Scientist Corner

How can we efficiently coordinate the growing network of prosumers who produce and consume energy? As future energy networks become more complex, they require a well-equipped toolbox. The algorithms discussed here are just two possible approaches, leaving significant room for further research to expand this set of tools. We also have a methodology that guides us in selecting the most appropriate tool for different applications. It does not matter if the tool is the best possible in theory… finding feasible solutions to keep decarbonizing our energy networks is what matters.

By Costanza Saletti, University of Parma

Published in Hot Cool, edition no. 7/2024 | ISSN 0904 9681 |

The decarbonization of heat is relevant to the energy transition.

One of the most urgent challenges of our times is mitigating climate change through reducing CO2 emissions from all human activities. To this end, the energy sector is transitioning to a fossil-free and more efficient system. While in the past, most efforts were dedicated to decarbonizing electricity, we have recently realized that we should also focus on decarbonizing heat.

District heating is a very efficient infrastructure for heat distribution in populated urban environments. However, most modern networks are still managed in an outdated way. Very often, the users connected to the network present a high demand for heat in the same period of the day, typically in the morning. This becomes a very high peak load that district heating providers must satisfy by operating the energy plants within the district heating supply area. These peak loads are typically met by expensive peak boilers fueled by natural gas or even oil.

A promising concept to reduce the high peak loads without changing the network infrastructure is called Demand-Side Management (DSM). It consists of using the connected buildings as heat storage devices, exploiting, for example, the heat capacity of the indoor air or the building envelope. It is possible to store heat within the buildings when cheaper or more convenient and then avoid supplying heat later while the building is cooling down. This technique, therefore, changes the profile of the building’s heat demand, enabling a lower cost or a more profitable operation.

The rise of prosumer-based energy networks

While the concept of DSM seems highly interesting in theory, there are still many challenges to its practical implementation. A critical point regards the complexity of coordinating all connected users in large district heating networks. This will become even more complex if we look at the future energy system, where heat is integrated with other energy forms such as electricity or sustainable fuels.

Imagine a district heating network where one or more connected customers have a photovoltaic plant installed on the roof. Or a solar thermal plant, a thermal storage, a heat pump, or even a combination of these and other units. Users of such kind are called “prosumers” because not only do they consume energy, but they also produce it with their own plants and are often not synchronized with their own consumption. With the rise of renewable energy generation and energy communities, prosumers will become more and more common in future heating networks.

So how can we coordinate all prosumers (and their plants) in a large system to achieve greater good? Is it possible to use DSM in large networks to reduce peak load and avoid using expensive boilers?

A novel set of tools to deal with this complexity

In this article, I introduce a set of tools that can help us deal with complexity in prosumer-based energy networks. These tools are based on optimization, a mathematical procedure that searches for the best possible solution to a problem according to a certain criterion. For example, we can look for the management strategy of a system that returns the minimum operating cost or the minimum emissions of CO2 over a period.

With these tools, the prosumers connected to a district heating network must collaborate. This means that they can implement DSM to shift their demand profile and reach the best performance of the overall system. The two algorithms that I present here differ in the way to solve this collaboration problem:

• The first tool is named centralized optimization. It considers the network as a whole, including prosumers and energy plants in the district heating supply area. It finds how to operate all buildings and all plants simultaneously to obtain the minimum system cost for the next two days. This is the optimal solution to the problem.
• The second tool is distributed optimization, leveraging continuous communication between different network parts. A distinct problem is solved for each individual prosumer, that searches for its own optimal management. At the same time, a supervisory agent asks the prosumers to modify their demand profile and decides to keep these changes only if they give a lower global cost. The problem is stopped after a certain number of modifications have been tested.

But why do we need two different strategies to pursue the same goal?

The case of my university campus to evaluate these new management strategies

To evaluate the performance of these two solutions, I applied them in a case study. I selected a portion of the University of Parma (Italy) campus, where I work, and assumed to add some energy plants to increase the complexity. In the case study, Figure 1 shows a district heating network supplying twelve prosumers and a central heating plant. This contains a biomass boiler (B1) for the base load and two natural gas boilers (B2 and B3) for the peak load.

The tools described before are compared with a baseline without coordination (meaning the prosumers only think of their own good). Figure 2 compares how the central station is controlled and how the required heat is supplied. It is clear that centralized optimization reduces the peak loads by more than half compared to the baseline. What is even more interesting is that it also completely avoids the use of expensive natural gas boilers.

However, a new issue arises. Since the problem is solved all at a time, a growing number of prosumers increases exponentially the time necessary to calculate the solution. On the other hand, distributed optimization returns a very similar solution: a slightly worse peak reduction (45%) but with a computational time that is less dependent on the number of users.

Since “an imperfect solution is better than no solution at all”, in very large cases, we can accept the use of distribution optimization if it solves the problem fastly. That’s why it is essential to have different tools able to operate in different conditions.

In conclusion…

Coordinating prosumers in large energy networks is feasible and helps us in the energy transition. The critical thing to remember is that complex future energy networks will require a well-equipped toolbox. Since those presented here are just two potential algorithms to achieve this goal, there is still room for much research to enhance the toolbox. We also have a methodology that guides us in selecting the most appropriate tool for different applications. It does not matter if the tool is the best possible in theory… What matters is to get feasible solutions in practice to keep decarbonizing our energy networks.

Figure 1. The case study is freely inspired by the Campus of the University of Parma, Italy. Twelve buildings with different distributed energy production units are included. The central heating plant comprises one biomass boiler and two peak boilers fueled by natural gas.

Figure 2. Management of the central power station in the baseline and with centralized optimization. The latter completely avoids using natural gas boilers (NG) and reduces operating costs.