WACS—Wide-Area Stability and Voltage

Control System: R&D and Online Demonstration

CARSON W. TAYLOR, FELLOW, IEEE, DENNIS C. ERICKSON, FELLOW, IEEE,
KENNETH E. MARTIN, SENIOR MEMBER, IEEE, ROBERT E. WILSON, SENIOR MEMBER, IEEE, AND
VAITHIANATHAN VENKATASUBRAMANIAN

Invited Paper

As background, we describe frequently used feedforward voltage support of a large power system. Features include
wide-area discontinuous power system stability controls. Then we synchronized positive sequence phasor measurements, dig-
describe online demonstration of a new response-based (feedback) ital fiber-optic communications from 500-kV substations,
Wide-Area stability and voltage Control System (WACS). The con-
trol system uses powerful discontinuous actions for power system a real-time control computer programmed in the G lan-
stabilization. The control system comprises phasor measurements guage, and output communications for generator tripping
at many substations, fiber-optic communications, real-time de- and 500-kV capacitor/reactor bank switching. The WACS
terministic computers, and transfer trip output signals to circuit software runs two algorithms in parallel.
breakers at many other substations and power plants. Finally,
we describe future development of WACS. WACS is developed as As background, we describe widely used emergency
a flexible platform to prevent blackouts and facilitate electrical controls termed Special Protection Systems (SPS). SPS
commerce. is based on direct detection of predefined outages, with
Keywords—Blackout prevention, emergency control, phasor high-speed binary (transfer trip) signals to control centers
measurements, power system stability, unstable limit cycle, voltage for logic decisions, and then to power plants and substations
stability, wide-area measurements and control. for generator tripping and capacitor/reactor bank switching.
Disadvantages of SPS include control for only predefined
I. INTRODUCTION events, complexity, and relatively high cost.
In contrast with SPS, WACS employs strategically placed
The Bonneville Power Administration (BPA), Portland,
sensors to react to the power system response to arbitrary dis-
OR; Ciber Inc., Beaverton, OR; and Washington State
turbances. WACS provides single discontinuous stabilizing
University (WSU), Pullman, are designing and imple-
actions or true feedback control. As true feedback control,
menting a Wide-Area stability and voltage Control System
the need for discontinuous action is determined and com-
termed WACS. WACS provides a flexible platform for
manded, the power system response is observed, and further
rapid implementation of generator tripping and reactive
discontinuous action such as generator tripping or capacitor
power compensation switching for transient stability and
bank switching is taken as necessary. The WACS platform
may also be used for wide-area modulation control of gen-
Manuscript received May 3, 2002; revised October 25, 2003. The work erators and transmission-level power electronic devices and
at Washington State University was supported in part by the Power System
Electric Engineering Research Center (PSERC) and in part by the Consor- for control center operator alarms and monitoring.
tium for Electric Reliability Technology Solutions (CERTS). We describe WACS benefits and describe large-scale sim-
C. W. Taylor and K. E. Martin are with the Bonneville Power Ad-
ministration, Vancouver, WA 98666 USA (e-mail: [email protected]; ulations showing the interarea stabilization of large distur-
[email protected]). bances by WACS.
D. C. Erickson is with Ciber, Inc., Beaverton, OR 97007 USA (e-mail: We also describe WACS design (measurement, communi-
[email protected]).
R. E. Wilson is with Western Area Power Administration, Lakewood, CO cations, control), and initial online implementation. Online
80228 USA (e-mail: [email protected]). testing results include statistics of communications delay
V. Venkatasubramanian is with the Department of Electrical and Com- from global positioning system (GPS) time-tagged sub-
puter Science, Washington State University, Pullman, WA 99165 USA
(e-mail: [email protected]). station measurements to GPS-timed receipt by the control
Digital Object Identifier 10.1109/JPROC.2005.846338 computer.

0018-9219/$20.00 © 2005 IEEE

892 PROCEEDINGS OF THE IEEE, VOL. 93, NO. 5, MAY 2005

 American synchronous interconnection are by small back-to-
back HVdc converter stations.
Long-distance interarea transmission lines charac-
terize the western interconnection. Major lines are 500-,
345-, and 230-kV. There are two 500-kV dc links: the
3100-MW 1360-km Pacific HVdc intertie from the Co-
lumbia River to Los Angeles and the 1920-MW 787-km
Intermountain Power Project link from Utah to southern
California. Hydro power predominates in the Pacific North-
west (PNW—British Columbia, Washington State, and
Fig. 1. Local and wide-area, continuous and discontinuous power Oregon). Large coal-based power plants predominate in the
system stability controls. Adapted from drawings by Dr. J. F. Hauer. eastern and southern portions of the interconnection. Most
generation in California is natural gas or oil based.
In spring and summer, with good hydro generation con-
II. POWER SYSTEM STABILITY CONTROLS
ditions, the dominant interarea power flows are from the
Power system stability controls are described in several PNW to California and also from coal-based generation in
books and reports [1]–[5]. Reference [6] is an early, but still the eastern areas (Wyoming, Utah, Arizona, and Nevada)
valuable, paper describing discontinuous controls. to California. The northern California portion of the Pacific
Fig. 1 shows a block diagram of the power system stability ac intertie comprises three series capacitor compensated
control environment. Power system stability encompasses 500-kV lines with nonsimultaneous rating of 4800 MW for
electromechanical (rotor angle) stability among groups of the three Oregon to California lines.
synchronous generators, and voltage stability involving load Large-scale power flow and transient stability simula-
response to disturbances [7]. RMS-type sensors are gener- tions of potential disturbances are necessary to determine
ally used—electromechanical oscillations and slow voltage transfer limits for many defined transmission paths. Power
variations amplitude-modulate the 50- or 60-Hz power flow (steady-state) simulations model over 10 000 buses
frequency waveforms. Electromagnetic transients are not of (nodes representing generation and load injection stations,
primary interest except for sensor filtering considerations. and transformer substations), requiring solving over 20 000
Most stability controls are continuous feedback controls nonlinear algebraic equations. Transient stability simula-
at power plants: automatic voltage regulator and power tion adds thousands of nonlinear differential equations. In
system stabilizer for generator excitation control, and prime defining simultaneous power transfer limits, major sensi-
mover control (speed governor). Controls are largely the tivities are shown on nomograms. Portions of nomogram
single-input–single-output type, designed via classical boundaries are limited by either first swing transient sta-
feedback control methods [1]. Additional local continuous bility, transient damping of oscillations, or postdisturbance
stability controls are at transmission-level power electronic voltage support criterion.
devices such as static var compensators. There are also local Controls described in this paper are oriented toward high-
discontinuous controls for reactive power compensation power transfers from the PNW and British Columbia to Cal-
(capacitor/reactor banks) switching and load shedding. ifornia, but are adaptable to other applications.
Installed wide-area stability controls are mainly based on
direct detection of selected outages. These emergency con-
IV. FEEDFORWARD WIDE-AREA STABILITY CONTROLS
trols are termed SPS or remedial action schemes. We describe
these controls in Section IV. The widely used SPS provide control for potential single
Advanced wide-area stability controls measure power and multiple-related outages identified in the power system
system response to disturbances. There are very few imple- planning process. Compared to the financial and permitting
mentations at present. We describe present-day projects in difficulties of transmission line construction, SPS are low
Sections V and VI. cost and easy to install. A large increase in power transfer
Discontinuous controls supplement the basic continuous capability is realized.
controls by relieving stress for very large disturbances, As generation and load increase without corresponding in-
providing a region of attraction and a secure postdisturbance crease in transmission lines, SPS controls proliferate [8]. At
operating or equilibrium point. Continuous controls then BPA, there are many schemes for a myriad of operating and
operate effectively over a smaller nonlinear range. disturbance conditions. Tens of millions of dollars have been
Wide-area controls—feedforward and feedback, contin- invested over many years. Additional schemes were added
uous and discontinuous—obviously have potential for im- following the cascading power failures in summer 1996 [9],
proved observability and controllability. [10].
Control actions are mainly for detection of transmission
outages, but also for some generation outages. The most
III. WESTERN NORTH AMERICAN POWER SYSTEM
complex scheme involves the Pacific ac intertie where
Fig. 8 in Section VI shows the western North American in- high-speed outage detection of around fifty 500-kV lines
terconnected power system. Connections to the eastern North is installed (detection at both line ends). Fault tolerant

TAYLOR et al.: WACS—WIDE-AREA STABILITY AND VOLTAGE CONTROL SYSTEM: R&D AND ONLINE DEMONSTRATION 893
programmable logic controllers are at BPA’s two control
centers: one near Portland and the other in Spokane, WA.
The most important control action is tripping of PNW
hydroelectric generators. There are few difficulties with
tripping hydro generators and they can be rapidly returned
to service. The generators are at the sending end of the
PNW to California power transfer path, with the generator
tripping braking remaining Northwest generators that are
accelerating relative to southwest generators. For outages
of either the Pacific ac or HVdc intertie, up to 2700 MW of
generation may be tripped.
Other control actions are energizing 500-kV series and
Fig. 2. Control center PDC.
shunt capacitor banks, and disconnecting shunt reactors.
BPA 500-kV shunt capacitor banks are in the 200–380 MVAr
range. after a major expansion of the HVdc intertie from two
Control actions take place as fast as 150 ms after the terminals to four terminals in 1989.
outage. The delay time includes detection time, commu-
nications to central logic, logic computer processing time, A. Phasor Measurements
communications to power plants and substations, and power Although various types of rms sensors may be used, digital
circuit breaker operating time. Communication of SPS acti- positive sequence, GPS-synchronized phasor measurements
vation signals use the same high-speed “transfer trip” used [4], [5], [13], [14] are most often considered for wide-area
for isolation of transmission line short circuits. At BPA these control. “Positive sequence” refers to transformation of an
are primarily frequency shift key audio tones over analog unbalanced set of three-phase voltages or currents into a set
microwave. Newer systems use digital messages over digital of positive, negative, and zero sequence “symmetrical com-
microwave or fiber optics (SONET). ponents,” where positive sequence is a set of three-phase
The consequences of SPS failure can be large-scale voltages or currents with equal magnitudes, 120 phase dif-
blackouts. Controls are clearly not as robust as additional ference, and normal phase rotation [5, Ch. 8]. In normal op-
transmission lines and must be highly reliable by design. eration without short circuits or individual phase outages, the
High redundancy in detection, communication, and logic phase voltages and currents are nearly equal to the positive
computers is required. sequence voltages and currents.
The complexity of the SPS is ever increasing. BPA has a Several manufacturers offer phasor measurement sensors.
full-time operator devoted to prearming (enabling) and mon- Typically, channels for multiple three-phase voltage and
itoring the many schemes. current measurements are provided. The positive sequence
Besides complexity, a shortcoming of preplanned event voltage and current phasors are computed and GPS time
driven control is that other disturbances may occur that have tagged once every two cycles, or in newer equipment, once
not been considered in planning. These may originate in every cycle of the power frequency (30- or 60-Hz data
other parts of the interconnected power system. rate for 60-Hz power frequency). Power system frequency
deviation from nominal is also computed, with GPS pro-
viding a precise time and frequency reference. There are
V. FEEDBACK WIDE-AREA STABILITY CONTROLS tradeoffs between response speed and filtering. The phasor
measurements are grouped, and data packets are transmitted
Feedback controls measure power system variables and to a central site where packets from several measurement
can respond to arbitrary disturbances. Control can be con- locations (substations) are organized by time stamp [15],
tinuous or discontinuous. [16]. Outputs from a “phasor data concentrator” (PDC) are
Stability control using remote signals are not new and networked to monitoring and control applications (Fig. 2).
are a simplified form of wide-area control. In 1976, BPA From the voltage and current phasors, applications may
implemented modulation of the Pacific HVdc intertie using compute active and reactive power.
active power and later current magnitude signals from a In coming years, phasor measurements will become more
remote substation on the parallel Pacific ac intertie [11], common as part of IT advances such as substation automa-
[12]. The continuous control damped electromechanical tion. The phasor measurements can be made available at
oscillations between groups of PNW generators and groups small cost as part of other substation measurements, for
of Pacific Southwest generators; the oscillation period was example, protective relaying [17].
around 3 s. Analog microwave communications and analog Networked phasor measurements are a key part of a
controls were used. The modulation was single input–four BPA/U.S. Department of Energy/Electric Power Research
outputs, the outputs being active and reactive power at the Institute (EPRI)/Western Area Power Administration pro-
northern HVdc terminal (rectifier operation) in Oregon and gram for wide-area measurement systems (WAMS) [5,
active and reactive power at the southern terminal (inverter Ch. 11.8], [18]. WAMS is valuable for power system iden-
operation) near Los Angeles. Modulation was discontinued tification, power system monitoring, control center state

894 PROCEEDINGS OF THE IEEE, VOL. 93, NO. 5, MAY 2005

Fig. 3. WACS block diagram.

estimation, and power system dynamic performance anal- Similar to feedforward controls (SPS), feedback dis-
ysis following disturbances—including large blackouts. continuous controls initiate a large stabilizing action that
improves first swing transient stability, reduces stress to
B. Continuous Wide-Area Controls improve oscillation damping, and provides a larger region of
attraction for a more secure postdisturbance operating point.
Continuous wide-area controls offer observability and
We next describe a specific wide-area discontinuous feed-
controllability benefits where conventional local continuous
back control (WACS) in development and demonstration.
controls have shortcomings. Possibilities include “wide-area
power system stabilizers” [19] and controls for powerful
transmission-level power electronic devices such as HVdc, VI. WACS
thyristor-controlled series capacitors, and static var compen-
sators. A. Overview
Wide-area controls are especially attractive for unusual
Fig. 3 shows a pictorial block diagram of WACS. Selected
system structures. Remote signals may augment control
existing phasor measurements are used for inputs, and ex-
using local measurements.
isting SPS transfer trip circuits are available for outputs. The
Because of increased control leverage and continuous
new development is the real-time controller.
exposure to adverse interactions, caution compared to local
Based partly on the 10 August 1996 cascading failure [10],
control is required. Communications latency is one concern.
the original BPA concept was to combine voltage magnitude
Dynamics mimicking electromechanical oscillations are
measurements with generator reactive power measurements
another [20], [21]. These may include sensor processing ar-
using fuzzy logic. The premise is that generator reactive
tifacts such as aliasing of network resonances or harmonics,
power measurements can be a more sensitive indicator of
or generator shaft torsional dynamics. Hydro plant water
insecurity than voltage magnitude—voltages can be near
column oscillations may appear to be electromechanical
normal but generator reactive power outputs near limits
oscillations. Extra monitoring and supervision of control is
indicate insecurity. R&D at WSU showed, however, that a
desirable.
voltage magnitude based control is faster and simpler for
transient stability [22]. Both methods are now used. Recent
C. Discontinuous Wide-Area Controls field experience has actually shown that the two methods,
Compared to continuous control, discontinuous control mag and mag algorithms, have similar speed. This is
tends to be safer—action is only taken when necessary. Dis- partly due to recent replacement of slow rotating generator
continuous control has similarities with biological systems field winding excitation equipment with modern thyristor
where stimuli must be above an activation threshold. exciters at two large power plants.

TAYLOR et al.: WACS—WIDE-AREA STABILITY AND VOLTAGE CONTROL SYSTEM: R&D AND ONLINE DEMONSTRATION 895
 References [23]–[25] describe recent WACS research and
development.
Twelve voltage magnitude measurements from seven
500-kV stations are used. Two stations are near the
Oregon–California border (Malin and Captain Jack), one is
in central Oregon (Summer Lake), and three are near the
Columbia River in northern Oregon or southern Washington
(John Day, Slatt, Ashe, and McNary). Fifteen generator
reactive power measurements at five power plant switching
stations near the Columbia River are used (Big Eddy, John
Day, Slatt, Ashe, and McNary). The hydro power plants
feeding into Big Eddy, John Day, and McNary comprise 18,
16, and 12 generators, respectively; two to four generators
are connected to a transmission line from the power plant to
the switching station where phasor measurements are made.
We designed WACS so that loss of measurements from a Fig. 4. Fiber-optic communications latency over 1 min, Slatt
PMU to PDC.
single location or even multiple locations will only slightly
degrade control. Measurements at widely spaced locations The SPS helps ensure postdisturbance voltage support for
(hundreds of kilometers) provide spatial averaging or fil- angle stability following generator overexcitation limiting,
tering against the aliasing effects discussed above. Spatial tap changing, and other slower actions. Many seconds are
filtering along with discontinuous control action biases the available for taking sequential feedback actions as necessary.
phasor measurement requirements toward fast response The conditions existing in northeastern Ohio preceding
rather than secure filtering. the 14 August 2003 blackout were exactly what the mag
algorithm caters to. Voltage magnitudes were mildly de-
B. Allowable Time for Control Actions pressed, but Cleveland-area generators were at or near their
For first swing transient stability, control action must reactive power limits [26]. Automatic load shedding by
be taken prior to the peak of the forward interarea angle control similar to WACS could have prevented the blackout.
swing—the sooner the better. For a simple second-order
C. Phasor Measurement Communications and PDC
undamped dynamic system with natural frequency of 1/3 Hz
(3-s period), the step response peak is at 1.5 s. The impulse BPA legacy communications is analog microwave. Trans-
response peak is at 0.75 s. Most disturbances are closer to mission of phasor measurement packets using modems has
a step response than an impulse (the rare three-phase short high latency (60–100 ms) and relatively high dropout rates.
circuit approaches an impulse, but opening the faulted line Thus, BPA-owned fiber-optic communications (SONET) are
provides a step response effect). Nowadays the frequency used for WACS. BPA has an extensive fiber-optic network,
of the Pacific intertie mode is around 0.25 Hz (4-s period), with links and terminal equipment still being added.
allowing more time for control action. The oscillation fre- Fig. 4 shows tests of fiber-optic latency of less than 26 ms
quency is even lower for high-stress operation. for a link from the Slatt switching station to a BPA control
For transient stability, control action should be completed center. The link uses direct digital transfer into SONET.
within around 1 s—especially for the less powerful capac-
itor/reactor bank switching. D. Real-Time Hardware and Software
The delay time for phasor measurement, fiber-optic com- We selected National Instruments’ LabVIEW Real-Time
munications, PDC throughput including wait time for slowly (RT) hardware and software [27]. The software is a true
arriving packets, transfer trip, and circuit breaker tripping dataflow language that prevents race conditions and allows
(of generators or shunt reactors) or closing (shunt capacitor for parallel tasking (multitasking and multithreading are
bank insertion) are approximately 3, 2, 2, 1, and 2–5 60-Hz supported). It has many needed programming features
cycles respectively, or around 10 cycles for tripping and 13 and library components, including data acquisition/pro-
cycles for closing (167 and 217 ms). Time for several con- cessing/output, TCP/UDP, signal processing, filtering,
trol execution loops, intentional time delay, and throughput math operations, execution logic and state machine, execu-
delay will be 67 ms or longer. Thus, it appears that control tion tracing and timing, display graphics, and fuzzy logic
action can be taken within 0.3 s after sufficient power system tools with graphical editors. The graphical code is largely
electromechanical response to the disturbance. For capac- self-documenting. Modular architecture aids the testing
itor/reactor bank switching, the local supervising voltage and certification of critical modules by designing virtual
measurement sensors will be responding during the same simulated “real-world” conditions around them.
time as the WACS measurements and processing. Controller development is done on a PC, and code is then
The more sensitive mag algorithm may operate fol- downloaded to real-time deterministic hardware and soft-
lowing longer time frame dynamics. In fact, need for the ware [27]. Fig. 5 shows the rack-mounted WACS hardware.
existing SPS action is determined by power flow simula- A full-featured host PC that connects via Ethernet to the
tion of a point in time several minutes after the disturbance. WACS RT engine is available to monitor, test, develop, and

896 PROCEEDINGS OF THE IEEE, VOL. 93, NO. 5, MAY 2005

Fig. 5. WACS real-time hardware.

upload software and to run model studies for tuning. This is

important, as the RT processor (engine) itself is always oper- Fig. 6. BPA 500-kV voltages for outage of two Palo Verde
ated “headless” (no monitor, keyboard, or mouse). We do use generators with existing controls, with Pacific ac intertie loading
of 4700 MW.
a small text display status monitor and a sequence of events
recorder. to improve transient damping. For growing oscillations, it
WACS parameters are managed using an .ini file. This will operate at some point for stabilization.
allows easy changes to the application for tuning the algo- The algorithm is fairly simple, based on 12 voltage mag-
rithms, defining station names, scale factors, etc. This is im- nitude measurements at seven 500-kV stations. A weighted
portant in avoiding the pitfalls of “hardwiring.” average voltage is computed from the 12 measurements, with
1) Control Execution Rate: The rate at present is 30 con- highest weight for measurements close to the Oregon–Cali-
trol executions per second (33.3–ms intervals), which is the fornia border where the voltage swings are usually greatest
same rate as the phasor measurement packets. The time for (Malin, Captain Jack, Summer Lake). A nonlinear accumu-
a control execution cycle is around 8 ms, allowing addition lator (integrator) computes volt-seconds below a threshold
of other features and moving to a future 60 packets per setting that is currently 525 kV for capacitor/reactor bank
second data rate. For security against inadvertent actions switching and 520 kV for generator tripping (normal voltage
during faults, measurement noise, and other disturbances, is around 540 kV). Accumulation is blocked for voltage re-
output must be above a threshold for two or more control covery. Control action results when the volt-second accumu-
executions. lation reaches a setpoint; also, the weighted voltage must be
2) Input Processing: Tasks include reading and decoding below 490 kV for generator tripping. The algorithm thus has
the data from the UDP Ethernet connection to the PDC and semblance to proportional–integral (PI) control. Beneficially,
data sanity checks for transmission errors, nonvalid data, faster operation results for more severe disturbances.
missing packets, and extra long latency. Close coordination 1) Critical Disturbance: A critical disturbance is near si-
with PDC data processing is required, and a customized multaneous outage of two nuclear generating units at a nu-
message of only the needed measurements is transmitted to clear power plant in Arizona or California. Such events have
WACS. For missing or corrupt data, there are two options: occurred several times, and Western Electricity Coordinating
one is to block the data and not use it in the algorithm, and Council (WECC) rules specify that cascading failure not re-
the second is to pad the data with the last valid packet for a sult from these outages. Multigenerator outages at large coal-
few control executions. The former is used currently. Both based power plants have also occurred.
fatal and nonfatal errors are managed and reports issued. The largest disturbance is outage of two nuclear generators
Also, the data is weighted and limit tested before being at the Palo Verde nuclear plant near Phoenix, AZ. The com-
passed to the algorithm subroutines. The WACS software bined power loss is around 2700 MW. The lost power is made
can also run in a library mode using simulation data or up by inertia power from rotors of all other generators, and
archived data from the PDC as inputs to the algorithms. This by response to decaying frequency by the speed governors
validates performance, facilitating certification. of other generators. About half of the response comes from
3) Output Processing: Following the algorithm compu- hydro generators in the northern half of the interconnected
tations, a pattern consisting of up to 32 isolated outputs is power system, resulting in a large increase in the Pacific ac
sent to a relay control stage, where several masking opera- intertie north to south loading. Instability (loss of synchro-
tions are done before passing to the transfer trip communi- nism between northern and southern generators) results if the
cations. Any fatal or nonfatal error will mask outputs, with initial intertie loading is high. There is no SPS for this outage.
a user-defined mask available to disable one or more outputs 2) Simulation of Two Palo Verde Generator Outages
either temporarily or long term or to disable the results of ei- Without and With WACS: The Appendix describes simula-
ther or both the fast or slow algorithms. tion methodology. Fig. 6 shows 500-kV voltage responses
for the two Palo Verde generator outages with existing
E. The mag Algorithm controls. The initial Pacific ac intertie loading is 4700 MW.
The largest voltage swing is at the Malin station near the
The voltage magnitude based mag algorithm provides Oregon–California border, and the performance is consid-
first swing transient stability stabilization and relieves stress ered marginally acceptable.

TAYLOR et al.: WACS—WIDE-AREA STABILITY AND VOLTAGE CONTROL SYSTEM: R&D AND ONLINE DEMONSTRATION 897
Fig. 7. BPA 500-kV voltages for outage of two Palo Verde Fig. 9. Malin voltage for outage of two Palo Verde generators
generators with WACS and with Pacific ac intertie loading of with partial failure of WACS generator tripping.
5000 MW.

Jack near Malin on the Oregon–California border. Simula-

tion results are nearly identical to the reference case, so are
not shown.
4) Partial Failure of Generator Tripping: Fig. 9 com-
pares the reference case to simulation of failure of generator
tripping at one of the two British Columbia power plants.
Stability is maintained, but performance is notably degraded.
The oscillation frequency decreases.
Other sensitivity cases lead to improved tuning of the
algorithm. We simulate other outages to verify expected
performance.

F. The mag Algorithm

The mag algorithm combines voltage magnitude mea-
surements and generator reactive power measurements using
fuzzy logic. Similar to the mag algorithm, we compute a
weighted average 500-kV voltage magnitude from 12 phasor
measurements at seven locations. More complicated is com-
putation of weighted average reactive power from 15 trans-
mission lines emanating from six large power plants. We
Fig. 8. Western North American interconnected power system compute active and reactive powers for these lines from the
showing disturbance location, and WACS input measurement voltage and current phasors.
locations and output action locations.
First, we estimate the number of connected genera-
tors—up to four generators per line may be connected on
Fig. 7 shows similar response with WACS at an initial lines from hydro plants. While this information may be
intertie loading of 5000 MW—a 300-MW gain. The main available at a slower data rate within a control center, we
WACS action is 916 MW of generator tripping at two hydro avoid interface and dependence on other data networks.
power plants in British Columbia 1.4 s after the outage. The number of generators is estimated based on the normal
WACS also inserted two 500-kV shunt capacitor banks loading range of individual generators. An error with esti-
1.2 s after the outage. The transfer trip circuits from the mation on one of the many lines is not serious.
BPA control center to the British Columbia power plants A normalization procedure follows, based on generator
and capacitor bank locations exist as part of BPA’s SPS. active/reactive power capability curves. We map the capa-
The higher oscillation frequency of Fig. 7 tends to indicate bility curves from the generator terminals to the transmission
improved stability. side where the phasor measurement sensors are located, ac-
Fig. 8 shows the wide-area nature: an outage in Arizona, counting for station service load, and generator step-up trans-
measurements in Oregon and southern Washington, and con- former impedance and tap ratio. Fig. 10 shows this one-time
trol actions in British Columbia. mapping for a large nuclear plant generator.
3) Partial Failure of Measurements: Using the Normalization results in reactive power output on a scale
5000-MW intertie loading with WACS case as reference, we of approximately 1 (generator controls allow large tem-
simulate failure of the most important phasor measurement porary and small steady-state operation outside limits). The
device. The failure is two voltage measurements at Captain transmission-side reactive power limits corresponding to the

898 PROCEEDINGS OF THE IEEE, VOL. 93, NO. 5, MAY 2005

Fig. 10. Active/reactive power capability curve for a nuclear plant
generator.
Fig. 12. Weighted average generator reactive power fuzzy set
input. Linguistic variables are NLq (negative large reactive power),
NMq (negative medium), Okq, PMq (positive medium), and PLq
(positive large).

Fig. 11. Weighted average voltage fuzzy set input. Linguistic

variables are VLOv (very low voltage), LOv (low voltage), OKv
(OK voltage), HIv (high voltage), and VHIv (very high voltage).
Fig. 13. Output (consequence) fuzzy set. Linguistic variables
active power are noted: and . We compute the are NLo (negative large output), NMo (negative medium output),
NSo (negative small output), ZEo (zero output), PSo (positive
normalized value from transmission side values as follows: small output), etc.

For example, for Fig. 10 with transmission side

MW and MVAr, .
A weighted average of the normalized reactive powers is
now computed. The individual weights are the product of the
generator or generator group MVA rating and a factor. The
factor is based on location and voltage support sensitivity.
We give higher value to generators providing more sensitive
control of transmission voltage by automatic voltage regu-
lator line drop compensation [28] or by automatic high side
voltage control. Fig. 14. Rule base. Output default value of zero if no rule is
active. Center of gravity defuzzification method is selected. The
We next combine the weighted average voltage magni- MAX–MIN inference method is used, which is explained in the
tude and the weighted average generator reactive power are Fig. 15 caption for rule 8.
using basic fuzzy logic. To date, linguistic variable tuning is
based on rules of thumb such as overlap cross points at 0.5. zero. The crisp output range is approximately 1, but elim-
Figs. 11–13 show the input and output fuzzy sets. Fig. 14 ination of unrealistic rule combinations result in a practical
shows the rule editor, and Fig. 15 shows the output con- 0–1 range.
trol surface as a function of weighted average voltage for A crisp output value above a threshold enters an accumu-
a weighted average generator reactive power input value of lator. With accumulator setpoint reached, capacitor/reactor

TAYLOR et al.: WACS—WIDE-AREA STABILITY AND VOLTAGE CONTROL SYSTEM: R&D AND ONLINE DEMONSTRATION 899
 Fig. 16. Northwest voltages for first three swings.

G. Tuning, Testing, and Monitoring

Fig. 15. Fuzzy controller voltage versus output characteristic for
reactive power input of zero. At the cursor point, the voltage input Since large disturbances are rare, algorithm verification
is 512 kV, the output is 0.30, and rules 8 and 13 are active. As and tuning is normally via offline large-scale simulation
shown, Rule 8 fires with value of 0.94, which is MIN of the LOv and (Appendix I). The real-time controller is tested by inputting
Okq linguistic variable values for 512-kV voltage and zero reactive
power inputs (Figs. 11 and 12). For the output variable PSo, the simulation results, and also by inputting archived phasor
area below 0.94 contributes to the center of gravity output (Fig. 13). measurements from actual events.
If another rule activates output variable PSo, the MAX value of the
two rules is used in the center of gravity computation.
H. WACS Response for a Large Real Event
At 07:40:56 on Monday 14 June 2004, a short circuit oc-
Table 1 curred near the Palo Verde Nuclear Plant west of Phoenix.
Power Flow Simulation Results, 4800 MW Intertie Power The fault was not completely cleared for almost 39 s! Ap-
proximately 4589 MW of generation tripped, at and near Palo
Verde in the southern part of the western North American in-
terconnection. All three Palo Verde units tripped.
Pacific intertie stability was threatened, but maintained.
With one line between Oregon and California out of service,
the intertie limit prior to the event was 3200 MW. North
to south intertie flow swung from the initial 2750 MW to
5500 MW and settled at 4500 MW several minutes later.
Malin and Captain Jack voltages near the Oregon–Cali-
fornia border swung from the initial 548 kV to 443 kV at
07:41:21.6. Operators and an existing “response-based”
scheme switched BPA series capacitors and shunt capac-
bank switching or generator tripping is commanded. Capac- itor/reactor banks during the swings and the subsequent
itor/reactor bank switching occurs first, with generator trip- intertie power increase. The power increase is from governor
ping only in more severe situations where capacitor/reactor action at PNW hydro plants, which carry large amounts of
bank switching is not available or not sufficient. spinning reserve.
The main tuning is of the input measurement weights and On 14 June, WACS was in a monitor mode at a labora-
the thresholds for output action—rather than the fuzzy sets. tory installation 5 km from the control center and PDC. Ad-
As a power flow program steady-state simulation example, justing for communications and PDC features that were not
we simulated a simultaneous transfer nomogram limit point. yet in service, WACS operated correctly on the forward angle
Table 1 shows the fuzzy logic inputs and crisp outputs for swing, before voltage swing minimum, as recorded by a se-
a predisturbance case, the double Palo Verde outage case quence of events recorder [25].
with existing controls, the outage case with WACS-initiated To further validate the WACS algorithms, we played
500-kV capacitor/reactor bank switching at three locations, archived data from the 14 June event into the WACS code
and the outage case with the capacitor/reactor bank switching on an offline personal computer. One parameter was retuned
and 500 MW of generator tripping. Also shown are key indi- to increase operating speed.
vidual inputs to the fuzzy logic: voltage magnitude at Malin Fig. 16 shows voltages that are WACS inputs and in-
and reactive power from John Day power plant line 1 (John cludes the weighted average voltage computed by WACS
Day reactive power has the high weighting because of loca- (same weights used for both algorithms). For the mag
tion and excellent voltage control performance). algorithm, the accumulator thresholds for capacitor/reactor
The crisp fuzzy logic output, perhaps rescaled to 0–100, bank switching are 525 kV and the thresholds for generator
could also be used for a thermometer-type voltage security tripping are 520 kV. Accumulator setting for capacitor/re-
index for control center operators. actor bank switching is 2 kV-s and accumulator setting for

900 PROCEEDINGS OF THE IEEE, VOL. 93, NO. 5, MAY 2005

 The 14 June massive lost of generation was well beyond
planning and operating reliability criteria. While unusual,
power systems are continuously exposed to unusual events.
If stress (Pacific intertie loading) was somewhat higher
on 14 June, instability would have occurred and caused
controlled islanding with massive generation/load imbal-
ance in the importing southern island that suffered the
initial 4600-MW loss. Either massive underfrequency load
shedding or a widespread blackout in California, Nevada,
Arizona, and New Mexico would have resulted. Stress would
have been higher later in the day as temperatures and load
increased.
Events like the 14 June event are exactly what WACS can
Fig. 17. V magQ fuzzy logic inputs and outputs. protect against. Reference [25] provides more details on this
event.
generator tripping is 4 kV-s. For generator tripping, the
weighted average voltage must also be below 490 kV. VII. WACS STATUS
For first swing stabilization, discontinuous control action
Since March 2003, WACS has been installed in a labora-
should occur before the voltage minimum at around 81.6 s
tory with real-time phasor measurement inputs from a PDC
after 07:40. This time is estimated for the real signal without
and recording of contact outputs. Based on the following, we
measurement delay (PMU timetags are at the last sample
have demonstrated “proof of concept,” and tuned and vali-
of the phasor computation window; the PMU uses a four-
dated the real-time controller hardware and software:
cycle moving average filter with phasors calculated over one
cycle). • large-scale simulations including playback of simula-
Using the archived data as input to the real-time code, the tion results into real-time code;
WACS mag algorithm output for capacitor/reactor bank • monitoring of real system performance over a two-year
switching occurs at 81.033 s after 07:40. Adding 170 ms for period;
communications, PDC, and circuit breaker delay, switching • playback of archived data into real-time code, particu-
would be at 81.203 s, or around 0.4 s before the real signal larly for the 14 June 2003 massive generation outage
voltage minimum. event.
WACS mag algorithm output for generator tripping oc- With successful R&D, BPA is considering a capital budget
curs at 81.233 s after 07:40. Adding 170 ms the delays, trip- installation for commercial use as described below.
ping would be at 81.403 s, or around 0.2 s before the real
signal voltage minimum. VIII. FURTHER DEVELOPMENT OF WACS
For the mag algorithm, Fig. 17 shows the weighted In parallel with the demonstration project, design require-
average voltage and weighted average reactive power that ments for permanent commercial implementation are being
are combined using fuzzy logic. The crisp (center of gravity) developed. Similar to SPS, very high reliability is required if
fuzzy logic output is also shown. High fuzzy logic output WACS is used to increase power transfer limits. Dedicated,
correctly occurs for the combination of low voltage and high latest generation phasor measurement sensors likely will be
reactive power output. Fuzzy logic output above a threshold required. A 60 packets per second data rate and a 60 times per
is accumulated. For capacitor/reactor bank switching the second control execution rate will probably be used. Control
thresholds are 0.40 per unit (p.u.). For generator tripping, the computers will likely be at both BPA control centers, and all
thresholds are 0.45 p.u. Accumulator settings are 0.05 p.u.-s communications will be self-healing (geographically sepa-
for capacitor/reactor bank switching and 0.2 p.u.-s for gen- rated fiber-optic or digital microwave rings). The PDC func-
erator tripping. tion may be incorporated in the WACS computers.
The mag algorithm WACS output for capacitor/reactor After experience with single discontinuous control ac-
bank switching occurs 81.033 s after 07:40 (same time as tions, bang–bang switching of capacitor banks can be
the voltage magnitude based algorithm). Adding 170 ms for considered for oscillation damping enhancement. Other
communications, PDC, and circuit breaker delay, switching control actions may be added, such as raising transmis-
would be at 81.203 s, or around 0.4 s before the real signal sion voltage schedules at power plants during emergency
voltage minimum. situations; see Fig. 3. Nonlinear modulation of a static var
WACS mag algorithm output for generator tripping compensator near Seattle, WA, using a voltage magnitude
occurs at 81.533 s after 07:40 (300 ms later than the voltage signal from Malin has been found effective [29] and could
magnitude based algorithm). Adding 170 ms for the delays, be implemented using WACS technology.
tripping would be at 81.703 s, or around 0.1 s after the real Many voltage and current phasor measurements plus
signal voltage minimum. While this will cause a larger back- system frequency measurements are available. Available
swing, the generator tripping reduces intertie loading and measurements increase year by year. Binary signals from
system stress. substations can be added to the phasor measurement packets.

TAYLOR et al.: WACS—WIDE-AREA STABILITY AND VOLTAGE CONTROL SYSTEM: R&D AND ONLINE DEMONSTRATION 901
From this data, WACS can provide monitoring and alarms for • Provides a combination of reliability increase and
control center operators. Possibilities include event detec- power transfer capability increase.
tion and monitoring of oscillation activity [30], [31, Ch. 6]. • Caters to uncertainty in simulation results used to de-
The software platform has extensive signal processing termine operating rules and limits.
tools, and poor damping following severe disturbances is • Future potential with cost reductions and further IT
relatively easy to detect. A challenge is to minimize false advances. Potential for application in meshed grid as
alarms—logic or artificial intelligence (AI) methods applied well as intertie corridors. Control inputs and outputs
to the large measurement base might be developed for this may be extended over a larger geographical area such
purpose. as the entire western North American power system.
Integration with other control center networks and appli- Moving from WAMS to WACS (wide-area measurements
cation may offer synergies. For example, online security as- to wide-area stability control) is a challenge in the new cen-
sessment simulations perhaps could be used for automatic tury.
tuning/learning/adaptation of WACS.
WACS is currently oriented toward improving stability of
a specific interarea power transfer path. Application in other APPENDIX I
networks may be straightforward. For example, in a load DYNAMIC SIMULATION METHODOLOGY
area, either wide-area voltage measurements or combined
voltage measurements and generator reactive power mea- Our large-scale simulations approximate the dynamic
surements could be used for reactive power compensation interactions between the measurement devices, communi-
switching or load tripping. Such control in northern Ohio cations systems, WACS actions, and power circuit breaker
could have prevented the 14 August 2003 blackout. operations. The simulation program with WACS models
Research is ongoing at WSU on further generalizing changes the topology of the modeled interconnection when
of wide-area controls to meshed networks where control set conditions are met during the simulation process. For
strategy (measurements, disturbance classification, and example, reactive power equipment is changed from ener-
generator or load tripping actions) is more difficult. WSU gized to deenergized. After any change to network topology,
is also researching theoretical aspects, and use of voltage the simulation process continues, with the effects of WACS
phase angles as control input. Appendices II and III describe actions fed back to the overall simulation.
this research. The simulations start from standard data sets (base cases)
compiled by the WECC. We use the General Electric
IX. CONCLUSION PSLF/PSDS power system simulation software to solve
the differential–algebraic–difference equations. WACS and
Automatic control experts state: “A modern view of con- other user-defined models are written in a language similar
trol sees feedback as a tool for uncertainty management” [32, to Basic or C. A numerical integration time step of 1/4
p. 1]. cycle of the 60 Hz frequency (4.17 ms) is used. This is a
Given the many changes in the electric power industry compromise between accuracy and computing time.
with increasing complexity and reduced investment in trans- Power system state variables and derived variables are
mission lines, the possibility and actual occurrences of large- calculated at every time step. Examples are substation bus
scale blackouts are a worldwide concern. Clearly, new means voltages, transmission line currents, and frequency. Bal-
to improve power system reliability and robustness are de- anced conditions are assumed so calculated qualities are
sirable. The addition of wide-area feedback control to fre- positive sequence values. Most controls are modeled within
quently used wide-area feedforward control is an effective the main numerical integration process (digital controls
additional layer of defense against blackouts [33], as well as other than WACS are approximated as analog controls).
facilitator of electrical commerce. We model the phasor measurements and substation to con-
WACS exploits advances in digital/optical communi- trol center communications as a combination of low pass
cations and computation. Specific advantages include the filter and pure time delay. The WACS algorithms of Sec-
following. tion VI are modeled as described, with computations every
• Control for outages and conditions not covered by two cycles or eight steps of the main program numerical in-
feedforward controls (SPS). tegration. If a setpoint for either algorithm is reached for two
• Potentially simplifies operations for changing system consecutive controller cycles, a global programming variable
conditions—currently, operators are required to reduce is asserted. This represents WACS broadcasting permission
power transfers when unstudied conditions are encoun- for generator tripping or capacitor/reactor bank switching.
tered. We also model actions taken at the substations and gen-
• Improved observability and controllability compared erating plants. One model deenergizes generation at selected
to local control. Discontinuous control reduces expo- generating plants. Propagation delays from the central site
sure to adverse interactions. and for circuit breaker operating times are modeled.
• Flexible, high reliability “open system” platform for Capacitor/reactor bank switching is locally “supervised.”
rapid, low-cost control and monitoring additions, in- The signal from the central site is one input to an AND gate.
cluding wide-area continuous control. The local substation bus voltage has to be below a setpoint

902 PROCEEDINGS OF THE IEEE, VOL. 93, NO. 5, MAY 2005

for a set time to satisfy the second AND gate input. We model
the supervision, and also the breaker operating times.
We compare and reconcile the programming used for sim-
ulations with the real-time code. Simulation data of WACS
inputs are formatted for playback into the real-time code.

APPENDIX II
THEORETICAL JUSTIFICATION FOR THE mag ALGORITHM
Section VI provides an overview of the WACS algorithms.
The objectives are: 1) the detection and 2) the mitigation of
severe disturbances in the western power system. Two algo-
rithms, namely, mag and mag , are currently used for
the real-time detection of critical large disturbances. Here we
discuss a theoretical justification for the mag algorithm.
The mag algorithm relates the size of the voltage dips
around the Pacific intertie lines during an event to the
severity of the disturbance. Specifically, when the voltages
stay below certain thresholds and when the voltage errors
accumulate above the prespecified values, different control
actions are initiated to mitigate the disturbance while it is
still evolving. We show that the voltage dips indicate the
proximity of the trajectory to the boundary of the region of
attraction of the current operating point. When the trajectory
stays inside the region of attraction, the system remains Fig. 18. Bifurcation diagram for the 10 August 1996 validated
transient stable, which is a primary objective of the WACS model [37].
control. On the other hand, if a trajectory is pushed outside
the region of attraction, the system becomes transient insta-
boundary of the region of attraction for MW transfer values
bility and the system separates into islands.
less than 4570 MW. For transfers below 4540 MW in Fig. 18,
Previous theoretical studies have shown that the region of
the ULC denoted “Outer ULC” anchors the transient stability
attraction is normally bounded by: 1) the stable manifolds1 of
boundary. At this ULC, the Malin bus voltage fluctuates from
unstable equilibrium points (UEPs) or saddle points; 2) stable
manifolds of unstable limit cycles (ULCs); and 3) segments a low value around 0.78 p.u. (390 kV) to a high value around
connected with singularities of the network equations [34]. 1.18 p.u. (590 kV). And, interestingly, in our simulations, the
Normally, the stable manifolds of UEPs are the only ones size of the outer ULC does not change appreciably when the
transfer decreases below 4500 MW. That is, the low voltage
studied in power system transient stability analysis such as
in the controlling UEP algorithm in the direct stability litera- excursion at the Malin bus along the ULC is usually just
ture [35]. Recent studies at WSU have shown that ULCs play below 400 kV for a wide range of transfer values. The same
crucial roles in anchoring the transient stability boundary in result was also observed in other WECC models in [36] and
was reported in the doctoral dissertation of J. Li at WSU in
the WECC models [36], [37].
In a recent paper [37], we proposed a novel algorithm for 2001.
computing the ULCs on the transient stability boundary in Based on these ULC computations, we conjecture that
large-scale power system models. The algorithm was used whenever the Malin bus voltage drops below 0.8 p.u.
(400 kV) during a contingency, the trajectory approaches the
for tracking the evolution of the unstable limit cycle on the
boundary of the region of attraction for a validated model transient stability boundary anchored by the ULC. Moreover,
of the 10 August 1996 western blackout. A summary of we can use the severity of the voltage dip at the Malin bus
the results is presented in the bifurcation diagram below. (or along the intertie corridor) during a disturbance to assess
Fig. 18 shows the loci of the amplitudes of the stable and how close the trajectory is from the ULC on the boundary
unstable limit cycles for the WECC model when the active of the region of attraction. That is, the smaller the voltage
power transfer across the intertie lines is slowly increased. swings along the corridor and the more the Malin voltage
The size of the limit cycles in Fig. 18 is illustrated by the dips along the trajectory stay above 0.8 p.u., the stronger the
high and low fluctuation values for the Malin bus voltage western system is with regard to transient stability.
(California–Oregon border) for the limit cycles. We employed this heuristic for assessing the dynamic se-
In Fig. 18, the equilibrium point becomes small-signal curity of the system by computing the severity of the intertie
unstable by undergoing a subcritical Hopf bifurcation for a corridor bus voltage dips in a dynamic security optimization
power transfer of 4570 MW. Specifically, ULCs anchor the algorithm proposed in [38]. We showed the heuristic to be
effective for improving dynamic security. The mag algo-
1Stable manifold of a UEP or a ULC denotes the set of all trajectories that rithm discussed in Section VI uses the same heuristic rule.
converge to that UEP or ULC respectively. We assume that the lower the voltages near the COI lines dip

TAYLOR et al.: WACS—WIDE-AREA STABILITY AND VOLTAGE CONTROL SYSTEM: R&D AND ONLINE DEMONSTRATION 903
during a disturbance, the closer the trajectory is to the tran- continues to decrease beyond a threshold, we would inter-
sient stability boundary, and hence the more severe the dis- pret that as a likely separation of Area that could be coun-
turbance is. Therefore, the measure of how much the intertie tered by load shedding in Area .
voltages dip toward 0.8 p.u. could be used as the heuristic for In our studies, we set the control trigger heuristics to be
triggering the WACS control actions in the mag algorithm. similar to the voltage error algorithm mag in Section VI.
In the case of phase angles, we subtract the moving average
APPENDIX III value (say, of the relative angle from the current
EXTENDING THE WACS ALGORITHMS USING value by defining . We then accu-
PHASE ANGLES mulate two integral error terms, denoted and , respec-
tively, to denote the speeding up or slowing down of Area
The current algorithms used in the WACS control are
with respect to the center of inertia reference frame. First, the
based on measurements of bus voltages and generator re-
term is the integral error for whenever
active power within the BPA network. At WSU, we are
continuously stays above a threshold, say . As in Sec-
developing new algorithms that detect and mitigate transient
tion VI, the accumulated error is reset to zero whenever
instability by utilizing the phase angle measurements of
the angle drifts below . When the error grows
critical generator bus voltages from across the entire western
above a prespecified value, say, , the Area is interpreted
system. Fast exchange of PMU measurements among WECC
to be speeding away from the rest of the system and a suit-
utilities is being pursued, and it is reasonable to assume the
able generation tripping may be initiated in that area. The
availability of system wide phase angle information (from
value of will be tuned in real-time based on the current
specific PMU locations) in the near future.
total generation and the current spinning reserve in Area .
The proposed algorithm extends the concept of the
That is, the smaller the current spinning reserve (relative to
voltage-based algorithm mag into consideration of the
the total generation) in Area , then the lower the threshold
phase angle measurements. At present, the algorithm ana-
value for . The computation of the error is then similar
lyzes the phase angles in two stages: 1) the angle stability
to accumulating the integral error of below a threshold,
within each control area and 2) the angle stability of the
denoted . When the error grows above the threshold
entire WECC system. The principle in each step is similar.
, load tripping in Area may be initiated to mitigate the
Let us assume the availability of the phase angle measure-
disturbance event. The details of the algorithm will be pre-
ments, say, , from a few key generating plants, say for
sented elsewhere together with simulation results.
in Area . Then, we introduce the notion of
the center of inertia angle reference for the area, say , by
the rule ACKNOWLEDGMENT
Many BPA engineers have assisted in the development of
WACS. Dr. Y. Chen contributed to the algorithms as a Ph.D.
candidate at Washington State University, Pullman.

REFERENCES
where denotes the current MW generation schedule at the
[1] P. Kundur, Power System Stability and Control. New York: Mc-
plant in Area . By increasing the number of angle measure- Graw-Hill, 1994.
ments within each area, we can improve the accuracy of the [2] C. W. Taylor, Power System Voltage Stability. New York: McGraw-
computation of the angle reference and we can also im- Hill, 1994.
[3] T. Van Cutsem and C. Vournas, Voltage Stability of Electric Power
prove the redundancy. Similarly, the center of inertia angle Systems. Norwell, MA: Kluwer, 1998.
reference for the entire system, denoted , can be computed [4] “Advanced angle stability controls,” CIGRE, Task Force 17, Advi-
with the rule sory Group 02, Study Committee 38, Paris, France, Brochure 155,
Apr. 2000.
[5] L. L. Grigsby, Ed., The Electric Power Engineering Handbook: CRC
Press/IEEE Press, 2001, ch. 11, Power System Dynamics and Sta-
bility.
[6] IEEE Committee Report, “A description of discrete supplementary
controls for stability,” IEEE Trans. Power App. Syst., vol. PAS-97,
no. 1, pp. 149–165, Jan./Feb. 1978.
[7] IEEE/CIGRE Joint Task Force on Stability Terms and Definitions,
“Definition and classification of power system stability,” IEEE
where is the total number of areas that are available in the Trans. Power Syst., vol. 19, no. 3, pp. 1387–1401, Aug. 2004.
control formulation, and denotes the current total gener- [8] P. M. Anderson and B. K. LeReverend, “Industry experience with
ation in Area . When the representative angle of an area special protection schemes,” IEEE Trans. Power Syst., vol. 11, no.
3, pp. 1166–1179, Aug. 1996.
continuously increases away from the center of inertia be- [9] C. W. Taylor and D. C. Erickson, “Recording and analyzing the July
yond a prespecified metric, we would heuristically interpret 2 cascading outage,” IEEE Comput. Appl. Power, vol. 10, no. 1, pp.
that Area is moving toward separation from the rest of the 26–30, Jan. 1997.
[10] D. M. Kosterev, C. W. Taylor, and W. A. Mittelstadt, “Model vali-
system. In this case, a suitable remedial action could be the dation for the August 10, 1996 WSCC system outage,” IEEE Trans.
tripping of generation in that area. Similarly, when the angle Power Syst., vol. 14, no. 3, pp. 967–979, Aug. 1999.

904 PROCEEDINGS OF THE IEEE, VOL. 93, NO. 5, MAY 2005

 [11] R. L. Cresap, D. N. Scott, W. A. Mittelstadt, and C. W. Taylor, [33] C. W. Taylor, “Improving grid behavior,” IEEE Spectrum, vol. 36,
“Operating experience with modulation of the Pacific HVDC in- no. 6, pp. 40–45, Jun. 1999.
tertie,” IEEE Trans. Power App. Syst., vol. PAS-98, pp. 1053–1059, [34] V. Venkatasubramanian, H. Schattler, and J. Zaborszky, “A tax-
Jul./Aug. 1978. onomy of the dynamics of large differential-algebraic systems,”
[12] CIGRE, Paris, France, CIGRE 14-05, 1978. Proc. IEEE, vol. 83, no. 11, pp. 1530–1561, Nov. 1995.
[13] A. G. Phadke, J. S. Thorp, and M. G. Adamiak, “New measurement [35] V. Vittal and A. A. Fouad, Power System Transient Stability Analysis
technique for tracking voltage phasors, local system frequency, and Using the Transient Energy Function Method. Englewood Cliffs,
rate of change of frequency,” IEEE Trans. Power App. Syst., vol. NJ: Prentice-Hall, 1991.
PAS-102, no. 5, pp. 1025–1038, May 1982. [36] J. Li and V. Venkatasubramanian, “Study of unstable limit cycles
[14] R. E. Wilson, “Uses of precise time and frequency in power sys- in power system models,” IEEE Trans. Power Syst., submitted for
tems,” Proc. IEEE (Special Issue on Time and Frequency), vol. 79, publication.
no. 7, pp. 1009–1018, Jul. 1991. [37] V. Venkatasubramanian and Y. Li, “Analysis of 1996 western Amer-
[15] IEEE Power Syst. Relaying Comm. (WG H-7, A. G. Phadke, ican electric black-outs,” presented at the Bulk Power System Dy-
Chairman), “Synchronized sampling and phasor measurements for namics and Control VI Conf., Cortina, Italy, 2004.
relaying and control,” IEEE Trans. Power Del., vol. 9, no. 1, pp. [38] Y. Li and V. Venkatasubramanian, “Coordination of transmis-
442–452, Jan. 1994. sion path transfers,” IEEE Trans. Power Syst., vol. 19, no. 3, pp.
[16] IEEE Power Syst. Relaying Comm. (WG H-8, K. E. Martin, 1607–1615, Aug. 2004.
Chairman), “IEEE standard for synchrophasors for power systems,”
IEEE Trans. Power Del., vol. 13, no. 1, pp. 73–77, Jan. 1998.
[17] G. Benmouyal, E. E. Schweitzer, and A. Guzman, “Synchronized
phasor measurement in protective relays for protection, control, and
analysis of electric power systems,” presented at the 29th Annu. Carson W. Taylor (Fellow, IEEE) received the
Western Protective Relay Conf., Spokane, WA, 2002. B.S.E.E. degree from the University of Wis-
[18] J. Hauer, D. Trudnowski, G. Rogers, B. Mittelstadt, W. Litzenberger, consin, Madison, in 1965 and the M.S. degree
and J. Johnson, “Keeping an eye on power system dynamics,” IEEE in electric power engineering from Rensselaer
Comput. Appl. Power, vol. 10, no. 4, pp. 50–54, Oct. 1997. Polytechnic Institute, Troy, NY, in 1969.
[19] I. Kamwa, R. Grondin, and Y. Hebert, “Wide-area measurement He joined the Bonneville Power Administra-
based stabilizing control of large power systems—a decentral- tion (BPA), Portland, OR, in 1969. He is now
ized/hierarchical approach,” IEEE Trans. Power Syst., vol. 16, no. Principal Engineer in Transmission Operations
1, pp. 136–153, Feb. 2001. and Planning, where he is developing wide-area
[20] J. F. Hauer and C. W. Taylor, “Information, reliability, and control in voltage and stability controls. Besides his BPA
the new power system,” in Proc. 1998 American Control Conf., vol. work, he consults and teaches power system
5, pp. 2986–2991. engineering. In 1986, he established Carson Taylor Seminars. He has
[21] J. Lambert, A. G. Phadke, and D. McNabb, “Accurate voltage phasor presented 72 seminars in 17 countries. He is the author of the book Power
measurement in a series-compensated network,” IEEE Trans. Power System Voltage Stability (New York: McGraw-Hill, 1994). The book is
Del., vol. 9, no. 1, pp. 501–509, Jan. 1994. translated into Chinese. Mr. Taylor has authored or coauthored many
[22] C. W. Taylor, V. Venkatasubramanian, and Y. Chen, “Wide-area sta- IEEE and CIGRE papers. His interests include power system control and
bility and voltage control,” presented at the Seventh Symp. Special- protection, system dynamic performance, ac/dc interactions, and power
ists in Electric Operational and Expansion Planning (VII SEPOPE), system operations and planning.
Curitiba, Brazil, 2000. Mr. Taylor is a Member of the U.S. National Academy of Engineering and
[23] C. W. Taylor and R. E. Wilson, “BPA’s Wide-Area stability and a Distinguished Member of CIGRE. He is Past Chairman of the IEEE Power
voltage Control System (WACS): overview and large-scale simula- Engineering Society (PES) Power System Stability Controls Subcommittee.
tions,” presented at the Sixth Symp. Specialists in Electric Opera- He is Convenor of three CIGRE task forces on power system voltage and
tional and Expansion Planning (IX SEPOPE), Rio de Janeiro, Brazil, angle stability.
2004.
[24] R. E. Wilson and C. W. Taylor, “Using dynamic simulations to de-
sign the wide-area stability and voltage control system (WACS),”
presented at the IEEE Power Engineering Soc. Power System Conf.
Exposition, New York, 2004. Dennis C. Erickson (Fellow, IEEE) received the
[25] C. W. Taylor, D. C. Erickson, and R. E. Wilson, “Reducing blackout B.S. degree in general engineering at the Univer-
risk by a wide-area control system (WACS): Adding a new layer of sity of Portland, Portland, OR, in 1968 and has
defense,” in Proc. 2005 Power System Computation Conf., submitted continued his education through postgraduate
for publication. programs and specialized courses.
[26] U.S.-Canada Power System Outage Task Force, Final Report on the During his tenure at the Bonneville Power
August 14, 2003 Blackout in the United States and Canada: Causes Administration, Portland, (1965–2000), he
and Recommendations, Apr. 2004. designed precision time dissemination and
[27] National Instruments [Online]. Available: http://www.ni.com frequency measurement systems for real-time
[28] D. Kosterev, “Design, installation, and initial operating experience power system control applications, and evaluated
with line drop compensation at John Day powerhouse,” IEEE Trans. research on free space optical and millimetric
Power Syst., vol. 16, no. 2, pp. 261–265, May 2001. communications systems. He was the Principal Engineer on BPA’s au-
[29] V. Venkatasubramanian, K. W. Schneider, and C. W. Taylor, “Im- tomatic overhead transmission line fault locator network and developed
proving Pacific inter-tie stability using existing static VAR compen- several major innovations in that field. He developed novel high voltage
sators and thyristor controlled series compensation,” in Proc. Bulk measurement systems employing electrooptic and magnetooptic dielectric
Power System Dynamics and Control IV—Restructuring, 1998, pp. sensors, and high precision wide bandwidth analog and digital fiber-optic
647–650. instrumentation links. From 2000 to 2002, he was with Quantum Controls
[30] J. F. Hauer and F. Vakili, “An oscillation trigger for power system Inc. as a Senior Technical Consultant. At present, he is with Ciber Inc.,
monitoring,” IEEE Trans. Power Syst., vol. 5, no. 1, pp. 74–79, Feb. Beaverton, OR, as a Chief Systems Engineer designing advanced power
1990. system control systems and data acquisition, dissemination, and analysis
[31] J. F. Hauer et al., “A Dynamic Information Manager for Networked systems. He holds one patent in the field. He is the author of over 20
Monitoring of Large Power Systems,” EPRI, Palo Alto, CA, Final technical papers, including papers on optical sensors and multifunctional
Rep. TR-112 031, 1999. data acquisition techniques.
[32] R. M. Murray. (2002, Jun.) Control in an information rich Mr. Erickson is the Founding Chairman of the Fiber Optics Subcommittee
world. [Online]. Available: http://www.cds.caltech.edu/~murray/ of the IEEE Power Engineering Society (PES) Power System Communica-
cdspanel/report/cdspanel-15aug02.pdf tions Committee.

TAYLOR et al.: WACS—WIDE-AREA STABILITY AND VOLTAGE CONTROL SYSTEM: R&D AND ONLINE DEMONSTRATION 905
 Kenneth E. Martin (Senior Member, IEEE) re- Robert E. Wilson (Senior Member, IEEE)
ceived the B.S.E.E. degree from Colorado State received the B.S.E.E. degree from the University
University, Fort Collins, in 1970 and the M.A. de- of Nebraska, Lincoln, in 1969, the M.S. degree
gree in mathematics from the University of Wash- in electrical engineering and the M.A. degree
ington, Seattle, in 1974. in mathematics from the University of Arizona,
He served in the U.S. Army from 1970 to 1972, Tucson, in 1973 and 1980, respectively, and the
primarily designing instrumentation at the Yuma M.E. and Ph.D. degrees in electrical engineering
Proving Ground. He joined the Bonneville Power from the University of Idaho, Moscow, in 1989
Administration, Portland, OR, in 1975 where he and 1992, respectively.
has worked with system protection, control sys- From 1980 to 1990, and again from 1994 to the
tems, telecommunications, and instrumentation. present, he worked for the Western Area Power
He is currently Principal Engineer in the Measurement Systems group. He Administration, U.S. Department of Energy, Lakewood, CO. From 1990 to
has authored or coauthored numerous technical papers and made technical 1992 he was the Washington Water Power Fellow at the University of Idaho.
presentations in several countries. His main interests are global positioning From 1992 to 1994, he was the first Nicholson Visiting Assistant Professor
system-based timing systems and wide area, real-time power system mea- of Electrical Engineering, University of Wyoming, Laramie. From 2002 to
surements, particular phasor measurements. 2003, he worked for the Bonneville Power Administration, Portland, OR,
Mr. Martin is a Registered Professional Engineer in Washington State. In via an interagency agreement. He has authored or coauthored many tech-
2003 he received the BPA Eugene C. Starr award for Technical Achieve- nical papers. He has studied the applications of precise time and frequency
ment. He is a Member of the IEEE Power Engineering Society (PES) Power in power systems since 1985. He is currently studying advanced control sys-
System Relay Committee and the Relay Communications Subcommittee. tems for the western North American power grid and advanced methods of
He chairs the Synchrophasor Standard working group. data display and methods to transport renewable energy into California.
Dr. Wilson is a Registered Professional Engineer in the state of Colorado.
He was a Member of the IEEE Power Engineering Society (PES) Power
System Relay Committee and the IEEE Working Group on Phasor Measure-
ments, and was Cochair of a task force that wrote one chapter in the IEEE
Special Publication Modeling and Analysis of System Transients Using Dig-
ital Programs (Pub. No. 99-TP-133-0, 1999).

Vaithianathan Venkatasubramanian received the B.E. (honors) degree in

electrical and electronics engineering from the Birla Institute of Technology
and Science, Pilani, India in 1986 and the M.S. and D.Sc. degrees in systems
science and mathematics from Washington University, St. Louis, MO, in
1989 and 1992, respectively.
He is a Professor of Electrical Engineering and Computer Science at
Washington State University, Pullman, WA. His research interests are
power system stability and control, and nonlinear system theory.

906 PROCEEDINGS OF THE IEEE, VOL. 93, NO. 5, MAY 2005