With regard to climate change, it is imperative to reduce greenhouse gas emissions. One solution approach is to increase energy efficiency in buildings. Buildings contribute a high share of the total global energy usage. As the rate of new building constructions is low, measures applicable to existing buildings become paramount. Before applying new approaches on a large scale, it is necessary to evaluate them in a representative, realistic environment. Living labs such as the Living Lab Energy Campus (LLEC) at Forschungszentrum Jülich (FZJ) facilitate innovative monitoring and control approaches in a real-world setting. In this work, we investigate the required steps for bringing sensor and control networks, comprising more than 1800 devices, into 18 existing and new buildings. This enables both room-level monitoring and control, as well as the integration of building-wide automation. By introducing an ICT infrastructure, we pave the way towards holistic approaches on a district level. We describe the workflows used for selected instrumentation variants and show first insights from the operation of the resulting infrastructure. We show that the investigated instrumentation variants exhibit similar characteristics; however, they affect control behavior differently. We emphasize that instrumentation should be planned in the context of existing infrastructure. Moreover, we present and evaluate sample measurements obtained from different buildings.

Model predictive control is a promising approach to robustly control complex energy systems, such as hybrid battery-hydrogen energy storage systems that enable seasonal storage of renewable energies. However, deriving a mathematical model of the energy system suitable for model predictive control is difficult due to the unique characteristics of each energy system component. This work introduces mixed integer linear programming models to describe the nonlinear multidimensional operational behavior of components using piecewise linear functions. Furthermore, this paper develops a new approach for deriving a strategy for seasonal storage of renewable energies using cost factors in the objective function of the optimization problem while considering degradation effects. An experimentally validated simulation model of the PHOEBUS Energy System is utilized to compare the performance of two model predictive controllers with a hysteresis band controller such as utilized for the real-world system. Furthermore, the sensitivity of the model predictive controller to the prediction horizon length and the temporal resolution is investigated. The prediction horizon was found to have the highest impact on the performance of the model predictive controller. The best-performing model predictive controller with a 14-day prediction horizon and perfect foresight increased the total energy stored at the end of the year by 18.9% while decreasing the degradation of the electrolyzer and the fuel cell.

Building control architectures are strongly limited by the systematic lack of measurements at user-relevant locations. This paper proposes a digital twin architecture grounded in Correlated Gaussian Processes (Corr-GP) that provide information in the form of pseudo-measurements. Tested with thermal and CO ₂ measurements collected from the field, close-to-person pseudo-measurements are provided based on the continuous input of remotely located measurement signals. In particular, detailed short-term as well as long-term results are provided for both temperature and CO ₂ digital twins. We show that the proposed approach is trainable on only a few days of measurements. This property makes the proposed approach especially useful in field applications, where alternative algorithms, such as, for example, neural network architectures, are not capable of dealing with small amounts of data. We demonstrate how to adjust the proposed approach to provide temperature and CO ₂ digital twins for the generation of pseudo-measurements. Within the given framework, we show how to utilize the proposed digital twin to couple multiple reference sensors to provide close-to-person pseudo-measurements. By extending the Corr-GP approach to a non-zero prior mean formulation, we show how to reduce the included information by the reference sensors. More precisely, the extended approach can be defined as a digital twin with only a single reference sensor. This enables a reliable long-term application by avoiding the need for retraining caused by changing seasonalities within the signal characteristics. That is, we show that the digital twin trained in summer can be operated in winter.