Projects per year
Abstract
The future energy system will rely on the production of renewable and lowcarbon energy sources, many of which are weatherdependent, such as wind and solar energy. To effectively implement thisweatherdriven energy system, there is a need to control energy demand in a manner that aligns with the fluctuating supply. This requires flexibility in various systems, such as adjusting indoor air temperatures in buildings, reactor temperatures in biogas production plants, and energy usage in wastewater treatment plants, among others, such as powertoX, district heating, and electric vehicles. Scheduling energy usage optimally, however, poses a significant challenge. Model predictive control (MPC) is a control method that accounts for future inputs, disturbances, and system dynamics, allowing for optimal scheduling of energy usage. This dissertation investigates modelling techniques for various energy systems and applies predictive control methods to quantify and improve control and flexibility concepts. A significant contribution of this research is the proposal of embedded disturbance models for optimal control problems, which provide the controller with continuoustime disturbance forecasts and potentially more filtering information.
This dissertation considers various models for flexible control of the indoor climate in buildings. Paper A introduces a nonlinear model describing the aggregated indoor air temperature of a building. Nonlinearities come from the radiator thermostats and the energy usage, which are difficult to consider as linear phenomenons. Paper B presents the results and findings of a control experiment carried out using the continuoustime model in Paper A. The results successfully showed that the building was able to shift its heat load in time and react to varying prices. Paper C is a simulation study of online control of the same building using the model in A. The paper reaches the same conclusion that the building is suited for control and that significant economic and energy savings are available. Paper D considers nonlinear ARXmodels describing the indoor air temperature in single rooms. These take various inputs and disturbances into account and are simple and fast to estimate and use—and are thus useful for control. The present dissertation also focus on modelling and forecasting of disturbances in control. Paper E introduces the concept of embedding a disturbance model in continuoustime into the formulation of the optimal control problem. This technique has the advantage of being able to describe the disturbances’ influence on the system in continuoustime (instead of e.g. zeroorder hold discretisations) and supply more information for the filtering of the system. Paper F uses this technique in a linearquadratic controller used to control the indoor air climate of a building. It demonstrates the potential improvements of this forecasting technique used in a quadratic controller compared to standard disturbance mitigation techniques and the tradeoffs between variation in inputs and the controlled system.
Choatic systems are dynamical systems governed by positive Lyapunovexponents. This means that the predictability is lost exponentially if the initial state is not known exactly (which is rarely the case). Paper G introduces a method for controlling a chaotic system into an arbitrary point on a Poincaré section. The proposed method consists in two steps: first of solving an optimal control problem to obtain a periodically applied control signal and afterwards applying an additional adaptive control.
This dissertation considers various models for flexible control of the indoor climate in buildings. Paper A introduces a nonlinear model describing the aggregated indoor air temperature of a building. Nonlinearities come from the radiator thermostats and the energy usage, which are difficult to consider as linear phenomenons. Paper B presents the results and findings of a control experiment carried out using the continuoustime model in Paper A. The results successfully showed that the building was able to shift its heat load in time and react to varying prices. Paper C is a simulation study of online control of the same building using the model in A. The paper reaches the same conclusion that the building is suited for control and that significant economic and energy savings are available. Paper D considers nonlinear ARXmodels describing the indoor air temperature in single rooms. These take various inputs and disturbances into account and are simple and fast to estimate and use—and are thus useful for control. The present dissertation also focus on modelling and forecasting of disturbances in control. Paper E introduces the concept of embedding a disturbance model in continuoustime into the formulation of the optimal control problem. This technique has the advantage of being able to describe the disturbances’ influence on the system in continuoustime (instead of e.g. zeroorder hold discretisations) and supply more information for the filtering of the system. Paper F uses this technique in a linearquadratic controller used to control the indoor air climate of a building. It demonstrates the potential improvements of this forecasting technique used in a quadratic controller compared to standard disturbance mitigation techniques and the tradeoffs between variation in inputs and the controlled system.
Choatic systems are dynamical systems governed by positive Lyapunovexponents. This means that the predictability is lost exponentially if the initial state is not known exactly (which is rarely the case). Paper G introduces a method for controlling a chaotic system into an arbitrary point on a Poincaré section. The proposed method consists in two steps: first of solving an optimal control problem to obtain a periodically applied control signal and afterwards applying an additional adaptive control.
Original language  English 

Publisher  Technical University of Denmark 

Number of pages  144 
Publication status  Published  2023 
Fingerprint
Dive into the research topics of 'Stochastic Disturbance Models A Next Generation of ModelBased Control'. Together they form a unique fingerprint.Projects
 1 Finished

Stochastic Disturbance Models: A Next Generation of ModelBased Control
Thilker, C. A. (PhD Student), Lindberg, K. B. (Examiner), Salom Tormo, J. (Examiner), Madsen, H. (Main Supervisor) & Jørgensen, J. B. (Supervisor)
01/06/2020 → 31/08/2023
Project: PhD