 Author
 Mohammad Moradzadeh (UGent)
 Promoter
 René Boel (UGent) and Lieven Vandevelde (UGent)
 Organization
 Abstract
 Power systems are nowadays becoming more and more interconnected, and controlled by several TSOs (Transmission System Operators), in order to ensure a reliable and economical supply and distribution of electric power. These (interconnected) electrical power networks are often considered as the most complex manmade dynamical systems ever. For example, according to the dataset provided by the ENTSOE (European Network of Transmission System Operators for Electricity) for static studies (calculation of the AC load flow), the European interconnected power grid consists of approximately 4300 buses, 6300 lines and 1100 transformers together with their loads, distribution systems and generations infeeds (in different voltage levels of 380 kV, 220 kV and 150 kV). The proper control of such a largescale interconnected power system is a very challenging problem due to the various continuous and discrete dynamics evolving in the system and their complicated interactions. Each local control agent (CA), corresponding to an area operated by one TSO, tries to achieve local improvement. However, it happens frequently that a local initiating disturbance in one area triggers some local control actions in its own area which in turn triggers further disturbances in the neighboring areas causing some undesirable control actions by their neighbors, and eventually a cascade of possibly wrong control actions lead the overall system to a final collapse. One important class of power system instability is voltage instability, which actually arises from the inability of combined generationtransmission systems to deliver the power requested by (dynamical recovery) voltagedependent loads. Such a voltage instability, if not corrected properly, due to a cascade of events, can eventually lead to voltage collapse (abnormally low voltages in a major portion of the system) often resulting in blackouts or separation of the system into separate unsynchronized islands. The societal impacts and financial costs/losses caused by blackouts are significantly huge. The voltage in electrical power systems is, in nature, a ``local" variable unlike frequency being a ``global" variable. This means that, in multiarea power systems, only areas that are electrically close together interact with each other for voltage, and there is no need to involve distant areas with negligible common interest in solving a local optimization problem. The latter promotes the decomposition approaches for voltage control, where the voltage control still remains a prerogative of the local utilities. This thesis focuses on longterm voltage instability  in the order of several minutes after a major disturbance. The driving force of such instability, following a disturbance, is the process of load restoration, where the dynamics of recovering loads directly as well as some control mechanism such as LTCs (Load Tap Changing transformers) indirectly (by restoring the distributionside voltages of the corresponding voltagedependent loads), try to locally restore the load powers to the predisturbance values. The longterm voltage instability often occurs when LTCs try to restore the distribution side voltages of the connected buses, while the maximum power that the transmission system can provide to loads is reduced by the reactive power capability limits of generators, mainly enforced by OXLs (Over eXcitation Limiters). It seems rather intuitive, then, to seek some way of anticipating what will be the future behavior of a power system, by employing controllers which can look ahead in time. The longterm voltage control becomes even a more complex and harder problem in largescale multiarea power system, each controlled by an independent TSO. The reason is that, for example, an arbitrary LTC move in one area can trigger undesirable LTC move(s), OXL activations in other areas, and such complicated global interactions may eventually lead to a blackout in the form of a voltage collapse. In order to avoid such a collapse in largescale multiarea power systems, the local control actions taken by each CA, must be coordinated with those of (adjacent) neighbors. This coordination requires communications between neighboring CAs. This thesis proposes an efficient distributed Model Predictive Control (MPC) paradigm which combines two concepts of ``lookingahead" and ``coordination". The proposed MPCbased control scheme relies on the communication of planned local control actions among neighboring CAs, each possibly operated by an independent TSO. Modelica, a free of charge objectoriented language, is used to develop a muchfasterthanrealtime simulator, providing an hybrid framework for effectively modeling and simulating power systems. Modelica facilitates the development of tools to generate very efficient codes for modeling of compositional physical systems such as electrical power networks, by relaxing the causality constraint of components, and focusing only on the topology of the overall system. In this thesis, the dynamic models for anticipation, are derived by considering each area as a hybrid dynamical system, using DAEs to describe piecewise continuous dynamics, and the set of events of hybrid automata representing the discrete logical controllers. This hybrid modeling framework captures the complex interactions between continuous and discrete dynamics. The ``lookingahead" voltage controller can anticipate, within the prediction horizon window, for example, the activation of OXLs, moving towards reaching the maximum physical tap limits for LTCs, and deviating too much from the prescribed voltage bounds for buses. The controller will then efficiently use these anticipations, by selecting a control sequence that does not cause the abovementioned constraint violations. The first input of the best control sequence selected by each local MPC, at each discrete time instant, will be applied to the local system until the next time instant, where the local optimization repeats again selecting the new best control action. Each CA, knowing a local model of its own area, as well as a reducedorder Quasi SteadyState (QSS) models of its immediate neighboring areas, and assuming a simpler equivalent PV model for the distant areas, performs a greedy local optimization over a finite window in time, communicating its planned control input sequence to its immediate neighbors only. The ``communicating" voltage controller enables each CA to coordinate its own local action with what its immediate neighbors are planning to do, assuming a QSS model for predicting how control actions of neighbors influence the interacting variables. The good performance of the proposed realtime modelbased feedback coordinating controller, following major disturbances, is illustrated using timedomain simulation of the wellknown realistic Nordic32 test system, assuming worstcase conditions. Robustness of the proposed method against measurement inaccuracies, modeling errors as well as the uncertainty of the load behavior has also been illustrated. This thesis considers two cases where, in the first reasonably sized network, a local CA, knows the complete model of the overall system, while, in the second realistic sized system, it employs reducedorder QSS models for immediate neighbors, and assumes a simpler equivalent PV model for the distant areas. Simulation results illustrates the significant achievements obtained when the proposed modelbased coordinating control is applied to different systems under some severe disturbances. This thesis compares the abovementioned simulation results with scenarios where a purely decentralized uncoordinated deadband control, as the current practice for LTCs, is applied, or where a decentralized uncoordinated MPC approach with no communication is applied. In this way it becomes possible to identify the two aforementioned distinct contributions of the proposed modelbased coordinating approach namely ``lookingahead" and ``communication", since the decentralized deadband approach lacks both anticipation and coordination, and the decentralized MPC approach ignores the communications with neighbors.
 Keywords
 model predictive control, Voltage control, coordination, optimization, load tap changing transformer
Downloads

PhD title page included.pdf
 full text
 
 open access
 
 
 2.53 MB
Citation
Please use this url to cite or link to this publication: http://hdl.handle.net/1854/LU3258526
 MLA
 Moradzadeh, Mohammad. Voltage Coordination in MultiArea Power Systems via Distributed Model Predictive Control. Ghent University. Faculty of Engineering and Architecture, 2012.
 APA
 Moradzadeh, M. (2012). Voltage coordination in multiarea power systems via distributed model predictive control. Ghent University. Faculty of Engineering and Architecture, Ghent, Belgium.
 Chicago authordate
 Moradzadeh, Mohammad. 2012. “Voltage Coordination in MultiArea Power Systems via Distributed Model Predictive Control.” Ghent, Belgium: Ghent University. Faculty of Engineering and Architecture.
 Chicago authordate (all authors)
 Moradzadeh, Mohammad. 2012. “Voltage Coordination in MultiArea Power Systems via Distributed Model Predictive Control.” Ghent, Belgium: Ghent University. Faculty of Engineering and Architecture.
 Vancouver
 1.Moradzadeh M. Voltage coordination in multiarea power systems via distributed model predictive control. [Ghent, Belgium]: Ghent University. Faculty of Engineering and Architecture; 2012.
 IEEE
 [1]M. Moradzadeh, “Voltage coordination in multiarea power systems via distributed model predictive control,” Ghent University. Faculty of Engineering and Architecture, Ghent, Belgium, 2012.
@phdthesis{3258526, abstract = {{Power systems are nowadays becoming more and more interconnected, and controlled by several TSOs (Transmission System Operators), in order to ensure a reliable and economical supply and distribution of electric power. These (interconnected) electrical power networks are often considered as the most complex manmade dynamical systems ever. For example, according to the dataset provided by the ENTSOE (European Network of Transmission System Operators for Electricity) for static studies (calculation of the AC load flow), the European interconnected power grid consists of approximately 4300 buses, 6300 lines and 1100 transformers together with their loads, distribution systems and generations infeeds (in different voltage levels of 380 kV, 220 kV and 150 kV). The proper control of such a largescale interconnected power system is a very challenging problem due to the various continuous and discrete dynamics evolving in the system and their complicated interactions. Each local control agent (CA), corresponding to an area operated by one TSO, tries to achieve local improvement. However, it happens frequently that a local initiating disturbance in one area triggers some local control actions in its own area which in turn triggers further disturbances in the neighboring areas causing some undesirable control actions by their neighbors, and eventually a cascade of possibly wrong control actions lead the overall system to a final collapse. One important class of power system instability is voltage instability, which actually arises from the inability of combined generationtransmission systems to deliver the power requested by (dynamical recovery) voltagedependent loads. Such a voltage instability, if not corrected properly, due to a cascade of events, can eventually lead to voltage collapse (abnormally low voltages in a major portion of the system) often resulting in blackouts or separation of the system into separate unsynchronized islands. The societal impacts and financial costs/losses caused by blackouts are significantly huge. The voltage in electrical power systems is, in nature, a ``local" variable unlike frequency being a ``global" variable. This means that, in multiarea power systems, only areas that are electrically close together interact with each other for voltage, and there is no need to involve distant areas with negligible common interest in solving a local optimization problem. The latter promotes the decomposition approaches for voltage control, where the voltage control still remains a prerogative of the local utilities. This thesis focuses on longterm voltage instability  in the order of several minutes after a major disturbance. The driving force of such instability, following a disturbance, is the process of load restoration, where the dynamics of recovering loads directly as well as some control mechanism such as LTCs (Load Tap Changing transformers) indirectly (by restoring the distributionside voltages of the corresponding voltagedependent loads), try to locally restore the load powers to the predisturbance values. The longterm voltage instability often occurs when LTCs try to restore the distribution side voltages of the connected buses, while the maximum power that the transmission system can provide to loads is reduced by the reactive power capability limits of generators, mainly enforced by OXLs (Over eXcitation Limiters). It seems rather intuitive, then, to seek some way of anticipating what will be the future behavior of a power system, by employing controllers which can look ahead in time. The longterm voltage control becomes even a more complex and harder problem in largescale multiarea power system, each controlled by an independent TSO. The reason is that, for example, an arbitrary LTC move in one area can trigger undesirable LTC move(s), OXL activations in other areas, and such complicated global interactions may eventually lead to a blackout in the form of a voltage collapse. In order to avoid such a collapse in largescale multiarea power systems, the local control actions taken by each CA, must be coordinated with those of (adjacent) neighbors. This coordination requires communications between neighboring CAs. This thesis proposes an efficient distributed Model Predictive Control (MPC) paradigm which combines two concepts of ``lookingahead" and ``coordination". The proposed MPCbased control scheme relies on the communication of planned local control actions among neighboring CAs, each possibly operated by an independent TSO. Modelica, a free of charge objectoriented language, is used to develop a muchfasterthanrealtime simulator, providing an hybrid framework for effectively modeling and simulating power systems. Modelica facilitates the development of tools to generate very efficient codes for modeling of compositional physical systems such as electrical power networks, by relaxing the causality constraint of components, and focusing only on the topology of the overall system. In this thesis, the dynamic models for anticipation, are derived by considering each area as a hybrid dynamical system, using DAEs to describe piecewise continuous dynamics, and the set of events of hybrid automata representing the discrete logical controllers. This hybrid modeling framework captures the complex interactions between continuous and discrete dynamics. The ``lookingahead" voltage controller can anticipate, within the prediction horizon window, for example, the activation of OXLs, moving towards reaching the maximum physical tap limits for LTCs, and deviating too much from the prescribed voltage bounds for buses. The controller will then efficiently use these anticipations, by selecting a control sequence that does not cause the abovementioned constraint violations. The first input of the best control sequence selected by each local MPC, at each discrete time instant, will be applied to the local system until the next time instant, where the local optimization repeats again selecting the new best control action. Each CA, knowing a local model of its own area, as well as a reducedorder Quasi SteadyState (QSS) models of its immediate neighboring areas, and assuming a simpler equivalent PV model for the distant areas, performs a greedy local optimization over a finite window in time, communicating its planned control input sequence to its immediate neighbors only. The ``communicating" voltage controller enables each CA to coordinate its own local action with what its immediate neighbors are planning to do, assuming a QSS model for predicting how control actions of neighbors influence the interacting variables. The good performance of the proposed realtime modelbased feedback coordinating controller, following major disturbances, is illustrated using timedomain simulation of the wellknown realistic Nordic32 test system, assuming worstcase conditions. Robustness of the proposed method against measurement inaccuracies, modeling errors as well as the uncertainty of the load behavior has also been illustrated. This thesis considers two cases where, in the first reasonably sized network, a local CA, knows the complete model of the overall system, while, in the second realistic sized system, it employs reducedorder QSS models for immediate neighbors, and assumes a simpler equivalent PV model for the distant areas. Simulation results illustrates the significant achievements obtained when the proposed modelbased coordinating control is applied to different systems under some severe disturbances. This thesis compares the abovementioned simulation results with scenarios where a purely decentralized uncoordinated deadband control, as the current practice for LTCs, is applied, or where a decentralized uncoordinated MPC approach with no communication is applied. In this way it becomes possible to identify the two aforementioned distinct contributions of the proposed modelbased coordinating approach namely ``lookingahead" and ``communication", since the decentralized deadband approach lacks both anticipation and coordination, and the decentralized MPC approach ignores the communications with neighbors.}}, author = {{Moradzadeh, Mohammad}}, isbn = {{9789085785675}}, keywords = {{model predictive control,Voltage control,coordination,optimization,load tap changing transformer}}, language = {{eng}}, pages = {{XIX, 122 [in multiple pagination]}}, publisher = {{Ghent University. Faculty of Engineering and Architecture}}, school = {{Ghent University}}, title = {{Voltage coordination in multiarea power systems via distributed model predictive control}}, year = {{2012}}, }