Abstract
Refrigeration is used in a wide range of applications, e.g., for storage of food at low temperatures to prolong shelf life and in air conditioning for occupancy comfort. The main focus of this thesis is control of supermarket refrigeration systems. This market is very competitive and it is important to keep the variable costs at a minimum and, if possible, offer products which have higher robustness, performance, and functionality than similar products from competitors. However, the multitude of different system configurations, system complexity, component wear, and changing operating conditions make optimal
tuning of controllers a difficult and time consuming task. These are also some of the challenges which make advanced model-based control difficult, and a model-based controller will often be tailored to a specific system. The focus in this thesis is therefore instead on development of data-driven control strategies with a higher plug and play potential.
One of the main control challenges in refrigeration systems is proper control of superheat for efficient and safe operation of the system. This task can be performed by an electronic expansion valve and requires two sensors, which traditionally are a pressure and a temperature sensor. In this thesis, a novel maximum slope-seeking (MSS) control method is developed. This has resulted in a control implementation, which successfully has been able to control the evaporator superheat in four widely different refrigeration system test facilities without using a pressure sensor. A single-sensor solution is thus provided, which either reduces the variable costs or increases the robustness of the system
by not relying on pressure measurements. MSS is an example of data-driven control and can be applied to a broad class of nonlinear control problems. The method utilizes the qualitative nonlinearity in the system and harmonic analysis of a perturbation signal to reach an unknown, but suitable, operating point.
Another important control task in refrigeration systems is to maintain the temperature of the refrigerated space or foodstuff within the desired/legislative requirement, e.g., to prevent possible deterioration of the foodstuff. Refrigeration systems are often dimensioned to be able to cope with the highest possible loads and the hottest temperatures during the year, while also taking into account extreme weather conditions. Overdimensioning is expensive, both in terms of the variable cost, but also due to possibly higher peak energy consumption costs. Further, load patterns could have some degree of repeatability on a daily basis, and the possible use of repetitive control and iterative learning control are therefore investigated in this thesis. As a result, learning-based precool strategies are proposed, which utilize the thermal storage capability in foodstuff to shift some of the peak load to less loaded hours. The precool time and period can continuously be updated based on data from previous days and the data-driven solutions are not based on models of the system, prior knowledge of load patterns, or weather forecasts and can
therefore easily be added to existing systems.
tuning of controllers a difficult and time consuming task. These are also some of the challenges which make advanced model-based control difficult, and a model-based controller will often be tailored to a specific system. The focus in this thesis is therefore instead on development of data-driven control strategies with a higher plug and play potential.
One of the main control challenges in refrigeration systems is proper control of superheat for efficient and safe operation of the system. This task can be performed by an electronic expansion valve and requires two sensors, which traditionally are a pressure and a temperature sensor. In this thesis, a novel maximum slope-seeking (MSS) control method is developed. This has resulted in a control implementation, which successfully has been able to control the evaporator superheat in four widely different refrigeration system test facilities without using a pressure sensor. A single-sensor solution is thus provided, which either reduces the variable costs or increases the robustness of the system
by not relying on pressure measurements. MSS is an example of data-driven control and can be applied to a broad class of nonlinear control problems. The method utilizes the qualitative nonlinearity in the system and harmonic analysis of a perturbation signal to reach an unknown, but suitable, operating point.
Another important control task in refrigeration systems is to maintain the temperature of the refrigerated space or foodstuff within the desired/legislative requirement, e.g., to prevent possible deterioration of the foodstuff. Refrigeration systems are often dimensioned to be able to cope with the highest possible loads and the hottest temperatures during the year, while also taking into account extreme weather conditions. Overdimensioning is expensive, both in terms of the variable cost, but also due to possibly higher peak energy consumption costs. Further, load patterns could have some degree of repeatability on a daily basis, and the possible use of repetitive control and iterative learning control are therefore investigated in this thesis. As a result, learning-based precool strategies are proposed, which utilize the thermal storage capability in foodstuff to shift some of the peak load to less loaded hours. The precool time and period can continuously be updated based on data from previous days and the data-driven solutions are not based on models of the system, prior knowledge of load patterns, or weather forecasts and can
therefore easily be added to existing systems.
| Originalsprog | Engelsk |
|---|---|
| Udgiver | |
| ISBN'er, trykt | 978-87-7152-032-3 |
| Status | Udgivet - 2014 |