8/22/2019 9A0F5d01
1/28
TOPICAL REPORT NUMBER 25 SEPTEMBER 20
Power Plant Optimization
Demonstration Projects
8/22/2019 9A0F5d01
2/28
Cover Photos:
Top let: Coal Creek Station
Top right: Big Bend Power Station
Bottom let: Baldwin Energy Complex
Bottom right: Limestone Power Plant
A report on our projects conducted under separate cooperative
agreements between the U.S. Department o Energy and:
Great River Energy
Tampa Electric Company
Pegasus Technologies
NeuCo. , Inc.
8/22/2019 9A0F5d01
3/28
Power Plant Optimization
Demonstration Projects
Executive Summary .......................................................................................4
Background: Power Plant Optimization ......................................................5
Lignite Fuel Enhancement Project ...............................................................8
Introduction ......................................................................................................... 8
Project Objectives ................................................................................................. 9
Project Description ............................................................................................. 10
Benefts ............................................................................................................... 11
Neural Network-Intelligent Sootblower Optimization Project ..................11
Introduction ........................................................................................................ 11Project Objectives ............................................................................................... 12
Project Description ............................................................................................. 12
Results................................................................................................................. 14
Conclusions ........................................................................................................ 14
Mercury Specie and Multi-Pollutant Control Project ...............................15
Introduction ........................................................................................................ 15
Project Objectives ............................................................................................... 16
Project Description ............................................................................................. 16
Anticipated Benefts ........................................................................................... 19
Demonstration o Integrated Optimization Sotware at the Baldwin
Energy Complex Project ..............................................................................19
Introduction ........................................................................................................ 19Project Objectives ............................................................................................... 20
Project Description ............................................................................................. 21
Anticipated Benefts ........................................................................................... 22
Conclusions ................................................................................................. 23
Bibliography .................................................................................................24
Acronyms and Abbreviations .......................................................................26
8/22/2019 9A0F5d01
4/28
Executive Summary
The Clean Coal Technology Demon-
stration Program (CCTDP) and the two
ollowing programsthe Power Plant
Improvement Initiative (PPII) and the
Clean Coal Power Initiative (CCPI)
are government and industry co-unded
programs. The goal o these programs is to
demonstrate a new generation o innovative
coal-utilization technologies in a series
o projects carried out across the country.
These demonstrations are conducted ona commercial scale to prove the technical
easibility o the technologies and to
provide technical and nancial inormation
or uture applications.
A urther goal o these programs is to
urnish the marketplace with a number
o advanced, more eicient coal-based
technologies that meet increasingly strict
environmental standards. These technologies
will help mitigate the economic and
environmental barriers that limit the ull
utilization o coal.
To achieve these goals, beginning in 1985
a multi-phased eort has been administered
by the U.S. Department o Energy (DOE)
National Energy Technology Laboratory
(NETL). The CCTDP, the earliest program,
initiated ve separate solicitations. The next
program, the PPII, sent out one solicitation,
and the CCPI has had two solicitations to
date. The projects selected through these
solicitations have demonstrated technology
options with the potential to meet the needs
o the energy markets while satisying
relevant environmental requirements.
This report describes our projects
aimed at improving or optimizing the
perormance o coal-red power plants. All
our projects are being conducted under the
CCPI and PPII programs. The rst project
deals with upgrading high moisture lignite
by partial drying to enhance its quality
and improve overall plant perormance.
The remaining three projects involve the
development o sotware that optimizes
overall power plant perormance or some
aspect o perormance by incorporating
eatures o articial intelligence (AI), a
decision-making capability that simulates
the human brain.
The Lignite Fuel Enhancement proj
is demonstrating improved plant per
mance by using waste heat to partia
dry lignite, which is normally high
moisture.
The Neural Network-Intelligent So
blowing (NN-ISB) project with t
Tampa Electric Company (TECO) B
Bend Power Station was intended
demonstrate improved eciency a
lower emissions o nitrogen oxid
(NOX) by using a computer-based neu
network to determine when sootblowi
is needed.
The Mercury Specie and Multi-Pollut
Control project with Pegasus Techno
gies is demonstrating the capability
optimize mercury speciation and cont
o emissions rom an existing power pl
using state-o-the art sensors and neu
network-based optimization sotware
NRG Texass Limestone Station.
The Demonstration o Integrated Optim
zation Sotware project at Dynegy Midw
Generations Baldwin Energy Compl
where NeuCo, Inc., is demonstrating t
integration o ve separate optimizati
computer programs to optimize over
power plant operation.
Great River Energys Coal Creek Station
8/22/2019 9A0F5d01
5/28
Background: Power
Plant Optimization
Overall optimization o a coal-red
power plant is a highly complex process.
One must rst decide what constitutes
optimal perormance. Obvious answers
include maximum thermal eciency,
lowest possible emissions, lowest possible
cost, readily marketable by-products, and
maximum system availability or power
generation. In reality, these goalsand
othersare interrelated. In some cases,however, these optimization goals are at
odds with each other. For example, high
excess air will result in better carbon
burnout and less carbon monoxide but will
also result in higher emissions o nitrogen
oxides (NOX). These interactions must
be kept in mind and addressed with any
optimization program.
There are a number o relatively xed
items that aect overall plant operation.
These include boiler design, cooling water
conditions, burner type, design steam
conditions, and environmental control
systems that capture and remove particulate
matter, sulur dioxide (SO2), NOX, and
mercury. Coal quality is also a major
actor that aects plant perormance. High
moisture and/or ash content decreases
eciency and increases wear and power
requirements on the pulverizers. High
sulur content results in more reagent
consumption and increased by-product
generation.
The benets o optimizing the overall
process o generating power rom coal are
signicant. Eciency is increased, total
maintenance costs are reduced, emissions
are decreased, and reliability is improved.
While the greatest benet can be achieved by
optimizing the overall operation, important
benets can also be achieved by optimizing
one or more o the actors that contribute to
the overall eciency o the plant.
Many optimizations can yield substantial
positive results. For instance, use o lignite
and sub-bituminous coals, which are high
in moisture, lowers the boiler eciency,
increases the load on the pulverizers, and
increases fue gas volume. Drying the coal
beore it is ed to the preparation system is
generally not practical due to the energy
required. Switching to a higher quality
coal, even i available, is oten not practical
either due to cost or to the act that a boiler
designed or a specic coal may not unction
as well with other coals. I such a switch
is made, the unit may need to be de-rated.So, i drying the coal can be economically
integrated into the overall power plant
process, potential benets are substantial.
There are several obvious systems that
can be optimized independently and result
in better perormance. Some involve simply
upgrading a specic piece o equipment.
Power Plant OptimizationDemonstration ProjectsTopical Report Number 25
8/22/2019 9A0F5d01
6/28
For example, a reurbished steam turbine
will improve heat rate and result in less uel
consumption per megawatt (MW). The
cost o electricity is then reduced, as are the
emission rates o some pollutants.
In some cases, optimizing one aspect o
boiler operation can have several benets.
For example, during boiler operation, ash
slowly builds up on boiler tubes. This causes
reduced heat transer to the boiler eed water
and steam which results in lower eciency
and higher NOX emissions. The buildup is
removed by blowing it o the tubes with
high-pressure steam. But when sootblowing
occurs, the electrostatic precipitator (ESP),
or baghouse, is temporarily overwhelmed
by the high particulate load at the inlet.
Sootblowing is traditionally done on a
set schedule rather than as needed. Thisresults in some boiler sections accumulating
excessive ash on the tubes while others,
having little ash buildup, are serviced when
it is not required. Optimized sootblowing
can solve these problems by using sensors
and articial intelligence (AI) sotware
to determine when a particular section o
the boiler needs to have the ash removed
rom the tubes, thus minimizing steam
consumption (and improving heat rate),
reducing the requency o a high particulate
load in the fue gas, and reducing NOX
ormation.
As one would expect, optimizing the
operation o multiple components normally
gives better results than optimizing one
aspect o the operation. Maximizing the
overall perormance o multiple pieces
o equipment does not normally have an
adverse eect on the other power plant
components. However, when using AI/
neural network systems to optimize multiple
aspects o power plant operation, care must
be taken to consider the possible negative
impact on other parameters. This can bestbe accomplished by designing the sotware
packages to communicate with each other
through a management sotware package.
This document describes our opti-
mization projects within the PPII and
CCPI programs. The ollowing are brie
descriptions o the our projects:
In the Lignite Fuel Enhanceme
project, Great River Energy has installed
ull-scale prototype dryer module to supp
one-sixth o the coal required or a 546 M
unit. Results to date have shown improv
perormance in overall operation o t
unit. In the next phase o this project, GrRiver Energy will design, construct, a
perorm ull-scale, long-term operation
testing on a complete set o dryer modu
to supply all the coal needed or the
operation o this unit.
The Neural Network-Intellige
Sootblower (NN-ISB) project with
Tampa Electric Company (TECO) Big Be
Power Station is complete. This proj
showed that the concept o using a neu
network system to optimize the sootblow
process is sound but that additiondevelopment and better equipment
needed. Mechanical problems with sens
and water cannons were encountered a
overall results were aected by these issu
However, some benet was obtained w
respect to stack opacity and nitrogenoxid
reduction.
In the Mercury Specie and Mul
Pollutant Control project, Pegas
Technologies will utilize state-o-th
art sensors and neural-network-bas
optimization and control technologies maximize the proportion o mercury spec
that are easy to remove rom the boi
fue. This project will demonstrate h
integrating sensors, controls, and advanc
analysis techniques into multiple ac
o plant operation can lead to improv
economics and environmental complianc
With the Demonstration of Integrat
Optimization Software project
Dynegy Midwest Generations Baldw
Energy Complex, NeuCo, Inc.,
integrating and optimizing their sotwaSCR-Opt, CombustionOpt, SootOpt
PerormanceOpt, and MaintenanceOpt
ProcessLink is the integration sotw
that coordinates these programs to achie
overall plant goals. The project is ongoi
as o this time and results to date appe
promising.
8/22/2019 9A0F5d01
7/28
Artifcial Intelligence
Articial intelligence (AI) is commonly dened as the science and engineering o makingintelligent machines, especially intelligent computer programs. Relative to applications with coal-red power plants, AI consists o aspects or considerations that deal with the ollowing:
Neural networks, which mimic the capacity o the human brain to handle complex nonlinearrelationships and learn new relationships in the plant environment
Advanced algorithms or expert systems that ollow a set o pre-established rules written in
codes or computer language
Fuzzy logic, which involves evaluation o process variables in accordance with approximate
relationships that have been determined to be suciently accurate to meet the needs o plantcontrol systems
Neural networks (NNs) are a class o algorithms that simulate the operation o biologicalneurons. The NN learns the relationships between operating conditions, emissions, andperormance parameters by processing the test data. In the training process, the NN developsa complex nonlinear unction that maps the system inputs to the corresponding outputs. Thisunction is passed on to a mathematical minimization algorithm that nds optimum operating
conditions.
NNs are composed o a large number o highly interconnected processing elements that workin parallel to solve a specic problem. These networks, with their extensive ability to derivemeaning rom complicated or imprecise data, can be used to extract patterns and detect trendsthat are too complex to be detected by either humans or other computer techniques. NNsare trainable systems that can learn to solve complex problems and generalize the acquiredknowledge to solve unoreseen problems. A trained NN can be thought o as an expert in thecategory o inormation it has been given to analyze. NNs are considered by some to be bestsuited as advisors, i.e., advanced systems that make recommendations based on various types odata input. These recommendations, which will change as power plant operations change, suggestways in which plant equipment or technologies can be optimized.
Advanced algorithms, on the other hand, are programmed to incorporate establishedrelationships between input and output inormation based on detailed knowledge o a specic
process. They are used by computers to process complex inormation or data using a step-by-step,problem-solving procedure. In particular, genetic algorithms provide a search technique to nd trueor approximate solutions to optimization problems. These algorithms must be rigorously denedor any computational process since an established procedure is required or solving a problem ina nite number o steps. Algorithms must tell the computer what specic steps to perorm andin what specic order so that a specied task can be accomplished. Advanced algorithms are nowpart o the sophisticated computational techniques being successully applied to power plants toincrease plant eciency and reduce unwanted emissions.
Fuzzy logic (FL), the least specic type o AI sotware, is equipped with a set o approximate rulesused whenever close enough is good enough. Fuzzy logic is a problem-solving control-systemmethodology that has been used successully with large, networked, multi-channel computers orworkstation-based data-acquisition and control systems. FL can be implemented via hardware,sotware, or a combination o both. Elevators and camera auto-ocusing systems are primaryexamples o uzzy logic systems. Fuzzy logic stops an elevator at a foor when it is within a certainrange, not at a specic point.
FL has proven to be an excellent choice or many control system applications since it mimicshuman control logic. By using an imprecise but very descriptive language, FL deals with input datamuch like a human operator. FL is very robust and provides a simple way to arrive at a deniteconclusion based upon vague, ambiguous, imprecise, or missing input inormation. However, whilethe FL approach to solving control problems mimics human decision-making, FL is much aster. TheFL model is empirically based, relying on operator experience rather than technical understandingo the system.
8/22/2019 9A0F5d01
8/28
Lignite Fuel
Enhancement Project
Introduction
The use o low-rank coals (lignite and
subbituminous) has seen a signiicant
increase in recent years. Because o the
low sulur content o such coals, many units
have adopted uel switching to meet sulur
emissions specications, and other units
have been built specically to burn low-
rank coals. However, a major disadvantage
o low-rank coals is their high moistu
content, typically 25 to 40 percent. Wh
such coal is burned, considerable ener
is required to vaporize the moisture
contains, thus raising the heat rate o
power plant and lowering its eciency.
Fuel moisture has many eects on u
operation, perormance, and emissions.
uel moisture decreases, the uels heati
value increases so that less coal needs to
red to produce the same electric pow
thus reducing the burden on the co
handling system. Drier coal is easier
convey as well, which reduces maintenan
costs and increases availability o the co
to the handling system. When the crush
coal is gravity-ed into bunkers, the dr
coal fows more readily than the wet co
causing ewer eed hopper bridging aplugging problems. Drier coal is easier
pulverize as well so that less mill power
needed to achieve the same coal nene
Finally, with less moisture in the uel mo
complete drying o coal can be achiev
in the mill, which results in an increas
mill exit temperature, better conveying
coal in the coal pipes, and ewer coal pi
plugging problems.
The mixture o pulverized coal a
air rom the pulverizers is combusted
the burners. With drier coal, the fatemperature is higher since there is le
moisture to evaporate. At the same tim
heat transer processes in the urnace
modied. The higher fame temperatu
results in a larger radiation heat fux to t
urnace walls. Also, drier coal results
less moisture in the fue gas, which chang
the radiation properties o the fame. T
change in the fame emissivity also ae
the radiation fux to the wall. With a high
fame temperature, the temperature o c
ash particles is correspondingly high
which could increase urnace ouling aslagging, reducing heat transer and resulti
in a higher fue gas temperature at the urna
exit. However, the reduction in coal fow r
as uel moisture is reduced also reduces t
amount o ash entering the boiler, which lea
to less solid-particle erosion in the boiler a
decreased boiler maintenance cost.
The Clean Coal Technology Program
The DOE commitment to clean coal technology development hasprogressed through three phases. The rst phase was the Clean CoalTechnology Demonstration Program (CCTDP) , a model o government andindustry cooperation that advanced the DOE mission to oster a secure andreliable energy system. With projects completed, the CCTDP has yieldedtechnologies that provide a oundation or meeting uture energy demandsthat utilize the vast U.S. reserves o coal in an environmentally sound manner.Begun in 195, the CCTDP represents a total investment value o over$.25 billion. The DOE share o the total cost is about $1.0 billion, orapproximately 0 percent. The project industrial participants (non-DOE)
have provided the remainder, nearly $2 billion.
Two programs have ollowed that have built on the successes o theCCTDP. The rst is the Power Plant Improvement Initiative (PPII), a cost-shared program patterned ater the CCTDP and directed toward improvedreliability and environmental perormance o the nations coal-burning powerplants. Authorized by the U.S. Congress in 2001, the PPII involves veprojects that ocus on technologies enabling coal-red power plants to meetincreasingly stringent environmental regulations at the lowest possible cost.Four projects have been completed and one is still active. The total value othese projects is $71.5 million, with DOE contributing $1.5 million or .percent.
The second program is the Clean Coal Power Initiative (CCPI), alsopatterned ater the CCTDP. Authorized in 2002, the CCPI is a 10-year
program having a goal o accelerating commercial deployment o advancedtechnologies to ensure that the nation has clean, reliable, and aordableelectricity. Total Federal unding will be up to $2 billion, with a matchingcost share by industrial participants o at least 50 percent. To date, twosolicitations have been completed and nine projects have been awarded orare in negotiation. These projects have a total value o approximately $2.billion. The DOE share is $5 million or 19.9 percent .
8/22/2019 9A0F5d01
9/28
The fue-gas fow rate rom a urnace
ring dry coal is lower than one ring wet
uel, and the specic heat o the fue gas
is lower due to its lower moisture content.
A lower fue gas fow rate also results in
a lower rate o convective heat transer.
Thereore, despite an increase in initial fuegas temperature with drier uel, less heat
will be transerred to the water or steam in
the boiler convective pass.
Drier coal is expected to lower the
temperature o lue gas leaving the
economizer and air preheater (APH).
APH perormance will also be aected by
changes in the ratio o air and fue gas fows
through the APH and changes in specic
heat. Improved overall process eciency
will result rom drier coal as the auxiliary
power decreases due to decrease in orceddrat, induced drat, and primary air an
power as well as decrease in mill power.
Previously, a number o proposals have
been advanced to dry low-rank coals prior
to combustion, but none o these eorts
has resulted in a successul commercial
operation. The two major problems with
drying schemes beore this have been the
cost o the energy required and the act that
low-rank coals become pyrophoric when
dried beyond a certain point. The Great
River Energy Lignite Fuel EnhancementProject overcomes these problems by using
waste heat to dry the coal and removing only
about 25 percent o the moisture, enough to
appreciably improve plant perormance but
not enough to cause handling problems.
Project Objectives
The objective o this project is to
demonstrate an economic process o
moisture reduction o lignite, thereby
increasing its value as a uel in powerplants. The project is being conducted
at the Great River Energys Coal Creek
Station in Underwood, North Dakota. The
demonstration activities ocus on using low
grade condenser waste heat and fue gas
in the plant to lower the moisture content
o the coal by about 10 percentage points
Aerial view o Great River Energys Coal Creek Station lignite power plant
(e.g., reduce the lignite moisture rom 40
to 30 percent). A phased implementation
is planned: In the rst phase, a ull-scale
prototype dryer module was designed or
operation o one o the pulverizers on one
o the two 546 MW units at the Coal Creek
Station.
The objectives o prototype testing were
to gain operating experience, conrm pilot
results, and determine the eect o air fow
rate, bed coils, bed depth, and coal eed
rate on dryer operation in order to optimize
perormance. The lessons learned rom the
prototype were incorporated into the design
o the dryers being installed in the second
phase. A total o our dryers will be built
or Unit 2. Although operating with wet
lignite requires seven pulverizers, six will
provide all the dried lignite required by theboiler.
Following successul demonstration
in the rst phase, Great River Energy is
designing and constructing a ull-scale,
long-term operational test on a complete
set o dryer modules needed or ull power
8/22/2019 9A0F5d01
10/28
10
operation o one 546 MW unit (our
dryers). The coal will be dried to a number
o dierent moisture levels. The eect
o coal drying on plant perormance will
be measured with respect to increase in
plant eciency and availability, reduction
in emissions, and improvements in plant
economics. The dryer design and operating
conditions will be determined or optimal
plant perormance.
Project Description
In response to the rst round o the Clean
Coal Power Initiative, Great River Energy
(GRE) submitted a proposal or a ull-scale
test o a lignite-drying technology that
they had been developing since the 1990s.
The previous work included bench-scale
research and development, eld trials, andpreliminary drying studies. These studies
convinced GRE o the technical easibility
and economic benets o lignite drying and
prompted the submittal o their proposal.
The Department o Energy evaluated and
selected their proposal, and a cooperative
agreement was awarded on July 9, 2004.
The project team or the Lignite Fu
Enhancement Project consists o GR
participant and site provider; the Elect
Power Research Institute, collaborat
Lehigh University, collaborator; B
Engineering, lignite handling; and Falk
Mining and Couteau Properties, lignsupplier. The project is sited at GREs C
Creek Station in Underwood, North Dako
Coal Creek Station is a mine-mouth pla
burning approximately seven million to
o lignite per year and consisting o t
546 MW, tangentially red Combusti
Engineering boilers. Steam is produc
at 2,400 psig and 1,000 oF with a 1,000
reheat temperature. The Coal Creek stati
has eight pulverizers per unit (seven acti
and one spare). The station has two sin
reheat General Electric G-2 turbines.
The gure at let provides a simpli
fow diagram o the lignite drying proce
Warm cooling water rom the turbi
exhaust condenser goes to an air hea
where ambient air is heated beore being se
to the fuidized bed-coal dryer. The cooli
water leaving the air heater is returned
the cooling tower. A separate water strea
is passed through coils in the fuidiz
bed-coal dryer (a two-stage dryer is us
to enhance heat transer). The purpose
these coils is to provide additional heat
the fuidized bed to reduce the amount air required. The dried coal leaving t
fuidized bed is sent to a pulverizer a
then to the boiler. Air leaving the fuidiz
bed is ltered beore being vented to t
atmosphere.
The technical aspects o the proj
are being implemented in two phas
The rst phase involved the constructi
and operation o a prototype dryer, a u
sized dryer with a maximum capacity
112.5 tons/hour (225,000 lb/hour). It w
designed to reduce the moisture conte
o lignite rom 38 percent to 29.5 perce
and improve the higher heating value ro
6,200 Btu/lb to 7,045 Btu/lb. The prototy
unit was ully automated and integrat
into the plant control system. The rst c
was introduced into the prototype dryer
January 30, 2006, and perormance testi
was carried out in March and April 2006
Schematic o Lignite Coal Drying Using Waste Heat From Condenser Water and Flue Gas
8/22/2019 9A0F5d01
11/28
Benefts
Firing drier coal results in improved
boiler eiciency and unit heat rate,
primarily due to lower stack loss and lower
auxiliary power (lower an, pulverizer,
cooling tower, and coal handling power).This perormance improvement will allow
greater electrical output with existing
equipment. Perormance o back-end
environmental control systems (scrubbers
and electrostatic precipitators) will also
improve with drier coal due to the lower
fue gas fow rate and longer residence time.
The reduction in required coal-fow rate and
modied temperature prole will directly
translate into lower emissions o NOX,
carbon dioxide (CO2), SO2, and particulates.
For units equipped with wet scrubbers,
mercury emissions resulting rom ringdrier coal would also be reduced. This is
due to reduced APH gas outlet temperature,
which avors the ormation o mercuric
oxide and mercuric chloride at the expense
o elemental mercury. These oxidized
orms o mercury are water-soluble and can
thereore be removed in a scrubber.
During testing o the prototype coal
dryer in 2006, at a eed rate o 75 tons/
hour (14 percent o total uel rate to the 546
MW unit), there were no major operating
problems. The moisture o the total coalwas reduced by only about 1.1 percentage
points. Yet there were signicant benets
in the prototype dryer operation or the
546 MW unit.
Perormance measures showed that
the lignite fow rate was reduced by 2
percent, pulverizer power was reduced by
3.3 percent, boiler eciency improved
0.5 percent (absolute), net unit heat rate
improved 0.5 percent, NOX emissions
decreased 7.5 percent, and SO2 emissions
decreased 1.9 percent. These results indicatethat there will be signicant improvements
in operations once the project is ully
implemented.
The potential market or GREs coal-
drying technology is quite sizeable.
There are 29 units with a total capacity
o 15.3 gigawatts (GW) that are burning
lignite directly, and another 250 units
with a total capacity o over 100 GW
burning Powder River Basin coal. I all
these units were to adopt coal drying,
the economic and environmental benets
would be quite large.
Neural Network
Intelligent-Sootblower
Optimization Project
Introduction
A neural-network-driven computer
system oers the potential to optimize
sootblowing in coal plant boilers, reduce
NOX emissions, improve heat rate and unit
eciency, and reduce particulate matter
emissions. Installed at the Tampa Electric
Company (TECO) Big Bend Power Station
in Hillsborough County, Florida, the
Pegasus Technologies neural network-
intelligent sootblowing (NN-ISB) system
was designed to be used in conjunction
with advanced instrumentation and water
cannons to prevent soot rom building up ina boiler. O the our coal-red units at the
power station, Unit No. 2 was selected or
installation o the NN-ISB control system.
Fired with bituminous coal, this wet bottom
pressurized Riley Stoker single-drum
radiant boiler has a total o 48 coal nozzles
on a single elevation, 24 on each side, ring
toward the center line o the urnace. The
nal project cost or the NN-ISB control
system (equipment/instrumentation, sotware,
testing, and reporting included) was $3.4
million, including a 27 percent Department
o Energy (DOE) cost share. Sotware costso a ew hundred thousand dollars were a
small part o the total cost. Project testing
o the NN-ISB was completed in December
2004, and the nal report on the system was
issued in September 2005.
Tampa Electric Big Bend Power Station
8/22/2019 9A0F5d01
12/28
12
Project Objectives
The project objective at the Big Bend
plant was to develop a neural-network-
driven system that could initiate, control,
and optimize sootblowing in response to
real-time events or conditions within the
coal-plant boiler rather than relying on
general rule-based protocols. The project
demonstrated and assessed a range o
technical and economic issues associated
with the sensing, management, display,
and human interace o sootblowing goals
as they relate to emissions and eciency
o a coal-red utility boiler. Specically,
this optimization process targeted reducing
baseline NOX emissions by up to 30 percent,increasing unit eciency by 2 percent, and
reducing particulate matter (stack opacity)
by 5 percent.
Project Description
This neural network project was
implemented under the Power PlantImprovement Initiative (PPII), a DOE
program designed to demonstrate plant
improvement technologies and processes
in commercial settings. At the time o the
award, this installation was the rst domestic
project to use neural network technology to
optimize the sootblowing process within
a boiler. Started in 2001 ater a series
o brownouts and blackouts had plagued
major regions o the country, the initiative
targeted new technologies that could help
coal plants boost their output and improve
their environmental perormance. The BigBend project was designed to be a ull-
scale demonstration o the neural-network-
driven technology on a large commercial
boiler, using state-o-the-art controls and
instruments to optimize boiler operation
and systematically control boiler slagging
and ouling.
Communications Architecture at Big Bend Power Station
In a coal-red boiler, the continui
buildup o ash and soot on the boiler tub
leads to reduced boiler eciency. I perio
ash and soot removal (sootblowing) is nperormed, this leads in turn to higher f
gas temperatures and ultimately to high
NOX ormation and reduced ecien
Thereore, cleaning the heat-absorbi
suraces is one o the most importa
boiler auxiliary operations. Typical
sootblowing uses mechanical devices
online cleaning o reside boiler ash and s
deposits on a periodic basis. Sootblow
clean by directing steam or water throu
nozzles against the accumulated soot a
ash on the heat-transer suraces in ord
to remove the deposits and maintain hetranser eciency. Basically, sootblow
consist o our components: a tube or lan
that is inserted into the boiler and carries t
cleaning medium, nozzles in the tip o t
lance to accelerate and direct the cleani
medium, a mechanical system to insert
rotate the lance, and a control system.
8/22/2019 9A0F5d01
13/28
Because it either uses steam that would
otherwise be used to generate electric
power or it requires energy or pumps or
compressors, sootblowing has a direct
impact on plant eciency. Thus, optimizing
sootblower operation is important in
maximizing unit eiciency. Typically,sootblowers operate on a specied timed
cycle or, alternatively, operation is initiated
by an operator who believes sootblowing
is needed. The purpose o an intelligent
sootblowing system is to decide when to
sootblow based on inormation rom boiler
instruments. The overall objective is to
sootblow when, and only when, necessary.
The Pegasus Technologies NN-ISB
control system uses a neural network to
model the characteristics o the boiler.
Designed to recognize patterns in inputdata, this network must be trained using
historical data beore it can associate a
particular pattern with a corresponding
plant state. Once this training has been
completed, the system can respond rapidly
to new inputs. An advantage o a neural
network is that i any inputs are aulty
the prediction capability degrades only
gradually compared to most other modeling
techniques.
The project installed at Big Bend Unit
No. 2 includes 16 heat fux sensors, 8 slagsensors, a heat transer advisor, acoustic
pyrometers, a sootblower control system,
an online perormance monitor (OPM),
and an advanced calibration monitor
(ACM). For the communications layout,
the combustion optimizing system and
intelligent sootblowing (ISB) sotware
were loaded into one computer. For this
application, the models were partitioned
so they could unction separately or work
interactively. This approach was important
since it permits upgrades to existing power
plants as well as applications to new boilers.Although the demonstration was carried
out on the hardware and sotware systems
developed or this project, the equipment
(including the distributed control system)
could be obtained rom any manuacturer.
Ater verication that the core elements
o the NN-ISB system were satisactorily
installed and operational, detailed model
tuning was completed. During this task,the unit was operated under a variety
o conditions, including some non-ideal
variations. This helped to dene acceptable
operating limits and constraints used by the
neural network while optimizing the system.
During system optimization, appropriate
adjustments were made to allow the system
to learn and to make recommendations
on Unit 2 operation, including both manual
or advisory (open-loop) and automatic
(closed-loop) operation. The advisory mode
provided recommendations to the operatorsand engineers, who used those results to
urther tune the system. This activity
also proved very valuable in assessing and
recording the perormance and status o the
new sensors and systems.
Pegasus Technologies NN- ISB Control System
8/22/2019 9A0F5d01
14/28
1
NRG Texass Limestone Power Plant in Jewett, Texas
Results
The automated closed-loop activation
o the sootblowers during this projectconrmed that neural-network, adaptive
sootblowing can benet eciency. There
was a clear improvement at low loads, with
the benet decreasing as the load increased.
During closed-loop operation o the NN-
ISB, Pegasus reported that eciency gains
were in the range o 0.1 to 0.4 percentage
points compared to baseline. Results with
open-loop operation were slightly lower.
With more operating experience, gains at
the high end o the load range should be
achievable.
NN-ISB closed-loop (automatic)
operation was shown to be better than
open-loop (non-neural network baseline)
operation. Other Pegasus results indicated
an improvement o 1.0 to 1.5 percent in
opacity or closed-loop compared to open-
loop operation during certain tests.
While it is reasonable to expect th
optimizing sootblowing would be benec
or NOX reduction (due to an improv
temperature prole in the urnace),
Big Bend project was unable to clea
demonstrate this. Supporting equipm
and material issues (e.g., unavailabilo the water cannons during the N
ISB tests, underperormance o much
the instrumentation) greatly limited
optimization sotware rom perorming
expected.
Prior to this project, sensors and contr
related to sootblowing were usually trea
as isolated systems. In contrast, the Big Be
NN-ISB system had the ability to understan
evaluate, and optimize the process as an ent
system with multiple, real-time objectiv
Integration o the sensors went well acommunication was established to the neu
network system with all sensors and eleme
o the project. The project demonstrated t
such systems can be linked together desp
the use o proprietary networks. Further
conrmed that the sensors can provide d
that can be correlated to achieve a set
objectives. Generally, the NN-ISB syst
appears to have merit and can improve boi
perormance.
Conclusions
The major conclusion rom th
project is that the Pegasus Technolog
NN-ISB control system is a sound id
with signicant potential. The Big Be
project successully demonstrated a neu
network, closed-loop operation on a u
scale boiler without causing unit ups
or violating any constraintsand it a
achieved operator acceptance. The NN-I
appears to provide generating compan
with an integrated solution that will ass
in optimal economic and environmenreal-time, online operation o a unit.
The NN-ISB is modular in design a
can be readily applied to a variety o pow
generating units. The solution architectu
and inrastructure are designed to all
ull or staged deployment, depending
8/22/2019 9A0F5d01
15/28
the needs o the generating company. The
technology applied throughout allows unit
fexibility (i.e., existing systems can be
integrated within the overall solution) and
is extendible (new modules/new equipment
can be readily modeled and incorporated).
In general, the project provided a testing
ground or several innovative measurement
devices and eedback on their operation
that may also lead to improved instruments.
Since some equipment and instrumentation
(e.g., water cannons, heat fux sensors,
slag sensors, and acoustic pyrometers) did
not ully operate as expected during this
testing, an additional project with improved
equipment and instrumentation may be
needed in order to ully quantiy all the
benets. Other project goals were also
achieved:
Promoted the use o coal by making
coal more uel-ecient automatically,
reducing all pollutants on a per megawatt-
hour basis. In addition, reducing NOXemissions should lower the resistance to
coal use or electrical generation.
Enabled rapid deployment into the
market. All coal-ired boilers employ
sootblowers which, in turn, require
control systems. Since current systems
cannot achieve the desired results
in sootblowing operations, a neural
network control system appears to oer
signicant advantages. Further, no new
hardware needs to be developed since the
hardware is o the shel and readily
available.
Expanded U.S. revenues through world-
wide market acceptance. The same rapid
deployment capability and acceptance by
domestic plants should apply to oshore
coal-ired boilers. Since the United
States is presently the world leader in
AI (o which the neural network system
is a subset) there should be minimal
competition rom oshore suppliers.
In summary, the project provided
valuable inormation on neural networks
and the positive results should encourage
other power plants to install these systems
to control sootblowing, improve boiler
eciency, reduce NOX emissions, and
improve other aspects o their operations.Although equipment and instrumentation
issues may have precluded the NN-ISB
project rom achieving all o its goals, the
project clearly demonstrated the validity o
using AI to control a major aspect o boiler
operation.
Mercury Specie and
Multi-Pollutant
Control Project
Introduction
Implemented under the CCPI, the proj
at the NRG Texas (ormerly Texas Genc
Limestone Power Plant in Jewett, Texas
designed to demonstrate the capability
optimize mercury speciation and cont
emissions rom an existing power pla
NRG Texas, with a generating capacity
more than 14,000 MW, has plants primar
based on ossil uels and is an importa
producer o electricity in Texas.
Perormed by Pegasus Technologi
Inc., a division o NeuCo, Inc., this demo
stration is occurring on an 890 MW util
boiler that uses 14,500 tons o coal p
day. The Pegasus technology provid
plant operators with the ability to asse
detailed plant operating parameters th
aect mercury capture eciency, over
heat rate, particulate removal, and fue g
desulurization (FGD) eciencies. The
data are also provided to a neural netwo
optimization system that controls pla
subsystems to provide the lowest possipollutant emissions, highest heat rate, a
least risk o environmental non-complian
all with minimal capital expenditure.
NRGs Limestone Power Plant
8/22/2019 9A0F5d01
16/28
1
Once demonstrated, this technology
is anticipated to have broad application to
existing coal-red boilers and a positive
impact on the quality o saleable by-
products such as ly ash. The project
began in April 2006, with perormance
testing targeted or December 2008. Thisestimated $15.6 million project will be
38 months in duration, with a DOE cost
share o 39 percent.
Project Objectives
On a large utility coal-ired boiler,
Pegasus Technologies is demonstrating
the ability to aect and optimize mercury
speciation and multi-pollutant control
using non-intrusive advanced sensor and
optim ization technologies. Plant-wideadvanced control and optimization systems
are being integrated into a coal-red, ste
electric power plant in order to minim
emissions while maximizing the ecien
and by-products o the plant. Advanc
solutions utilizing state-o-the-art sens
and neural-network-based optimizati
and control technologies are being usto maximize the portion o the mercu
vapor in the boiler fue gas that is oxidiz
or captured in particle and chemical bon
resulting in lower uncontrolled releases
mercury.
This neural-network-based control a
optimization system gathers data rom c
composition, combustion gas compositio
mercury species, eed rates, etc., and u
this inormation to optimize power pl
operations. The greatest advantage o neu
networks in power plants is their abilitygeneralize rom previous inormation a
develop possible similar patterns or utu
use. Such intelligent control is expected
improve mercury capture by over 40 perce
reduce NOX emissions by 10 percent, redu
uel consumption by 0.5 to 2.0 percent, a
improve operating fexibility.
Project Description
The estimated 48 tons o mercu
emitted annually by domestic coal-rpower plants is about one-third o t
total amount o mercury released annua
rom all human activities in the Uni
States. Mercury emissions take a num
o chemical ormsor speciesincludi
the pure element, as part o a gaseo
compound, or bound to particulates
fue gas. Certain mercury species, su
as mercury that is adsorbed onto fy-a
particles or bound in the FGD, are relativ
easy to remove rom fue gas. Adjust
certain parameters during combusti
can optimize the speciation process amaximize the mercury captured in parti
bonds. This results in greater capture
mercury and lower uncontrolled releases
Control System Schematic or NRG Texas Limestone Power Plant
8/22/2019 9A0F5d01
17/28
The NRG Texas demonstration power
plant is equipped with a tangentially-red
boiler that uses a blend o 70 percent Texas
lignite and 30 percent Powder River Basin
subbituminous coal, which are known to
emit relatively high levels o elemental
mercury under routine combustionconditions. Pegasus Technologies will
apply sensors at key locations to evaluate
the mercury species (elemental and
oxidized mercury), develop optimization
sotware that will result in the best plant
conditions to promote mercury oxidation
and minimize emissions in general, and
use neural networks to determine the
optimization conditions.
The unit is equipped with a cold-side
ESP rated at approximately 99.8 percent
particulate removal eciency and a wetlimestone FGD system rated at approximately
90 percent SO2 removal eciency. Both
devices are capable o high mercury-capture
eciency, especially when the mercury is in
an oxidized state rather than an elemental
vapor state.
Using a neural network to aect and
optimize mercury speciation and multi-
pollutant control, the non-intrusive advanced
sensor and optimization technologies will
act as a highly trained operator, making
decisions on inputs to the process bymeasuring and learning the outputs. By
using AI and simulation technologies,
Pegasus will minimize the use o raw
material resources and pollutant emissions
while simultaneously optimizing the
operating capabilities o the plant.
This project involves the installation and
demonstration o sensors and optimization
sotware in six separate technology
packages. While the modular design is
transparent to this project, it is important to
the uture marketing o this system becauseo the fexibility needed with utilities to
include or exclude a particular module
based on either the existing equipment or
budget or a specic plant. Many o the
sensors and optimizer technologies that will
be installed are utilized across the modules;
thereore, they have been included under
the module in which they are most used.
The technology packages or this project
include the ollowing:
The intelligent uel management system
(FMS): The FMS is composed o the
Pegasus Combustion Optimizer system,
the Ready Engineering CoalFusion
system, and a Sabia elemental analyzer.
The mercury specie control system:
This system includes the boiler areaoptimization, Pegasus virtual online
analyzers, and various sensors. Mercury
emissions will be measured through
continuous emission monitors.
The advanced ESP optimization
system: The ESP optimization system
is composed o a carbon-in-ash virtual
online analyzer, a carbon-in-ash sensor,
and Pegasus ESP optimization sotware.
The advanced ISB system: The ISB
system is made up o Pegasus ISBsotware. This module has been
previously demonstrated.
The advanced FGD optimization system:
The FGD System is composed o Pegasus
FGD optimization sotware.
Key locations where sensors will be applied to evaluate mercury specie s
8/22/2019 9A0F5d01
18/28
1
The intelligent plant (unit optimization):
This is the Pegasus i-Plant Optimization
System that will contain a simulator and
will arbitrate among the solutions or the
above systems. This system will interace
with users through a commercially
available computer.
Each technology package includes
non-intrusive sensors and the appropriate
sotware needed or data acquisition,
optimization, and integration with the
overall neural network. In using this
approach, all acets o coal-red power
plant operation will be optimized by
balancing the inputs and outputs o the plant
within a realm o multiple constraints. The
intended result is to improve the eciency
o plant operations while operating within
regulatory and commercial constraints.
During the rst o three perormance
phases, sensor installation, sotware system
design, and baseline operating metric testing
will be completed. Instruments or instrument
technology packages to be installed include
a coal elemental analyzer (part o the uel
management system), mercury sensors,
coal fow sensors, laser-based urnace gas
speciation sensors, online carbon-in-ash
sensor (located in the ESP), communications
links or data acquisition and control, and
related computers, controllers, and Pegasusoptimization products.
Baseline testing will be perormed
to establish comparative data or the
operational testing that will ollow in
Phase 3. Ater initial baseline testing,
parametric testing will be perormed to
exercise various combinations o control
variables to determine their eect on
mercury speciation and by-product
generation and to determine overall plant
perormance. These data will be used in
Phase 2 to adjust the neural network oroptimization control.
During Phase 2, sotware installation,
data communications modication, and
distributed control system modication
will be achieved. The test plan data and
historical data (i applicable) will be
evaluated to conrm that no irregularit
exist prior to model development. A
extraneous data (e.g., calibrations) a
eliminated rom the data set, operati
issues and constraints will be review
as part o urther model developme
Control models will be developed characterize the eect o control variab
on the operational characteristics o t
boiler, mercury speciation, and by-produ
generation. Models will be created th
accurately and robustly represent the ee
that changes in the unit have on the outp
to be optimized. Beore the control mod
are implemented in an online system, ofi
simulation will be perormed. The mod
will then be evaluated and demonstrat
to Limestone Power Plant operators a
engineers so their input can be used
nalize the behavior o the models.
Pegasus uses pre-designed and custo
methods or constraining the models und
various design and operational limitatio
These are dynamic constraints that fuctu
with load, number o burners in service, r
o change, etc. Ater the initial modeli
is completed, a shorter series o tests w
be conducted. These will involve setti
up operational parameters to veriy
predictive capabilities o the neural netwo
model and to assure that the model h
been properly trained. During this perithe models will be coarse-tuned. Cont
loops will rst be tested one at a time a
then as groups to deal with the individu
loop characteristics beore dealing with t
interactive characteristics.
At the end o Phase 2, a decision w
be made whether to initiate work und
Phase 3 or to conclude the project a
the successul demonstration o close
loop operability or neural networks a
controllers.
Phase 3 plans include demonstrati
and validation o all systems as well as
comparison o the test results with the proj
objectives. Extended mercury and mu
pollutant testing will be conducted. T
technology packagesthe uel managem
system, combustion and mercury cont
8/22/2019 9A0F5d01
19/28
system, ESP system, ISB system, FGD system,
and intelligent plant (i-Plant) systemwill
all be demonstrated during closed-loop
operation. Operator and engineering training
will also be conducted during Phase 3.
Anticipated Benefts
In this project, Pegasus Technologies and
NRG Texas are attempting to put together
all o the required best-o-class articial
intelligence and simulation technologies
to prove that mercury speciation and
multi-pollutant reduction benets can be
measured, optimized, and controlled. I
successul, Pegasus will demonstrate the
capability o sophisticated control processes
and advanced sensor technologies to
simultaneously reduce harmul emissionso mercury and increase plant eciency.
Increased control o SO2, NOX, and
particulate matter should also result, along
with a reduction in water usage. Since
these technologies are designed to control
and optimize all major acets o power plant
operations, the demonstration is expected
to provide the capability to maximize plant
eciency or electricity production while
reducing mercury emissions. This project
is also expected to address concerns that
higher mercury concentrations in existing
by-products, such as ash, may adverselyaect the commercial value o those by-
products.
This project should demonstrate an
operating environment that simultaneously
oers higher-than-average compliance with
environmental requirements and better
control o emissions, resulting in both a
smaller risk o non-compliance to the utility
and minimization o capital expenditures.
In general, the project is expected to
demonstrate how integrating sensors
and advanced controls into a total plantsolution can lead to improved economics
while being environmentally compliant.
The technologies being demonstrated are
expected to have widespread application
since they can be directly retrotted to the
existing coal feet or integrated into uture
new plant designs.
Demonstration
o Integrated
Optimization Sotware
at the Baldwin EnergyComplex Project
Introduction
As part o the CCPI, sophisticated
computational techniques are being applied
to an Illinois coal-red power plant to show
how new technology can increase power
plant eciency and reliability and reduce air
emissions. NeuCo, Inc., o Boston, MA, is
designing and demonstrating an integratedonline optimization system at the Dynegy
Midwest Generation power plant located
in Baldwin, IL. The Baldwin Energy
Complex (BEC) consists o two 585 MW
cyclone-red boilers with selective catalytic
reduction (SCR) and a 595 MW tangentially
red boiler with low NOX burners.
Dynegy Midwest Generations Baldwin Energy Complex
8/22/2019 9A0F5d01
20/28
20
The ve system optimization modules
being developed include cyclone combustion,
sootblowing, SCR operations, overall unit
thermal perormance, and maintenance
optimization. This project builds on the
NeuCo proprietary ProcessLink technology
platorm. Power plants operate in manydierent conditions and plant processes are
highly complex and interrelated. The goal
o optimization is to continuously assess and
adjust (or provide actionable advice about) the
settings o the many variables aecting plant
perormance so that the optimal balance o
plant emissions, uel eciency, capacity, and
reliability is achieved. The total cost o this
45-month project is estimated at $19 million,
including a 45 percent DOE cost share.
Project Objectives
The overall objective o applyi
integrated optimization sotware is
improve the emissions prole, ecien
maintenance requirements, and pla
asset lie or coal-based power generatiin order to extend the use o abund
coal resources in the United States in
environmentally sound manner. In gener
sotware optimization oers seve
advantages to power plants, includi
the ability to control key parameters on
consistent basis, compensate or chang
in coal quality, optimize controls to m
specic plant objectives, and help
understanding the available data and
use or improved operations. The proj
at BEC will demonstrate and quant
the environmental and emissions beneassociated with deployment o a u
integrated set o sotware solutions
optimization o plant perormance or co
red power generation.
Because retrots, repowering, mod
cations, and new technologies are stead
increasing the complexity o modern pow
plants, an integrated process-optimizati
approach is required to maximize equipm
perormance and minimize operating cos
Optimization solutions are now availab
or a variety o power plant control systemlinking these systems together will provi
overall plant-level optimization that
expected to yield additional benets.
Thereore, the primary objective
this project is to demonstrate integrati
o existing controls and control system
sensors, and computer hardware w
advanced optimization techniques at BE
and to link the individual optimizati
modules through the NeuCo ProcessLink
platorm. Collectively, these modules
expected to provide optimization solutioor this 1,765 MW coal-red power pla
by reducing emissions, increasing pla
eciency, and increasing the availability
the plant or power generation.
The Baldwin Energy Power Plant at sunset
8/22/2019 9A0F5d01
21/28
Project Description
NeuCo is designing and demonstrating
an integrated online optimization sotware
system or the Dynegy Midwest Generation
power plant using advanced computational
techniques that are expected to achieve peakperormance rom the three coal-red units
at the energy complex. NeuCo is using its
ProcessLink technology platorm o neural
networks, expert systems (heuristics), rst
principle models, advanced algorithms, and
uzzy logic to maximize perormances rom
the power plant combustor, soot removal
system, and emission controlsthe rst time
that all o these modules have been integrated
into a computerized process network.
Five separate modules are being
designed and demonstrated by NeuCo atBEC and then integrated to provide unied
plant optimization.
CombustionOpt uses neural network-
based optimization, model predictive
control, and other technologies to
extract knowledge about the combustion
process, determine the optimal balance
o uel and air fows in the urnace,
and respond to changing conditions.
CombustionOpt directly adjusts the
distributed control system to more
consistently position the dampers, burner
tilts, overre air, and other controllable
parameters at their optimal settings or
given sets o conditions, objectives, andconstraints. This module should reduce
uel consumption and NOX emissions; it
should also improve carbon monoxide
control, reduce opacity, and improve loss
on ignition.
SootOpt optimizes sootblowing to
reduce adverse ouling conditions
and unplanned outages that soot and
sootblowing can cause. It also expands
Planned Integration Concept or the Baldwin Energy Complex
eiciency improvements, reduc
emissions, and drives boiler-cleani
actions toward optimal plant heat ra
emissions, and reliability goals. T
SootOpt adaptive neural network mod
identiy the equipment and actions m
eective or achieving plant ecienreliability, and emissions objectiv
and then bias control activity towa
those objectives. The neural mod
work within boundaries dened
expert rules to ensure all applicable un
specic constraints are considered.
SCR-Opt uses neural-network-bas
optimization, model predictive contr
and other technologies to make t
operation o the SCR as eicient
possible. SCR-Opt is expected
minimize ammonia usage and reduNOX emissions.
PerormanceOpt uses a rigoro
rst-principles-based thermodynam
model o the boiler and steam cyc
to conduct both what is and wh
i simulations o unit operatio
Continuously monitoring the actu
versus the expected perorman
levels o key equipment and proce
Control Room at the BaldwinEnergy Complex
8/22/2019 9A0F5d01
22/28
22
conditions, PerormanceOpt detects
when perormance deviates rom what
is achievable under current operating
conditions and calculates the impact o
that deviation so that remedial actions
can be prioritized. PerormanceOpt is
expected to improve heat rate, steamtemperatures, and generating capacity.
MaintenanceOpt helps engineers
manage the entire lie-cycle o reliability,
capacity, and eciency problems more
eciently and eectively. It uses neural
network technology to constantly search
or gaps between actual and expected
behavior across a broad range o
process and equipment health variables.
Its powerul diagnostics knowledge
base also helps to lter alse alarms,
determine the root causes and correctiveactions o identied problems, and aid
in problem resolution and tracking.
MaintenanceOpt is expected to increase
annual power output and assist in
providing lower costs o electricity to the
consumer.
These advanced computational capa-
bilities will be used to comprehensively
optimize a variety o systems within BEC by
using existing control technologies and then
linking these systems to each other. This
innovative project will provide solutionsthat use system-speciic optimization
applications as data sources and actuators.
In general, the overall architecture o this
control platorm is designed to permit
lexible deployment strategies. Rather
than requiring that all data and logic reside
on a single computer, the service model
allows applications to leverage networked
computational resources. The application
architecture is built around interoperable
services that should result in more ecient
plant operations. The planned integration
concept is shown below.
This integrated optimization sotware
project at BEC consists o two phases.
Phase I, which has been completed, entailed
the development and installation o initial
versions o each o the ve optimization
modules, as well as their integration through
the ProcessLink platorm to address theull scope o plant operations and relevant
system interactions.
Extensive operating experience will be
required to quantiy the benets associated
with control system optimization. The goal
during Phase I was to establish each system
and demonstrate its role in unied plant
optimization. Phase I activities ocused
on developing, deploying, integrating,
and testing prototypes or each o the ve
optimization modules; identiying and
addressing issues required or the modulesto integrate with plant operations; and
systematically collecting and assimilating
eedback to improve subsequent module
releases.
The goal o Phase II is to improve
upon the sotware installed and tested in
Phase I and to perorm rigorous analysis
o operating data in order to quantiy the
benets o the integrated system. Phase
II entails quantication o results at BEC;
renement o the sotware installed and
demonstrated in Phase I to support additionalcommercial releases o the ve products;
installation and beta testing at BEC; and
commercialization o the solutions, taking
into account both what is learned at BEC
and eedback systematically incorporated
rom other operators o U.S. coal-red
power generation plants.
During both phases, best practices
iterative sotware development methods
will be applied toward integration,
ull-scale demonstration, and eventual
commercialization o these ve solutions. Allsystem sotware engineering, applications
engineering, and systems integration will
proceed through a multi-step, iterative
process that supports a structured, modu
approach to determining sotware a
hardware requirements and unction
denitions along with various desig
development, test, installation, and start
activities. This iterative developme
process is specically designed to depla commercially viable product as soon
possible, while at the same time applyi
host-site eedback and what is learn
toward maximizing unctionality a
benets in subsequent releases.
Anticipated Benefts
This optimization initiative is expect
to reduce NOX emissions by 5 percent a
increase thermal eciency by 1.5 perce
The increased thermal eciency is expectto reduce emissions o CO2, mercury, a
particulates. Ammonia consumption
expected to be reduced by 15 perce
accompanied by a one-year extension o t
SCR catalyst. The optimization initiat
is expected to also result in improv
power plant capacity and reliability whic
in turn, is expected to increase net annu
electrical power production by 1.5 perce
Consumers should benet through low
electricity costs.
The NeuCo ProcessLink architectuoers plant operators a highly fexib
control system platorm. Optimizat i
modules can be designed and applied
individual subsystems in a plant, leveragi
existing sensors, actuators, and network
computational resourcesand then linki
them to other individual subsystems
provide overall plant-level integration
controls responsive to plant operator a
corporate criteria. This integrated proc
optimization approach will likely be
important tool or plant operators as pla
complexity increases through retrot arepowering applications, installation o n
technologies, and plant modications.
8/22/2019 9A0F5d01
23/28
Conclusions
The U.S. Department o Energy Clean
Coal Technology programs continue to
sponsor projects that develop technologiesor optimizing power plant operation.
These technologies help keep the cost
o electricity low, reduce emissions, and
conserve our uel supply.
The Lignite Fuel Enhancement project
demonstrates a technology that reduces
the moisture content o low-rank coals,
which results in a number o benets to
power plant operation. Coal consumption
is reduced, thereby reducing CO2 and other
emissions. Parasitic power requirements
are also reduced. When the technology isapplied to new plants, capital costs will be
reduced in several major subsystems, such
as SO2 removal, pulverizers, and cooling
towers. This technology achieves these
benets using only waste heat to remove the
moisture rom the uel.
While the Neural Network-Intelligent
Sootblower project did not reach all o its
goals, the NN-ISB control system was ound
to be a sound idea with signicant potential.
The project successully demonstrated a
neural network, closed-loop operation on aull-scale boiler without causing unit upsets
or violating any constraintsand also
achieved operator acceptance.
Although not yet completed, the
Mercury Specie and Multi-Pollutant
Control project and the Demonstrationof Integrated Optimization Software
project have demonstrated the ability o NN
and AI technologies to provide signicant
economic, operational, and environmental
beneits to power plant operation. The
demonstrated technologies are applicable
to all types o coal-red boilers and do not
require the purchase o major equipment.
Given the benets and relatively low cost,
these types o technologies are likely to nd
a ready market.
The continuing development o sot-ware to control overall power plant
operation, or selected aspects o it, has
shown substantial progress in optimizing
power plant perormance. These sotware
packages allow operators to more easily
stay within their emission limits while
improving power plant eiciency and
lowering the cost o power production.
While doing so, the participants have
produced products that are expected to
nd a substantial market both in the United
States and abroad.
8/22/2019 9A0F5d01
24/28
2
Bibliography
Bullinger, Charles, Mark Ness, and Nenad Sarunac. Coal Creek Prototype Fluidized
Bed Coal Dryer: Perormance Improvement, Emissions Reduction, and Operating
Experience, Paper presented at 31st International Technical Conerence on Coal
Utilization and Fuel Systems, Clearwater, Florida, May 21-25, 2006.
Gollakota, Sai. Great River Energy Project Benets Presentation. Paper presented on
February, 2007.
Great River Energy. Great River Energy Lignite Fuel Enhancement Proposal to DOE
Solicitation DE-PS26-02NT4142 8Public Abstract, August 1, 2002, http://www.
netl.doe.gov/technologies/coalpower/cctc/ccpi/proposal-pd/greabs.pd.
Ness, Mark, and Charles Bullinger. Pre-Drying the Lignite to CREs Coal Creek Statio
May, 2005.
NeuCo, Inc. Technical Progress Reports #11 to the U.S. Department o Energy, Oce o
Fossil Energy, National Energy Technology Laboratory. Cooperative Agreement DEFC26-04NT41768. September 30, 2006.
NeuCo, Inc. Technical Progress Report #12 to the U.S. Department o Energy, Oce o
Fossil Energy, National Energy Technology Laboratory. Cooperative Agreement DE
FC26-04NT41768, February 5, 2007.
Tampa Electric Company and Pegasus Technologies, Inc. Project Perormance and
Review: Neural Network Based Intelligent Sootblowing System, Tampa Electric
Company Big Bend Unit #2. April 2005
U. S. Department o Energy. Clean Coal Technology Programs: Program Update 2006,
September 2006, pp. 3-54, http://www.netl.doe.gov/technologies/coalpower/cctc/ccp
pubs/2006_program_update.pd.
U. S. Department o Energy, Oce o Fossil Energy, National Energy Technology
Laboratory. Advanced Coal Conversion Process Demonstration: A DOE
Assessment, April 2005, http://www.netl.doe.gov/technologies/coalpower/cctc/cctd
bibliography/demonstration/pds/rsbud/NETL-1217_as%20sent%20to%20OSTI.pd
U.S. Department o Energy, Oce o Fossil Energy, National Energy Technology Laborator
Big Bend Power Station Neural Network-Intelligent Sootblower Optimization.
Project Brie, Sept. 2005. http://www.netl.doe.gov/technologies/coalpower/cctc/PPI
bibliography/demonstration/environmental/neural/bigbenddemo.pd
U.S. Department o Energy, Oce o Fossil Energy, National Energy Technology
Laboratory. Big Bend Power Station Neural Network-Intelligent Sootblower
Optimization. Project Fact Sheet, Dec. 2006. http://www.netl.doe.gov/publicationsactsheets/project/Proj233.pd
Contacts or Participants
in CCT ProjectsDavid Wroblewski, Senior VP
Development
Pegasus Technologies100 Seventh Avenue, Suite 210
Chardon, OH 020-25-779
John McDermott, Vice President,
Product ManagementNeuCo, Inc.
200 Clarendon Street, HancockTower, T-1Boston, MA 0211
Charles Bullinger
Great River Energy
275 Third St., SWUnderwood, ND 557-959
Mark Rhode
TECO Energy
P.O.Box 111Tampa, FL 01
1-22-152
8/22/2019 9A0F5d01
25/28
U. S. Department o Energy, Oce o Fossil Energy, National Energy Technology
Laboratory. Big Bend Power Station Neural Network-Sootblower Optimization: A
DOE Assessment, DOE/NETL-2006/1234, June 2006. http://www.netl.doe.gov/
technologies/coalpower/cctc/PPII/bibliography/demonstration/environmental/neural/
BigBendSootblowerPPA_Final_061306.pd
U.S. Department o Energy, Oce o Fossil Energy, National Energy Technology
Laboratory. Cleaner Air by the Numbers. TechLine, March 17, 2004. http://www.
netl.doe.gov/publications/press/2004/tl_ccpi_neucoaward.html
U.S. Department o Energy, Oce o Fossil Energy, National Energy Technology
Laboratory. Demonstration o Integrated Optimization Sotware at the Baldwin
Energy Complex. Project Brie, 2006. http://www.netl.doe.gov/technologies/
coalpower/cctc/ccpi/bibliography/demonstration/environmental/ccpi_demo.html
U.S. Department o Energy, Oce o Fossil Energy, National Energy Technology
Laboratory. Demonstration o Integrated Optimization Sotware at the BaldwinEnergy Complex. Project Fact Sheet, Dec. 2006. http://www.netl.doe.gov/
publications/actsheets/project/Proj221.pd.
U.S. Department o Energy, Oce o Fossil Energy, National Energy Technology
Laboratory. DOE, Tampa Electric Sign Agreement to Add Intelligent Computer
System to Florida Power Plant. TechLine, August 21, 2002.TechLine, August 21, 2002. http://www.netl.doe.gov/
technologies/coalpower/cctc/PPII/bibliography/demonstration/environmental/neural/
neural_tech.pd
U.S. Department o Energy, Oce o Fossil Energy, National Energy Technology
Laboratory. Mercury Specie and Multi-Pollutant Control. Project Brie, 2006.
http://www.netl.doe.gov/technologies/coalpower/cctc/ccpi/project_bries/CCPI_
Pegasus.pd.
U.S. Department o Energy, Oce o Fossil Energy, National Energy Technology
Laboratory. Mercury Specie and Multi-Pollutant Control Project. Project Fact Sheet,
Dec. 2006. http://www.netl.doe.gov/publications/actsheets/project/Proj372.pd
U.S. Department o Energy, Oce o Fossil Energy, National Energy Technology
Laboratory. Pegasus Project Selected as Part o Clean Coal Power Initiative.October 28, 2004. TechLine. http://www.netl.doe.gov/publications/press/2004/tl_
ccpi2_pegasus.html
Contacts or U.S.
Departmento Energy CCT Program
John Rockey, Project ManagerNeural Network-Intelligent
Sootblowing Optimization
National Energy TechnologyLaboratory
10 Collins Ferry RoadP.O. Box 0
Morgantown, WV 2507-00
Sai Gollakota, Project Manager
Lignite Fuel Enhancement
National Energy TechnologyLaboratory
10 Collins Ferry RoadP.O. Box 0
Morgantown, WV 2507-000-25-151
George Pukanic, Project Manager
Demonstration o IntegratedOptimization Sotware
National Energy TechnologyLaboratory2 Cochrans Mill Road
P.O. Box 1090Pittsburgh, PA 152-090
12--05
Michael McMillian, Project ManagerMercury Specie and Multi-Pollutant
ControlNational Energy Technology
Laboratory
10 Collins Ferry Road
P.O. Box 0Morgantown, WV 2507-000-25-9
8/22/2019 9A0F5d01
26/28
2
To Receive Additional
Inormation
To be placed on the Departmento Energys distribution list oruture inormation on the Clean
Coal Demonstration Programs,the projects they are nancing,
or other Fossil Energy Programs,please contact:
John L. Grasser, DirectorOce o Communication
FE-5/Forrestal BuildingU.S. Department o Energy
1000 Independence Ave., SW
Washington, DC 2055202-5-0
202-5-51 [email protected]
David J. AnnaOce o Public Aairs Coordination
U.S. Department o EnergyNational Energy Technology
Laboratory
P.O. Box 1090
Pittsburgh, PA 152-09012--12--195 ax
Acronyms and Abbreviations
ACM Advanced calibration monitor
AI Articial intelligence
APH Air preheater
BEC Baldwin Energy Complex
CCPI Clean Coal Power Initiative
CCTDP Clean Coal Technology Demonstration Program
CO2 Carbon dioxide
DOE Department o Energy
ESP Electrostatic precipitator
FGD Flue gas desulurization
FL Fuzzy logic
FMS Fuel management system
GRE Great River Energy
ISB Intelligent sootblowing
MW Megawatt
NETL National Energy Technology Laboratory
NN Neural network
NN-ISB Neural network-intelligent sootblowing
NOx Nitrogen Oxides
OPM Online perormance monitor
PPII Power Plant Improvement Initiative
SCR Selective catalytic reduction
SO2 Sulphur dioxide
TECO Tampa Electric Company
8/22/2019 9A0F5d01
27/28
8/22/2019 9A0F5d01
28/28
National Energy
Technology Laboratory
1450 Queen Avenue SW
Albany, OR 97321-2198
541-967-5892
2175 University Avenue South, Suite 201
Fairbanks, AK 99709
907-452-2559
3610 Collins Ferry Road
P.O. Box 880
Morgantown, WV 26507-0880
304-285-4764
626 Cochrans Mill Road
P.O. Box 10940
Pittsburgh, PA 15236-0940
412-386-4687
One West Third Street, Suite 1400Tulsa, OK 74103-3519
918-699-2000
Visit the NETL website at:
www.netl.doe.gov
Customer Service:
1-800-553-7681
U.S. Department o Energy
Oce o Fossil Energy
Printed in the United States on recycled paper