February 2017

True Analytics™ - Energy Savings, Comfort, and Operational Efficiency
Ecorithm - Cloud-Based Analytics Software

(Click Message to Learn More)

Main Benefits of Model Predictive Control

MPC is not a single strategy, but a class of control methods with the model of the process explicitly expressed in order to obtain a control signal by minimizing an objective function subject to some constraints.
Jiří Cigler
Feramat Cybernetics

Jan Siroky

Jan Široký
Research Team Leader
Energocentrum Plus

New Products
Distech Controls
Site Search
Reliable Controls
Past Issues
Control Solutions, Inc
Blue Ridge Technologies

Model predictive control (MPC) algorithms applied for a control of building technologies have gained a lot of attention in last years. Various research studies, as well as real installations, proved its potential for energy efficient operation leading to energy savings in the range up to 30%. The highest potential of this method is considered to be in i) systems with energy storages/accumulation (e.g. thermal storage tanks, TABS systems, PCM), ii) systems with multiple sources of energy and complex interconnections (typically containing renewables - microgrids), iii) systems in environment with dynamic energy prices or with strict power consumption limits, iv) demand response applications, etc.

This article will present a case study of the MPC applied for a room temperature control equipped with TABS emission systems and with power consumption limit that has to be fulfilled.

MPC Principle

MPC is a method for a constrained control which originated in the late seventies and early eighties in the process industries (e.g. oil refineries, chemical plants). MPC is not a single strategy, but a class of control methods with the model of the process explicitly expressed in order to obtain a control signal by minimizing an objective function subject to some constraints. In building control, one would aim at optimizing the energy use or cost subject to comfort constraints.

During each sampling interval, a finite horizon optimal control problem is formulated and solved over a finite future window called a prediction horizon. The result is a trajectory of inputs and states into the future satisfying the dynamics and constraints of the building while optimizing some given criteria. In terms of the building control, this means that at the current point in time, heating/cooling, etc. plan is formulated for the next several hours to days, based on predictions of the upcoming weather conditions.  Predictions of any other disturbances (e.g., internal gains), time-dependencies of the control costs (e.g., dynamic electricity prices), or of the constraints (e.g., thermal comfort range) can be readily included in the optimization.

The first step of the control plan is applied to the building, setting all the heating, cooling and ventilation elements, then the process moves one step forward, and the procedure is repeated at the next time instant.  This receding horizon approach is what introduces feedback into the system, since the new optimal control problem solved at the beginning of the next time interval will be a function of the new state at that point in time and hence of any disturbances that have acted on the building.

The MPC is a perfect mean for regulating energy consumption peaks because the controller foresees the predicted energy consumption and knows the power limits and thus it optimizes the consumption profile in order to satisfy the consumption limits. Therefore, the end customer saves not only on the costs for the commodity but also on the costs for reserved capacity.

In the control hierarchy of the building automation system, the MPC is typically on the top representing an energy optimization layer. The MPC sends demands (e.g. setpoints on temperatures in zones or accumulations) to a common BMS or to a building control system, which consists a Programmable Logic Controller (PLC).

Case Study

An application of the MPC to a university building (Figure 1) in Prague will be described. The Czech Technical University (CTU) building is composed of four five-floor blocks, three eight-floor blocks and four-level intermediary parts among the respective blocks. All the blocks have the same construction and way of use. This provides us with a unique opportunity to compare different control techniques under the same weather conditions since we can use different controllers in different blocks at the same time.

Figure 1: The CTU Building 

Figure 1: The CTU building

The CTU building uses a Crittall type ceiling radiant heating and cooling system (nowadays, the system would be called TABS – thermally activated building system). In this system, the heating (or cooling) beams are embedded into the concrete ceiling that enables the utilization of the thermal capacity of the building.

As mentioned, the MPC is also suitable for power peak reduction. This will be illustrated on the CTU building data. We will focus on a particular heat exchanger. There are three heating zones connected to one heat exchanger (HEX), and there also exists a power consumption limit for the HEX (i.e. a quarter-hour power maximum). A traditional control system is not able to optimally solve this problem; it can only contain some heuristic rules. The predictive control is able to have the consumption under the limit and fulfill the room temperature setpoints (i.e. thermal comfort).

The figure below (Figure 2) shows a situation when a change of the setpoint (denoted as ref) was coming after the Christmas setback when the room temperatures Ti,j in the building were kept at 19°C. The predictive controller plans to start the heat supply Qj two days before the change of the setpoint, and as can be seen from the graph, the consumption limit was always under the limit Qmax. The traditional control system would not meet the temperature setpoint or would break the consumption limit.

Notice also the progress of the supply water temperature shown in the middle figure, which shows very unusual behavior, but for this situation, these profiles were the optimal ones.

It is also important to set the power limit and the prediction horizon of MPC properly. To do so, we used the model that the MPC uses for the optimization, and we evaluated this scenario: a transition from the room temperature 19°C to 22°C, the outside temperature -10 °C, the heating for maximum power over the time of the prediction horizon. Based on this simulation, we obtained the most suitable prediction horizon equal to 36 hours.

Figure 2: Example Data

Figure 2: Example data from one HEX in the CTU building

On the studied building, the MPC is in operation since 2008 and controls 18 heating circuits with the TABS emission system. Implementation of the MPC was one of the last steps in modernization, which started as early as 1996 by successive façade insulation improvements, improvements in the HVAC system technologies with higher efficiencies that went hand in hand with the implementation of proper basic control strategies. The effort is shown in Figure 3 below, where the consumption of the whole building is depicted over the years.

Figure 3: History 

Figure 3: History of the energy consumption of the CTU building. There was a major modernization from 1995 to 2007 (mainly vapor heating system removal and building façade insulation). Energy savings in 2008 were achieved after application of MPC.


Optergy In this article, we have shown how the MPC can be applied for the regulation of slow reacting heating systems where there exists a quarter hour power limit. The traditional control strategies solve this problem hardly because they control either supply water temperatures or mass flows in the circuits but not the heating power.

Feramat Cybernetics and Energocentrum Plus have integrated the MPC control system into the product portfolio as the MPC maturity has already reached the production level. The CTU building heating system is operated by MPC from 2008. Significant energy savings were achieved thanks to MPC during this period (USD 300 000).

Parts of this article are based on the journal paper

About the Authors

Jiří Cigler is CEO of company Feramat Cybernetics ( He gained Ph.D. at the Department of Control Engineering of Czech Technical University in Prague with topic model predictive control applied for HVAC systems. During the Ph.D studies, he was also part of OptiControl team at ETH Zurich. Now, Jiří is fully concentrating on the application of the advanced control algorithms in practice through the Feramat company.

Jan Široký is leader of research team of Energocentrum Plus company. Jan has gained his Ph.D. at the University of West Bohemia, Department of Cybernetics.  During his studies he undertook an internship at ETH Zurich. The main activity of Energocentrum is operation and servicing of energy equipment, technological equipment of buildings and measurement and regulation. Energocentrum has its own research and development department. This team develops software products for measurement and control (e.g. cloud based SCADA system Mervis - ). Research team is focused on advanced control and monitoring algorithms for buildings.


The S4 Group
[Click Banner To Learn More]

[Home Page]  [The Automator]  [About]  [Subscribe ]  [Contact Us]


Want Ads

Our Sponsors