AAU logo

Reports

Here is a list of reports made by students at Aalborg University since 2010 (those related to the refrigeration lab)

Absorption Cycle Heat Pump Control, June 2014

Students: Bjarni Geir Pétursson, Christian Køcks Lykkegaard
Supervisors: Kristen M. Nielsen, Tom S. Pedersen, Kasper Vinther

Abstract: Green Energy is a widespread and highly topical subject within nearly all fields of engineering. Sønderborg CPH and Sønderborg District Heating are not exceptions and operate as pioneers in Denmark within the area of district heating by means of absorption cycle heat pumps.While a field trip to Sønderborg ascertains major issues of the internal operation of the heat pump, a large network of various stakeholders works alongside to develop a simulation tool, FlexIQ, which intends to provide economic benefits for the owners and solve given issues in its current mode of operation. The specific setup in Sønderborg is thoroughly analyzed and finds merely two valves available for true control, which is control of the district water, whereas internal controllability of the refrigerant is left for simulation purposes solely. Sufficient models of the heat pump are developed to provide fundamental necessities for the FlexIQ tool. The models establish the basics for a model predictive control scheme taking in evaporator district water mass flow as disturbance which indeed is the true disturbance of the heat pump in Sønderborg. A developed SISO control topology allows heavy advantages in form of disturbance rejection, prediction features of low pressure side set-points, optimality and compliance of the true system constraints. The controller is easily extended to include disturbance prediction. A MIMO controller provides an optimal coefficient of performance (COP) while taking in absorber and condenser district water mass flows as control signals. An improvement is found to be 1.52% compared to a performance function with no COP punishment.


Superheating Control, June 2014

Students: Edouard Merle, Christos Sansaridis
Supervisors: Seyed Ehsan Shafiei, Palle Andersen

Abstract: This master project was aiming at controlling the superheat, that is a phenomenon that happens in the evaporators of refrigeration systems, designing a controller for the Expansion Valve that lets the refrigerant going through the evaporator. A mathematical model of the evaporator was derived to make some further simulations and obtain temperature changes while modifying certain parameters. This model based on moving boundaries representation, gave some positive results in simulations but could be improved to take all the exterior interactions into account. The superheat controller was therefore simple and generic but was then modified further to meet some optimality and robust requirements that should be ensured when dealing with public technology. Both some simulations and tests could have been run as Aalborg University owns a refrigeration laboratory. This results could then have been compared to the simulations done with our model-based control.


Aggregation of supermarkets in a smart grid, June 2013

Students: Rasmus Pedersen, John Schwensen
Supervisors: Benjamin Biegel, Jakob Stoustrup

Abstract: We consider control strategies for aggregation of a portfolio of supermarkets with the purpose of acting in the electricity balancing market. Supermarkets are able to shift the power consumption in time by sub-cooling the contained foodstuff. It is shown how the flexibility of an individual supermarket can be modeled for an interface consisting of an on/off activation signal; initially based on a simulation model but later verified through tests on a full scale system. Several control strategies for managing the portfolio to follow a power reference are presented and compared: MPC with boolean variables, a relaxed version and a PI controller with a dispatch algorithm. The predictive controller has a high performance but is computationally impractical while the dispatch/PI strategy has lower performance but requires little computational effort and scales well with the number of supermarkets, while the relaxed solution falls somewhere in between. Large scale simulations are conducted based on high-fidelity supermarket models of which 400 are aggregated together. The large-scale simulation shows how a portfolio of 400 supermarkets can be used for upward regulation of 900 kW for a two hour period. The possible services of an aggregator is investigated with suggestion for specific regulation products. Additional, it is shown that; it by optimizing power consumption according to market prices can save an average of 2.34% on the cost of power. Different approaches to improve on these results have been made, mainly to enhance the flexibility by changing the display case controllers from air temperature to food. Without sensors available for this measurement a novel estimation method, based on actual experiments, is presented to overcome this. Using the food temperature showed an increased actuation of compressor and thus increased the range of response of the portfolio. Finally, an alternative interface was designed and tested, using and exchanging far more knowledge about the supermarket. This was based on creating a flexibility model of the display cases which a VPP used to generate a power reference that a local supervisor controller would follow. The performance of this interface was demonstrated through simulations.


Adaptive control methods for superheat control in refrigeration systems, June 2013

Students: Martin Østergaard Jørgensen, Brian Bach Nielsen, Henrik Steffensen, Anders Overgaard, André Sekonda
Supervisors: Tom Søndergaard Pedersen, Kirsten Mølgaard Nielsen, Roozbeh Izadi-Zamanabadi

Abstract: This project has been made in collaboration with Danfoss and deals with investigation of two adaptive controller methods for implementation on a superheated refrigeration system. The system uses a classical vapour-compression process and consist of a evaporator, valve, compressor and condenser. In this project a model only knowing the evaporator and valve is used. An analysis of the system is made and a moving boundary model from opening degree on the valve to the superheat in the evaporator is derived. Tests reveal, that the model is of too low order to be able to embrace the non-linearities in the system. Parameters are found for different operating points and these models are used for further work. Using the knowledge gained from the analysis, adaptive controllers are developed and implemented using the two methods, online Maximum Extremum Seeking for tuning of a PI controller and Simple Adaptive Control. The performance of these controllers are tested against a normal PI controller, which is tuned for a specific operating point. The tests reveals, that both methods are capable of optimizing the disturbance rejection. The SAC control scheme showed the best disturbance rejection compared to the PI with static parameters and the MES-optimized PI controller.


Energy optimal control of combined compressor-pump system for regulating brine temperature in a chiller system, June 2013

Students: Dragos Adrian Urian
Supervisors: Henrik Rasmussen, Kasper Vinther, Roozbeh Izadi-Zamanabadi

Abstract: In this project the design and implementation of a Model Based Controller (MBC) and an Extremum Seeking Controller (ESC) for a chiller system is described. This project is motivated by the need of obtaining an energy optimal controller of combined compressor-pump system for regulating brine temperature in a chiller system. The project is proposed by Danfoss, thus a close connection between the control strategies and the sensors used in commercial chillers systems has been maintained. A model of the chiller systems is derived such that the brine temperature is dependent on the frequency of the compressor and the brine mass flow. Because the dynamics in the vapor compression cycle are much faster than the dynamics in the brine circulation, the model of the vapor compression cycle is considered static. With the model computed, in order to observe the behavior of the system, a simulation benchmark has been developed. After creating the model and the simulation benchmark, a Model Based Controller and an Extremum Seeking Controller were designed and implemented. The purpose of the two controller was to drive the input signals to values that maximize the efficiency. The first proposed control strategy is Model Based Control, a mathematical method that uses the plant model in order to design the control strategy. The technique relies on some simplifying assumptions in order to derive a mathematically tractable representation. Because of this, system dynamics or effects could have been overlooked. The second proposal investigated in this project is Extremum Seeking Control, a model free optimizing controller. The method uses a sinusoidal perturbation that finds the input signals which meet the load requirements and minimize the power consumption. It has been shown that Model Based Controller and Extremum Seeking Controller can be effectively implemented on a chiller system. Extremum Seeking Controller has the ability to drive the inputs to the optimal values without requiring the model of the system and can be considered as an edge advantage above the Model Based Controller.


ILC based cost reduction strategies applied on supermarket refrigeration systems, February 2013

Students: Dragos Adrian Urian
Supervisors: Roozbeh Izadi-Zamanabadi, Torben Green, Henrik Rasmussen, Kasper Vinther

Abstract: The design of a supermarket refrigeration system that is able to adapt to the energy prices is presented in this paper. A supermarket refrigeration system with multiple display cases that share a common compressor rack is considered. The design and implementation of an Iterative Learning Control (ILC) for a supermarket refrigeration system is used. The motivation of the project is to decrease the energy cost while maintaining the temperature constraints and help balancing the smart grid energy source. Throughout the project the display cases share a common evaporation temperature. The cooling capacity is chosen with respect to the most loaded display case in order to deliver the minimum cooling capacity and achieve the maximum efficiency. A benchmark for hybrid system control of a supermarket refrigeration system is used as a model. The model is adapted with respect to the required specifications. Further the objective is to develop a control strategy that is able to adapt to the energy prices and reduce the total cost of operation. Real time energy prices and power measurements have been used for the spectral analysis that shows a periodicity of approximately 24 hours. In the project the available control strategies for a real refrigerating system were analyzed and adapted to fulll the requirements. After the controllers of the system were implemented, a frequency ILC was designed. The purpose of the ILC is to cool the foodstuff when the energy prices are low and use the thermal mass stored in the foodstuff to reduce the cooling when the energy prices reach peak values. The simulation results show that there is a total energy savings of around 10% in the display cases after only 6 ILC iterations. It can be concluded that the Frequency ILC can be successfully implemented on a supermarket refrigeration system. This project was supported by Advanced Engineering-RA-DP, Danfoss A/S, DK-6430 Nordborg, Denmark.


Model Predictive Control of a Refrigeration System in a Smart Grid, June 2012

Students: Chiara Corazzol, Rasmus Pedersen, John Schwensen, Senthuran Sivabalan
Supervisors: Jakob Stoustrup, Seyed Ehsan Shafiei, Kasper Vinther

Abstract: This project deals with implementation of a refrigeration system into a smart grid. The objective is to investigate the flexibility potential of such systems, where flexibility is seen as their ability to store and release energy. Two different methods of implementation are examined. One based on an indirect control strategy, where the refrigeration controller receives predictions of future power prices and based on this runs the system in an optimal manner. The other method is based on a direct control strategy, where the refrigeration system receives predictions of future power consumption references and in return, feeds back a generic flexibility model to a virtual power plant (VPP). The VPP then calculates the power reference for the refrigeration system, based on knowledge of several connected nodes. The refrigeration system applies the desired power consumption, given it is feasible. With the communication based on a priori insight, a control scheme being capable of handling this is utilized in form of model predictive control (MPC). In order to employ MPC, two different models of the AAU test refrigeration system with distributed controllers are derived. One based on references to a local evaporation pressure controller used as input. The other based on compressor speed as input. During verification of the models it was discovered that there are certain discrepancies between measurements, which influenced the quality of the models. In the indirect setup both models were tested with the MPC. Both tests and simulations indicates that a refrigeration system is capable of offering flexibility, by lowering display case temperature when power price is low and later lower power consumption when price increases. A critical discovery is the inability to control the evaporation pressure with the distributed controller and therefore this model was dismissed for the direct control method. In the direct setup the system is fitted to a generic "bucket" model, well suited for the VPP. To close the loop a simple VPP is designed, based on the idea of minimizing power plant production. When testing the small grid setup, two nodes are connected to the VPP, one being the actual system the other a virtual one. The test substantiates the assumption of a refrigeration system being implemented as a flexible consumer in a smart grid.


Iterative Learning Control Applied on a Refrigeration System, June 2012

Students: Thomas Hansen, Claus Thy Henningsen, Kevin Sanchez Sancho, Ayub Tashkilot, Dragos Adrian Urian
Supervisors: Henrik Rasmussen, Kasper Vinther

Abstract: In this project the design and implementation of a Iterative Learning Control(ILC) for a refrigeration system is described. This project is motivated by the need of using more cooling capacity even if the refrigeration system can not physically provide it. The system needs to predict when the disturbance might appear based on the previous iterations and cool in advance. Thus the system should be able to detect a pattern in the load changes created by customers. The control objective is to maintain the temperature inside a display case within a certain temperature range, even though a large and repetitive disturbance is added to the system. Throughout the project a small scale refrigeration system from Danfoss A/S was provided as an experimental platform. First the model is derived and its parameters are estimated through measurements on the experimental platform. Because the model has a nonlinear behavior the parameters are linearized in a operating point and all the experimental measurements are done within a range of that point. In the project, several controller design methods have been investigated. The final implemented controller is a PI controller with a Modified Smith Predictor. After creating the model and designing the controller for the system, a Frequency ILC was designed and implemented on the experimental setup. The purpose of adding an ILC is to increase the performance of the system, when exposed to a large load disturbance. An experiment was designed to test the ILC. The experiment was done on three controller types, one where no feed-forward was added to the system, one where a manual designed feed-forward was added to the system and one where the ILC computed the feed-forward. The experimental results shows that there is a maximum error improvement of 36% with the manual feed-forward and even a better improvement of 43% with the ILC only after 5 iterations compared to the system with no feed-forward added. It can be concluded that the Iterative Learning Control scheme can be effectively implemented into a refrigeration system.


Generic Superheat Control of Evaporators using One Sensor and One Actuator, June 2011

Students: Casper Hillerup Lyhne, Erik Baasch Sørensen
Supervisor: Henrik Rasmussen, Kasper Vinther, Roozbeh Izadi-Zamanabadi

Abstract: This master thesis deals with generic superheat control of an air conditioning system using one sensor and one actuator. The system is analyzed and a model of the evaporator is derived that extends a conventional evaporator model. The extended model improves the understanding of superheat phenomenon in evaporators, which are unevenly filled. A model based superheat controller, which utilizes all parameter knowledge, is constructed and tested on the test setup. The controller showed positive results, and showed the feasibility of the control structure. The superheat controller was simplified to a generic controller that only measures the output temperature of the evaporator, and does not need knowledge of the system parameters beforehand. The controller was tested on two different test systems and showed positive results on both systems. Feed forward was also included in the controller to help improve the performance of the controller with regard to disturbance rejection.


Modeling of Microchannel Refrigeration System, June 2010

Students: Casper Hillerup Lyhne, Erik Baasch Sørensen, Lars Jespersen, Marc Westring Krogen
Supervisors: Henrik Rasmussen, Lars Finn Sloth Larsen

Abstract: The project deals with the modeling of an air conditioning system with a micro channel condenser. The system was analyzed and a low order model of the system was derived. The parameters of the model were fitted to data from experiments on the physical system. Furthermore a thermal camera was used to estimate the length of the superheat zone in the evaporator and the condenser. The model was validated with measurements from the system. It showed, that the model for the valve is not sufficiently accurate and may have to be refined. Measurements with the thermal camera did however show, that during a cycle of the valve the length of the superheat varied approximately 15%. The validation showed, that the model of the pressure matches the measured data, and the outlet temperatures to a lesser degree. It was judged, that the model was usable as a basis in the design of a controller. The goals for a controller were analyzed. The controller should minimize the length of the superheat zone. An idea to achieve an even filling of the tubes was to use an unknown input observer to detect if overflow occurs. The observer was implemented to detect a valve error to test the principle.