You are on page 1of 6

Voltage InstabiIity/Collapse An Overview

Rodolfo J. Koessler Power Technologies, Inc. Schenectady, New York


Abstract
Voltage instability and voltage collapse are a frequent concern on heavily stressed power systems. This paper contains an overview of what, in the author's view, are the main aspects to be considered when addressing such problems. Also included in the paper is a review of some of the analytical tools that are available for voltage stability analysis, and a discussion of the fundamental physics behind the phenomena.

It is clear from Figure 2 that although it is true that voltage reductions will indeed lead to an initial increase in reactive imports, a point is always reached where further reductions in voltage will not only fail to increase reactive imports but, instead, will lower the reactive support to the load.
This is because, whereas reactive power imports are linearly proportional to voltage gradient, the relation between reactive losses and current is quadratic and, consequently, so is their relationship with voltage. A point is therefore reached where losses overwhelm voltage gradient induced imports, thus setting a limit on the reactive power support that transmission can provide a system, at whatever voltage the receiving end operates, and independently of the resources that are available at the sending-end of the system. This limit decreases as the MW loading of the transmission increases, potentially leading to transmission becoming a consumer, rather than a supplier, of reactive power.
OUT

The Problem
The cause behind most voltage stability problems is a lack of sufficient reactive power resources to satisfy the reactive component of loads and losses.

Figure 1 Insufficient Reactive Resources


It is clear, for example, that in an isolated system deficient in reactive power sources, such as that on the left-hand side 01 Figure 1, there is no steady-state solution, whatever the system operating voltage.

El 3
c
U

vr

ne

L-zA;
VOiiAE

NO LOAD

WE

c
U
W

<

[L

Most modern power systems, however, are interconnected (riight-hand side of Figure 1) , and serve their loads by a combination of local resources (generators, shurit capacitors, SVCs, etc.) and imported reactive power. There is a widespread conception, however, that there is alwavs a low enough voltage, at which the necessary reactive power will flow into such a system. This i:j not true. Consider for example, the case in Figure 2, where a radial system is fed over a long transmisssion line. The figure shows net reactive power flowing from the line to the load, as a function of load voltage, ancl for different MW flow levels.

Figure 2 Reactive Transfer Capability over a Long Transmission Line


This explains why voltage problems are common in large urban centers, with significant imports from lower cost, cleaner generation sources. It also characterizes voltage problems as potentially local, as opposed to the typical widespread nature of angular/frequency dynamics.

Complexities
Although the above description addresses the fundamentals of most voltage stability problems, their effective analysis and solution constitutes a significant challenge to many planners and operators. This is because of, among other reasons:
0

ways of analyzing the same phenomena. Operating conditions in the VP plane may be "transformed" into the QV plane, and vice versa. See, for example, points A, B and C in both figures. The VP curves (Figure 3) are the result of a series of load flow solutions with gradually increasing power transfers. V is the voltage at a critical bus, P is the power transfer across a specified interface. There is a transfer level beyond which the power solution fails, shown in Figure 3 by an increase in solution mismatches. This is the "knee" of the VP Curves. Most utilities plan the network operating point to the left of this knee following disturbances. A stable operating point is usually determined by requiring a large enough "power margin" between the operating point and the point of instability.

Dimensionality. Whereas anguladfrequency dynamics are normally limited to those of power plants, any bus in the system, from bulk transmission to distribution, is a potential candidate for voltage instability. Local Nature. Voltage instability can occur in the midst of an otherwise perfectly healthy system. Contingencies that are critical to voltage stability may not be the same than those critical to angular stability. Data Requirements. In-depth analysis of voltage instability may require information not normally available to the system planner. This includes: overexcitation limiter characteristics, composition of the load (motors, thermostat heating, etc.), on-load tap changer (OLTC) characteristics, location of shunt compensation relative to OLTCs, etc. Understanding and Application of Analytical Tools. The myriad techniques available for voltage stability analysis (VP and QV Curves, Optimal Power Flow, Modal Analysis, and LongTerm Dynamic Simulation, to name a few) can become quite overwhelming to the analyst, who may lose sight of the underlying nature of the problem .

"04
1.03

c<
.
S a l e Area

1.02

1.01 -

CAPSl

,Posl-Dislurbance ( 4 6 8 MVAR C A P S l

I
0.99

0.98

t1
2000

Posl-Dislurbance

Mismatches

0.97
1900

2100

2200 2300 Transler ( M W )

2400

2500

Figure 3 VP Curve Example


Q (MVAR)

700 r

Steady State Analytical Tools


The best way to start a voltage stability investigation is with a conventional power flow, with a constant power load model, and reactive power limits on generators. Base cases with loads and/or transfers of 5 to 10% above peak levels should be employed to guarantee adequate margins to voltage stability. If, after comprehensive contingency analyses, no divergent cases are detected, it is likely that voltage stability is not a problem. On the other hand, contingencies that fail to converge are candidates for further analyses. Two of the most popular conventional power flow-based tools that can be used for such analyses are VoltagePower (VP) and Reactive-Voltage (QV) curves. Their application is illustrated in the curves of Figures 3 and 4, respectively, which were derived from a highlystressed 345 kV system, before and after a critical stuck-breaker contingency. These are two alternative
O L 955
I

965

975

Figure 4 - Q V Curve Example


Another useful measure of voltage stability is provided by the QV curves shown in Figure 4. The curves are of two different types:

985 995 1005 Voltaoe lo U I

1015

1025

0 (MVAR) 700 -

reactive "demand" curves, which show the reactive in,jections at a bus or group of buses necessary to sustain a certain voltage. The results are given as a function of voltage for different system loads, before and after the disturbance. reactive ?wpply" curve, which indicates the Mvars produced by shunt compensation as a function of bus voltage. In Figure 4, this curve is the result of the Q = V2 x Y formula for fixed shunt resources, but can also be a straight line for SVCs operating within limits.

600

300 F

The intersection of the appropriate reactive supply and demand curves determines the operating point of the system. If the system is stressed beyond its steadystate limit, no such intersection occurs. As with the VP curves, the QV curves may be obtained by repeated application of a load-flow program under slightly modified boundary conditions. The analysis is carried out at a critical bus in the troubled area with a reactive source simulated at this bus. Generator Mvar limits are ignored and scheduled voltage is relatively high. The system is solved and the source reactive output is logged. Scheduled voltage is then slightly reduced, say, by 0.01 pu. The system is solved again and new source output is logged. The procedure goes on reducing the source's scheduled voltage further until a decrease in scheduled voltage results in an increase in reactive output. This is the "knee" of the QV Curve. The reactive requirements at this "knee" are then compared to the available reactive resources. If resources are higher than requirements, the system is voltage stable and the difference is the "reactive power margin". If the margin is negative, the system is voltage unstable. Among the many uses of QV Curves is the assessment of the need for dynamic sources of reactive compensation. Taking as an example the case depicted in Figures 3 and 4, one means of increasing the ~ 2 1 0 0 MW transfer limit is to increase fixed.shunt compensation to a total of, say, 600 or 700 Mvars. Such compensation, however, would most likely result in excessive pre-contingency voltages. What is required is a variable, fast source of reactive power, quickly responding to changes in loading and contingencies. One such source is an SVC, whose steady-state characteristics may be overlaid on the Q V Curves as in Figure 5. The figure illustrates how the SVC can maintain voltages within a narrow band for a wide variety of operating conditions. The figure also
suggests, however, that it must be adequately sized (in

0'
.955 .965

,975

,985

.995

1.005

1.015

1.025

Voltage (P.u.)

Figure 5 SVC Application


Although very useful and widely popular, VP and QV curve analyses have the disadvantage that they require previous knowledge of the problem characteristics, such as location of critical buses and interfaces. The techniques are also meant (QV curves in particular) for analysis of fairly localized problems. Optimal Power Flows (OPF), on the other hand, are not limited by such constraints. Give an OPF a load-flow that fails to converge, and, with well-chosen objective functions and constraints, the program will, in one solution, provide an answer to questions such as:
0
0 0

Least-cost allocation, for candidate buses, of shunt compensation to attain marginal stability. Least-cost generator MW redispatch to attain stability. Least-cost load-shedding

Through sensitivity analysis the program will also indicate the effect of other control variables on the objective function, for example, the effect of a tap change on reactive deficit. OPF is therefore a very powerful tool in the analysis of voltage instabilities, particularly under "unexpected" contingencies. As such, they are likely to play an increasingly significant role, not only in planning departments, but also in dispatch centers as an aide to operators.

Slow and Fast Dynamics


Although very useful, the above steady-state tools fail to fully address the physical process of voltage instability and collapse, and consequently do not provide answer to questions such as: What indeed happens to a power system when its load-flow model fails to converge ? This requires consideration of the "slow" and "fast" dynamics of voltage stability.

terms of Mvar capability), lest it become an expensive capacitor.

113

A power system is voltage-unstable when it fails "to maintain voltage so that when load admittance is increased, load power will increase" as defined by the IEEE System Dynamic Performance Subcommittee. Following a drop in voltage, and thus load, loads restore power levels by increasing their effective admittance. This increase in admittance can be due to many factors, such as a change in tap ratio on distribution transformers, a slip increase on induction motors, or a reduction in off-times on thermostatcontrolled loads. For a voltage-unstable system, the increase in admittances leads to further deterioration of system voltages with little or no improvement in load power levels. Hence, a condition of positive feedback arises where loads attempt in vain to restore required power levels by steadily increasing admittance levels. At the bulk transmission level, this results in a progressive decay in voltages, the speed of which is determined by system, transformer tap changer, and load characteristics. This performance is characterized as one of "slow dynamics".
These slow dynamics may or may not lead to a steadystate equilibrium:
0

Self-restoring load characteristics, including the effects of load tap changers and thermostatically-controlled loads.

Load Center Example To illustrate the application of dynamic simulations, a mayor load center was simulated. The center is fed through local generation, two relatively long 400 kV transmission lines, and several 220 kV transmission lines. The subtransmission voltage level is 132 kV. The total load is 4800 MW, 2100 MW of which is imported through the transmission system. Load flow studies indicate a voltage instability (i.e., non convergence) following loss of one of the 400 kV circuits. The voltage instability is the combined result of increased reactive losses on the transmission system and an exhaustion of local reactive reserves. Optimal Power Flow (OPF) investigations indicated the need for an additional 150 Mvars in shunt compensation to attain power flow solution. The normal stability model was augmented to simulate maximum excitation limiters and load tap changers on distribution transformers. Loads on transformer secondaries were initially set to 100% constant current for the real part and 100% constant admittance for the reactive part of the load. Figures 6 and 7 show simulation results for loss of the 400 kV circuit. Figure 6 shows voltage traces for a representative bulk transmission (220 kV) bus voltage, a subtransmission (132 kV) bus voltage (primary side of the "distribution" transformer), and a distribution bus voltage (secondary side of the "distribution" transformer) .

If the loads' power-restoring mechanisms are constrained to a limited range, voltages may stabilize at a level where operation is possible, although precarious. Voltage-sensitive loads controlled by tap range-limited distribution transformers are a possible scenario for such performance.

If, on the other hand, loads show little sensitivity


to voltage, or are self-restoring with time, voltages may deteriorate to levels where "fast

dynamics" phenomena are triggered. These phenomena, which include motor stalling, line trip and angular instability, will generally lead to very low voltage levels and to significant disruptions in system operation. This latter performance is generically described as "voltage collapse".

1.2

,
Distribution Voltage

1.2

1.1
3
v

tr
0

1
0.9

'5
a m
+

Dynamic Simulation
Slow and fast dynamics are best investigated with longterm dynamic simulations, which in addition to the normal stability models of generators, governors and excitation systems, include representation of phenomena that are critical to the mechanisms o f voltage instability and collapse, such as:
0

L 132 kV Voltage
0.8 0
10

0.8

20 30 Time (Minutes)

40

Maximum excitation limiters on generator excitation systems, which limit the long-term reactive output to levels within those determined by the unit's rated field current values.

Figure 6 Loss of 400 kV Circuit


Also shown in Figure 6 is a trace showing tap ratio performance. The figure clearly demonstrates that increases in tap ratio are accompanied by slight or no

1/4

improvements in distribution voltages and by major declines in transniission and subtransmission voltages. The main reason for the decline in transmission voltages is suggested in Figure 7. This figure shows field voltage and terminal voltage for a representative generator in the load center. Following loss of the transmission line, reactive demands on the generators exceed steady-state capabilities. These demands are exacerbated by the increase in load apparent admittances as a result of tap changer action. Following the timing-out of maximum excitation limiters, voltage setpoints are lowered in an attempt to maintain field currents within acceptable bounds.

Cascading stalling of motors, particularly large industrial motors directly connected to the subtransmission system. Stalling could be triggered by an attempt to start a large motor under already depressed voltages. Line trip due to low voltagelhigh current operation. Automatic or manual trip of an overloaded generator. Monotonic steady-state instability due to low voltages and reactive-limited generators.

o
o

1.2
3 P
d)

__

1.2
1.1
v

.05

2.8

cn
-1
Distnbuhon Voltage

n
0

2
i i

' L

I !
a,

I! I
Q

tlt 0.9
0.8
0
10

- __

. _

12.6
40

10

20 30 Time (minutes)

0.8
20 30 Time (Minutes)

40

Figure 7 Excitation Limiter Performance


This tug-of-war between loads and generation would lead to ever decreasing voltages were it not for the limited tap range on distribution transformers. Figure 6 shows that once transformer taps reach their limits, load voltages will drop, thus providing relief to the system by reducing both reactive demand and the amount of imported MW. A new steady-state condition is arrived at. All local generation is at a reactive limit. Voltages, no longer under the control of operators, are determined by loads. The close relationship between the new steady-state reached and the load characteristics is further illustrated by the voltage traces in Figure 8. These traces correspond to the same disturbance as in Figure 6 except that load was assumed to be composed of 60% small motors and 40% constant current real parVconstant admittance reactive part load. Because motor power corisumption has low sensitivity to voltage, voltages stabilize at much lower levels. In addition to affecting the quality of service to customers, operation at such low voltages exposes the system to a variety of secondary disturbances, which in turn can lead to system collapse (i.e., the "fast dynamics" of voltage instability). For example:

Figure 8 Loss of 400 kV Circuit with 60% Motor Load


Radial Svstem Examde In the load center example above, the relatively slow 20 to 40 minute dynamics are for the most part the result of field overload capability in generators, which helps "buff er'# the process. In many instances, however, voltage instability arises in radial systems, with no local generation at hand. The radial system example in Figure 9 corresponds to a 15O-kilometer, 115 kV double-circuit transmission system connected between two strong systems. Loads are distributed throughout the 115 kV system and are served mainly by System "A". Reactive requirements are shared by systems "A" and "B". The 115 kV system has no generation. Due to relay malfunction, the 115 kV system is separated from System "B".Hence the 115 kV transmission system and associated load become radial to System "A'. Load flow investigations indicate that after separation the 115 kV transmission system is no longer voltage stable. Q-V curve analyses indicate reactive deficits of at least 30 Mvar.

1/5

consist of small motors. System performance is now governed by the fast dynamics of motors stalling, which leads to an almost immediate voltage collapse.

\
n
v)

''' 2
,--. 3
a ,
0)

Motor Speed

0.9
a ,

-0.7

-.

0.6
0.4
Distnbution Voltage --

f
2
0 .c

2 0.5

0.3

0.1 0 0

3 4 5 Time (Seconds)

Figure 9 Schematic of Radial System Example


The contingency was also analyzed with time simulation. As in the load center case, dynamic modelling was augmented to represent maximum excitation limiters on generators and tap changers for loads.

Figure 1 1 Separation from System "B" with 60% Motor Load

Conclusions
This paper summarizes the concepts and techniques that, in the author's opinion, are necessary for a thorough review of voltage instability and collapse. The techniques range from the simpler contingency analyses to the more comprehensive dynamic simulation. System performance is shown to be affected by both the fast dynamics common in normal stability investigations and the slow dynamics of tap changers and maximum excitation limiters. Other effects such as load self-restoration, automatic capacitor switching and AGC action may also play a significant role. The simulations show a variety of possible outcomes to voltage instability. Some of them are slow in development and thus provide an opportunity for operator action. Others may lead to system collapse within minutes or seconds following a disturbance.

1.1
13
v

1.1

S
v

0 I > 0.9
0.8

. & ' I. 4. . . . . . . . . . . . m
v)

I+\/_ _

- TapRatio

.............

115 kV Voltage
Distribution Vollage

.......

..

.I
4

CL
0

I1 z
0.9

'3

+----J

0.8

2 3 Time (Minutes)

Figure 10 Separation from System "B"


Figure 10 shows voltage traces and transformer tap performance for a case with a 100% constant current real part and 100% constant admittance reactive part load model. The traces indicate a steeper voltage decay than in the reactive-deficiency example of Figure 6. (Note: the time scale of Figure 10 is one tenth of that in Figure 6). This is the result of a weaker system and the lack of local generation. Without the temporary overload capability of generators, the rate of decay in system voltages is dictated by transformer tap delay settings. Figure 11 corresponds to the same disturbance as in Figure 10, except that 60% of loads are assumed to

Biography Rodofo J. Koessler (M'86) received the degree of


Engineer of Electromechanics from the University of Buenos Aires, Argentina and the M.E. degree in Power Systems from the Rensselaer Polytechnic Institute, Troy, NY in 1979 and 1982, respectively. From 1978 to 1981, he was with Servicios Electricos del Gran Buenos Aires, working with the Power Systems Planning Department. He joined Power Technologies, Inc. in 1985 and is presently a Senior Engineer in the Utility System Performance unit. Most of his work at PTI has been in the area of dynamic performance and model development. He is a Senior Member of the IEEE-and its Power Engineering Society.
0 1997 The Institution of Electrical Engineers. Printed and published by the IEE, Savoy Place, London WC2R OBL, UK.

1/6

You might also like