Implementation and test results of a wide-area measurement-based controller for damping interarea oscillations considering signal-transmission delay

R. Majumder, B. Chaudhuri and B.C. Pal Abstract: The paper describes the implementation of a centralised control-design scheme in a real-time laboratory-based dynamic simulator. A centralised multivariable-control algorithm is designed employing remote feedback signals considering delay in signal transmission. The transmitted signals can be used for multiple-swing-mode damping using a single controller. Such an algorithm is numerically intensive owing to the high controller size and the predictor that takes care of the delay in signal transmission. The performance of the proposed controller is demonstrated by practical implementation using a rapid-prototyping controller mounted on a PC-ATX. Results of the experimental tests are shown for a detailed model of the power system implemented using Linux PC-based multiprocessor technology.

Introduction

Interarea oscillations [1] are inherent in large interconnected power systems. These oscillations often suffer from poor damping. In a practical power system, the number of dominant interarea modes is often larger than the number of controllable devices available to control them. Research attention has therefore been focussed on designing new control structures to improve the damping of multiple swing modes. With rapid advancements in wide-area-measurement systems (WAMS) technology, the transmission of measured signals to a remote control centre has become relatively simpler. Employing phasor-measurement units (PMU), it is possible to deliver the signals at a speed of as high as 30 Hz sampling rate [2, 3]. It is possible to deploy the PMUs at strategic locations of the grid and obtain a coherent picture of the entire network in real time [3]. Notable work from Kamwa [5] suggests that wide-area control can be 1012 times more effective than local decentralised control of wide-area oscillations. However, the cost and associated complexities restrict the use of such sophisticated signal-transmission hardware on a large commercial scale. As a more viable alternative, the existing communication channels are often used to transmit signals from remote locations. The major problem is the delay involved between the instant of measurement and that of the signal being available to the controller. This delay can typically be in the range of few hundreds of milliseconds
r The Institution of Engineering and Technology 2007 doi:10.1049/iet-gtd:20050493 Paper rst received 8th November 2005 and in nal revised form 15th February 2006 B. Chaudhuri and B.C. Pal are with Department of Electrical and Electronic Engineering, Imperial College London, Exhibition Road, London SW7 2BT, UK R. Majumder is with the School of Information Technology and Electrical Engineering, The University of Queensland, St. Lucia, Qld 4072, Brisbane, Australia E-mail: b.pal@imperial.ac.uk IET Gener. Transm. Distrib., Vol. 1, No. 1, January 2007

depending on the distance, protocol of transmission and several other factors. Therefore, a predictor approach is adopted in this work to formulate the damping-controldesign problem for a power system having a signaltransmission delay of 0.75 s. A H1 controller is designed by solving the problem using linear-matrix inequalities (LMIs) with additional pole-placement constraints [5]. Once the designs and simulations have been performed, the next credible step would be to implement the control scheme and verify the closed-loop behaviour of the entire power system in real time. One concern, however, is the accessibility to the actual system for validating the performance of the controllers in real time. It is extremely difcult to build even a prototype of an actual power system in the laboratory or a manufacturing site. For obvious reasons, it is rarely permissible to perform such validation tests in the eld. From technical as well as economic prospective, it is thus desirable to have a dynamic-system emulator which can emulate physically the dynamic behaviour of the power system in real time. Several projects reported in [57] described simulation results. A serious bottleneck, which often hampers effective real-time implementation of controllers designed by H1 optimisation techniques is that they tend to have a large size and possibly stiff state equations. This is more so for time-delayed systems as the resulting controller is the feedback combination of the designed controller and the predictor [5]. Therefore before the controllers in actual plants are commissioned, the experimental verication of the designed control algorithm is very important. First implementation and experimental verication of a robust exible ACtransmission-systems (FACTS) controller is described in this paper. The dynamic behaviour of the power system is emulated in real time in a real-time station (RTS) on which the designed control algorithm is tested using a rapidprototyping controller (RPC). The hardware interface between the two platforms is in the analogue domain through digital-to-analogue and analogue-to-digital converter (DAC/ADC) modules, so that it is virtually
1

impossible for the controller to distinguish between the actual plant, and the emulated plant, and same holds for the plant as well. In this way, the costly proposition of building a large prototype power system in the laboratory for testing purpose can be avoided. 2 Predictor-based control design for dead-time system In dead-time systems, measured feedback signals take a certain amount of time to reach the controller. Controlling such time-delayed systems is difcult. Figure 1 shows the generalised control-design setup. The delayed system is modelled as Ph s P sesh , where e sh is the block to represent the delay in feedback measurement with a dead time h 4 0. The general control setup for a system having an output delay is shown in Fig. 1, where ! P11 s P12 s 1 P s P21 s P22 s

uncontrollable response that is likely to govern the H1 performance index. One possible way of achieving this is to introduce a uniform delay in both the paths (path 1 and path 2), as shown in Fig. 4. There are two steps in achieving this. First, the delay blocks (e sh) at points 1 and 2 need to be shifted to point 3 by introducing a suitable predictor block in parallel with K. Secondly, a delay block needs to be introduced in path 1. The rst step is achieved by introducing a Smith predictor block Z(s) P22(s) P22(s)e sh, as shown by the dotted box in Fig. 3. The second task of bringing a delay in path 1 is achieved while forming the generalised plant prior to control design. The presence of the predictor block Z and the delay in path 1 ensures that the responses (through path 1 and path 2) governing the performance index is delayed uniformly, as shown in Fig. 4.
Path 1

e-sh

P11 Z= P22 -P22 e-sh Z

z P

S
3

P12

y K

e-sh
Path 2

P21

y
-sh

u ym K
Fig. 3

P22

-sh

e
Fig. 1

Introduction of Smith predictor and delay block

Control setup for dead-time systems

Path 1

The closed-loop transfer-matrix from d to z is: T zd s P11 P12 Kesh I P22 Kesh 1 P21 . This suggests that there exists an instantaneous response through the path P11 (path 1 in Fig. 2) without any delay. An equivalent structure is shown in Fig. 2.
Path 1

e-sh

P11

e-sh

P12

y K S

P21
Path 2

P11
1
Fig. 4

P22
Uniform delay in both paths

P12

y K

-sh

P21

Path 2

P22
Fig. 2

-sh

Equivalent representation of dead-time systems

It can be seen that, during the period t 0h after d is applied, the output z is not controllable, since it is determined only by P11 and d with no response coming through the controlled path (path 2 in Fig. 2). This means that the H1 performance index kTzd k1 is likely to be dominated by a response that cannot be controlled, and this is not desirable. Designing a controller for such systems is rather difcult [8]. Smith predictor is the rst effective tool for tackling such control problems [9]. The primary idea is to eliminate any
2

A predictor-based controller for the dead-time plant Ph(s) P22(s)e sh consists of a predictor Z P22 P22e sh, and a stabilising compensator K, as shown in Fig. 5. The predictor Z is an exponentially stable system such that Ph+Z is rational, i.e. it does not involve any uncontrollable response governing the H1 performance index. Let us consider a generalised delay-free plant given by 2 3 A B1 B2 2 Ps 4 C1 D11 D12 5 C2 D21 D22 ~ can be formulated using the unied The generalised plant P Smith predictor method. The steps for arriving at the nal ~ are detailed in [5]. The nal form of the expression for P
IET Gener. Transm. Distrib., Vol. 1, No. 1, January 2007

e ym

-sh

d e
-sh

revealed that the system had three critical interarea modes i h si and (li si 7 joi). The damping ratio zi p 2 2
oi frequency fi 2 p Hz are shown in Table 1. The objective is to damp these modes by designing a supplementary damping controller for the TCSC. Appropriate feedback stabilising signals were chosen for each mode using the modal observability analysis, see [7] for details. The open-loop plant is constructed using the linearised system matrix A, the input matrix B corresponding to the output of the TCSC and the output matrix C corresponding to the measured signals. si oi

P(s)

+ v + yp K(s)
Fig. 5 Smith-predictor formulation

Z(s) P
aug

(s)

Robustness validation by linear analysis

generalised plant is as follows: 3 2 1 B1 B2 Eh A 0 7 6 A h 6 0 Anc 0 e nc Inc V 1 B1 0 7 7 6 7 6 ~ 6 7 ! P 7 6 0 7 6 C1 0 D 12 C V 1 7 6 Inc 5 4 D12 0 C 2 Eh 0 3 The related H1 problem as stated in (3) considering the generalised plant formulated above, was solved in Matlab using the LMI approach with pole placement as additional constraints [5] for ensuring minimum damping. The LMI Control Toolbox available with Matlab has been used to perform the necessary computations. In a large power-system model P22 is of large order, making it difcult to construct Z. It will be easy to carry out design with the reduced plant model and then validate the performance against the full model. 3 Study system

The damping action of the designed controller was examined under different types of disturbances in the system. These include changes in power ow levels over key transmission corridors, change of type of loads etc. The damping ratios of the critical interarea modes under these operating conditions are summarised in Fig. 7. Note that in Fig. 7 CI, CP and CC stand for constant-impedance-, constant-power- and constant-current-type loads. The damping is found to be highly satisfactory in all the cases. 5 Implementation in real-time platform

Simulation has long been recognised as an important and necessary step in development, design and testing of FACTS controllers for mitigation of interarea oscillation. Recent advances in both computing hardware and sophisticated power-system-component modelling techniques have signicantly increased the application of real-time digital simulation in the power-system industry. Testing controllers with such hardware-in-the-loop (HIL) digital simulators is demonstrated in this paper. Historically the method of performing dynamic studies utilises simulators made up of scaled-down power-system components. Each component is physically connected to
Table 1: Interarea modes of the study system
Dumping ratio z 0.0626 0.0435 0.0554 Frequency (Hz) 0.3913 0.5080 0.6232

The control design and simplication exercise are carried out on a 16-machine, 5-area study system, shown in Fig. 6. A thyristor-controlled series capacitor (TCSC) is installed in the system to strengthen the transmission corridor between New York power systems (NYPS) and area 5. An eigenvalue analysis of the linearised model of the system

AREA#5

New England Test System

G3 G7 G5 07 23 04 G6 06 68 22 21 G4 05 20 65 64 19 66 67 52 37 27 G9 09 29 24 57 56 03 62 02 63 58 G2 13 59

New York power system

16 G13

G16

y
43 39 45 35

60 17 G12 12 36 34 61 33 32 30 11 G11 38

50

K u TCSC

18

51

AREA#4

49 46 31 10 G10 48 40

G15 15

55

42

28 26 08 G8 25 01 G1 54 47

53

41 14 G14

remote signal links

AREA#3

Fig. 6

16-machine ve-area study with TCSC

3

IET Gener. Transm. Distrib., Vol. 1, No. 1, January 2007

0.25
Damping ratio Damping ratio

0.3 0.25 Open Loop Close Loop 0.2 0.15 0.1 0.05 Mode 1 Mode 2 Mode 3

0.2 0.15 0.1 0.05 0

Mode 1

Mode 2
Inter-area modes

Mode 3

100

500

700

900

Power flow from NETS to NYPS (MW)

0.25 0.2
Damping ratio

0.2 0.16 Mode 1 Mode 2 Mode 3

Damping ratio

0.15 0.1 0.05 0

0.12 0.08 0.04

Mode 1 Mode 2 Mode 3

CI

CC + CI CP + CI Dynamic
Types of load

60-61

53-54
Line outage

27-53

Fig. 7

Robustness validation through damping ratio

measured signals from system

6 RT-LAB real-time station and rapid-prototyping controller

6.1

RT-LAB real-time station

Fig. 8

A schematic overview of the RT-LAB real-time station, used for the real-time simulation of the study system, is depicted in Fig. 8. The RTS is a PC running on the realtime operating system RedHawk RT-Linux. It has a dual Xeon processor of 3.2 GHz. Both the processors share a common memory, as shown in Fig. 8. The separation of the tasks involved in simulating the power-system dynamic behaviour is depicted in Fig. 8. The computational tasks are distributed as follows: CPUl of the dual-CPU computer handles the differential equations describing the dynamic behaviour of the generators, associated excitation systems and the FACTS device (TCSC in this case). CPU2 simultaneously solves the network equations to connect the generators with the network. It has been noticed that the most computationally expensive task is to solve the set of network equations to calculate the complex bus voltages from the admittance matrix and the set
4

Schematic diagram of RTS with system-task separation

of complex bus currents. The inversion of the admittance matrix is avoided by using LU factorisation. The LU factors of the admittance matrix is precomputed and stored. Owing to the change in series compensation of the TCSC, the admittance matrix must be updated dynamically. An optimal ordering of the TCSC bus is done to keep the computational burden or updates of admittance-matrix to a minimum. In the current HIL application, the CPU in charge of solving the differential equations associated with the generators, excitation system and TCSC also controls the FPGA I/O card that sends the measured signals from the plant and reads the control signal generated by the rapid-prototyping controller. The digital signal generation and sampling are both obtained using 10 ns resolution. The FPGA card, built around the Xilinx Virtex-II Pro, also
IET Gener. Transm. Distrib., Vol. 1, No. 1, January 2007

control input

the next in a manner similar to that in real system. This analogue-simulation technique forms the basis of transientnetwork analysers (TNA) and high-voltage DC (HVDC) simulators. However, owing to the size and the cost constraints, it is difcult to build even a prototype of an interconnected power system in the laboratory. An alternative method, which could be applied to test the designed controller, is based on the mathematical representation of the dynamics of the system rather than a scaled-down version of the physical components. The algorithms necessary for software-based electromagnetic-transient simulation were rst described by H. Dommel in 1969 [10] and have been utilised in several well known programs suchas EMTP and EMTDC [11, 12]. One of the most signicant disadvantages of software simulations relates to the speed at which they operate. Unlike analogue simulators which operate in real time, most digital-simulation systems operate in nonreal time. Real-time operation implies that an event in the system which lasts for 1 s can be simulated on the simulator in exactly 1 s. Any external hardware could be connected with a system simulated in real time with in-built DAC/ADC modules. Therefore, the controller under test, whether implemented in a low-cost dedicated microcontroller or in analogue circuitry, could be interfaced with the real-time digital simulator.

Real time station

Ethernet connection to command station

CPU2

task#2
Network algebraic equation

Shared memory CPU1

of generators, excitation system and TCSC

task#1
Opal FPGA (digital I/O) Opal FPGA (D/A) bus PCI

Signal Conditioning

controls fast 16-bit D/A and A/D converters. The sampling time used in the real-time simulation is 1 ms. CPU1 takes around 300450 ms to solve the differential equations as stated above, and the computational time for CPU2, which solves the algebraic equations simultaneously, is almost 800 ms. Therefore no overruns are detected during the realtime operation. The RTS is a multiapplication real-time platform suitable for hosting a wide variety of dynamic applications and, unlike other realtime environments, is not conned only to power systems. Bearing in mind the cost associated with this state-of-the-art technology, and its ability to support a wide variety of dynamical-systems applications, over a variety of disciplines, it makes the facility a very cost-effective technology.

6.2

RT-LAB rapid-prototyping controller

A rapid-prototyping controller (RPC) is a PC having a Pentium 4 processor running at 3.2 GHz under the RedHawk RT-Linux operating system with the capability of handling I/O interactions in real time. Once the controller is designed and realised in its statespace form, it can be implemented in the RPC. The computational time for the RPC to calculate the control signal is about 60 ms, which is well within the sampling period. In practical situations, the controller could also be implemented in any dedicated lowcost microcontroller. The internal architecture and the I/O interfacing of the RPC and the RTS are almost identical. The efcacy of the designed robust controller is proved in real time by implementing it on RPC.

Fig. 9

Schematic diagram of RTS and RPC

6.3

HIL conguration of RTS and RPC

A Windows host command station is used to set up various test, scenarios or evaluate controller performance for different system disturbances, operating conditions and system/ controller parameters. The host command station and both the real-time digital simulators interact between themselves through a 100 Mbit/s ethernet connection. The schematic diagram of the RTS and the RPC is shown in Fig. 9. If necessary, the computational tasks can be distributed across several PCs to decrease the simulation time step or to simulate more complex systems. Intercomputer communication systems supported by RT-LAB are FireWire 800 Mbits/s as well as SignalWire, which is an FPGA-based fast serial communication link capable of delivering up to 1.25 Gbit/s transfer rates, with a latency of 200 ns. In the RTS platform, the input model is developed in MATLAB Simulink. A graphical user interface, called

RT-lab main control, is available for managing the communication between the RTS and the command station. Real-time workshop (RTW) of Matlab interfaces Simulink and this hardware platform. The RT-lab main control could be used to build real-time code, and to download and execute this code on a Xeon processor of the RTS through the Ethernet link. Any signal from the model can be bought outside through a built-in DAC and any physical signal can be fed back to the system using a built-in ADC. A robust FACTS controller is a three-input, one-output controller which has been implemented using RPC. The same RT-lab main control is used to generate code for the controller and to download it to RPC, for real-time execution, through the Ethernet link. The inputs to this controller are deviation in the active power ow through the transmission line connecting bus 5145, 1816 and 1317. These signals are generated in the RTS and are interfaced with the controller through on-board DACs of RTS, as shown in Fig. 10. All three measured signals are delayed. These delays are implemented by using transport-delay blocks in the RTS. The control signal is computed by the RPC at every sampling instant and fed back to the simulated power system by a DACADC combination, as shown in Fig. 10.

Transport delay

DAC DAC DAC Measured signals

flows

ADC

TCSC

Power system 16 machine 68 bus inter-connected system

Transport delay Transport delay

Dual Xeon 3.2 GHz processor real time station (RTS)

Control input

Hardware coupling between two platforms

ADC DAC Robust FACTS controller ADC ADC

Pentium 4 3.2GHz Processor rapid prototyping controller (RPC)

Fig. 10

Closed-loop conguration of RTS and RPC

5

IET Gener. Transm. Distrib., Vol. 1, No. 1, January 2007

Experimental results

The performance of the designed controller implemented on RPC has been evaluated under various operating conditions, using the emulated power systems on RTS. The
1.5

deviations in active power ow through the transmission line from RTS were sampled and fed to the controller. The cotrol signal is fed back to the TCSC control input.

7.1

Validity check

0.5

The validity of implementation of the nonlinear-powersystem model on the RTS is rst established by a number of identical tests performed on the RTS-implemented power system and the same system modelled using MATLABSimulik on a PC. A representative set of results is shown in Fig. 11 and Fig. 12. A three-phase to-earth fault is created at 1 s at bus 27 and the fault is cleared after 80 ms followed by opening of one of the tieline connecting buses 29 and 53. Figure 11 shows deviation in active power ow in the transmission line connecting bus 51 and 45 in MatlabSimulink and Fig. 12 shows the same parameters with the system implemented in RTS. The dynamic behaviour of the study system in thus emulated in real time with reasonable accuracy using the RTS.
0 5 10 time, s 15 20 25

7.2

Case studies

Fig. 11 Dynamic response of the system, Matlab Simulink environment

Fig. 12 Dynamic response of the system, real-time station, dual Xeon Processor in real time

One of the most severe disturbance stimulating poorly damped interarea oscillations is a three-phase fault in one of the key transmission corridors. For temporary faults, the circuit breaker autorecloses and normal operation is restored; otherwise, the faulted section would be taken out. There might be other types of disturbance in the system suchas change of load characteristics, sudden change in power ow etc. which are less severe than faults and are not considered here. To evaluate the performance and robustness of the designed controller, simulations were carried out corresponding to some of the probable fault scenarios in the New England Test System (NETS) and New York Power Systems (NYPS) interconnection. There are three transmission corridors between NETS and NYPS connecting buses 60 and 61, 53 and 54 and 27 and 53, respectively. Each of the corridors connecting buses 60 and 61 and 53 and 54 consists of two tielines and the corridor connecting buses 27 and 53 consists of one tieline. The outage of one of these lines considerably weakens the transfer capacity of the

Fig. 13

Dynamic response of the system

a G1G15 fault at bus 60 with autoreclosing b G1G15 fault at bus 27 followed by opening of one of the lines connecting buses 27 and 43 c G1G15 fault at bus 63 followed by opening of one of the lines connecting buses 63 and 64 d G1G15 fault at bus 60 followed by opening of one of the lines connecting buses 60 and 61
6
IET Gener. Transm. Distrib., Vol. 1, No. 1, January 2007

lies the potentiality of a unied Smith-predictor approach (USP) for power-system-damping control design involving nite delay in signal transmission. 8 Conclusions

This paper has demonstrated the implementation and test results of the unied Smith-predictor technique. Because of structural complexity of such predictor-based control, the validation of the control algorithm in real time is an important issue. A key feature of this paper is the realisation of a multimachine power-system model which is capable of providing results in real time on a commercially available real-time simulator (RTS). The designed control algorithm is implemented in a rapid-prototyping controller and the coupling between RTS and RPC is done through a DAC/ ADC module in the analogue domain. The real-time multimachine model has opened the door to many possible application areas, including the provision of a exible test bench for the development and testing of physical controllers and network compensators. 9
Fig. 14 Dynamic response of the system without considering delay

Acknowledgment

The authors thank Corporate Research Center ABB, Switzerland and EPSRC, UK for funding this research. 10 References

a G1G15 fault at bus 60 with autoreclosing, without control b G1G15 fault at bus 60 with autoreclosing, with control

corridor. The following disturbances were considered for simulation. A three-phase solid fault for 80 ms (5 cycles): (i) at bus 60 followed by autoreclosing of the circuit breaker; (ii) at bus 53 followed by outage of one of the tielines connecting buses 53 and 54; (iii) at bus 53 followed by outage of the tieline connecting buses 27 and 53; and (iv) at bus 60 followed by outage of one of the tielines connecting buses 60 and 61. The response of the relative angular separation of G1 with respect to G15 is shown in Fig. 13. To demonstrate the drawback of the conventional H1 design with a delay-free plant, a separate controller was designed for the TCSC without considering delay in the design stage. The design was carried out as described in [7]. The controller was found to be acceptable both in time and frequency domain for a delay of up to 0.1 s. For larger time delays, the performance began to deteriorate. The simulation results following the same disturbance with a delay of 0.75 s was carried out and are shown in Fig. 14. It is clear that the controller performance deteriorates considerably if the delay is more than the time period corresponding to the dominant interarea modes and is not taken into account in the design stage. Therefore, inclusion of the delay in the control synthesis is signicant and there

1 Pasebra, J.: Analysis and control of power systems oscillation, ! Special Publication 38.01.07, Vol. Technical Brochure 111 CIGRE 2 Kamwa, L., Grondin, R., and Hebert, Y.: Wide-area measurement based stabilizing control of large power systems a decentralized/ hierarchical approach, IEEE Trans. Power Syst., 2001, 16, (1), pp. 136153 3 Heydt, G., Liu, C., Phadke, A., and Vittal, V.: Solutions for the crisis in electric power supply, IEEE Comput. Appl. Power, 2001, 14, (3), pp. 2230 4 Kamwa, I., Heniche, A., Trudel, G., Dobreseu, M., Grondin, R., and Lefebvre, D.: Assessing the technical value of FACTS-based bide area damping control loops. Proc. IEEE PES General Meeting, San Francisco, USA, 2005, pp. 16361645 5 Chaudhuri, B., Majunder, R., and Pal, B.: Wide area measurement based stabilizing control of power systems considering signal transmission delay, IEEE Trans. Power Syst., 2004, 19, (4), pp. 19711979 6 Chaudhuri, B., Pal, B., Zolotas, A.C., Jaimoukha, I.M., and Green, T.C.: Mixed-sensitivity approach to HN control of power system oscillations employing multiple facts devices, IEEE Trans. Power Syst., 2003, 18, (3), pp. 11491156 7 Chaudhuri, B., and Pal, B.: Robust damping of multiple swing modes employing global stabalizing signals with a TCSC, IEEE Trans. Power Syst., 2004, 19, (1), pp. 499506 8 Zhong, Q.C.: HN control of dead time systems based on a transformation, Automation, 2003, 39, pp. 361366 9 Smith, O.: Closer control of loops with dead time, Chem. Eng. Progress, 1957, 53, (5), pp. 217219 10 Dommel, H.W.: Digital computer solution of electromagnetic transient in single and multiphase networks, IEEE Trans. Power Appar. Syst., 1969, 88, (4), pp. 388399 11 Meyer, W.S., et al. (Eds.): EMPT rule book (Bonneville Power Administration, System Engineering Portland, Oregon, 1980, revised 1992) 12 Nayak, O., Irwine, G., and Neufeld, A.: GUI enhances electromagnetic transient simulation tools, IEEE Comput. Appl. Power, 1995, 8, (1), pp. 1722

IET Gener. Transm. Distrib., Vol. 1, No. 1, January 2007