeat is a spin-off out of this institute and Peter, the CEO of our Aachen branch, also co-authored this paper. Furthermore, this paper is a result of our joint research project TransUrban.NRW in which we continue to collaborate. Nevertheless, we think this is a very interesting paper with much practical relevance for future 5GDHC networks, so we hope you agree with our choice of this paper for this newsletter issue (and for more on 5th Generation district heating we recommend our newsletter issues 3 und 9).

The paper presents a mathematical model to find the optimal network temperature for a 5GDHC network supplying 17 buildings with heating and cooling. The supply plant contains an air-sourced heat pump and a compression chiller, while electricity is generated on site through PV with support from the electric grid. The authors show that for this setup, the optimized network temperatures reduce the operation costs compared to both reference cases of constant network temperatures and of a network operation with free-floating temperatures. Furthermore the take-aways of this research include the finding that keeping the network temperatures low enough for direct cooling without use of chillers in the buildings has the largest potential for cost savings. If this is not possible, the optimal network temperature depends strongly on the heating-to-cooling demand ratio. At the same time, the authors conclude that heat losses (or gains) have a moderate influence on the optimal network temperatures.

Operation of the 5GDHC network

In order to put the research results into context, we think it's important to summarize the system setup and assumptions made in the presented paper. As a use case for the presented optimization methods, the authors assume a 5GDHC network supplying a research campus containing 17 buildings with heating and cooling (for the ongoing discussion about the terms 4GDH and 5GDHC see our newsletter issue 9). One important feature of this district is that it includes not only offices, laboratories, and a canteen, but also 2 data centers which account for a baseload cooling demand also in winter. Together with heat demands in summer for providing domestic hot water, the resulting demand structure has overlapping heating and cooling demands year-round, which is a generally favorable setup for a bi-directional 5GDHC network.

For the network temperatures, the authors assume a possible temperature range between 6 and 40 °C. Furthermore, the difference between the warm and cold lines in the network is assumed to be a constant 8 K. To make sure that the buildings can meet their heat demand with the assumed requirement of 60 °C in the building heating system, each building uses a heat pump integrated into its substation which draws low-temperature heat from the network and raises the output temperature. For the cooling, the authors consider both the case where a heat exchanger can provide direct cooling if the network temperatures are sufficiently low and the case where the building utilizes a chiller which uses the network as a heat sink. And as the demand structure includes simultaneous heating and cooling demands in the network, waste heat from the cooling of one building can be used as source heat for another building's heat pump, thus reducing the overall heat input to the network.

On the supply side, the authors assume that the network can be supplied with heat from a central air-sourced heat pump while for cooling, excess heat can be removed by a compression chiller. Both these generators are connected to a central thermal water storage. From there, the warm line of the bi-directional network is connected to the top of the storage and the cold line to the bottom. Thus, the storage can act as a passive buffer when the network switched between net heating and net cooling mode. As the thermal supplies both rely on electricity as input, the authors also consider the local PV generation and the connection to the electric grid in their optimization to arrive at a holistic view of the system efficiency. We think that this holistic perspective, together with modeling the thermal inertia of storages and network, is a particular strength of this paper. As a result, the optimization can consider effects like raising the network temperature by feeding in heat from the central heat pump when there the local PV generation offers cheap electricity and letting the network temperature decrease when heat generation is expensive.

Optimal network temperatures for 5GDHC

To demonstrate the network temperature optimization, the paper shows results for 3 representative months: January as a reference period in which the buildings mainly need heat from the network. July as a period in which the network mainly provides cooling to the buildings by removing excess heat. And March as an intermediate period in which heating and cooling is provided to the buildings in similar amounts and there is a high potential for balancing effects and synergies in the bi-directional network.

To evaluate the network temperature optimization, the resulting operating costs are compared to 2 benchmark cases. As the simplest mode of operation, the first benchmark is a network operating at constant network temperatures. For this case, the authors consider constant temperatures of 6 °C in the cold line and 14 °C in the warm line as well as the combinations 14 °C/22 °C and 22 °C/30 °C. And as a second benchmark, the paper considers a free-floating network temperature between the limits of 6 °C and 40 °C. This means that the central supply adds heat to the network only when the minimum temperature of 6 °C is reached and excess heat is removed to keep the network below the maximum temperature of 40 °C. In between, the network is allowed to passively let the network temperature change according to the heat drawn and fed back by the buildings.

For the optimization results, the authors can show that their optimization achieves improvements over both benchmarks for most reference months while it never performs worse than the benchmarks. In general, the results seem to suggest that the cost savings achieved by the optimization are greatest when the network provides not only heat, but also cooling to the buildings. For the scenario in which the buildings are assumed to have a heat exchanger for direct cooling, the optimization always chooses network temperatures below 14 °C so that direct cooling is possible and no additional electricity is needed to operate chillers in the buildings. If this is not possible, the optimization chooses low network temperatures in winter and high network temperatures in summer. This has the additional benefit that heat "losses" from the network to the surrounding soil are actually "gains" of cooling potential, thus reducing the need to actively remove excess heat in the central chiller.

From a high-level perspective, we think that this paper not only shows good ideas and promising methods for optimizing network temperatures in 5GDHC networks. The results also indicate that in general, such networks can benefit significantly from adjusting the network temperature during the course of the year. The optimal temperatures then seem to depend largely on the installed system components as well as on the demand structure of the buildings. While the paper shows a real-time-ready approach for optimizing the network temperatures in a model predictive control application, we think these and similar methods can also be used to determine feasible boundaries for network temperatures with varying set points over the course of a year also in early planning stages as crucial support for the design process. We look forward to further explore such ideas in our ongoing research project TransUrban.NRW, also collaborating with the authors of this paper.

Further information on optimal control of 5GDHC

For more information on the TransUrban.NRW research project we recommend to visit the official project website. Furthermore, we chose this paper as a very recent and relevant contribution to optimization of 5GDHC networks, but we also recommend other recent papers by Marco and his co-authors, e.g. on Quantifying Demand Balancing in Bidirectional Low Temperature Networks and A novel design approach based on mathematical optimization for 5GDHC.

The next issue of our newsletter will be released on January 5, 2022. Until then we already wish you happy holidays, a happy new year 2022, and thank you for your interest in our newsletter!