Description of the Building TSoS Use Case

The Local4Global methodology was applied to the Building Use Case as one application for the final product. The methodology was implemented and evaluated at the E.ON ERC main building of RWTH Aachen University in Aachen, Germany, which is a non-residential building and every-day-used as a university building. The building contains offices, conference rooms, CIP-pools, laboratories etc. that are equipped with different systems for controlling the room climate. The energy supply system is designed for a maximum of efficiency using several techniques for cooling, heating and power generation as well as to maximize the integration of renewable energy from a set of energy suppliers like geothermal fields, photovoltaics and a gas-fired CHP and condensing boilers. The entire energy system represents an efficient but complex TSoS that requires an optimal control strategy to exploit the energy saving potential to the full.


The Local4Global building Use Case consists of a down-scaled TSoS due to the evaluation reasons and project limitations. The building TSoS consisted of two different types of constituent systems: there were 6 offices and 7 conference rooms, different in use, equipment, and size. Local4Global was implemented in all rooms overtaking the control function that is currently performed by the Building Management System (BMS). All constituent systems are interconnected via the distribution systems and capable to shift loads from one distribution path to another and, thereby, from one energy source to another. For evaluating the impact of the control strategy, further measurements  were needed to be installed for observing and evaluating the TSoS highly in detail. Information on the thermal comfort and the compliance with all thresholds due to the room temperature and CO2 concentration as a reference for air quality, the operation of the constituent systems and their subsystems, the control behavior as well as all energy flows will be gathered as data base for evaluation and comparison. Therefore, different control strategies, for instance the current one as a base case and Local4Global as a test case, can operate in parallel using different but similar rooms at the same environmental conditions or in series if the particular room configuration is more important than external influences. Therefore, both a simulation environment and a real test bed was available for the experiments.


The optimization objective that is pre-defined for the Building Use Case is to maximize the integration of renewable energy, based only on local information. For each distribution path, the ratio of renewable energy contained in the particular energy flow is dynamically changing and constrained to power limits. This is the only global information that has to be available for the control system as the optimization problem has to be defined for optimal control approaches.


The second Use Case of Local4Global concerns the climate control of buildings equipped with their own renewable energy generation elements (such as e.g., photovoltaic arrays, wind turbines, geothermal energy, etc) along with a collection of automatic control elements for affecting the building thermal characteristics such as e.g., automatically control the HVAC system set points, window and blind openings, heat pumps and central radiator set points, etc. The problem at hand is a quite challenging problem where the control system attempts to exploit “as much as it can” the renewable energy so as to reduce the demand for non-renewable energy (coming from the grid) or during time-slots of low-cost tariffs, while maintaining user comfort (i.e. making sure that the building occupants are “satisfied” with the in-building temperature and other thermal conditions). One frequently adopted procedure for developing control systems for building climate control is to first model the building dynamics using one of the existing building modelling tools (e.g., TRNSYS, EnergyPlus, Modelica, etc) and then develop a model-based control using the model for the building dynamics. Unfortunately, such an approach suffers from several disadvantages:

  • First of all, developing and maintaining a reliable model for the building dynamics is “very expensive”: a tedious and time-consuming effort is required to develop, calibrate and validate such a model. Moreover, such a model should be continuously tuned so as to take care of changes in the building infrastructure. All these factors render the “luxury” of developing a reliable model for building dynamics prohibitive for most of the buildings (in some cases, the cost of developing and constantly tuning the building model will be more expensive than all the other operating costs of the building!).
  • Moreover, even if a very accurate and reliable model for the building is available such a model must be accompanied by an expensive sensor infrastructure in order for the model to be able to accurately predict the building dynamics performance: real-life experience has shown that the building models may produce quite inaccurate results or even completely fail in cases where they are not accompanied by a complete set of real-time sensor measurements: for instance, if there are no sensors for detecting whether a room is occupied or not or whether a window is open or not, the overall model may produce totally erroneous predictions.
  • Last but not least, even if the above two problems/shortcomings were not present, model-based approaches face the so-called “curse-of-dimensionality” problem: even if the perfect model and a complete sensor infrastructure are used, the current state-of-the-art in control system design is unable to compute the best (optimal) controller as this is a classical NP-complete problem.

Due to the above-mentioned shortcomings, the vast majority of control system implementations in real-life buildings concern buildings that were equipped with a complete and expensive sensor infrastructure and, moreover, a tedious and elaborate modelling procedure was employed before the development and implementation of the control system. Apparently, these control systems are not “transferrable” to everyday buildings where the requirement for deploying an expensive and complete sensor infrastructure and for developing an accurate and reliable model for the building dynamics, render the deployment of such control systems prohibitive.


Within Local4Global, we developed an efficient control system that is easy- and inexpensive-to-deploy in everyday buildings: no expensive infrastructure or modelling tools and effort will be required. The approach for doing so is the same as in the case of the Traffic TSoS Use Case: by treating the overall building system as a TSoS comprising constituent systems which based only on local information, learn, evolve and self-organize so as to optimize the global TSoS performance. To better appreciate the ambition and objective of the Local4Global system in the case of Building TSoS consider an everyday apartment building or a neighbourhood of buildings: in each of the apartments (or, buildings in the neighbourhood), the respective local control system takes decisions based only on local information, i.e., based on the temperature and humidity inside and outside (ambient) the apartment (resp. building) as well as on information regarding the apartment’s/building’s energy consumption cost. No information about what is happening in the neighbouring apartments or buildings is provided. Similarly, for energy influencing systems affecting all apartments (buildings) such as e.g., the heat pump or the central radiator system in the apartment buildings, the respective control mechanisms (e.g., thresholds for activating/deactivating the system) use only local information (e.g., demand per apartment and total energy consumption cost).

To test and evaluate the Local4Global system, the E.ON ERC Main Building, located in Aachen Germany, served as the Local4Global Building Use Case. The main reason for choosing this building to serve as the Local4Global Building Use Case is the fact that the building is equipped literally with all possible renewable energy generation elements and energy/thermal influencing elements that may found in a real-life building. Moreover, the overall building control infrastructure comprises a quite complex hierarchical system with different control elements affecting the building thermal and energy consuming performance at different levels and in a quite complex manner. Furthermore, it has to be emphasized the fact that the building contains a large number of rooms and offices with totally different characteristics and purposes (laboratories, office rooms, conference spaces, data centre rooms, etc). Finally, it has to be emphasized the fact that the building is subject to severe and abrupt weather as well as occupant behaviour changes.

Use Case Site: E.ON ERC building, Aachen, Germany

The building site is situated in Aachen, which is a city in the west of North Rhine-Westphalia, Germany. Figures 3 and 4 provide an impression of the building’s façade and floor plans. The façade is designed in a grey-black metallic coating with a unitized curtain walling. The all over appearance is simple and quite artless but representative and presentable.

The building is the new headquarter of the E.ON Energy Research Center at the RWTH Aachen University. The building particularly integrates the following usable areas: office rooms and staff facilities (e.g. computer rooms, in the following called CIP-Pools), seminar and conference rooms, laboratories, common areas as well as areas for LAN- and server equipment. Offices, conference rooms, common areas and CIP-Pools have demand for cooling, heating and ventilation dependent on the outside weather conditions. LAN- and server-equipment has permanent cooling demand. Laboratories have volatile and unpredictable demand for process heating and process cooling.

building2Heating loads occur during winter while during transition times and summer cooling loads has to be regarded. As it can be seen in the figure above, some CIP-Pools and conference rooms are placed in the middle of the building. Due to internal loads, cooling demand can also occur during winter, most widely independent from outside weather conditions. The design constraints for the energy concept are kept with thermal comfort following EN 13779, indoor air quality 2 (IDA 2), with a temperature spread between 20 and 26 °C, which must not be exceeded for more than 50 hours per year.

Constituent Systems:

The building control automation system is distributed to different types of controlled constituent systems (shown in the Distribution level of Figure 5):

  • A total of 64 active chilled beams provide fresh air to laboratories in the basement of the building. These units efficiently cool and heat the laboratories according to their individual demand. The implemented active chilled beams’ control strategy can be freely programmed according to flexible demands resulting out of the use of each laboratory. At the moment the ventilation air flow and the temperature are used as set points per lab to activate/deactivate the beams.
  • Two sorption supported AC units that humidify, dry, cool and heat fresh air in order to satisfy the time-depending demand. They are shifting heat to cold via an open sorption process and use an adaptive heat recuperation. Currently, a sophisticated rule-based control strategy is implemented in order to decide which operation mode appropriate, that uses twelve adjustable set points per unit.
  • A total of 18 rooms are air-conditioned by displacement ventilation systems that provide appropriate and controllable ventilation as well as control the inlet air temperature depending on the exhaust air temperature. The parameters of this control strategy as well as the control strategy itself are freely programmable.
  • A total of four separate concrete core activation zones that serves as a base load cooling and heating system in office rooms with a very high inertia. The control strategy of the CCA can be freely adjusted. At the moment a heating curve is implemented. Four each zone, 11 control parameters can be set.

A total of 86 facade ventilation units (one per small office, two per large office). Their purpose is to maintain a certain room temperature, that can be adjusted by the user or automatically controlled and to not exceed a certain threshold of VOC (volatile organic compounds) and of CO2. They are able to cool, heat and ventilate. Further the ventilation units are equipped with waste heat recuperation with adjustment possibility of the recuperation ratio. The sensor information they receive is room temperature, user presence, user room temperature feedback, CO2 and VOC ratios, humidity and outdoor temperature to mention the most important out of 40 control parameters.building3

Specific TSoS Challenges:

  • Use of local information to globally optimize TSoS performance (energy consumption cost subject to user comfort satisfaction). Each of the constituent systems receives only local information (e.g., each facade ventilation unit receives information about energy consumption and temperature, humidity, etc conditions only from the office the unit is located)
  • Constantly changing dynamics (due to weather conditions and occupants’ behaviour/patterns). It has to be emphasized the presence of severe and abrupt changes both in outdoor weather conditions and in the occupants behaviour (especially in meeting rooms). In many cases such abrupt changes cannot be predicted.
  • Combination of continuous (thermal dynamics), discrete (activation/deactivation of energy consuming systems, room/office occupancy) and highly stochastic (the human factor and the weather conditions) characteristics. Extremely highly complex and discontinuous dynamics.
  • Emergent behaviour. This is probably the most challenging problem the Local4Global system had to face in this Use Case: each of the constituent systems is attempting to optimize its own energy consumption/thermal requirements which, however, are affected by the actions of the neighbouring constituent systems (e.g., due to exchange of thermal energy between neighbouring offices). This, in turn, leads to the need for an emergent behaviour where some of the constituent systems operate even when, ostensibly, there is no need for them to operate (e.g., a facade ventilation unit should be operating in an office where there are no occupants and will not be occupants coming in the near future, so as to reduce the energy transferred/lost from its neighbouring offices through the walls). Another example is the case of variable tariffs in energy consumption for the Use Case: Apparently, the constituent systems attempt to use as much energy as possible during the time-slots that energy is cheaper (low-demand time-slots). This, in turn, may lead to an emergent behaviour where the demand – and thus the tariffs – increase significantly during these timeslots and renders the use of variable tariffs virtually useless.
  • Highly heterogeneity involving a very complex structure, topology, hierarchy and interplay between different types of constituent systems.
  • Complex hierarchy and topology as depicted in Figure 5, e.g., some of the constituent systems perform at the central level (concrete core activation), some of them at the office level (facade ventilation units).
adminEfficient Building TSoS Use Case