network’
M. Reza
1,∗,
A. O. Dominguez
2,
P. H. Schavemaker
1,3,
A. Asmara
4,
F. A. Viawan
5and
W. L. Kling
1,31Electrical Power System Laboratory,
Faculty of Electrical Engineering, Mathematics and Computer Sci-ence, Delft University of Technology,
Mekelweg 4, 2628 CD, Delft, the Netherlands Fax: +31 15 278 1182 E-mail: m.reza@tudelft.nl ∗Corresponding author
2Electrical Engineering Department, University of Vigo,
36310 Vigo (Pontevedra), Spain 3TenneT TSO,
TenneT B.V.,
Utrechtseweg 310, 6812 AR, Arnhem, the Netherlands 4Ship Production,
Marine and Transport Technology, Delft University of Technology, Mekelweg 2, 2628 CD, Delft, the Netherlands
5Division of Electric Power Engineering, Chalmers University of Technology,
Gothenburg, Sweden, S-412 96 Gothenburg, Sweden
Abstract: This paper presents the concept of an “empty network” and shows how the power balance can be controlled in such a system. In this study, an “empty network” is defined as a transmission system in which no rotating mass is present. All generators are connected to dis-tributed systems and ’hidden’ behind power electronic interfaces. One generator creates a neat 50 Hz voltage that serves as a frequency refer-ence for the other generators. Consequently, a power imbalance cannot be detected in the classical way, as an altered system frequency.
the voltage deviations to detect the power imbalances are proposed and discussed.
Keywords: ‘empty network’, power balance control, distributed generation Reference to this paper should be made as follows: Reza, M., Dominguez A. O., Schavemaker P. H., Asmara A., Viawan F. A. and Kling W. L. (xxxx) ‘Controlling the power balance in an ‘empty network’, Int. J. of
Energy Technology and Policy, Vol. x, No. x, pp.xxx–xxx.
Biographical Notes: Muhamad Reza obtained his B.Sc. from Ban-dung Institute of Technology (ITB), Indonesia in 1997 and M.Sc. from Delft University of Technology (TU Delft) in 2000, the Netherlands, both in Electrical Engineering. He is currently pursuing Ph.D. in the Electrical Power System (EPS) laboratory, TU Delft, in the main frame-work of Intelligent Power Systems.
Alejandro O. Dominguez is a MSc student in the Electrical Engineer-ing Department, University of Vigo, Spain. Within Socrates-Erasmus Program, he spent 1 semester in the Electrical Power System (EPS) laboratory, Delft University of Technology (TU Delft) for performing research on the topic of “Empty Network”.
Pieter H. Schavemaker obtained his M.Sc. in Electrical Engineering from the Delft University of Technology in 1994 and he obtained his Ph.D. in Electrical Engineering from the Delft University of Technology in 2002. Since 1996 he has been with the Power Systems Laboratory where he is currently Assistant Professor. His main research interests include power system transients and power system calculations.
Andi Asmara received the B.Sc. and M.Sc. degrees from Bandung Institute of Technology, Indonesia in 1996, and Dortmund University, Germany in 2005, respectively. He worked as an Automation Product Engineer at Schneider Electric Indonesia from 1996 to 1999 and from 1999-2002 he worked at Klockner-ROH Joint Operation. Since 2005, he is a PhD Student at the Ship Production, Delft University of Technology, Delft, The Netherlands.
Ferry A. Viawan received the B.Sc. and M.Sc. degrees from Bandung Institute of Technology, Indonesia in 1996, and Chalmers University of Technology, Sweden in 2003, respectively. He worked as a Power System Engineer at PT Caltex Pacific Indonesia from 1996 to 2003, where he worked on operation, planning and protection of a transmission and distribution system. Since 2004, he is a PhD Student at the Division of Electric Power Engineering, Department of Energy and Environment, Chalmers University of Technology, Gothenburg, Sweden.
1 Background
Nowadays, natural and artificial constraints limit the expansion of centralized large power plants and a shift towards an extensive use of distributed generation (DG) - small, decentralized/distributed power generators that are connected to distribution network - appears.
The implementation of DG turns the current passive distribution network into an active one. This active distribution network does not only consume power, but it also generates power and supplies it to the transmission system [1-3]. In this way, power can be transferred from one distribution network to another. When we reflect further on this issue to the extreme, we could imagine that at a certain time, all centralized power plants are shut down and the electrical power generated by the DG in distribution networks is sufficient to meet the total demand of the grid. In other extreme, we can also imagine that the three-phase ac transmission systems are no longer used and those distribution networks are interconnected by dc transmission systems instead [4].
In conventional large power plants, the generators, i.e. synchronous genera-tors, operate at fixed speed and thereby with a fixed grid frequency. DG, however, presents a quite different characteristic. For example, the voltage generated by variable speed wind power generator, photovoltaic generator and fuel cells can not be directly connected to the grid. The power electronic converters play an im-portant role to match the characteristic of DG units and requirements of the grid connections [5].
For a stable operation of a power grid, there should always be a balance between generated power on one side and consumed power (plus losses) on the other. For example, in an isolated grid with power generated by a synchronous generator, the power imbalance causes the generator to accelerate (or decelerate) and alter the grid frequency. The increasing (or decreasing) power generated by the generator as a response of the altered system frequency will balance the power and bring the frequency back to the reference frequency. Many researchers try to adopt the classical synchronous generator control to an isolated grid supplied from power electronic interfaced DGs. For example, in [6], a control scheme based on droop frequency concept (of a synchronous generator) to operate inverters feeding an isolated system is presented. This concept is further explored in different modes of operation in [7].
Figure 1 The illustration of an ’empty network’ model
3
3
3
In this paper, the usage of the voltages to detect a power imbalance in the empty network is further explored by using different control schemes. While in [8], the power balance is maintained by using one generator that can compensate the whole system, which is uncommon in practice; in this paper, controlling the power balance in the system by using multi generators, where each generator has their own controller, is presented. Three different control schemes are developed and tested to examine whether or not the power balance in the empty network can be controlled.
2 Power Balance
Classically, a power system is characterized by a relatively small number of centralized power plants that are based on large synchronous generators. These generators are connected directly to the grid so that there is a coupling between the generator rotor speed and the power balance in the system.
The fundamental equation that governs the rotational dynamics of the syn-chronous generator is the swing equation [9]:
2H ωs
d2δ
dt2 = Pa= Pm− Pe. ([pu]) (1)
where, ωs s is the synchronous speed in electrical units [rad/s], H is the inertia constant (the stored kinetic energy in MJ at synchronous speed per machine rating (Smach) in MVA) [s], δ is the angular displacement of the rotor [rad], Pm is the shaft power input minus rotational losses [pu], Pe is the electrical power crossing the air gap [pu], Pa is the accelerating power [pu], and t is the time [s].
It can be seen from equation (1) that any imbalance in active power (Pe6= Pm) will result in non-zero accelerating power (Pa 6= 0), i.e. the rotor of the synchronous generators will either accelerate (d2δ
dt2 > 0) or decelerate (d 2δ
dt2 < 0).
the power system varies. To maintain the power balance in a power system, the generators are equipped with turbine speed governor that monitors the turbine-generator speed and adjusts the input from the prime mover in response to changes in this frequency.
3 ‘Empty Network’ Model 3.1 Basic assumptions
To decouple the changes of the voltages with the changes of the reactive power in an empty network, several assumptions are applied on the model of the empty network as the following:
• The distribution networks are equipped with reactive power sources. The reactive power source is sufficient to fulfill the reactive power load within the distribution networks.
• A reactive power control system is assumed to be applied within each dis-tribution network, locally. This control system is responsible to absorb the reactive power load changes within the distribution network.
• The (active) distribution networks are connected to the transmission system via power electronic interfaces. The power electronic interface is assumed to permit only active power to flow (bi-directional).
• The reactances of the transmission lines are compensated, in such a way that they behave like resistive lines.
3.2 Active Distribution Network Model
This paper focuses on the controlling of power balance in an empty network, on the transmission system level. Therefore, by taking the previously-mentioned assumptions (see section 3.1), the active distribution networks are considered as the following:
• the (distributed) generators are connected via power electronic converters and they generate only active power. The generators are initially set to balance the active power demand.
• the loads demand active and reactive power. They are modeled as constant impedance and constant power. Electrical motors (and the corresponding inertias) are hidden behind power electronic interfaces from the transmission system; therefore, electrical motors, if any, are assumed to be included in the constant power model of the loads.
• the reactive power is supplied by the dedicated reactive power sources. The reactive power sources are preset to balance the reactive power demand. • a power imbalance is simulated by changing the active power demand of the
Figure 2 The ‘empty network’ model used in the simulation of controlling the power balance '* '* '* 3 3 3
reactive power imbalance is not simulated. The consequences to the active distribution network model are:
– only active power will flow between the distribution networks and the transmission system. The power electronic interfaces that connect the distribution networks to the transmission network are then not included, and
– reactive power control systems that are responsible to balance the re-active power changes locally within the distribution networks are not included. The reactive power sources are then modeled as shunt devices. Thus, in the simulation of controlling the power balance, the empty network can be modeled as shown in figure 2.
3.3 (Distributed) Generator Model
In the empty network, three converter connected generator models are used to perform different function, i.e.:
• a constant voltage source. This model is used to represent one generator that provides the voltage and frequency reference for the other generators in the system. Figure 3 shows the representation of this generator, where Us is the constant (reference) voltage with a fixed frequency, and Zsis a source impedance.
Figure 3 Constant voltage source model
Figure 4 Constant current source model
ig
Ut
Zs
Figure 5 Controlled current source model
ic Controller
Ut
• a controlled current source ic. This model is used to represent one or (more) generator that serves as the ’slack’ generator. The slack generator that will either supply or absorb any deficit or surplus of active power in the system. This generator is assumed to have no current limiter and to be equipped with sufficient (energy) storage. The representation of this generator is shown in figure 5.
A remark should be given that power electronic interfaces that drive the output of converter connected (distributed generator) basically represent a voltage source converter, the mostly used converter nowadays [5,10]. Yet, the use of constant current sources to represent converter-connected generators in this simulation is supported by the following assumptions:
• a converter is actually a voltage source converter, but it behaves like a current source, so that current source models can substitute voltage source converters when simulations are made in large systems [11].
A constant current source that generates a current in phase with the (termi-nal) voltage represents a P Q-source that generates constant active power (P ) and zero reactive power (Q = 0), as long as the terminal voltage of the gen-erator is constant.
• In practice, a converter is equipped with a current limiter. When the terminal voltage drops, the converter will supply less active power. Thus, the use of a constant current source corresponds to a converter whose current is limited to the nominal value (in practice, 100% up to 120% of the nominal value). 4 Basic Controller Model
The basic functionality of the controller-block in figure 5 is highlighted hereun-der:
• In figure 6, three types of converter-connected generators are implemented. • In steady state, the current of the voltage source should be zero (is = 0), so
that there is no voltage drop across the impedance Zs.
• Any power imbalance should be eliminated by controlling the current ic that is generated by the controlled current source. This controlled current source represents the ’slack’ generator. When, for example, the active power con-sumption of the load Pload increases, the current flowing to the load (iload) will rise. Both generators that are modeled as current sources do not react (yet), and the voltage source will start to supply active power in order to balance the power. Thus is increases and causes a voltage drop over Z, so that Utdecreases. This voltage drop will be detected by the controller of the controlled current source. As a result, the controlled current source will sup-ply more active power (i.e. it injects more current ic) until the power balance is restored.
The controller that is used in the basic controller model is a proportional-integral controller (PI controller) that is a common feedback loop component in industrial control applications.
To verify the basic controller model, a load jump will be simulated to cause a power unbalance in the system shown in figure 6. The system voltage is set at 10 kV. The load demands 60 MW of active power (Pload) initially. The load is equally divided in constant impedance and constant power. The constant current source supplies all the initial power demand. The currents generated by the constant voltage source (is) and the controlled current source (ic) are thus equal zero. A load jump is applied by increasing the load modeled as the constant impedance with 30 MW.
Figure 6 Basic controller idea applied at a 1-bus test system Ut Controller US ig ic Pload Z iload ig is ic 1
Figure 7 The transitories of the voltages when a 30 MW load jump is applied at bus-1 of the system shown in figure 6
Figure 8 The transitories of the currents when a 30 MW load jump is applied at bus-1 of the system shown in figure 6
0 2 4 6 8 10 0 1 2 3 4 5 6 7 8 9 10 Current (kA) Time (s) i g i load i c i s
It can be seen in these figures that, following the load jump, the power balance is restored when the ’extra’ power demand is supplied by the ’slack’ generator (see the increasing current ic and active power P‘slack0 in figures 8 and 9). The bus voltage is also restored to a steady-state value that lies within a +/-5% margin of the nominal voltage.
5 Empty Network Controls
Practically, a power system consists of more than one bus. In this case, the system is decoupled and no longer linear, due to the dependency of one bus with another. In addition of that, the voltages at the buses throughout the system are practically not the same (see Section 2). Consequently, applying the basic control model of Section 4 for each bus gives potential difficulties, since the basic control model is a linear control system.
In this section, three control systems are proposed. These control systems are defined as:
1 Stand-alone master controller
2 Decentralized-controller with single reference 3 Decentralized controller with hysteresis
To verify these control systems, a simple test system that consists of 3 buses is defined as shown in figure 10. Tables 1 and 2 show respectively the component parameters used and the load flow settings and computed results in the 3-bus test system. Note that Gref refers to the reference (‘master’) generator. Gj refers to the constant-power generator (at bus-j). Loadjrefers to the load (at bus-j) and Cj refers to the reactive power source (at bus-j). Tjrefers to the transformer (at bus-j) and TLjkrefers to the transmission line (between bus-j and bus-k). In table 1, Sbase denotes M V Abaseof the test system [MVA]. UHV and UM V denote the system high-and medium voltage levels [kV]. R, XLand B denote the resistance [pu], reactance [pu] and susceptance [pu] of the transmission lines. XT denotes the transformer reactance and Z denotes the impedance between the reference ’master’ generator and bus-1.
Table 1 Component parameters used in the 3-bus test system
Description Parameter Value Unit
System base Sbase 100 MVA
Figure 10 A simple 3-bus test system (empty network)
5.1 Stand-alone master controller
The most simple way to overcome the non-linearity problem is by using a stand-alone master controller. In this approach only one basic controller is applied and connected to one of the bus of the system. In this case, the controller only reg-ulates the voltage of one bus and let the system comes to balance using its own connectivity.
Figure 11 shows the implementation of the stand-alone master controller in the test system (empty network). The ‘slack’ generator (that is represented by the
Table 2 Load flow settings and computed results in the 3-bus test system
Description Parameter Setting Computed Setting Computed
[MW] [MW] [MVAr] [MVAr] Generation Gref 0.0 -0.5 0.0 -6.5 G1 60.0 59.9 0.0 2.8 G2 60.0 60.4 0.0 2.8 G3 60.0 60.4 0.0 2.8 Load Load1 60.0 59.0 30.0 31.8 Load2 60.0 59.7 30.0 32.2 Load3 60.0 59.7 30.0 32.2 Shunt device C1 0.0 0.0 15.0 15.0 C2 0.0 0.0 15.0 15.3 C3 0.0 0.0 15.0 15.3
Figure 11 The implementation of the stand-alone master controller in the test system (empty network)
controlled current source) is implemented at bus-1. In this controller scheme, only one ‘slack’ generator is implemented in the test system. The generator should handle the active power imbalance occurs in the system. Note that GCj refers to the controlled-power/‘slack’ generator (at bus-j).
A load jump is applied by increasing the constant power load at bus-2 (the load is modeled as constant power) with 30 MW. Figure 12 shows the transitories of the active power where all the 30 MW power (of the load jump) is supplied by the ‘slack’ generator. The power balance is restored and all system parameters are back to stable steady state values.
When the stand-alone master controller scheme is implemented, the problem of ‘different signal’ of voltages for the control input is eliminated by only using one controller at bus-1. In this way, the controller uses only one voltage point as its reference signal. However, challenge will be that one generator should compensate for the power balance in the whole system. In the following Sections, some other proposals concentrate on dividing the ‘slack’ generators.
5.2 Decentralized-controller with single reference
Figure 12 The transitories of the active power when a 30 MW load jump is applied at bus-2 of the system (empty network) shown in figure 11
0 2 4 6 8 10 −10 0 10 20 30 40 50 60 70 80 90 100 Active Power [MW] Time [s] G ref GC1 Load1 , Load3 Load 2 G 1 , G2, G3
with single reference is proposed here. In this approach, the ‘slack’ generator in each bus has its own controller but the reference signal is common for all of them, and it can be taken from any voltage point.
Figure 13 shows the implementation of the decentralized-controller with single reference in the test system (empty network). Each of the ‘slack’ generators (rep-resented by the controlled current source) is implemented at bus-1, -2 and -3. The generators (altogether) should take care the active power imbalance occurs in the system. One control signal is used by all generators, that is the voltage at bus-1 (U1). Note that GCj refers to the controlled-power/‘slack’ generator (at bus-j).
A load jump is applied by increasing the constant power load at bus-2 (the load is modelled as constant power) with 30 MW . Figure 14 shows the transitories of the active power. The 30 MW power (of the load jump) is supplied by the three ‘slack’ generators at bus-1, -2 and -3. Each generator supplies the requirement power in balance, 10 MW. Also, all system parameters are back to stable steady state values.
In the same manners with the first approach, by implementing the decentralized-controller with single reference, the problem of ‘different signal’ of voltages for the control input is eliminated by only using one controller at bus-1. The difference is that this control signal is used for all three controllers. With this approach the control generator is no longer centralized, however there is one aspect that should be considered. This approach needs a communication link between each controller to transfer the reference data signal. It might happened that not all system has this luxury.
Figure 13 The implementation of the decentralized-controller with single reference in the test system (empty network)
Figure 14 The transitories of the active power when a 30 MW load jump is applied at bus-2 of the system (empty network) shown in figure 13
5.3 Decentralized controller with hysteresis
The third proposed approach uses three decentralized controllers; one controller is applied at each bus. As already described above, the linear controller cannot works perfectly on the non-linear decoupled system. The symptom that happens by applying the basic controller over the test system is that on the steady-state region the controller oscillates. This is due to the system that is decoupled and the reference signal that is not exactly the same.
To prevent this symptom, hysteresis on the controller input is applied. It means that the controller stops regulating the system whenever the value of the voltage lies within the hysteresis boundary. The used of hysteresis is allowed as long as the width of it is less or equal than the tolerance of the voltage, which is 1% of the nominal value.
Figure 15 shows the implementation of the decentralized controller with hys-teresis in the test system (empty network). The ‘slack’ generators represented by the controlled current sources) are implemented at bus-1, -2 and -3. One ‘slack’ generator is implemented in each bus. In this controller scheme, each ‘slack’ gen-erator uses the voltage of where the gengen-erator is implemented. The gengen-erators (altogether) should take care the active power imbalance occurs in the system. The generator GCj is infed by the control signal the voltage of bus-j (Uj).
A load jump is applied by increasing the load at bus-2 (the load is modeled as constant power) with 30 MW. Figure 16 shows the transitories of the active power. The 30 MW power (of the load jump) is supplied by the three ‘slack’ generators at bus-1, -2 and -3. Each generator supplies the power requirement in balance, 10 MW. All system parameters are also restored to stable steady state values.
By applying this approach, the power balance of the system can be achieved. The effect of hysteresis is that there is a steady state error on the voltage, but while the steady state value is smaller than the tolerance of the voltage value, it can be accepted. Thus, the voltage regulation is maintained. It might be not the optimal solution, but the system still operates on the specification. The main advantage using this approach is that the requirement of the communication link between controllers is eliminated.
5.4 Remarks
It is possible to improve the performance of the system by tuning the control parameters of the controllers. However, parameter tuning is mostly practical, that is, the tuned parameters can be used only for that particular system. Therefore that is not done in this paper.
Each control approach that is proposed in this paper has its own advantages and disadvantages among each other. The stand-alone master controller is the most simple one, it is easy to implement and its performance is good. The main disadvantage of this approach is that it uses a single generator to supply the required power.
The second approach, the decentralized-controller with single reference can over-come the single generator problem, with the price that it requires the availability of a communication link between controllers to transfer the reference signal data.
de-Figure 15 The implementation of the robust controller in the test system (empty network)
Figure 16 The transitories of the active power when a 30 MW load jump is applied at bus-2 of the system (empty network) shown in figure 15
centralized controller with hysteresis, since this approach uses split generator for each bus and in addition of that the three controllers are totally independent with each other. Thus, the needs of communication link between controllers can be eliminated.
6 Conclusions
In this paper, the method of controlling power balance in an ’empty network’ by using voltage deviation as control signal to detect the power imbalances has been presented. In addition design of different control system has been presented. The simulation results indicate that the power balance in the empty network can be controlled by using the proposed method.
So far the proposed method is developed with several simplified assumptions and the method is tested in a simple three bus system. Those simplifications should be eliminated in future work to see the applicability of this method on more realistic power systems. Further, this preliminary work on controlling the power balance in empty network concept is focused on how the system responses to the load change in normal conditions. Hence, the response of the system on an abnormal condition, such as a fault, is a remaining open question that is beyond the scope of this paper. Acknowledgements
This research has been performed within the framework of the research program ’intelligent power systems’ that is supported financially by SenterNovem. Senter-Novem is an agency of the Dutch Ministry of Economic Affairs.
References and Notes
1 Provoost F., Ishchenko A., Jokic A., Myrzik J.M.A. and Kling W.L. (2005), ‘Self Controlling Autonomous Operating Power Networks’, 18th Int. Conf. on Electricity Distribution, CIRED 2005, Turin, Italy.
2 Lasseter R.H. (2002), ‘Microgrids’, 2002 IEEE Power Engineering Society Winter
Meeting, New York, USA.
3 Meliopoulos A.P.S. (2002), ‘Challenges in simulation and design of µ-grids’, 2002 IEEE
Power Engineering Society Winter Meeting, New York, USA.
4 Bayegan, M. (2001), ’A vision of the future grid’, IEEE Power Engineering Review, December 2001.
5 Blaabjerg, F., Chen, Z., and Kjaer, S. B. (2004), ’Power electronics as efficient interface in dispersed power generation systems’, IEEE Transactions on Power Electronics, Vol. 19, No. 5, September 2004.
6 Chandorkar M.C., Divan D.M. and Adapa R. (1993), ‘Control of Parallel Connected Inverters in Stand Alone sc Supply Systems’, IEEE Transaction on Industry
Applica-tions, Vol. 29, No. 1, 1993.
8 Reza, M., Dominguez, A. O., Schavemaker, P. H., and Kling, W.L. (2005), ’Main-taining the power balance in an ‘Empty Network”, 2005 IEEE Conferene on Future
Power Systems.
9 Kundur P. (1994), Power system stability and control, McGraw-Hill Inc., New York.
10 Slootweg, J.G., de Haan, S.W.H., Polinder, H. and Kling,W.L. (2002), ‘Modeling new generation and storage technologies in power system dynamics simulations’, 2002
IEEE Power Engineering Society Summer meeting, Chicago,Illinois, USA.
11 Morren, J., de Haan, S.W.H. and Ferreira, J.A. (2004), ’Model reduction and control of electronic interfaces of voltage dip proof DG units’, 2004 IEEE Power Engineering
Society General Meeting, Denver, Colorado, USA.
