Skip to content

Controls

Detailed specification of aedifion.controls' control algorithms.

Reliability and robust control

edge-device-communication
Figure 1: Edge Device communication schematic

The energy systems of a building are essential and must function properly 24 hours a day, 365 days a year. They need to be able to function robustly and withstand communication issues and sensor outages by possessing reliable fallback strategies. Our Edge Devices are equipped with state-of-the-art algorithms able to detect all of these issues and resetting cloud-based control algorithms, handing over the control responsibilities to the locally implemented automation system. This way we can ensure that whatever happens, the building will continue to operate reliably.

Available algorithms

MaxPhi

Description

This function optimizes the energy-efficiency and therefore the cost-efficiency of energy distribution systems with distributed consumers and a central volume flow control actuator. Some examples are:

  • air handling units (AHU) with variable air volume (VAV)-Boxes
  • district heating systems
  • underfloor heating systems
  • borehole heat exchanger fields
  • etc.

The internal control function chart of each consumer in a standard system normally looks like the following figure. There is a volume flow \(Q\) which is influenced by disturbances \(z\) and should have the desired value \(Q_{soll}\). Therefore the difference between \(Q\) and \(Q_{soll}\) is calculated and is the input of the internal controller. The internal controller itself has a controller output to influence the medium volume flow, in this case, the angle \(\varphi_{soll}\) for the actuator:

closed-loop-function-chart
Figure 2: Function chart of a closed-loop circuit of the subordinate, internal consumer's controller

The volume flow from all the distributed consumers is generated centrally with a volume flow control actuator, for instance, a pump or ventilation unit. In standard systems, they are controlled by a constant pressure or volume flow setpoint. Depending on this given plant state the distributed controllers reduce this volume flow to the respective setpoint, which is needed to fulfil the desired state. Because the system's pressure- or volume flow setpoint is selected for a full-load operation mode, in a partial-load operation mode, where the distributed actuators are (partially) closed, the overall system is not working efficiently.

Here comes this higher-level control algorithm to place. The inputs are the continuously retrieved feedback signals of all the distributed actuators. With the maximum of those feedbacks, the setpoints of the superordinate volume flow control actuator are being calculated.

Below is a schematic function chart of a combination of two distributed closed-loop controllers as seen already above and a superordinate control of the overall medium volume flow, controlled via the speed of the volume flow control actuator. With \(\varphi_{01}\) provided as the maximum feedback angle of both closed-loop circuits, the change in rotation speed is calculated, which affects the volume flow of the two loops. Because of the combined system, the current angle of each loop affects also the volume flow of the other loop. This example extends with the number of distributed consumers but the principle is clear.

control-loop-two-consumers
Figure 3: Example of the control loop with two consumers

Example and real values

In this section, the implementation of the MaxPhi-Algorithm to a real test bench is described. At the E.ON ERC, an institute of the RWTH Aachen, there is a borehole heat exchanger field, where every probe has its own volume flow sensor and volume flow actuating valve and, therefore, can be used as a hydraulic test bench. Two redundant pumps provide the overall volume flow.

Within the scope of testing this algorithm, every probe was provided with its own volume flow setpoint. As you can see the volume flows of the probes are roughly constant, while the speed of the pump is decreased and the feedback values of the "consumer's" actuating angles increase until the maximum reaches the setpoint of 95%.

maxphi-algorithm-example
Figure 4: Example of the MaxPhi-Algorithm tested on a borehole heat exchanger field

Due to the opening of the actuating valves, the differential pressure of the distribution grid decreases. In the pump performance chart (figure 5) you can see the state before (red) and after (green) the activation of the MaxPhi-Algorithm. It is possible to operate the grid at 75% pump-speed instead of 90% while fully fulfilling the consumer's needs, meaning providing the needed volume flow. As you can see in the below diagram the positive outcome of the MaxPhi-algorithm is the saving of 1.5 kW energy consumption. Assuming a 50% operation time this is equivalent to roughly 6.6 MWh per year.

pump-performance-chart
Figure 5: Pump performance chart before and after activation of MaxPhi

Temperature Spread Regulator

Description

This application can be applied to water-bearing thermal distribution systems, such as heating and cooling circuits. In conventional distribution systems, the pump is operated statically, regardless of the actual user demand or the actual heat requirement of the heating circuit. The Temperature Spread Regulator .controls application closely monitors the inlet and outlet temperatures in the system in order to determine whether the system is running efficiently. Deriving the necessary intelligence from historical data, a determination is made about how large the temperature gradient within the thermally activated component should be in order to justify pump operation.

thermal-control-loop
Figure 6: A schematic of a typical water-bearing thermal control system, including the temperature sensors, a pump and a valve

Based on the derived thresholds, the operating time of pumps is reduced, increasing the efficiency and reducing the mechanical wear of the entire system.

With absolute minimum invasiveness and no negative impact on user comfort, more than two-thirds of the electrical energy required to operate the pump can be saved over the course of one year. The controller will be operated in the cloud and no retrofitting is required in the existing DDC.

The only actuator which is actively influenced is the pump itself. All other mentioned data points are used for monitoring purposes.

Example and real values

The application was demonstrated on a real concrete core used for room temperature control for heating as well as in cooling mode. In normal DDC operation, the pump of this unit is permanently switched on so that the concrete core is continuously flowed through, regardless of the actual heating or cooling requirements. After commissioning the .controls application, the following behavior was observed: It can be seen that the pump (light blue) only runs for a fraction of the total time, only operating for short intervals in order to flush the system. In this case, the operating time could be decreased by 75%. Larger energy demand is noticeable during the morning hours between 6 and 8 AM. After this point in time, the demand slowly decreases until the pump is ultimately switched off. In the diagram below, no direct influence of the periodic pump operation on the room temperature measurements can be seen, thereby, the user comfort has remained unimpaired.

TSR-Roomtemperature
Figure 7: The influence of the pump operation reduction on the room temperature

Weather Predictive Controller

The Weather Predictive Controller is an application that is used to control the operation of heating and cooling systems in a weather predictive manner, thus reducing their energy consumption and improving user comfort at the same time. Heating and cooling system controllers usually work by controlling the water temperatures in front of the thermally activated component in accordance with the outside temperature. This relationship is based on a pre-defined heat curve like the one given in the figure below.

thermal-control-loop
Figure 8: A heat curve showing the inlet temperature control of a concrete core during cooling operation

Heating and cooling systems are often systems with high thermal inertia. This means that it may take many hours before a change in water temperature has an effect on the room temperature. As a result, the room temperature that is actually desired is always reached a few hours too late and the desired temperature can never be achieved. As a result, the user experiences suboptimal room temperatures and thus an impaired thermal comfort. Frequently, further measures are taken in response, in order to achieve a more comfortable room temperature. These include, for example, the operation of an air conditioning system, electric heating or the simple opening of a window. In effect, inefficient room temperature conditioning systems are used to compensate for the delays in the actually desired and efficiently working control system, which leads to energy waste and, therefore, to increased costs.

The aedifion .controls application, the Weather Predictive Controller, solves this problem. By evaluating historical system data and combining it with current weather forecasts, the thermal control circuits of a building can be optimized to the extent that the desired room temperatures are available exactly when they are desired. This provides the occupants with increased thermal comfort and eliminates the need for additional temperature compensation measures, resulting in less energy waste and reduced cost.

The exact quantification of the cost reduction strongly depends on the system specifications (in general: the more thermal inertia is inherent to the system, the more potential for savings) and the user behaviour. However, investigations into this matter have shown that energy savings of up to 35% are certainly achievable.

Supervisory Model Predictive Control

Supervisory Podel Predictive Control (MPC) is considered a high-potential control approach for buildings and energy systems. It is based on a model of the controlled system that is used by an algorithm to determine the most favorable control decision for a given aim such as energy savings or thermal comfort improvement. However, the effort of providing a good model is slowing down the spread of MPC in the building sector. Our approach reduces the effort for model creation. We apply automatic system identification algorithms (e.g. SID) to the data of small subsystems. The reduced dimensionality of small subsystems increases the quality of the model fit.
Our implementations of distributed MPC algorithms (BExMoC and LC-DMPC) allow combining these small models into one application that yields an optimal result for the total system. We conducted the first tests in the research project NextGenBAT and aim for releasing the algorithm in Q2 2021.

Further Information

More control algorithms will be described here. If you wish to implement your own algorithm or want us to implement it, feel free to contact us.


Last update: 2020-11-27