A significant amount of the total energy consumption in industrialized countries is used for building operation. Heating, Ventilation and Air Conditioning (HVAC) systems are today controlled by simple PID or rule-based controllers. In the last several decades, substantial progress has been made in the areas of Model Predictive Control (MPC) and weather prediction. The idea in MPC is to repeatedly solve optimization problems online that are based on a model of the system in order to calculate control inputs that minimize some performance measure evaluated over a future horizon. With MPC, weather and occupancy forecasts can be taken into account to optimize the energy efficiency of an HVAC system.

The building dynamics are in general highly affected by disturbances, e.g. weather and occupancy. However, all buildings have a natural thermal storage capacity, which can be used to save energy. If the weather were to be known in advance, the building could be "prepared": one could for example let the solar radiation heat up the structure of the building if it is going to be cold during the next days or prevent the sunlight from entering the building if the weather is going to be hot.

To achieve these effects, we propose to combine weather and occupancy forecasts within an MPC framework. It is expected that such an approach will significantly increase the energy efficiency of HVAC systems while still providing for the comfort of occupants.

  

The goal of building climate control is to guarantee occupant comfort, i.e. to keep the room air temperature level within a defined range or comfort band (sometimes also the air quality, humidity, illumination, etc.). The range itself is defined by international standards (SIA 382/1, ISO EN 7730, ASHRAE 55- 2004), which also allow for this range to be violated from time to time by a specified amount. We seek to exploit this flexibility to decrease the energy demand by formulating so-called chance constraints in the MPC formulation, i.e. constraints that have to be fulfilled with a specified probability.

In the last decades, substantial progress has been made in the quality of weather forecasting, because of increased global data assimilation, improved numerical models and the availability of more computation power. Still, weather forecast data are uncertain and this uncertainty is expected to significantly impact the performance of the controller. Thus, in order to improve building climate control with MPC, we seek to incorporate forecast uncertainty into the controller. The weather forecast data are supplied by MeteoSwiss, whose expertise is very important for understanding and modeling these uncertainties. The modeling of occupancy uncertainty will be done by the Institute of Integrative Biology, ETH Zurich in collaboration with EMPA.

In this project there are two parts that have to be addressed. In the first part, it has to be understood how HVAC systems are affected by uncertainties, how the uncertainties look like and how they can be incorporated in the controller. The second part is the theoretical part, where the results of Robust MPC and Stochastic MPC are used and a new robust stochastic framework is developed.

Further literature:

A detailed list of literature about this topic can be found on the official project homepage.

_____________________________________________________________________________________________________

Efficient Ventilation: Control of underfloor ventilation systems

Maintaining a good indoor climate is one of the main tasks of modern HVAC installations. Indoor temperature and air quality e.g. CO₂ concentration have to be kept within given bounds while at the same time reducing the amount of energy used for conditioning and delivering fresh air. A variety of different ventilation concepts have been proposed in the literature and many of them are actually implemented in buildings.

One can distinguish between two principal ventilation concepts:

Mixing Ventilation:

Incoming air is delivered with a high momentum at ceiling level, causing a mixing of incoming fresh air with polluted air in the room. This concept is employed in most buildings in the US and in Europe.


Fig: Principles of mixing ventilation.

Displacement Ventilation:

In contrast to classical mixing ventilation, where incoming air is delivered with a high momentum at ceiling level, causing a mixing of incoming fresh air with polluted air in the room, in displacement ventilation fresh air is delivered directly into the occupied zone with a low-velocity. The supply air temperature is chosen slightly below the desired room temperature. This pre-conditioned air is ventilated into the occupied zone of a room at a low level, usually through a raised floor configuration using diffusers. By extracting air from the space at ceiling level, an overall floor-to-ceiling air flow pattern is produced. The upward movement of air in the room takes advantage of the natural buoyancy of heat. As air gets heated it rises into the region above the occupied zone and carries contaminants upward. The result is a higher concentration of contaminants in the warm stratified air near the ceiling and a lower zone of only marginal polluted air below. Displacement ventilation systems aim to minimize mixing of supply air with room air, instead maintaining conditions in the occupied zone as close as possible to that of the conditioned supply air.

A related ventilation concept is the so-called underfloor ventilation where the air is delivered with a slightly higher momentum at floor level. This results in a smoother transition between warm air at the ceiling and cold air at floor level and leads to a higher perceived comfort for the occupants.


Fig: Principles of displacement ventilation.

Modeling and Experiments:

Modeling of displacement ventilation in order to describe the behavior of the system as well for controller design and testing is the core challenge in this project. Models of different complexity are considered, ranging from a quasi-steady state description of the stratification to a full dynamical model based on partial differential equations. Verification of the model structure and parameter identification requires data gained from actual experiments. These experiments are done in collaboration with the BP Institute for Multiphase Flows, University of Cambridge, England as well as with Siemens Building Technologies, Zug, Switzerland which allows the use of latest IT technology and knowhow.

For a first analysis, small scale, so called salt bath modeling techniques have been performed at the BP Institute, University of Cambridge, England.


Fig: Salt-bath modeling of a displacement ventilation system.

Full scale test are performed in an HVAC Laboratory in Zug in collaboration with Siemens Building Technology. For this purpose, a test room is equipped with diffusors for underfloor ventilation and a variable air volume (VAV) ventilation system. The effect of single and multiple heat loads is investigated and data is extracted for the purpose of verification and parameter identification.


Fig: Test room at Siemens Building Technology in Zug.

Sensor and actuator positioning:

As seen in small scale experiments and discussed already in other publications (see [2]), the position of sensor and actuator hardware in the room plays a fundamental role in the design of ventilation systems and can influence the performance of the system significantly. One aim of this research is to utilize the freedom in the positioning of sensors and actuators given by modern wireless technologies. A holistic approach incorporating these decisions already in the controller design phase is aimed at.

Estimation / Detection:

In underfloor and displacement ventilation systems it is important to know the mode the system is operating in, i.e. if the system is in "displacement mode", that is with stratification of air or, maybe due to disturbances or an increased heat load, in "mixed ventilation mode". A dynamic model as well as a description of all possible steady state configurations is the basis for the design of an estimation scheme that will be developed in this part of the project. A repeated detection of a false mode in a room can indicated a bad topology, a failure of an actuator or a disadvantageous tuning.

Controller design:

The main control objective in ventilation control is to maintain a certain indoor air quality while at the same time keeping the energy efficiency high. Indoor air quality is defined in terms of specifications on desirable and tolerable bands on air temperature, concentration of pollutants like CO2 and air velocity. The controller needs to maintain displacement ventilation despite of disturbances and to adjust e.g. ventilation rates if necessary in a VAV underfloor ventilation scenario. Disturbances are e.g. varying heat and contaminant loads, changes in the topology (like an open door or window) or varying outdoor conditions.

Acknowledgment:

This project is funded by NCCR-MICS, a center funded by the Swiss National Science Found