This action might not be possible to undo. Are you sure you want to continue?
J.Machowski J.W.Bialek S.Robak J.R.Bumby
Indexing terms.' Synchronous generators, Excitation control system, Lyapunov's method, Nonlinear system equations
=
Abstract: An optimal excitation control strategy for a synchronous generator is derived using Lyapunov's direct method and the nonlinear system equations. The control strategy requires neither phase compensation nor washout circuits characteristic of standard power system stabilisers (PSS). By using a nonlinear system model, the control strategy is optimal over a wide range of rotor angle and swing frequency changes. The excitation control system required to implement the control strategy is hierarchical and has a different structure to the traditional excitation control system with PSS. In the proposed structure a primary controller damps quickly any power swings using the synchronous EMF as the feedback signal. A secondary controller maintains constant generator voltage by adjusting the reference value of the synchronous EMF fed to the primary controller to that required by the actual operating conditions. The proposed excitation control system has been tested on a single generatorinfinite bus bar system with the simulation results showing excellent damping of power swings over a wide range of operating conditions whilst retaining good voltage control.
=
=
K M =
=
=
= =
Telo, Tqo V
V, Vs
= = =
qaxis component of the synchronous internal EMF proportional to the field winding selfflux linkages (i.e. proportional to the field current itself) excitation voltage d and qaxis components of the transient internal EMF armature current and its d and qaxis components gain of the controller inertia coefficient electromagnetic airgap power and mechanical power supplied by the prime mover opencircuit d and qaxis transient time constants Lyapunov function voltage at the generator terminals and the infinite busbar voltage and qaxis synchronous and transient reactance of the generator
d
List of symbols
A bar on top of a symbol denotes a phasor or a complex number (e.g. V, S) Bold face denotes a matrix or a vector (e.g. Y) A 'hat' on top qf a symbol denotes the final equilibrium value (e.g. 8) E = synchronous internal EMF (i,e. voltage behind the steadystate reactances Xel and x./) Ed = daxis component of the synchronous internal EMF
©TEE,I998 lEE Proceedings online no. 19982182 Paper first received 17th December 1997 and in revised form lst June 1998 J. Machowski and S. Robak are with the Warsaw University of Technology, Instytut Elektroenergetki, Koszykowa 75, 00662 Warsaw, Poland lW. Bialek and lR. Bumby are with the University of Durham, School of Engineering, Science Site, South Road, Durham DHI 3LE, UK
lEE Proc.Gener. Transm. Distrib., Vol. 145, No.5, September 1998
= total d and qaxis synchronous and transient reactance between (and including) the generator and the infinite busbar = reactance of the transformer, transmission line and the system, respectively = rotor speed and the synchronous speed = w  Ws = rotor speed deviation = rotor angle with respect to the synchronous reference axis defined by v" 1 Introduction
Synchronous generators are used almost exclusively in power systems as a source of electrical energy. The generator is supplied with real power from a prime mover, usually a turbine, whilst the excitation current is provided by the excitation system shown schematically in Fig. 1. The excitation voltage E( is supplied from the exciter and is controlled by the automatic voltage regulator (AVR). Its aim is to keep the terminal voltage V equal to the reference value Vre!' Although the AVR is very effective during normal steadystate operation, following a disturbance, the generator is in the transient state and the AVR may
537
have a negative influence on the damping of power swings [IJ. As power swings cause the terminal voltage to oscillate, the reaction of the A VR is to force field current changes in the generator which, under certain conditions, may oppose the rotor damping currents induced by the rotor speed deviation flcu. This socalled negative damping may be eliminated by introducing a supplementary control loop, known as the power system stabiliser (PSS), also shown in Fig. 1. The task of the PSS is to add an additional signal Vrss into the control loop, which compensates for the voltage oscillations and provides a damping component that is in phase with flo).
In this paper a novel approach to the design of the control system is proposed, based on Lyapunov's direct method, Traditionally, Lyapunov's direct method has been used in power system analysis to evaluate stability margins and in realtime dynamic security assessment [7, 8]. These applications require accurate system modelling and careful choice of Lyapunov function to achieve accurate results. Recent work has shown how Lyapunov's direct method can also be used very effectively for deriving control strategies for FACTS devices using a nonlinear model of the power system [1, 9, 10]. In such applications Lyapunov's method is used to point the way to the required control situation, and simpler Lyapunov functions can be tolerated than those necessary for assessing stability margins. A similar approach is followed in this paper where Lyapunov's direct method is used to derive a control strategy for the excitation control of a synchronous generator. 2 Design method
Fig.1
v
Functional diagram of excitation control
Considerable research effort has been devoted to the problem of PSS design. More detailed information and additional references can be found in [13]. Generally, the properties of a particular PSS depend on the choice of the input quantities q, with the most commonly used quantities being: speed deviation I1cu, generator real power P, frequency deviation Sf, the transient EMP E', and the generator current I. As each of these signals has its advantages and disadvantages, the PSS is often designed to operate on a number (usually two) of these input signals [1, 2]. Fig. 2 shows the block diagram of a PSS, which uses real power as the input signal.
VPSSmax
2.1 Design approach using method
ivepunov's direct
_f
Lyapunov's direct method is concerned with assessing the stability of a dynamic system described by a set of nonlinear equations of the form x = F(x). The point is the equilibrium point if it satisfies the equation F(x) = O. Lyapunov's direct method is based on finding a suitable scalar function Vex) defined in the statespace of the dynamic system that is positive definite and has a stationary minimum value at the equilibrium point x (i.e. for any flx .. 0 it holds that I1x) > The point is stable if the derivative )} = dV/dt is negative semidefinite along the trajectory x(t) of equation X = J:<'(x)(i.e. when V :5 0). The point x is asymptotically stable if the derivative V is negative definite along the trajectory x(t) (i.e. when V < 0).
x
x
Vex 
Vex»~.
washout
phase compensation
vPSSmin
Fig.2
Power system stabiliser with real power as input signal
To produce the necessary control signal, the PSS usually has a washout block and one or more phase compensation blocks. The washout block is a highpass filter with a time constant high enough to allow signals associated with the SPeed oscillations to pass through unchanged. The phase compensation blocks provide the appropriate IeadIag characteristic to compensate for the phase shift between the exciter output and the generator airgap torque. As the PSS design is usually based on the linearised system model, the phase compensation blocks typically provide efficient compensation only for a narrow range of swing frequencies. Thus the actual settings of the PSS parameters should depend on the generator load, impedance of .the transmission network and the voltage characteristics of the local loads. As these parameters may vary quite considerably during system operation,. usually a cornpromise
Fig.3
tem
has to be made. Moreover, interactions between generators equipped with PSS may cause a particular PSS to damp oscillations in one area of the system but excite oscillations in another area or excite intcrarca oscillations. Many research centres continue their efforts towards developing improved PSS designs. Among the approaches used, it is worth mentioning attempts to usc adaptive systems [4] or artificial intelligence tools [5,6].
538
A negative value of the derivative V means that function Vex) decreases with time tending towards its minimum value. As the minimum value of Vex) is at the equilibrium point, the trajectory x(t) tends towards the equilibrium point X. This is illustrated in Fig. 3 where point Xo = x(t = OT) denotes a nonzero initial condition lying beyond the equilibrium point X. It is important to note that the higher negative value of V, the faster the trajectory xCI) tends towards the equilibrium point x.
lEE Proc=Gener. Trunsm. Distrib, Va!. 145, No.5, September 1998
Consequently, a given control strategy is optimal in the Lyapunov sense if it maximises the negative value of V at each instant of time, With these observations it is now possible to define a design approach that comprises three stages: (i) find an appropriate Lyapunov function Vex) for the system that is an explicit function of the control variables, (ii) select a control structure that maximises the negative value of V at all points along the trajectory, (iii) select such locally available signals which can be used to execute the chosen control structure. This last point is important as quite often in PSS design a control structure is proposed which relies on measuring the rotor angle {) in real time. However, measuring the rotor angle with respect to some synchronously rotating reference is not easy in a multimachine system. 2.2 Lyapunov function for generatorinfinite busbar system Figs. 4 and 5 show a generatorinfinite busbar system and the corresponding phasor diagram. For simplicity, only the series reactance of all the modelled elements is considered. The generator is assumed to have both steadystate and transient saliency (i.e. Xq ¢ Xd and X'q ¢ X'd) and is represented by the fourthorder model detailed in the Appendices, Section 8.1 [1]. The transformer reactance XT, the line reactance XL and the system equivalent reactance Xs have been added together to form the equivalent transmission reactance X = XT + XL + Xs so that the total daxis reactances are Xd = X + Xd and X'd = X + X'd, while the total qaxis reactances are Xq = X + Xq and X'q = X + X'q.
where Vk is a kinetic energy term, Vp is a potential energy term, and VE'd and VE'q are two additional terms to account for flux decrement effects on the d and q axes. All four terms are defined in the Appendixes, Section 8.2, eqns. 28, 29, 30 and 31. Section 8.2, eqn. 45, shows that the derivative of the Lyapunov function given by eqn. 1 can be expressed as:
V = Vk + Vp + VE, + VE,
i1 .
=fj[ejej1
1
dE' 1 [dE'] rl/(J d/
d
2
::y1 [dE'] d/
2
(2) where e( = EITdu and {3, yare the parameters defined in Section 8.1, eqn. 22. Lyapunov's stability theorem states that the post fault system is stable, provided that the time derivative of the Lyapunov function along the system trajectory is negative semidefinite (i.e. V ,;; 0). Assuming constant excitation voltage (Ef = constant, i.e. AVR switched off), ef = (if and the first component of eqn. 2 vanishes giving:
V =  [~
(d!~
r
+~(
d!~
r]
<0
(3)
sient state
Fig.4
Schematic
0@+i Hi e
L diagram of generatorinfinite
Thus the derivative V is negative semidefinite and the system is stable. Whether or not V ,;;0 and the system is stable when the excitation voltage Ef is not constant (i.e. when the reaction of the AVR to the disturbance is taken into account) depends on the chosen control strategy. Having obtained a suitable Lyapunov function in eqn. 1, the next stage of the design process is to devise the necessary control strategy that not only ensures that eqn. 2 is negative definite but also maximises the negative value of V at all instants. 3
Control strategy
busbar system
in tran
The control strategy is based on the observation illustrated in Fig. 3 that the higher the negative value of the derivative V, the faster the system trajectory tends towards the equilibrium point. 3.7 Maximisation of (1/) by controlling E, Eqn. 2 shows that the derivative V depends on what is happening in both the direct and quadrature axes. This may be expressed by rearranging eqn. 2 as: (4) Section 8.2, eqns. 49 and 53, show that: , lid
= ,
i~
1
,
1
Xq
:~'
I
I
r jTdx'd
Tqo Xq
(Ed)
2
<0
(5)
I
i~d
state
E'd
'd
busbar system in transient
(6)
Fig.5
Phasor diagram of generatorinfinite
The Lyapunov function used in this paper comprises four terms: (1)
fEE Proc.Gener. Transm. Distrib., Vol. 145. No.5, September 1998
V d is always negative semidefinite, while V q may be positive or negative depending on the excitation control strategy chosen. Note that Vd does not depend on the excitation control as there is no EMF induced in the d axis due to the rotor excitation current. In other words, V d corresponds to the natural damping produced by currents induced in the rotor q axis.
539
Now, let us analyse Va' The first component of eqn. 6 is always negative and, similarly as in eqn. 5, corresponds to the natural damping produced by currents induced in the rotor d axis. The second component of eqn. 6 is proportional to the change in both the excitation voltage and the synchronous EMF. The control strategy should therefore, aim to make this component as highly negative as possible at each instant of time during the transient state. This can be achieved by applying the following excitation control strategy:
postfault equilibrium pomt and, importantly, Eq defines the postfault steadystate synchronous EMF. As this EMF depends on the required terminal voltage, it is natural to assume that the generator terminal voltage is regulated by a proper setting of = Eq ref' This can be achieved if the A VR acts with a classical voltage feedback control loop. The resulting excitation control system is shown in Fig. 7. Because Et is the postfault s~eadystate value of Ef, it is assumed that Ef  Jiq Eqrcf
e,
__c
(EjEf)=K(EqEq)
where K» 1 is the controller strategy eqn. 6 becomes:
(7)
gain. With this control
. Vq
= ,1
1
I
TdoXdxd
(K
+ 1) ("  Eq )2 ::; 0 (8) Eq
.
and the Lyapunov function V decreases at a rate equal to the sum of eqns. 5 and 8, that is:
.lm f
limiter i>i~' 1+'& E1 E
1
V
=
Vq
+ Vd
1 K +1 ( ~)2 1 1 +f! x x' (Ed) [T' X x' EqEq
do d d qo q q
Fig.7
Block diagram a/proposed
excitation
controlier
=
2l
(9)
sO
This derivative is independent of the network reactance X as xd = Xd + X and = X'd + X so that xd  X'd = Xd  X'd' Similarly, Xq  X'q = Xq  X'q. This means that the control strategy defined by eqn. 7 assures the same positive damping regardless of the values of the transmission network parameters. This is an important feature of the proposed excitation control strategy.
3.2
Structure of the control system
strategy defined by eqn. 7 can be rewritten (10)
The control as:
Fig. 6 shows the block diagram of the controller required to execute this control strategy. The controller ensures a fast return of the system trajectory to the equilibrium point, which is equivalent to strong damping of the power swings. It is important to note that the controller contains neither a phase compensation block nor a washout block typical of traditional PSS systems. This has been achieved by using Eq as the feedback signal rather than other quantities (such as frequency, real power, power angle etc.) used traditionally in PSS.
In the control system shown in Fig. 7 the output signal is the excitation voltage E; It is assumed that the exciter contains no significant time lag (as would be the case for the static exciter). If the exciter has a significant time constant (e.g. it contains cascaded DC generators), an additional control block that compensates for this must be connected to the regulator output This is necessary as the control strategy defined by eqn. 10 requires changes in the excitation voltage Ef (proportional to the changes in the synchronous EMF Eq) to be made without any delay. The feedback signal used in the control strategy given by eqn. 10 is Eq, the quadrature componeruof the synchronous EMF. By definition, Eq is proportional to the field current if so that field current can be used as a feedback signal instead. However, it can be shown [1] that a disturbance causes also a fundamental frequency component to appear in &" Thus, to obtain a signal proportional to Eq, it is necessary to filter out the fast 50Hz (or 60Hz) component from the field current. This can be done by using a lowpass filter (e.g. of the Bessel type).
3.3
Discontinuous excitation control
Fig.6
Block diagram a/proposed
stabilising controller
It is now necessary to link the stabilising controller, shown in Fig. 6, with the main, steadystate, voltage controller (A VR). Within Lyapunov's stability theorem the equilibrium values of Ef and Eq correspond to the
540
The control strategy defined by eqn. 10 has been derived using Lyapunov's direct method. However, Lyapunov's stability theory is valid for autonomous system only (i.e. when the system is timeinvariant and function F(x) is not explicit with time [7, 8]). Consequently the system response is due to some nonzero 'initial condition' Xo ;0' x and no further disturbances are considered. As the system trajectory tends from the initial condition towards the final equilibrium point, the control strategy defined by eqn. 10 maximises the speed with which function V decreases so that the system returns to the equilibrium point as quickly as possible. In the case of a power system, the nonzero initial condition may be caused by a shortcircuit, tripping of a network element etc, It is important to appreciate that the derived control strategy is optimal after the disturbance has taken place but may not be at all optimal during the disturbance itself. Consider, for example, a shortcircuit. A shortcircuit causes the generator synchronous EMF to increase suddenly [1], and the controller shown in Fig, 7 will try to reduce the excitalEE Proc.Oener. Transm. Distrib.
Vol.
145, No.5, September 1998
tion voltage. Unfortunately, what is really required during this period is an increase in the excitation voltage [1]. Some traditional PSS systems react in a similar way and, to prevent this, traditional AVR + PSS systems arc often equipped with an additional discontinuous excitation control circuit [2]. This circuit bypasses the PSS during the shortcircuit forcing an increase in the excitation voltage up to its ceiling value. A similar solution can be used in the proposed controller. Fig. 7 shows an additional relay SHC, which disconnects the Eq signal during a shortcircuit, thereby causing a large positive regulation error !!Eq = Eq• This error forces the excitation voltage to reach its ceiling value quickly. The relay SHC closes again when the speed deviation I1w reverses its sign (i.e. during the back swing).
4 Hierarchical control
The structure of the proposed control system shown in Fig. 7 is different to the traditional structure of AVR + PSS system shown ir, Fig. 1. In the traditional solution, the AVR is the main controller, while the PSS is a supplementary control loop. The structure of the proposed controller is hierarchical and of the masterslave type. The slave controller (primary level) is the stabiliser based on the derived control strategy. Its task is to damp quickly any power swings, ensuring that the system reaches the steadystate as quickly as possible. The master controller (secondary level) is the AYR. Its task is to set the reference value Eq ref for the PSS. The referepee voltage Vref for the AVR can be set manually or automatically from a tertiary level. The task of this tertiary level control is to maintain the optimal system voltage profile as defined by the economic dispatch of reactive power. The overall structure of the proposed hierarchical AVR + PSS system, including the tertiary control level, is shown in Fig. 8.
This hierarchical excitation control scheme is very similar to the hierarchical structure used in automatic generation control (AGC) [1, 2]. In hierarchical AGC the primary controller is the turbine governor, Its task is to control the turbine speed and the mechanical power Pm by acting on the control valves. The reference power is set by the secondary controller, whose task is to control frequency and tieline power flows. The tertiary AGC controller controls the setting of the primary controller to maintain the economic dispatch of generation. Table I compares the hierarchical structures of the AGC and the proposed overall generator control system. Comparison between the generator and the turbine control is quite revealing. In both cases the primary control is concerned with maintaining the stability of the turbinegenerator unit, the secondary control is concerned with how the unit impacts the network, while the tertiary control is concerned with the economic dispatch. The aim of the primary control to maintain the stability of the generator and the turbine. For the generator, the proposed stabiliser controls Eq to damp quickly power swings which threaten the generator stability. For the turbine, the turbine governor controls the mechanical power Pm as a response to the frequency changes, with the droop characteristic of the turbine governor ensuring stable operation [1]. In both cases the controlled variable (E for the generator and Pm for the turbine) is controlled indirectly by regulating the control variables, the excitation voltage Ef and the valve position cv, respectively. The secondary level is concerned with how the turbinegenerator unit impacts the network. The aim of control is to keep constant the network quantities (i.e. the generator terminal voltage V and the frequency f). These network quantities arc controlled indirectly by setting the reference values for the primary controlled quantities: Eq ref for the generator and Pm ref for the turbine. The tertiary level is concerned with the economic dispatch. The economic dispatch of reactive power is achieved by setting the reference value at the secondary level (i.e. the voltage at the generator busbars Vrej). The economic dispatch of real power is achieved by setting the reference value at the primary level Pm ref 5
Simulation results
Fig.S tem
Proposed
excitation
controller shown as hierarchical control 'ys
Simulation results have been obtained for a generatorinfinite busbar system using the data detailed in Appendix 10. When deriving the control strategy given
Table 1: Comparison of hierarchical control of generator and turbine
Control level Primary Goal of control maintain stability of the unit by controlling: by regulating Secondary maintain reference value of the network quantity: by regulating the reference
value of:
Generator synchronous EMF Eq excitation voltage Ef terminal voltage V the primary level Eq rot reactive power the secondary level Vrof
Turbine mechanical power Pm valve position Cv frequency f the primary level Pm ref real power the primary level Pm ref
Tertiary
enforce economic dispatch of: by regulating the reference value of:
lEE Proc.Gener. Transm. Disrib .• Vol. 145. No.5. September 1998
541
by eqn, 10, the fourthorder generator model was used with all the system resistances neglected. For the simulation, the sixthorder generator model has been used, thereby including subtransient effects [1]. Generator and network resistances have also been included. Also included in the generator model is a thyristorcontrolled static exciter, having both a buck and boost facility with a ceiling voltage of ±8pu. As the time constant of the firing circuits is very short, this exciter is simply modelled by a constant gain and its excitation limits. The exciter is also assumed to be supplied from a compensated supply so that its excitation function is not impaired during the fault period. The block diagram of the proposed controller is shown in Fig. 7 with the gain of the primary level stabilising controller set at K = 15. The secondary level AVR controller is responsible for setting the reference value Eq ref and in the simulations a PIDbased secondary regulator has been used. The parameters of this controller have been selected to ensure an almost aperiodic settling of the generator voltage to a new steadystate value. As the design principles involved in selecting the parameters for this new control structure are very different to those used in traditionalA VR + PSS systems, the principles of optimising the parameters of this secondary regulator will be the subject of a future paper.
150
1.2
1.0 :i
Q_
~
08
0.6
0
I
2
t, S
4
6
10
o
2
t,s
4
6
100 ::i
Fig. 10 tion
Response of sample system to nonzero initiai rotor angle deviaconstant lif_ standard rss proposed controller
ci 2:
co 50
...
150~
0
2 t,
S
4
6
1,5
::i o,
£
1\ 50
""
::i d,
~
0.5
.
o
, i,
V,
I,
1.5
0
2 t.s
4
6
2
t,
4
S
6
Fig.9
constant E,
standard
Response of sample system to nonzero initial rotor angle deviation
1'00
proposed controller
To study the performance of the proposed controller, simulations have been performed that cover a wide range of possible generator loading conditions. Comparisons have also been made with a conventional exciter equipped with the standard PSS shown in Fig. 2. The values of the PSS parameters used are given in Appendix 10. This PSS is tuned to give 'good' performance over a wide range of operating conditions,
542
t, S
6 in network
Fig" 11
~fa~J~~~ ~SS proposed controller
Response of sample system to shortcircuit
lEE Proc.Gener. Transm. Distrib., Vol. 145.. No. 5, September 1998
rather than being tightly tuned to one specific operating condition. In all the tests performed, the proposed excitation control system produced fast damping of power swings and small oscillations in the terminal voltage following the disturbance. For illustration, Figs. 9 and 10 show the effect of a nonzero initial rotor angle deviation, whilst Figs. II and 12 show the controller performance following a shortcircuit in the
1.2
::i ci
~
2 t,
S
4
6
10
5
::i ci.
..i
.,
  ...  1
I I
I.,!
,\
.:;f
0
,rv
J
5
10
0
2
t, S
4
6

Fig.12

~a~d~~~~s proposed controller
Response of sample system to shortcircuit in network
network. The dashed lines correspond to the case when the excitation voltage was held constant (AVR switched off), the solid lines to the proposed controller and the dotted lines to the standard PSS. For the initial nonzero rotor angle deviation, Figs. 9 and 10, the proposed controller brings the system into the steadystate after only one swing with almost no backswing. For the shortcircuit in the network, Figs. 10 and 11, the situation is similar. During the shortcircuit, the discontinuous control (switch SHe in Fig. 6) forces the excitation voltage up to its ceiling. After fault clearing, the proposed controller takes over bringing the system very quickly to the steadystate, again in an almost aperiodic way. The backswing of power, angle and voltage is almost negligible. In both cases the postfault performance of the proposed controller is substantially better than the standard PSS. The worst performance aspect of the proposed controller is the initial recovery of the terminal voltage following the fault, Figs. 10 and 12, and, in this respect, the performance of the conventional AVR with standard PSS is slightly better. However, this voltage recovery depends primarily on the control action taken during the fault itself. As explained in Section 3.3, the proposed controller is optimal only after the fault has been removed. What action the controller should take during the fault period is a different question, and the introduction of the relay SHe in Fig. 7 was a first step in improving this part of the control action. Note that, although the standard PSS brings the real power to its steadysteady value relatively quickly, the regulation of the rotor angle and the voltage takes longer and there is still a small error after 6 seconds. This is due to using real power as the only input signal for the PSS. By comparison, the proposed controller brings all three signals quickly to their steadystate values. To further quantify the benefits of the proposed controller, simulations have heen performed over a range of inductive and capacitive loading conditions. Table 2 shows the results of integrating the absolute value of error in power angle, real power and terminal voltage
in the network
Integral of deviation of: voltage
Table 2: Simulation
Prefault loading generator p= Q=
results for a shortcircuit
Type of AVR + PSS system conventional proposed
of the
Transmission line 50km
r;
s
rotor angle
real power p 0.573 0.315 0.564 0.222 0.328 0.211 0.308 0.193 0.336 0.193 0.319 0.173 0.306 0.149 0.295 0.136
V
0.122 0.108 0.126 0.162 0.113 0.079 0.095 0.082 0.114 0.083 0.102 0.089 0.150 0.095 0.125 0.101 543
o,
0.434 0.240 0.538 0.349 0.310 0.150 0.293 0.184 0.344 0.157 0.330 0.183 0.605 0.206 0.444 0.205
100km
conventional proposed
P= 0.7Pn Q= 0.7Qn
50km
conventional proposed
100km
conventional proposed
P= 0.7Pn Q = 0.45Qn
50km
conventional proposed
100km
conventional proposed
P= 0.6Pn Q=O.lQn
50km
conventional proposed
100km
conventional proposed
lEE Proc.Gener. Transm. Disuib., Vol. 145. No.5. September 1998
over a 5 second period for two transmission line lengths of 50km and 100km (the base case). In all cases the lower value of integrated error indicates the better performance. In this Table the loading condition P = Pm Q :=: Qm and 100 km long transmission line correspond to Figs. 11 and 12. In all cases, the proposed controller outperforms the conventional A VR ·1 PSS except for the voltage error corresponding to Figs. 10 and 12. As explained above, it is expected that adjustments. to the control during the fault period will improve this situation. 6 Conclusions
10 MACHOWSKI, J, and NELLES, D.: 'Optimal modulation controller for superconducting magnetic energy storage', Int. I Elect. Power Energy Syst., 1994, 16, (5), pp. 291300 11 AYLETT, P.D.: 'The energy integral function of transient stability limits of power system', Proc. lEE, 1958, 105, (2) 12 GLESS, G.B.: 'Direct method of Liapunov applied to transient power system stability', IEEE Trans., 1966. PASS5, (2) 13 KAKIMOTO, N., OHSAWA, Y., and IIAYASHI, M.: 'Transient stability analysis of multimachine power systems with field flux decays via Lyapunov's direct method', IEEE Trans., 1980, PAS99, (5), pp. 18191827
8
Appendices
In this paper an optimal excitation control strategy has been derived for the nonlinear generatorinfinite busbar system. By using the Lyapunov's direct method in conjunction with an energytype Lyapunov function, the optimal control strategy has been derived, which maximises the speed with which Lyapunov function decreases, thus maximising energy dissipation in the system. The resulting proportional controller dispenses with the need for leadlag correcting circuits or washout circuits characteristic of the standard PSS system. By using the nonlinear system model, the control strategy is optimal over a wide range of rotor angle and swing frequency changes. It also achieves the same positive damping, regardless of the values of the transmission network parameters. The resulting overall structure of the excitation control system is completely different to that of the standard A VR + PSS system. In the standard systems the main voltage controller is the AVR and the damping of power swings is achieved by a supplementary control loop (PSS). In this paper a masterslave structure for the excitation control system is proposed, where the damping of power swings is achieved by the fast slave controller controlling the generator synchronous EMF. The master voltage controller is much slower and its task is to set the reference value of the synchronous EMF. Simulation results on a singlemachineinfinitebusbar system have shown that the proposed excitation control is very effective in damping power swings caused by a variety of disturbances in the generatorinfinite bus bar system. 7 References
The system model is that of the singlegeneratorinfinite busbar shown in Fig. 4, with a fourthorder generator model being used to formulate the Lyapunov function. Full mathematical details of this model can be found in [1] and only those equations used in the evaluation of the J .yapunov function are quoted here. The model takes into account transient effects in both d and q axes with the differential equations determining the generator behaviour heing:
8. 1 System model
d5
dt
= 6w
r;  r.
q
(11 )
( 12)
dDoW 1'vl Tt=
, dE~ T, =E,E J ,10 dt T' dE~ =E
qo
(13)
dt
d
(14)
where
(15) x :r  x' Erl=__.'i_E'+~V~sin5
x~ a x~ and
(16)
r, = PE' (5)
E~Vs = _'_ sin 5 + EI V, _d_, x~
cos
5
_s_ ~~q__
V2 2
a'.

sin
x~x~
25
(17)
The coordinates of the equilibrium point are denoted by using a 'hat' on top of a variable so that: Do~
MACHOWSKI, J., BIALEK, J.W., and BUMBY, J.R.: 'Power system dynamics and stability' (John Wiley, Chichester, 1997) 2 KUNDUR, P .. 'Power system stability and control' (McGraw Hill, 1994) Task Force 07 of Advisory Group 01 of Study Committee 38: 'Analysis and control of power system oscillations'. CIGRE 1996 4 CHEN, G.P., MALIK, OP, HOPE, GS, QTN,YR, and XU, G.Y.: 'An adaptive power system stabiliser based on the selfoptimising pole shifting control strategy', IEEE Trans., 1993, EC8, (4), pp. 6JY645 ELMETWALLY, K.A., and M~LIK, O.P.: 'Fuzzy logic power 6
~?stem sta~11iser' Lb.E Proc., Oener. Tr ansrn. Distrib., 1995, 142, > (3), pp. 2T!281 ABDELMAGID, Y.L., BETTAYEB, .'Vr., and DAWOUD,
=
0;
Pe(b) = Pm.;
Eq = Ef;
Ed =
0 (18)
Dividing both sides of eqns. 13 and 14 by the corresponding time constants Tdo and T'qo and substituting Ed and Eq given by eqns. 15 and 16 yields:
O;E1 [ x' rl dE' __ d _
dt where
q

,OlT s cos u v
V~ SIn u s . 5:J
sJ
(19) (20)
9
M.M.: 'Simultaneous stabilisation of power systems using genetic algorithms', lEE hoc. Gener. Transm. Distrib., 1997,144, (I), pp. 3944 PAl, ..'\II.A.:'power system stability, Analysis by the direct method of Lyapunov' (North Holland Publishing Company, Amsterdam, 1981) PAVELLA, M., and MURTHY, P.G.: 'Transient stability of power systems, theory and practice' (John Wiley, 1994) MACHOWSKI, r., and NELLES, D.: 'Power system transient stability enhancement by optimal control of static VAR compensators', Int. J. Elect. Power Energy Svst., 1992, 14, (5), pp, 411421
l
qo
E
x~
E"'d
+ ry
x~
(21 )
Xci.
E
0;=
~
I Tdo '
= T' )
j
544
16'6' Proc.Gener. Transm: Distrib., Vol. 14), No.
September 1998
Eqns. 19 and 20 together with eqns. 11 and 12 form the nonlinear system state model where the state variables are (/lw, 0, E'q, E'd)' At the equilibrium point, both sides of eqns, 19 and 20 are equal to zero:
0=
er [!!:...E' q _l_Vs x~ x~
E' [I "y
Xq Xq
cosJ]
,]
(23) (24)
o= ~
E~ + I Vssin 6
when adding eqns. 19 and 23, and eqns. 20 and 24, gives:
dE' ___J£ dt
~ :, [E~~ E~]
q
~ _1_ Vs [sin 6 ~ sin (26) x~ These equations are equivalent to eqns. 19 and 20 but additional values corresponding to the equilibrium point have been introduced. This formulation will be useful when applying Lyapunov's stability theory. 8.2 Development of the Lyapunov function If flux decrement effects are neglected, the system model is simplified to the classical secondorder model in which the generator is modelled by the swing eqns. II and 12 and a constant transient EMF E' behind the transient reactance X'd' An energybased Lyapunov function for such a system was derived in [11, 12] and corresponds to the total system energy made up from the sum of the system kinetic and potential energy with respect to the equilibrium point. For the thirdorder model (i.e. when eqn. 14 is neglected), a Lyapunov function was derived in [13], where the additional component added to the total system energy is proportional to (E'q  E' )2. In this paper the fourthorder model is used and the Lyapunov function requires a further component proportional to (E'd  E'd)2 giving the proposed Lyapunov function as; V where
t
8]
to the kinetic and potential energy of the system at any point (l'lw, 8) calculated with respect to the equilibrium point (Aw == 0, 8) with their sum being the wellknown Lyapunov function for the classical system model [11, 12]. The two additional components given by eqns. 30 and 31 take into account the changes in the transient EMF in both the d and q axes. To use Lyapunov's stability theorem it is necessary to: (i) prove that function V defined by eqn. 27 is positive definite in the vicinity of the equilibrium point; (ii) derive a control strategy which makes the time derivative V negative semidefinite (V ::$ 0) along the system trajectory, At the equilibrium point, the righthand sides of eqns. 28,29, 30 and 31 are all equal to zero. The components Vk, V£, and V£'d are all positive definite as they are proportio~a1 to the state variables squared. To prove that the remaining component, Vp given by eqn. 29, has a minimum at the equilibrium point 8 note that as WpffJ/j = (Pm  Pe) and Pm == PeCD), fJVpffJt'i :::: 0 at 8 ::::8. Whether this is a maximum or minimum depends on the sign of the second derivative:
respect to the equilibrium point (t :::: 0, 8, 0 0, E'q, E'd)' The first two components, Vk and Vp, correspond
l:.w ::::
(32)
which is equal to the generator synchronising power coefficient. The synchronising power coefficient must be positive below the maximum, pullout, value of the powerangle characteristic, otherwise the steadystate stability conditions are not met. This proves that Vp has a minimum value equal to zero at D :::: D and is positive definite around it Consequently, the function defined by eqn. 27 is positive definite around the equilibrium point, thereby satisfying the first condition of the Lyapunov stability theorem. The time derivative V is equal to the sum of the time derivatives of the components given by eqns. 28, 29, 30 and 31, each of which can be calculated using:
V=
=
dV dt
=
ax!
8V dx! dt
+
av dx.
aX2 dt
T
+ ' .. +
8V dXn 8Xn dt (33)
[gradVf x
=
[gradV1 F(x)
= t\ + Vp + VE~ + VE~
.a.w
/ Md(llw)
(27)
The derivative of eqn. 28 is:
V=
Vk
=
JM(d~W)dt=
00
= ~M(llW)2
(28)
(29)
k
aVk d~w = M.6wd~w 8.6w dt dt
(34)
.a.w=o
8
d~W] .6w:::::+ [Pm  Pel.6w == [Mill
where eqn. 12 has been used in the last transformation. For eqn. 29 the use of eqn. 33 yields: . avp d6 Vp == a6 dt
Vp
=  / [Pm ~ Pe(6)1 d6
,\
+ aE~ dt + aE~ dt
avp dE~
avp dE~
( 35)
(30) VE'
d
eqn. 17 is used to obtain the partial derivatives:
1,,"1[ =: ':::',
2, x~
,]2 E~ ~ E~
(36) (31)
In these equations the values taken at any point along the trajectory (t, 8, l'lw, E'q, E'd) are calculated with
lEE Proc.Gener. Transm. Distrtb .• Vol. 145, No.5. September 1998
av ~ = +
aE~
p
6 /
sr, ~d6
8
aE~
= ~Vs
x~
[_ ,coso  cos6,]
(37)
545
8~ ,
ee:a
=
which substituted
.'
+ ,, do aE.' o
5
J8~,
5
=
+ s",ino  sino A] x'
q
~[
d
(38)
.Vd _ 1 ldE~J '/ dt
2
(48)
dR/dt
into eqn. 35 gives:
~0) 
Substituting y from eqn. 22 and the derivative from eqn. 14 into eqn. 48 gives:
~ [Pdt  Pel
Vs [ ,] dE~ x' cos 6  cos <5 dt
Vd
= _~
__ 1_,_,
Xq
(Ed)2
v, [sin e + ____':.
x~
dE~  sin d dt
~]
Tqc Xq 
s: 0
(49)
(39)
given (40), \
Following a similar procedure by eqns. 30 and 31 gives:
for the functions
V/
Eq
=
avp;~ dE~ a£~ dt
= ~2_[E'
;3
_ j;;1]
q q
dE~
dt
.This term correspondsentirely to the natural damping produced by electromagnetic effects in the rotor quadrature axis. In contrast V g' eqn. 47, depends explicitly on the excitation control via the term ref  J1 and implicitly via the derivative dE'/dt. Substituting values of ef and f3 from eqns. 21 and 22 into eqn. 47 gives:
e
" aVH~ dE,.~ e dE~ VE, ==E,E (41) d [)E~ dt r x~ '" d ell To further transform the components of'eqn. 40, an auxiliary equation is derived by multiplying hoth sides of eqn. 25 by liP dE'q/dt to give:
1[,
~!]
Vq= 
~dO
Xd
~
xd
rT~od!~]rT~oel~~
0=

(EfEf)]
(50)
At the equilibrium rewritten as:
point dE'qldt = 0 and cqn, 13 can be ffrom eqn. 13 gives:
73 ' d/)
1 (' dE' \
2,
= j3[ej ef]d/
Vs +
1
dE'
1 . ()c_ x~ [FY'q f3
"E/]
xa
[cos e  coso
~l
q
dE~ dt
E
Eq
(51)
which, when subtracted
dE~ ' dt
dE~ Tz! ._ dt
(0
=
,(.  E. E. J
J
c:)
for

(E
q
 Eq
A
')
(52) eqn. 50
(42)
is identical
As the second component in this equation to the righthand side of eqn: 40:
Substituting this finally gives: 1
equation 1
dE'/dt
into
VE~ == 73 [ej ~
.
1
ill  fj dt·
Q '
dE, ~
1 (dE~)
2
T~O »«  x~
dl A similar procedure is used to transform eqn. 41 but now the auxiliary equation is obtained by, multiplying both sides of eqn. 26 by l/y dE'dldt to finally give:
v, +xd
[cos 5 ~ cos O'..] dE'
(43)
+ ,~,a,~,x
0
(
Ef  Ef)
(Eq  Eg)
(53)
8.3
Simulation data
,',E~ v
1 elE~ =  :;.( '(it
')2   [.sm 5  "sin :0] ill Vs elE~
U
(44)
All the reactances and resistances are in p.u. and the time constants in seconds. For per unit calculations Sbase= 4261fVA was assumed, System: Shortcircuit MVA equal to ] 5000 MV A at 220kV giving the equivalent system impedance
Importantly, the derivative Vp expressed by eqn. 39 contains identical components as the derivatives expressed by eqns. 34, 43 and 44. This is due to there being an exchange of energy during transient state between the potential energy Vp, the kinetic energy Vlo and the remaining components VBa, VE'd' Summing all four components gives: .
23 '' 0.0024 + )0.024
Transmission line: length 100km, ZL ~ 0.0384 + )0.285 Stepup transformer: Zr  0.0006 + )0.12
S; = 426 MVA,
Vn = 22/250kV, cosrpn
X"d
T'qO
V = Vii; + Vp + VE' + VE'
q
d
Generator: 426MYA, Vn 22kY, ~n = 6.45 s, M = Tm1ws = 0.0205pu
2 _~.
s, 
0.85, 0.235, 1.095,
=~,ler _
;..i'
e,] elE~ _ ~[dE,~] eli. ,8 elt .
J
')'
[dE, ~] dt
2
R = 0.0016,
TdO
= 9.2,
T"dO
Xci = 2.6, = 0.042
X'd X"q
0.33, 0.29,
~5)
Xq
and the system is stable provided that rn eqn. 45 i' s; O. This depends on the signofthe first term in eqn. 45. . Eqn. 45 can be further. transformed by analysing separately the 4. andqaxes i.e. by lIsrng: (46) where . 1 [dE~ elt
=
2.48,
X'q
=
0.53,
T"qO = 0.065
Vq
=
73
12 1 J + 73 [Ef

q
dE'
dt
(47)
PSS: Tp = 0.03s, T, = 5s, T[ = 0.05s, T2 = 7s, Kpss = 120, VPSSmin = 1.75, VPSSmax = 1.75 Predisturbance generator loading was assumed to be: S= 426MVA, V = l.OS, cosrp = 0.85, all measured at the uppervoltage terminal of the transformer. Shortcircuit clearing time was assumed to be 150ms. The discontinuous relay SHe was assumed to open at the instant of fault and close after 180m8.
546
1};H ProcGener. 'lransm. Distrib .. Vol. 145, No.5, September 1998
This action might not be possible to undo. Are you sure you want to continue?
We've moved you to where you read on your other device.
Get the full title to continue reading from where you left off, or restart the preview.