Predictive control of indoor climate and energy consumption at FER

Could you briefly tell us what kind of a project you started within the Laboratory and when did that story begin?

In fact, it is a series of research projects on the topic of predictive control of heating, cooling and ventilation systems in a building. The planning of the research itself started in approximately 2008, but the first project started to be implemented in 2013. Within it, we equipped two floors of our skyscraper building at FER with devices that enabled us to centrally control the indoor climate in all rooms of these two floors. This was the ENHEMS-Buildings project (2013 - 2015) implemented in cooperation with the Croatian Meteorological and Hydrological Service, financed by the EU pre-accession funds.

One important legacy from that project is also the equipping of the main meteorological stations in the Republic of Croatia with instruments for measuring solar irradiance and the establishment of meteorological support for predicting solar irradiance, which is very important for predictive control in buildings. From that project until today, numerical forecasts have been generated for the location of FER's skyscraper, including solar irradiance forecasts. This helps us to experimentally investigate predictive control in the building, and every 6 hours we receive new forecasts for the next 72 hours. It was followed by the 3CON project (2014-17) of the Croatian Science Foundation related to the hierarchical management of different subsystems of the building, and then by the international project 3Smart (2017 - 2019). 
 

Mathematical algorithms are essential in planning and monitoring energy consumption

How will mathematical calculations generate energy savings?

If we are talking about a building, it is a dynamic system, which means that the temperature and other variables decisive for the building's indoor comfort is the result of actions on the building over a period of time in the immediate past. On the other hand, some conditions important for the building can be predicted, such as the outside temperature or solar irradiance or the occupancy of certain rooms. Therefore, by planning the operation of the building, for example, one day in advance, which is based on the mathematical model of the building, the predicted conditions and mathematical optimization, it is possible to use the computer to select the action that will consume the least energy. Designing this action practically requires a computer because in a typical office building decisions are made on how to control several hundred devices at once.

The essential difference compared to the current realizations of building automation systems is that the decision to act on the building's indoor climate system is based not only on current measurements but also on predictions.
 

It is important to emphasize that the building user's cost for the operation of the building's indoor climate system is actually directly optimized – it correlates a lot with energy consumption, but it is not identical to it. This is especially related to tariff models for different energy sources, which can be used to the full extent by using predictive control, and also to a new, slowly emerging need–demand response.
 

Predictive control means that energy from volatile renewable sources will be better used

As part of the project, you have taken several buildings where you are doing measurements, and one of the main demonstration buildings is that of the Faculty of Electrical Engineering and Computing. What have you discovered so far?

The 3Smart project (Smart Building – Smart Grid – Smart City) is an international project financed by the Interreg Danube programme, initiated and coordinated by FER. As part of that project, 7 buildings in the Danube Region were equipped with sensors and a data collection system that enabled us to create mathematical models of the buildings mentioned and investigate how their operation can be improved through predictive control.

At FER's building, a step further was made - it was possible to perform predictive control in real-time and to have a feedback effect through the existing automation system on the indoor climate in the building (370 fan coil units in 250 zones of the building, chiller, thermal station). Thus, we were able to verify that it is truly possible to control the building with predictive control - a doctoral thesis was prepared on this very topic and we demonstrated the continuous work of predictive control on one floor of the building over a one-year period.

Let's go back to the different buildings at 3Smart – two were in Croatia (the old administration building of Croatian electricity utility and the FER skyscraper), two were in Austria (the retirement & care centre and the elementary school in Strem), one in Slovenia (the elementary school in Idrija), one in Hungary (the administrative building of the electricity distribution company for eastern Hungary in Debrecen) and one in Bosnia and Herzegovina (the electricity utility company building in Tomislavgrad).

By carrying out the installation of the necessary sensors and performing data collection, we were able to process that data, using the procedures we developed, to arrive at mathematical models of the buildings themselves. This allowed us to conduct an analysis of how much better such a concrete building could function by applying predictive control under different conditions of external temperature and solar irradiance. I would like to briefly return to one of the previous answers - the key is how much money can be saved in operation compared to the classic automation system, which as a rule is designed to react only to the current situation in the building. By using predictive control, it is possible not only to use different tariffs for energy during the day but also to participate in demand response. As part of the 3Smart project, FER's laboratory which deals with energy grids was also engaged. In energy grids, the flexibility of consumption becomes very important due to the growing share of volatile renewable energy sources present (the security of power supply in the grid must be ensured by the synchronous operation of energy production and consumption, not only on the side of energy production as it was traditionally done). Therefore, using predictive control it was possible to test the way of operating with flexibility in energy consumption, which is separately rewarded and therefore reduces the cost of the building for energy use.
 

The results we obtained are very interesting and encouraging and were published in several scientific papers in highly-ranked journals (Applied Energy, IEEE Transactions on Energy Conversion) – we saw that in buildings with an inert heating/cooling system, the significant potential arises from the demand response because the primary heating/cooling device can be switched off almost completely for some short periods when flexibility in energy consumption is required. By operating predictive control on some outdated heating systems, such as one-pipe heating system, we could achieve very high thermal energy savings of as much as 30%. As far as the FER skyscraper building is concerned, with the demand response functionality included the savings for typical days are 10% in heating and 18% in cooling.
 

How realistic is the commercialization of predictive control in buildings?

Why do you think that the developed software will be commercialized? What concrete benefits will the users achieve? And how should energy be stored?

Many research centres in the world deal with systems for predictive control in buildings, and they are certainly also dealt with by large global corporations engaged in building automation. So far, the installation of such systems in buildings is non-serial and non-routine and requires a much higher level of expertise compared to classic building automation systems, where the algorithms are reduced to simple if-then rules that are mostly built in during the production of the automation equipment itself, so adjustments made in the field are very simple or none.

Professional solvers of convex optimization problems, which we rely on when solving the building optimization problem we set, are reliable and fast, but they are still complex (and expensive) algorithms. All of these are aggravating circumstances for the wide implementation of the technology of predictive control of systems in a building.

However, only with predictive control, it is possible to realize the demand response of the building's indoor climate control system in such a way that the comfort for users is fully preserved – this is not possible with reactive control and classic automation. With the tendency for high percentages of energy supply provided from intermittent renewable energy sources, where storage systems are also very important, demand response is a key instrument for maintaining the security of energy supply. As buildings make up a very significant part of the total energy consumption (previously 40%, now 30%), realizing demand response in them will be very important for the energy transition to clean energy sources, and this is where I see the necessity for predictive control systems in buildings.

In addition, we tried to build the entire predictive control system in a modular way, which also means that some of its parts (zone management, zone monitoring, various forecasts, management of the battery storage in the building) could be commercialized separately, as modules of the entire solution, which is gradually being introduced to the commercial side and one module pull the other over time.
 

The basic benefit for users is the reduced cost of running the building with retained comfort. The key is, of course, how much this cost is reduced because that tells whether it is worth investing in such a system or not.

Regarding the question of energy storage, the building is a dynamic system and can temporarily store part of the energy in the building structure itself, and part in different storage systems, either thermal or electrical. Exactly by using predictive control, it is possible to use these storage systems smartly or optimally to reduce costs.
 

The main advantages of predictive control

What basic consumer habits must change in the future in order to maximize the benefits of smart systems like this?

It is crucial that this system should be invisible to the user of the building and should not impair the comfort of using the building; therefore, consumer habits, at least as far as this system is concerned, should not change. However, there are also possibilities for its extension in advising users when it would be advisable to use energy on different devices in the building, closely related to tariff models and demand response and local production of energy from renewable sources. In this case, we can talk about behaviour change supported by predictive control to reduce costs.

Let's repeat what is the basic difference between a smart building as it currently exists on the market and a smart building as your laboratory is preparing?

The essential difference compared to the current building automation systems is that the decision to act on the building's indoor climate system is based not only on current measurements but also on predictions. At the same time, decisions are made through mathematical optimization directly minimizing the economic cost while preserving comfort, and it is not about simple if-then decisions that are implemented in typical building automation systems today.

How is it possible that in these days when the market is fighting hard for every educated individual, you have put together a team of about 40 employees in the Laboratory? What is their and your main motive in choosing this particular profession?

I would say that it is about the fact that challenging jobs at this level of complexity are very rare in the industry in Croatia, and the reason for this is clear – the results are very risky and uncertain and require long-term development. Senior researchers among these 40 or so employees, who operationally lead certain fronts of research in the Laboratory, like the challenging problems we solve and choose to stay in the Laboratory despite the fact that in the industry with their competencies the salary could be twice as high. Essential is also the motive to create something new and useful and to leave a lasting mark on it on the international level.

Of course, there is another side of the coin - if there are private companies that would be willing to engage in this development by investing for a longer period and attracting the best experts (we all know such an example in Croatia in the field of electric mobility) - surely the Laboratory could support them in this on both sides benefit. So far, we have used the possibility of cooperation with the Klimaoprema company through co-financing from European funds (PC-ATE Buildings project). It is excellent that this resulted in the development of a system for receiving and storing data from the building on the side of the Klimaoprema company, to which the advanced predictive management services that we developed through the project can be added. This has an extremely great potential in the energy transition ahead of us. It is certainly necessary to increase the number of experts who are familiar with the method of predictive control in buildings already through their academic education, and the Energy-efficient Buildings Control course as part of FER's new graduate study programme is a step in that direction.

The main motive is certainly solving technical challenges and pushing the boundaries, but it is essential that they are solved within the framework of scientific projects in order to ensure the financing of these activities and maintain the human potential that will lead to the commercial take-off of the developed solutions.