d reducing CO2 emissions from buildings is increasingly the focus of current research (United Nations Environment Programme 2021).

The Living Lab Energy Campus (LLEC) at Forschungszentrum Jülich pursues the goal of researching solutions for energy efficiency in buildings and presenting them to the public. Highly networked demonstrators show various research projects, for example on photovoltaics or energy storage. In addition, two projects are attempting to minimize energy consumption by integrating innovative technologies and approaches. The central key here is the collection and analysis of building data. Based on this, modular solutions are being designed that can be transferred to other buildings following successful testing. One focus of optimization in this regard is the development of intelligent control systems for controlling energy consumption systems in some buildings and the campus energy demonstrators. In order to counter the high heterogeneity of the buildings and parts of buildings, a large number of influencing variables must be considered. At the same time, user behaviour is being researched and used as an active means of minimizing consumption in another building. The LLEC enables the development and application of such solutions in a representative and realistic environment that allows research results to be validated. In administrative buildings in particular, influencing the intrinsic motivation of people can be essential, as it has been proven that they do not feel responsible for their influence on energy consumption as long as they do not bear the costs. Another aim of the project is to promote knowledge transfer and raise awareness of the issue of energy efficiency in buildings. The LLEC offers training courses, workshops and information events for experts, students and the interested public in order to raise awareness of energy-efficient construction and disseminate specialist knowledge.

ghting, thermal discomfort and potential user disturbances caused by high-frequency shading. The control system takes into account tuning parameters and comfort limits. It should be noted that the use of MPC currently requires customized modeling for each individual building, but simplifications are also being worked on in this project (e.g. easy-to-create and transferable Modelica modules). With regard to the influence of the weather, different effects such as outside temperatures / heating requirements, wind forces and solar gains were included in the control algorithms.

Independently of the MPC approaches, the researchers are working on the integration of software applications that make it possible to view energy consumption data and provide users with real-time recommendations for behavioral changes (Ubachukwu et al. 2023). The "Energy Dashboard" offers employees and visitors to the LLEC an overview of heating and electricity consumption in real time at room level, among other things. "JuControl" goes one level further and enables users to map their personal preferences for the indoor climate and intervene in the control system. A gamification approach is embedded in the application, which enables comparison with other people/teams. Feedback to users is provided via a traffic light system that indicates energy consumption. The traffic light is evaluated on the basis of defined limits, measured in kilowatt hours. In addition, the application reflected an "energy reference value" that simulates what the energy consumption of the entire LLEC campus would have been if everyone had had a corresponding usage profile (ideal value). In this sense, the project differs from conventional approaches to evaluating positive environmental impacts. While the savings are normally set in reference to the status quo consumption, the reference at the LLEC asks about the ideal state that could be achieved. This is because the potential for energy savings in individual rooms can vary greatly depending on the location, type of use and personal experience of the users. These influences are taken into account by defining an ideal. The system for user involvement is supplemented by a leaderboard (in the "JuPower" application), which recognizes teams with the lowest energy consumption. Incentives for energy-efficient user behavior are also created through social interaction and "game rewards".

In order to ensure the correct measurement of data, EnOcean sensors were installed in all participating buildings to measure air quality, CO2 concentration, temperature and humidity. The data was processed using M-Bus. KNX control elements and BACnet systems were used for the automatic control of lighting, heating and sun protection. Across the entire LLEC campus, studies were carried out on the integration of over 1800 devices in 18 buildings using an information and communication platform (EnOcean: 1676 sensors and 108 actuators, KNX: 48 sensors and 36 actuators) (Althaus et al. 2022).

is designed to make maximum use of solar heat gains and daylight in order to minimize energy consumption for heating and lighting. The measurements were able to confirm the quality of the simulations. More energy-efficient room temperature profiles were measured. However, due to a lack of meters in the individual rooms/parts of the building, it was not possible to carry out a precise comparison of the energy consumption. Quantitatively, the LLEC can confirm the high accuracy of the white-box MPC model. The Root Mean Square Error (RSME) between the measured and estimated room temperature was 0.49 °C. The efficient operation of the control system was also demonstrated. In comparison with a conventional control approach (0.47 kh/d), the thermal discomfort remained largely stable with a value of 0.53 kh/d.

In terms of user involvement, it should be noted that various measures (dashboards, JuControl, JuPower) have been used by more than 1,300 employees since 2020. To evaluate the effectiveness of user involvement, a regression model was set up based on the historical data before the measures were implemented, which describes the heating energy demand as a function of the outside temperature (fit quality: R2 = 0.91). This model was then also applied to the period with implemented measures for user involvement and compared with the actual energy consumption of the building. It was shown that, compared to the reference period, the heating demand increased less strongly depending on the outside temperature and the average temperature of the pilot building could be reduced by 1 °C on average. These effects contributed significantly to the measured reduction in consumption of 11.1 MWh per year (- 16.7 %), which those responsible attribute to the more efficient use of the heating system and more efficient user behavior due to the same system parameters (energy status of the building and the heating system). With a gas-fired power plant for heating (assumed emission factor of 200 g CO2e/kWh primary; primary energy factor natural gas = 1.1), this corresponds to a saving of approx. 2.44 t CO2e per year. Corrections were made for differences in the outside temperature between the two periods, but other influencing variables were not taken into account (e.g. wind, solar gains). The comfort level of the people was not negatively affected. The project managers point out the additional potential, as only a fifth of the offices in the building were activated for the measures to integrate users.

The negative environmental impacts of the installed components were not balanced for either the MPC or the user integration. With the (component) infrastructure found at the LLEC, the difficulty in assessing the negative impacts lies primarily in the allocation of the components, as many other processes and applications use the same components in addition to the digital applications described here. It is therefore difficult to allocate them precisely to individual digital applications. However, a large number of battery-operated components had to be installed for the individual room measurements (> 1800 components on the campus in total). As all of these components are constantly exchanging data with the digital applications or accessing them, a correspondingly high amount of data is required. For the user integration measures alone, 364 million data records were recorded and processed in around 3 years (approx. 74 GB in total or approx. 25 GB/a). The researchers assume that the amount of data could be reduced by approx. 25% with an even more practical implementation. On the one hand, because the number of data points could be reduced by approximately this amount. Secondly, because after successful testing in a research context, fewer redundancies in the measuring points and less data granularity would be required. Around 300 MB of data is processed daily on an OpenStack server for the MPC. Only a fraction of this is used to record measurement data from the rooms (approx. 5 MB per day; 15 minutes of data on room temperature, the status of window and door contacts and the number of people in the room). Taken together, user integration and the MPC require virtual resources for approx. 124 GB per year, which would mean estimated CO2 emissions of approx. 339 kg CO2e (see estimate of the negative effects of data transmission and processing here ).

Finally, the project sees various opportunities for future research projects. As MPC models promise high savings potential, but at the same time require a great deal of effort for individual calibration, there is a great need for the development of easily transferable MPC approaches for commercial purposes. The intended linking of the two applications of the project (MPC & user integration) is also aimed at simplified transfer to other buildings. This means that efficiency-enhancing customization options are still possible within the MPC cost function. As both aspects were developed on a modular basis, a future linkage is conceivable and makes sense, for example to test the effectiveness of the MPC under the influence of user-defined comfort settings or customized recommendations for action to the users. The researchers also cite the need for more comprehensive tests in larger buildings, including integer decision variables and the flow temperatures of heating water volume flows (central heating systems) as further areas of development.