SCADA/EMS History
Table of Contents
Supervisory Control Systems
Pilot Wire
Stepping Switches
Relay/Tones
Solid State
Vendors
Telemetry
Current Balance/transducers
Pulse Rate/variable frequency
Selective
Continuous Scanning
Automatic Data Logger
Vendors
SCADA
Computer master
Remote Terminal Units
User Interface
Communication channels
Vendors
Automatic Generation Control (AGC)
Recorders
Servos
Mag amps/ operational amplifiers
Computer Controlled Analog
Vendors
System Operation Computers
SCADA
Forecast and scheduling functions
Configurations
Vendors
Energy Management Systems
Network Analysis Functions
Configurations
Consultants
Vendors
Projects
Supervisory Control Systems
Pilot Wire Systems(1940's and earlier)
Supervisory control in electric utility systems evolved from the need to operate
equipment located in remote substations. In the past it was necessary to have personnel
stationed at the remote site, or send a line crew out to operate equipment. The first
approach was to use a pair of wires or a multi-pair cable between the sites. Each pair of
wires operated a unique piece of equipment. This was expensive but justified if the
equipment needed to be operated often or in order to restore service rapidly.
Stepping switch systems
It soon became obvious that if they could multi-plex one pair of wires and control several
devices, the process would become much more efficient. However in doing this it would
be necessary to have very good security, because the affect of operating the wrong piece
of equipment could be severe. During the 30's the telephone company developed
magnetic stepping switch’s for switching telephone circuits. The electric utilities used
this technology as a early form of supervisory control. However to assure security they
developed a select/check/operate scheme. This scheme had the master station send out a
selection message, which when received by the remote would cause it to send out a
corresponding check message for the selected device.
If this was received by the master station it would indicate this to the dispatcher who
would then initiate a operate message which when received by the remote would cause
the device to operate. The remote would then send a confirmation message to the master
station to complete the action. This select before operate scheme has been used in
supervisory control systems for many years and forms of it are still in use today.
Relay/Tone systems
Relay systems used telephone type relays to create pulses that were sent over a
communication channel to the remote. They typically used a coding scheme either based
on the number of pulses in a message, or a fixed message size with combinations of short
or long pulses. Westinghouse in conjunction with North Electric Company, developed
Visicode supervisory control based the pulse count approach. I am most familiar with this
equipment since I spent over five years doing installation and trouble shooting on it. Most
of the installations were with electric utilities, some were pipelines, gas companies and
even done an installation in the control tower of O Hara airport in Chicago, to control the
landing lights. Visicode used two time delay relays to create pulses, and timing chain of
relays to count them. It used the select/check/operate scheme for security. There were
two sizes the “junior” used one select/check sequence, but the larger used a select/check
sequence for a group and a second for the point. Thousands of this equipment went in
service during the years from 1950 until about 1965. I worked on it so often that I could
trouble shoot it often by just listening to the relay sequence, in fact more than once I was
able to solve their problem over the telephone. It was very reliable and had very few false
operations, although it could happen. One I investigated was caused because the
equipment was installed in a diesel generating station with the master station cabinet
beside the generator. The was a lot of low frequency vibration, and the battery connection
was loose. As the battery connection would make and break with the vibration, the master
was sending out a constant stream of pulses 24 hours a day. The remote was going crazy,
and it finally hit the right combination for a false operation.
General electric also had a supervisory control system that was a competitor of Visicode.
It operated similarly but used a fixed length message type and relied on two different
pulse lengths for the coding. Both used the select before operate scheme.
Control Corporation supplied a supervisory system that used combinations of tones for
the coding. They were created with a “Vibrasender” and received with a “Vibrasponder”.
In 1957 I was installing a Visicode system in Watertown South Dakota with the Bureau
of Reclamation and a Control Corp field engineer was there installing one of their
systems also. We spent a lot of time together at lunch and in the evenings. It was fun
listening to him trouble shoot because he put a speaker on the line and listened to the
tones. He turned my name over to their personnel department who contacted me and I
went to Minneapolis to interview, but since I had only been at Westinghouse for two
years I decided to stay. Ironically I went with CDC ,who had bought Control Corp, 13
years later.
Solid State (60's and 70's)
Westinghouse developed their solid state supervisory system around 1960. It was called
REDAC. It used a fixed word length format with a checksum character with each
message. It still used the select/check/operate security technique. GE came out with a
similar system about that time that I believe was called GETAC. Control Corporations
solid state system was called Supertrol. These systems were basically solid state versions
of their previous systems.
Vendors
The predominate vendors during this evolution of supervisory control were
Westinghouse, General Electric, and Control Corporation. There were European vendors
such as Siemens but they were not a significant factor in the US market.
Telemetry
Current Balance/transducers
Telemetry in electric utilities began like supervisory control because it became necessary
to obtain readings of power values like volts, amps, watts etc. from remote locations. As
power systems became larger and more complex and the need for central operation
developed, telemetry was required. The first approach was to utilize a pair of wires and
pass a current proportional to the reading. This first required a transducer to convert the
power system value to a DC voltage or current proportional to the power system quantity.
The first transducers for voltage or current were relatively simple since the AC value
from metering transformers could be rectified to create a DC value. Watt and Var
readings were more complex however, and the first successful transducers commonly
used for these were Thermal Converters. These units used instrument transformer inputs
from voltage and current readings to heat a small thermal element which was then
detected by thermocouples to create a proportional voltage. Later on, a different type of
transducer was invented which used the Hall Effect to create the proportional voltage.
Unfortunately these transducers could not be applied directly to a communication channel
consisting of a pair of wires.
The variations in resistance of the long communication channel ,with temperature,
prevented a consistent accurate reading. Therefore a telemetry transmitter was required
that would convert the transducer voltage to a constant current source that would be
nearly independent of the variations in resistance. This was done with a current balance
type of telemetry transmitter. It was first implemented using vacuum tubes and later on
transistors.
Pulse rate/Variable frequency telemetry systems.
In order to transmit telemetry across other types of communication channels such as
power line carrier, and to improve accuracy, telemetry schemes were devised using pulse
rate and variable frequency methods. The pulse rate equipment generated a pulse rate that
between the minimum rate and the maximum rate was proportional to the telemetered
value. The variable frequency systems worked similar, as the telemetered value was
proportional to the range between the minimum and maximum frequency. While at
Westinghouse in the late 50's I worked on the development of their system, which was
called Teledac. The solid state system generated a signal with a frequency that varied
from 15 to 35 cycles, proportional to the telemetered value. There was another pulse
based system that was used during these years that was supplied by the Bristol Company.
They used a mechanical variable length pulse that was generated by a rotating serrated
disk. The value in this scheme was proportional to the pulse width. This equipment was
used more in gas and pipeline systems because of it’s slow time response. All of these
systems had the advantage of being able to send the signal over a variety of
communication media. They were typically used over pilot wire, power line carrier, and
micro-wave.
Selective Telemetry systems.
Since it is usually only necessary to read these values periodically, and to reduce the
number of communication channels needed, it became necessary to multiplex several
readings over a single communication channel. This need was consistent with the
requirements of supervisory control. So the usage of selective telemetry over supervisory
control became common. The dispatcher could take his hourly readings by selecting the
readings on a supervisory/telemetry system.
Continuous Scanning systems.
Since the equipment was by now implemented with solid state technology, and in some
cases on dedicated communication channels, it was possible to continuously scan the
values. With no mechanical parts involved the gear would not wear out. Various vendors
developed continuous scanning systems with a solid state master system.
Automatic data loggers.
Automatic data loggers relieved the dispatcher from the tedious task of recording all the
readings each hour. Early versions came before solid state and continuous scan telemetry,
although they were unusual until the solid state or computer master evolved.. I installed a
Automatic Data Logger in 1958 at the dispatch office of Omaha Public Power District,
that was a good example of these early attempts. This system used stepping switches at
each end to select the points. The telemetry scheme used was Bristol pulse duration. Once
a point was selected it would take at least three pulses before the telemetry would settle
down, so that had to be added to the logic in the field. The telemetry receiver was a
Bristol Metamaster which was a big mechanical beast that had a shaft position output and
a Gianini shaft position analog to digital converter, was coupled to it. The digital output
of the converter was typed out on a IBM long carriage typewriter. It was a mess to get
working but it did the job. However because of all the complexity and electro-magnetic
components, it was difficult to maintain and was replaced after a few years. Solid state
equipment and computer based master stations replaced this type of equipment.
Vendors
The vendors of early telemetry systems included Westinghouse, General Electric, and
Control Data who were extending their supervisory control business. A few other vendors
were telemetry suppliers like Bristol, and as the equipment turned to solid state
continuous scanning, several new vendors such as Pacific Telephone and Moore
Associates appeared.
SCADA System(1965 and later)
The term SCADA for Supervisory Control and Data Acquisition Systems, came into use
after the use of a computer based master station became common. By the middle 60's
there were computers that were capable of real time functions. Westinghouse and GE
both built processors that could be used by this time. The Westinghouse computers were
called PRODAC and GE processors GETAC. The complexity of the logic required to
make a hardwired master that would provide all the necessary functions of a SCADA
system was so great, that advantage of using a computer became apparent. Other
computer companies came out with computers that were applicable to master station
SCADA included Digital Equipment Company and Scientific Data Systems. The
functions now included scanning data, monitoring the data or status, and alarming for
changes, displaying the data, on digital displays, and later, on CRT displays and periodic
data logging.
Most SCADA systems worked on a continuous scan basis with the master sending
requests for data and the remotes only responding. A few systems ,mostly in Europe, had
the remote issue data continuously without a master station request. This was possible if
the communications channel was full duplex. (Could send and receive simultaneously).
Remote Terminal Units
The remote station of the SCADA systems took the nomenclature, Remote Terminal
Units (RTU’s), during the 60's. RTU’s were supplied primarily by the SCADA vendor,
however later there developed a market for RTU’s from vendors other than complete
SCADA vendors. RTU’s used solid state components, mounted on printed circuit cards
and typically housed in card racks installed in equipment cabinets, suitable for mounting
in a remote power substations They needed to operate, even if the power was out at the
station, so they were connected to the substation battery, which was usually 129volt DC..
Since most RTU’s operated on a continuous scan basis, and since it is important to have
fast response to control operations in event of a system disturbance, the communication
protocol had to be both efficient and very secure. Security was a primary factor, so
sophisticated checksum security characters were transmitted with each message, and the
select/before/operate scheme used on control operations. The most common security
check code used was BCH, which was a communication check code developed in the
60's. During the 60's and 70's most RTU communication protocol’s were unique to the
RTU vendor i.e. proprietary.
Because of the need for both very high security and efficiency, common protocols used in
other industries such as ASCII were not used. In order to allow different types of RTU’s
on a SCADA system there was a effort to standardize protocols undertaken by the IEEE.
The development of the Micro Processor based communication interface at the master
station solved some of the compatibility problems however, as it now became possible
for the master station to communicate with RTU’s of different protocols as the micro
processor based communication interface could be programmed to handle the different
protocols. The basic structure of a RTU consisted of the communication interface, central
logic controller, and input/output system with analog inputs, digital inputs, control digital
outputs and sometimes analog control outputs. They were typically supplied in 90 high
steel cabinets, with room for terminal blocks for field wiring to substation equipment.
Large SCADA systems would have several hundred RTU’s.
User Interface(Man/machine Interface)
Early SCADA systems used a wired panel with pushbuttons for selecting points and
doing control operations. Data was displayed using digital displays, and alarms were
listed on printers or teletypewriters. A major evolution came during the late sixties when
the Cathode Ray Tube, CRT, was applied to the Man/machine interface (MMI). The
original versions were black and white character mode CRT’s. They offered the ability to
display data and status in tabular format, and to list alarms as well as acknowledge them,
and clear them. A keyboard was used instead of the pushbutton panel for interacting with
the system. At first there was resistance from the dispatchers to the new methods. They
felt they were too complicated and too “computer like”, although it did not take them
very long to appreciate the added flexibility and capability. Then the Limited Graphic
color CRT’s became available in the early seventies. These CRT’s had a graphic
character set that would allow presenting one line diagrams of electric circuits with
dynamic representation of circuit breakers, switches etc. Power system values such as
volts, amps, megawatts etc can be presented on the one lines drawings and periodically
up-dated. Up-date rates were typically in the area of five seconds. In the 80's Full Graphic
CRT’s became available. Now one-line diagrams became much more sophisticated as
you could pan across a large diagram or zoom in to a substation to present more detailed
data. Control operations were accomplished by selecting the device on the diagram,
confirming the selection visually and initiating the control operation. The
select/before/operate scheme still prevailed. Selection was done by moving the CRT
cursor to the device on the diagram. There were several devices used to move the cursor
that evolved over the years. Early systems used a keypad with directional arrows, then a
track/ball was used were the dispatcher rolled the ball to position the cursor. Another
approach was to use a light pen that was used to touch the screen and b ring the cursor to
that position. Finally the computer “mouse” that is so common today evolved as the most
useful, although lots of dispatchers resisted it for the same reasons they resisted CRT’s.
With the alarm lists being presented on CRT’s alarm printing became just a record
keeping function and so the printers were now in the computer room instead of the
dispatch office, and were usually line printers. Mapboards have been used in Dispatch
offices for many years. They are typically mounted on the wall in front of the dispatchers.
They serve various purposes but one of the most important is to keep the dispatchers
constantly familiar with the power system. In the early days they were static boards,
showing lines, stations, and salient equipment. Sometimes they were manually updated
with tags or flags to equipment out of service or on maintenance. As computer based
SCADA systems proliferated, it became possible to make these boards dynamic. They
could show the status of devices such as circuit breakers and, using digital displays,
indicate line flows and voltage. By the middle seventies these boards became very
elaborate and expensive. However the capability of the Full Graphic CRT’s, by the late
seventies and early eighties reduced the need for a elaborate dynamic board, and they
became less common, instead going with less dynamics on the board and making more
use of the CRT.
Communication Channels.
In the early days the most common communications media was a pair of dedicated wires.
These were sometimes laid by the utility, and sometimes provided by the telephone
company. A modem was provided by the SCADA vendor that used audio range
frequency, and was frequency modulated to connect to the channel. For a while in the late
sixties, power line carrier was used. Although the limited number of frequencies limited
the usage. In the later seventies and eighties the most common channel became utility
owned micro- wave equipment. It was expensive but had the advantage of providing a
large number of channels.
Vendors:
The vendors for SCADA systems evolved from the vendors of supervisory control. These
included Westinghouse, General Electric and Control Corporation. Early Westinghouse
supervisory control was a combination of Westinghouse personnel and people from
North Electric Co. In the sixties a few people from both operations split off and joined a
company in Melbourne Florida, called Radiation to develop SCADA systems. This
company evolved into Harris Systems. A company in San Jose California, Moore
Associates began making digital telemetry equipment evolved into a SCADA vendor
during the seventies. This company later became Landis&Gyre.
Automatic Generation Control (AGC)
AGC Evolved from the use of telemetry and recorders to record and display data such as
tie line flows and generation values. Dispatch offices in the fifties, and earlier, typically
had recorders that showed all major generation MW values and tie line flows. They often
used retransmitting slidewires on the recorders to add the tie line flows together and
display Net Interchange on a separate recorder. Also total generation was often summed
up using a similar method from the various generation recorders. System frequency was
also typically displayed on a recorder, and measured by a frequency transducer. By using
a slidewire on the Net Interchange recorder and comparing it to a Net Interchange setter
consisting of a digital dial driven potentiometer, a value representing Net Interchange
Error was developed and displayed.. This error signal was then applied to a mechanical
device that would generate pulses of either a raise or lower direction. These pulses were
then transmitted to the various generating plants, and applied to a governor motor control
panel that would drive the governor motor and hence raise or lower generation. The
control pulses would move the generating units to bring the Net Interchange Error to
zero. This was the beginning of AGC. Frequency error was developed from the frequency
recorder, and a desired frequency setter. Frequency error was applied through a frequency
bias setter calibrated in MW/0.1 cycle. This allowed the control system to control
interchange and contribute to intra-system control of frequency. By the late fifties, there
were other vendors that got involved in AGC, primarily Westinghouse and GE. It was at
this time I moved out to Pittsburgh and joined the Westinghouse group that was
developing analog AGC. The technology at this time used a combination of recorder
retransmitting slide wires and Magnetic Amplifiers. During this period the concept of
Incremental cost and Economic dispatch was refined. The incremental cost curves were
implemented with resistors and diodes, such that when a analog voltage was applied, the
output was a nonlinear curve representing the incremental cost curve for the unit or plant.
Economic Dispatch is based on the fact that a unit once on the line, can be loaded with
respect to the other units on line according to its incremental cost. However it’s
commitment to be put on line, is a function of production cost, start-up cost, and other
factors. There were several large AGC/EDC systems built by the late fifties, a few even
included large “B” matrix systems to compensate Economic Dispatch for transmission
losses. These systems were large, expensive and complicated, using hundreds of Mag
Amps, recorders, servo mechanisms, etc. There is a large justification for the complexity
and expense of using an automatic Economic Dispatch System with transmission loss
compensation. If these systems can result in a savings of a few percent in a large utilities
production cost, it would amount to several million dollars a year. By the late 50's/early
60's solid state operational amplifiers became available. These made developing the
analog computing methods needed by these systems much easier. There were hundreds of
systems shipped in the 60's using this technology. These systems were usually similar in
control configuration. They were a type 1 control system. The systems Area Control
Error, value was developed by comparing net interchange to desired interchange, and
adding in a frequency bias term. Later a Time Error Bias was added in. The ACE value
was then filtered and processed. It was then added to total generation to develop a total
requirement value that was allocated to the individual generating units by the Economic
Dispatch circuitry. This developed a Unit Generation requirement that when compared to
actual generation, developed a unit error signal which was usually used to create control
pulses which were transmitted to the plant and applied to the governor motors to adjust
generation in response to the error. By the middle 60's there were enough advances in
real time oriented digital computers that the Economic Dispatch function could be done
on the computer. Economic Dispatch was typically done at approximately three minute
periods, while the other portion of AGC ( typically called Load Frequency Control) must
execute as often as every few seconds. Thus was developed the Computer Controlled
Analog type of system where Economic Dispatch with transmission loss compensation
was done in the computer and the result output on a digital controlled set-point to the
analog LFC system. There were dozens of systems like this shipped in the late 60's.
About this time the digital computers with real time operating systems evolved to the
point where the entire AGC algorithm could be implemented on the computer. By this
time I had been in the Systems Control group at Westinghouse for ten years working on
mostly analog systems, but was fortunate to work on the earliest of the all digital systems.
Also at about this time I left Westinghouse to go with Control Data Corporation who had
been a SCADA vendor but wanted to develop AGC. Within the year we had a AGC
system at CDC and our first order for a System Operation Computer.
Vendors:
AGC evolved from recorders and re-transmitting slidewires in the 40's and 50's. The
primarily vendor at the start was Leeds and Northrup, although by the late fifties both
General Electric and Westinghouse had developed analog systems. These three were the
primary vendors through the 60's. After the 60's Control Data became a active vendor,
and by the early70's Harris Controls, TRW, IBM, and others became involved.
System Operation Computers:
By the late sixties the term System Operation Computers (SOC) became common. These
systems combined the SCADA, and AGC functions, with the forecast and scheduling
functions. The SCADA function was still the most important and basic function. It
became more sophisticated with RTU’s capable of more sophisticated functions such as
local analog data monitoring, local Sequence of Events Reporting (down to a few
milli-seconds), and local display and logging functions. There were even a few cases of
substation computers taking over the RTU function. One of the first, was at PG&E, a job
I was fortunate to work on. Also computer based AGC became common, with the
algorithm becoming more capable, such as reallocating generation limited by
maneuvering limits. The major addition was the inclusion of the forecast and scheduling
functions.
Forecast and scheduling functions.
The three major functions involved are system Load Forecast, Unit Scheduling or Unit
Commitment, and Interchange Negotiation. These functions are so important to the
evolution of System Operation Computers that I will go into some detail about each.
Load Forecast: In order for a utility to schedule the generation it needs to meet the load
each day it needs a accurate forecast of the system load. Since correct scheduling of
system resources can result in the savings of hundreds of thousands of dollars each day to
a large utility (more on this later), a accurate load forecast is essential. During the 40's
and 50's the load forecasting function was generally done by a very experienced
scheduling person in the system operation group. They used patterns of loads that had
actually occurred for different types of days recently. They then considered the detailed
weather forecasts for the scheduling period, usually derived from the National Weather
Service, as well as any events scheduled for the forecast period, such as ball games, fairs
etc. They then put together a hourly forecast for the scheduling period, usually two or
three days. Typically the forecast was put together the night before and adjusted early in
the morning. As the load grew during the day it typically became necessary to adjust the
forecast, however the adjustments hopefully were minor since much of the Unit
Scheduling had already occurred by then. A experienced accurate forecaster was highly
regarded. In the 60's the computer became fast enough to take over the forecasting
function. Load Forecast programs were adaptive in that they used load data from the past
several years to create a forecast based on the day of the year. This forecast was then
adjusted according to the weather forecast, and various scheduler entered data. The
programs typically ran a forecast for a forecasting period up to a week, although only the
first several days were accurate enough for use. This is a very complex function and the
early programs often had so much complexity that, on the computer of the time ,they
would take longer than the forecast period to run (i.e. to run a 24 hour forecast could take
30 hours). Simplifications were made and the programs results then compared over time
with the forecasts made by the experienced operator who had been forecasting. The
computer generated forecast were as good or better, for the second or third day but rarely
beat the scheduler for the next day. However they were consistently reasonable and they
didn’t depend on your forecaster working past his retirement date. The load forecast
programs of the 70's and 80's became ever more sophisticated and accurate.
Unit Scheduling (Unit Commitment). It is necessary for a Utility to schedule enough
generation to be on line at any time to meet the load plus the spinning reserve
requirements. For a large utility this is a complex task. Each unit has a wide variety of
constraints, such as start-up crew scheduling time. Time to minimum load, ramp loading
time once on line, minimum on line time, etc. The units need to be scheduled hour by
hour over the scheduling period, meeting all the constraints, and the load. Since many of
the constraints are over several hours, the schedule for one hour affects the next few
hours. For any hourly load level, there may be 20 combinations of units that will meet the
load. Cost are incurred for each unit that is committed., such as crew start-up costs, unit
start-up costs, fuel production cost, maintenance costs etc.. The scheduling task is to find
the combination of units for each hour of the scheduling period that meets all the
constraints, and is at the minimum over all cost. Since there may be twenty valid
combinations to meet the load each hour, and each must be dispatched and coasted, there
are thousands of unique paths through the scheduling period of several days. However
there is only one optimal path, or in other words the one with the lowest over-all cost. It
is very important to find the right schedule because the difference between the best
schedule and others may be tens of thousand dollars a day , or millions in a years time for
a large utility. Before the use of computers this task was done by a generation scheduling
group, more or less manually, and based on experience. By the late 50's there were
computer programs developed to help with this problem. The early Unit Commitment
programs attempted to find the optimum path by investigating all combinations. The
problem was the running time of the program was longer than the scheduling period, (the
running time for a 24 hour schedule, might be 40 hours or more). Obviously the problem
had to be simplified. Engineers developed the technique of Dynamic Programming which
studied the combinations for the current hour and a few in the future and past to pick a
optimum path. Studies showed the path was not quite as good as checking all paths, but it
was close and much better than could be done manually. There were many variants of
this approach developed in the 70' and 80's, and each required a lot of computing
capability, but were major cost justifications for a larger System Operation Computer.
Interchange Negotiation. During the 40's and 50's most utilities were connected with their
neighbors, and entered into interchange agreements to buy and sell power. There were
two types of contracts, short term(usually next hour), called Economy A, and longer term
(several hours to up to six months) Economy B. Economy A or/and economy B contracts
were often available at any time and the decision on whether to enter in a contract to buy
or sell was complex. The Economy A contracts were easier to evaluate as they primarily
only affected the load level of the units on line. Therefore they could be evaluated by
using a study version of the Economic Dispatch function to determine the system
production cost with and without the contract. Since these opportunities occurred often a
separate program was developed called Economy A Interchange Negotiations. This
program would be run often during the day to evaluate contract opportunities, or to
propose contracts to the neighboring utilities. However evaluating or proposing Economy
B interchange contracts was a much more complex task. These contracts directly affected
the Unit Commitment schedule, and therefore had to be evaluated including unit
commitment. A interchange contract has different constrains than a generating unit so it
cannot just be considered as another potential unit to commit. To properly evaluate each
contract you have to run unit commitment with and without the interchange. If there are
many potential contracts this can result in a lot of runs of the function. To alleviate the
possible long run times of the function, several versions of the Economy B Interchange
Negotiation program utilize a simplified version of unit commitment. Another
complication that happens with some utilities, is the need to include Hydro generating
units. Hydro units have a completely new list of constrains, such as dam outflow, river
level, pond level scheduling etc. They require including a function to optimize hydro with
respect to thermal usage. A few utilities even
have some large units that can be used as pump storage units, this adds a whole new level
of complexity, as now not only do you have to figure out a way to commit a variety of
units, but when to pump and when to generate with the pump storage units.
During the 70's, and later, these programs became more and more sophisticated. They
were a major justification for a large computer system. They were relatively easy to
justify because you could take actual scheduling done manually and compare it to
corresponding runs of the programs and compare the production costs. Every time one of
these studies were done the results showed the programs would save millions of dollars
over a years time.
Configurations:
The most common control system configuration for these early SOC systems, consisted
of a dual computer configuration, with dual ported process I/O, and dual ported
communication interfaces. Because of the requirement for very high availability for the
critical power system control functions, complete redundancy was required. One of the
two process control computers acted as the primary processor in the configuration. The
primary system periodically passed all time critical data to the secondary, in a process
called “check-pointing.” This was done usually over a high speed data link between the
processors. Data was check-pointed as soon as it was received by the primary system. It
was required to complete a fail-over to the secondary system for a primary system failure,
in a matter of seconds. The requirement was to achieve a “bump-less” transfer from the
dispatchers view. Periodic functions such as AGC and SCADA would not be
significantly interrupted. It would not be necessary to complete man/machine functions
such as display call-ups in process, although it was necessary to continue execution of
long running time programs. To achieve these reliability requirements meant all process
I/O, man machine equipment, and communication interfaces, had to be dual ported and
connected to each computer. One of the most important and complex functions of these
systems was Configuration Control, which monitored the system, initiated fail-over, and
allowed operator interaction with these functions. The computers used were process
control capable processors with real time operating systems. By the late 60's the user
interface had evolved from push-buttons to CRT’s. First the CRT’s were character format
black and white units, and were used in conjunction with lighted push-button panels. By
the early 70's, colored CRT’s with a limited graphic character set, came into use. These
could display dynamic one line electric system diagrams. For Cursor control, the
trackball was used in conjunction with the keyboard cluster. Dispatchers were reluctant to
give up lighted push-button panels for keyboards, but quickly adapted to them for the
greatly increased functionality.
Vendors:
In the late 60's the major vendors for SOC systems were General Electric, Westinghouse,
and Leeds and Northrup. CDC, IBM, Harris Controls and TRW became a factor by the
middle 70's. GE used their GETAC 400 series computers, Westinghouse their PRODAC
500 series, and L&N used the Scientific Data Systems SDS 900 series. CDC in the 70's
used their 1700 series, IBM their 1800 series and Harris and TRW used various
machines.
Energy Management Systems
Network Analysis Functions
The addition of network Analysis functions in the middle 70's brought on the
nomenclature Energy Management Systems, (EMS). Dispatchers always had a need for a
fast load flow calculation that they could use to investigate requests being made to allow
clearances of equipment. They needed to know what the affect would be on the system, if
certain equipment was taken out of service. In the 40's and 50's, the only help for them
were engineering studies done by the system planning group using analog network
analyzers, or later on early power flows. These studies could not be done rapidly and
were therefore not of great value for daily operations use. In the middle sixties, there was
a lot of work to develop a faster load flow program for the computer. A load flow
calculation for a large network (i.e. 300 nodes, 700 branches for example), is a difficult
calculation, as the network is complex, involving both real and imaginary values. A direct
solution of the network equations could take several days on the computers available in
the sixties. A iterative approach was developed that greatly reduced the computing time.
The first commonly used method was a “Gaus Siedel” method which involved solving
the equations for each node or “bus” and then iterating through the network. This
typically required several hundred iterations for a solution, but was still much faster than
a direct solution. Later a method know as “Neuton Raphson” was developed which
developed a change matrix relating the change in bus injections to line flows. Iterating
with this method resulted in solutions with four to six iterations. By the late sixties it was
possible to get load flow programs that would solve in 10 to 15 minutes, thus making
them useful to Dispatchers. I was involved in an interesting project in the late sixties
attempting to get a fast power flow for dispatchers use it was a hybrid load flow.. We
developed a analog representation of the network, using operational amplifiers. We then
used the computer to output bus injections to the analog model and read in the line flows
with analog inputs. The analog calculation solved the IYV portion of the calculation and
we iterated through the computer for the bus injections. It took about five or six
iterations, and we were able to get solutions in about 10 seconds for a 300 bus system.
However it was difficult to maintain the analog and digital hardware and the system was
not very reliable. Also the digital computers were rapidly getting faster and faster by the
late sixties making a all digital approach the best choice. I was presenting a paper on the
hybrid load flow at a conference in Minneapolis, when CDC contacted me and I ended up
going with them. The Dispatchers Load Flow was a main part of the Network Analysis
functions but there was another function that became a cornerstone of these systems. It
involved the concept of on line security analysis. The idea was to build a model of the
power system, using load flow techniques, then up-date the model with real time data so
that it closely resembled the actual system in real time. The model then could be
periodically modified with possible contingencies that could happen to the system and
each of these would be checked to report any failures. Possible failures then could be
handled as further contingencies, until a major outage is predicted. If all contingencies
that were feasible or likely, and significant were checked, it would result in the dispatcher
being aware of the current risks to the system. The dispatcher could then make changes to
avoid these risks. It was desired to operate the system such that no first contingencies
would cause further trouble.
It was difficult to use a conventional load flow as the model however, as the telemetered
data did not ordinarily include all the bus injection data as is necessary to up-date a load
flow. Instead the telemetered data included primarily line flows. To help resolve this
problem, a group of engineers at MIT led by Fred Schweppe developed the State
Estimator. This program took the telemetered data base and matched it to a load flow
calculation, using a least squares method, to create a model of the system that could use
all of the telemetered data. It even used redundant values, and indicated which
measurements seemed to be out of line or erroneous. Thus not only did the function
recreate the needed model, it served to verify the telemetered data base. Know that a
model was created that was consistent with a load flow solution and could be checked
with contingencies using a load flow. However for a large system there may be hundreds
of valid contingencies to check, and it was needed to check them as often as every fifteen
minutes. A very high-speed load flow was needed. This resulted in much work to develop
a higher speed load flow. Since the power system network model matrix is sparse (I. E.
not nearly a line between every bus), sparse matrix solution techniques were developed to
speed up the solution. Another technique that was developed by Brian Stout. Mark Enns
and others, was the Decoupled Power Flow where the real and reactive portions of the
solution were separated during the iterations. The result was simplified load flows that
solved in seconds for large networks and allowed the function of Security Analysis to be
accomplished. Later additions to this grouping of functions included the Optimal Power
Flow which allowed production cost to be minimized considering line losses, security
constraints etc.
Configurations;
EMS configurations were made up of various numbers of pairs of computers.
Redundancy is required in order to achieve a high availability. A typical specification
requires 99.8% or better availability for critical functions. One processor was generally
the on-line processor, and the other standby. The standby processor is kept -up-to-date by
a “checkpoint” process that run at frequent intervals. Some vendors configurations were
made up of as few as two main computers while others used as many as six dual
processors of different sizes. One technique used by CDC and others was to use smaller
process control oriented computers for the “front end” functions such as SCADA and
larger computers more capable of engineering functions such as network analysis, and
forecast and scheduling functions. Often the User/Interface functions were also
distributed to two or more separate computers. It was not un-common to see
configurations with up to a dozen computers. The configuration control software was
very sophisticated to control these large distributed processor configurations. In the late
80’s and 90’s, very powerful server oriented processors became available, and the trend
towards distributed processing increased. Some configurations now used thirty or more
of the processors.
Consultants:
As these systems became larger and more complex, it became more and more difficult for
a Utility to specify and purchase a system. Starting in the late sixties to early seventies
there developed a group of consulting companies who were capable of helping the utility
specify, purchase and in some cases implement these systems. The consultants developed
expertise that was difficult for the utility to acquire since the Utility went through the
process of acquiring a new system only ever 10 years or so, and the consultants worked
on several systems a year. Most of the consulting companies in this business evolved
from personnel who worked for vendors, as that was were the detailed work was initially
done. One early consulting firm was MACRO, which was formed mostly from people
that left Leeds and Northrup. ECC was formed from people that left Westinghouse,
EMCA from people that left CDC. Other smaller firms such as Stagg, were from people
that used to work at GE. Systems Control Incorporated, was associated with CDC, and
then became a consultant, and later a vendor. A split off from SCI became ESCA, who
again developed from a consultant to become a vendor. In the later years, late 90’s and
later, there was much consolidation in these consulting firms, and many were bought out
by European companies. ECC and Macro were purchased by KEMA, a Dutch
company and ESCA by a French company.
Vendors:
Systems in the class to be called EMS developed in the seventies.
The EMS
market developed rapidly in the seventies and seemed to peak during the late eighties. At
it’s peak, in the eighties it was not un-common for there to be a dozen bidders on a EMS
specifications, and the spec. could have come from as any as eight different consultants.
Westinghouse , General Electric and Leeds and Northrup evolved into the market from
their involvement with System Operation computers as did Control Data Corporation.
CDC became driver in the market because of their Cyber computers offered the large
scale computing requirements of the network analysis functions. Harris Controls got
involved in the middle seventies, and IBM was a factor several times during the EMS
heaviest activity, becoming active and then fading out. TRW was a major vendor trough
out the period. During the energy crisis of the late seventies, and subsequent reduction in
government defense contracts several defense contractors tried to gain a foot hold.
Boeing and North American Rockwell took several large contracts each during this
period, but left quickly when they found out utilities were not as apt to fore-give cost
over-runs as the government was. Other vendors who were active for a while were
Stagg Systems, and Systems Control Inc Later in the period a split off from SCI, ESCA
became a major factor in the market. Their were close to a dozen vendors active by the
late eighties. By the early nineties there was consolidation in the market as it could not
sustain that many vendors. Westinghouse and GE left, and foreign vendors began to buy
out some of the US vendors. This was at a time, during the energy crisis, when electric
utility load decreased, and traditional US utility apparatus vendors like Westinghouse,
GE etc either got out of the business or combined with foreign vendors. ESCA was
bought out by the French, and CDC by the Germans. By the mid nineties their were fewer
vendors, ESCA, Siemens (CDC), Harris, and other foreign vendors. However as time
went on and the market consolidated, there were some smaller companies that were split
offs of other companies who became active. Two of these that evolved from CDC are
OATI and OSI.
Projects
These projects were large and expensive. Some large EMS systems cost over 40 million
dollars. And they would go on for five or more years. A typical project would take up to
a year for the specification to be released and, the negotiation period with the vendors
often lasted a year before the contract was awarded. Then the project detailed definition
took another year or so, and implementation and testing up to three years. After the
system was shipped, the installation and testing often went on for several years.
Therefore the total involvement between the utility, consultant and vendor on a project
could last as long as ten years, and involve four or five consultants, 10 to 15 utility
personnel and up to fifty vendor people. At different stages of a project, consultants and
utility personnel lived at the vendors site and later vendor people stayed at the utility
location. Therefore their developed a close relationship between the various project
staffs that lasted for long periods of time. As time went on during the project, all of the
people involved usually took ownership of the project and they all worked together for a
satisfactory conclusion. Similar government projects usually resulted in large delays and
cost over runs, however because of the mutual involvement this was usually avoided on
these projects.