Industrial Gas Turbine

Performance Improvements Through

Advanced Controls & Modeling

Tim Healy
April, 2009
1
The Difference Between…

…Failure,
… and Success,

… Often Rests Heavily on The Control System

2
 Increasing Generation Diversity Requires
Increasing Flexibility From All Sectors
There Exists Significant
Opportunity To Improve
Nuclear

Thermal
Performance & Emissions
In The Thermal Sector
Through Advanced Control
& Modeling
Nuclear Gas Cleaner Coal
Renewables

Biomass Wind Solar Hydro

3
Thermal Sector Remains A Very Large Part of
The Generation Portfolio
Projected World Electricity Generation by Fuel
Trillion Kilowatt-Hours
35
Coal
Natural Gas
30
Liquids
Renewables
25
Nuclear

20

15

10

0
2005 2010 2015 2020 2025 2030
Source: History: Energy Information Administration (EIA), International Energy Annual
2005 (June-October 2007), Projections: EIA World Energy Projections Plus (2008) 4
A Dramatically Revised Outlook for ‘09
2009 economic outlook
8.7%
Last year’s outlook (April 2008) 8.2%
Current outlook (January 2009)
6.4% 6.5%
5.1%
4.7%
3.2%
2.2%
1.7% 1.6% 1.7%
1.4%

-0.5% Russia Middle China India

-1.5% East
World -2.5% -2.3%
USA Eurozone Japan

World real GDP growth slowed from about 4% in 2006 and

2007 to 2.4% in 2008, expecting -0.5% in 2009
Source: Global Insight Outlook, April vs. December 23, 2008
24
Outline

Industrial Gas Turbines Short-Course

• Legacy Control Algorithms
• Model-Based Control for Fuel Flexibility
• The Road Ahead

6
A Sense of Power

x 10

Ford Shelby GT500

~500 SHP

GE Evolution Locomotive GE 9H Industrial GT Engine

~5000 SHP ~500,000 SHP (combined-cycle)

x 10 x 10

GE-90 Aircraft GT Engine

~50,000 SHP 7
 Gas Turbine Plant - Simple Cycle
Fuel
Air
3
1
Comb 4
2
Stack
Gen Comp Turb

Heat Source 3
T
COMBUSTION

GT
TEMPERATURE

BRAYTON GAS CYCLE

ON

4
2
COMPRESSI

K
STAC

1
ENTROPY S

8
 Gas Turbine & Steam Turbine - Combined Cycle
9

HRSG
8 7

Integrated Combined Cycle

ST Gen

10
Pump Cond
6
5
Fuel
Air
1 Comb 3 4
2
Gen Comp GT

Heat Source 3
T
COMBUSTION
GT
TEMPERATURE

BRAYTON GAS CYCLE

4
ON

GAS
2 UST 9
COMPRESSI

HRSG EXHA
ST

8 RANKINE
STACK
7 STEAM CYCLE
5, 6 CONDENSER 10
1 Heat Sink
ENTROPY S
9
 Industrial Gas Turbine Overview
Fuel Combustor
Flow Exhaust
Inlet
Flow Turbine Flow
Compressor

Shaft

10
Can–Annular Combustion Systems

Cross-Section Through
One Chamber

Chamber Arrangement
on Gas Turbine

Multiple Fuel
Nozzles

11
Industrial Gas Turbine Operability
(Also Known as Control Requirements)

Hot Gas Exhaust

Fuel Path Frame
System Durability Durability
Operability
Compressor
Aero-
Mechanics
Combustor
Flashback
(Flameholding)
Compressor
Surge
Combustor
Emissions
(NOx, CO, UHC)
Power Combustion
Output Dynamics Combustor
Lean Blow-
Optimal Out (LBO)
Auto-
Efficiency
Ignition

12
Outline

• Industrial Gas Turbines Short-Course

Legacy Control Algorithms
• Model-Based Control for Fuel Flexibility
• The Road Ahead

13
Typical Industrial Gas Turbine
Sensor/Effector Suite

Inlet Bleed Heat (IBH)

Compressor Fuel Splits

Inlet Guide Effectors
Actuator stroke feedback and some
Vanes (IGV) Total Fuel Flow (Wf)
fuel system pressures not shown

Ambient Pressure Generator Inlet Exhaust Pressure Drop

Power Pressure Fuel Temperature
Ambient Drop Exhaust Temperature
Temperature
Generator Inlet Temperature Compressor
Losses Discharge
Inlet Humidity Pressure

Sensors Compressor
Discharge
Temperature

14
Sensor-Based Control Approach
P3
3 P 2=
Maximum Cycle
Temperature Ideal Brayton Cycle
Temperature

Expansion
e

Isentropic
ssur
2 tan t Pre on Work Output (T3 − T4 ) − (T2 − T1 )
s i
Con at Addit η Cycle = =
Compression

He Heat Added (T3 − T2 )

Isentropic

P 1= P4
γ γ −1 γ γ −1
4 P2  T2  P T 
=  = 3 =  3  Problem:
1 P1  T1  P4  T4 
Entropy (1−γ ) ( γ −1)
Desire To Control T3,
T 
η Cycle = 1 −  3  But T3 is Not Measured
Comp  T4 
Turb
1 2 Comb 3 4 Higher T3 = Higher ηCycle

Turbine Efficiency
∆T
η turbine =
∆T'
Solution:
T4 T3 = f ( T4 , PRc )
Correlate T3 to a T3 =
   for assumed ηt and PRc ~= PRt
Measured Variable 1 -η turbine 1 − 1 
  ( )
P3 γ −1 γ 
  P4 

T3 = f ( T4 , ηt , PRt )
15
Indirect (Schedule-Based) Boundary Control
Fuel

Splits
X ~ Tx Splits
• Pre-Programmed
X
T4 Control Schedules

MINIMUM
T4_req + Wf / IGV • Field-Tuned For
PRc P+I
- Performance &
T4_max T4
Operability
PRc

Characteristics
• Simple • No Explicit Accommodation Of Machine
(Easily Understood and Verified)
Deterioration
(New & Clean / Mean Machine Assumption)
• Approximate Boundary Protection
(Accommodates Worst-Case Condition)
• Coupled Effectors Prohibit Optimization
(Part-Load Exhaust Temperature & Fuel Splits)
• Poor Accommodation Of Ambient/Fuel
Variation
(Impact to Emissions, Combustion Dynamics, LBO Margin)

16
Outline

• Industrial Gas Turbines Short-Course

• Legacy Control Algorithms

Model-Based Control for Fuel Flexibility
• The Road Ahead

17
 Gas Fuel Composition Variation
98
Trinidad 14 Abu
96 USA Dhabi

Composition Variation
Content [%]

12
94

Content [%]
Methane

Norway
92 10
Nigeria

Will Increase As More

Ethane
Algeria Oman
90 Algeria Qatar
8 Qatar
Malaysia Malaysia
88 Abu Oman

LNG Is Injected Into

Dhabi 6 Norway
86 Nigeria

84 4 Trinidad
USA

Pipelines
82
2
80
US (Typical) Abu Dhabi Algeria Malaysia Nigeria Norway Oman Qatar Trinidad
0
Geographic Origin US (Typical) Abu Dhabi Algeria Malays ia Nigeria Norway Oman Qatar Trinidad

Geographic Origin

4.5 1500
Malaysia
4

Wobbe Index
Content [%]

3.5 Oman Oman

1450 Abu Malaysia
Propane

Nigeria
3 Dhabi Nigeria Qatar
Qatar
2.5 Algeria
Norway
Norway 1400
2 Trinidad
Abu
USA
1.5 Dhabi
Algeria
1350
1
USA
0.5 Trinidad

0 1300
US (Typical) Abu Dhabi Algeria Malays ia Nigeria Norway Oman Qatar Trinidad US (Typical) Abu Dhabi Algeria Malaysia Nigeria Norway Oman Qatar Trinidad

Geographic Origin Geographic Origin

HHV Wobbe Index

WI = Sg

LHV
MWI = S g ⋅T
Modified Wobbe Index

HHV, LHV Fuel Higher/Lower Heating Value [BTU/Scf]

Sg Fuel Specific Gravity

T Fuel Temperature [°R]

18
 What Is At Risk?

Gas turbine operability concerns due

to composition variation:

Addressed by gas • Auto-Ignition

fuel specification
(given expected variation,
• Flashback
not an issue for most pre-
• Emissions (NOx, CO) Addressed today
mixed combustion systems) by manual tuning
• Combustion Dynamics
(given expected variation,
• Blow-out potentially a very serious issue)

Tuning is required to protect against fuel

composition variation

19
Gas Fuel Composition Rate-of-Change

NP
Significant & rapid NG

shifts in “Null-Point”
are possible NG

NP

NG
LNG

• Rate and frequency of pipeline

composition changes will increase NP
LNG

• An automatic tuning process is Fictitious

required to support continuous & region /
LNG pipeline
reliable operation

20
Legacy Solution –
Closed-Loop MWI With Fuel Temperature

IP Feedwater
Dual GCs Control

Performance
Heater

Characteristics
• Costly • Limited Authority
(Dual Gas Chromatographs) (Performance Heater Capability)

• Low-Bandwidth • Sub-Optimal Efficiency

(GCs & Fuel Heat Exchangers) (Any Off-Nominal Fuel Temperature)

21
Direct (Model-Based) Boundary Control
+_
Limit Scheduling +_ Loop IGV
+_ Selection
+_
Loop Wf (SISO vs. MIMO:
+_
+_
Selection
Fuel
Industrial GT System
+_
+_
Loop Splits Coupling & Time Scale
+_
Selection Does Not Demand
MIMO Control,… Yet)
0 . 16 Physics-Based
 
W (6 . 394 * SH )  Boundary Models Virtual
3 . 95 * e
  .006*(Tfl −Tfl ref ) ARES - Parameter
NOx@ 270
T3 
 
O2 =
 NOxref * e Sensors
 P3 * e
1 . 25 15 %
 Estimation
− 9.5( SH − SH ref )
*e *Q

Engine Model

Characteristics
• Robust / Flexible / Expandable • Accommodation Of Machine Deterioration
(Additional Boundaries / Loops) (Adaptive Model Ensures Accurate Virtual Sensors)

• Direct Boundary Protection • Implicitly De-Coupled Effectors

(Physical Space of Boundary) (Automatic Performance Optimization)

• Good Accommodation Of Ambient / Fuel

Variation
(Manages Emissions, Combustion Dynamics, LBO Margin)

22
Adaptive Real-time Engine Simulation (ARES)
Model
• Non-Linear Component-Level Cycle Model
• Optimized for Real-Time Operation
Filter
• Extended Kalman Filter Formulation
• On-Line Jacobian & KF Gain Calculation
• Re-configurable for Fault Accommodation
• Avoids Parallel ‘Linear Model’ Process
Measured Measured
Inputs Outputs
On-Line Partial Derivative Calculation
u y
Estimated xˆ , yˆ
Outputs
u ARES - Parameter
ŷ prt
Partial
Deriv.
ARES - Parameter ŷ +
x̂ prt
Estimation
Calc.

Estimation _ Engine Model

ŷ ext
On-Line Filter Gain Calculation
Engine Model
Extended
Outputs P = a ⋅ P ⋅ aT + Q (Covariance of
Prediction Error)
a, J
“State”
Estimate
K s = J ⋅P⋅ JT + R (Covariance of
Q, R
Residual)
K = P ⋅ J T ⋅ s −1 (Gain Matrix)
x̂
+ P = P− K ⋅J ⋅P (Covariance of
Prediction Error)

+
P
Z-1 Z-1 23
Model-Based Control Adapts Well To
Environmental / Fuel Variation
e1 Fuel_

max
NOx e2 Fraction
Control

min
x (target) eNOx eNOx
+_

NOx
Environment
Limit Scheduling +_

Loop-In-Control
+_
GT

Structure
+_
+_
Effectors
+_
+_
+_
+_
+_

CDM
Sensors

ARES - Parameter
Estimation
Physics-Based Virtual
Sensors
Boundary Models
Engine Model
NOx @15%O2 =
f ( Tflame, Humidity, Tflame, Tfire,
Fuel_Fraction ) W2, etc.

24
Integrating Models, Sensors, & Algorithms
Physics-Based Boundary Models Adaptive-Model Approach
1.5

15
[ppm@15%O2]
[psi]

14 Site A
Site B
13 Site C Closed-Loop
Dynamics

Site D
12 Site E (10% C2)
Control
11
XR
+_
1.0

10
NOx

9
Boundary
Predicted

X
Predicted

8
7
Sensor
6
0.5

5
0.5 1.0 1.5 X̂ Boundary
5 6 7 8 9 10 11 12 13 14 15
Measured Model
Measured Dynamics [psi]
NOx [ppm@15%O2]

( Performance
Impact )
Design ( Small
Center Performance Load
Impact ) Runback

( No Performance Fuel
Performance optimization
Impact )
Fuel
Temp.
through hierarchical
Fraction
application of effectors
Expected Wobbe
LNG Range

25
Model-Based Control Performance
Field Test Closed-Loop Simulation
6%
MWI Reaching Combustor

56 400

Wobbe Index (WI)

Fuel Temperature [degF]
54 350 4%

Change [%]
52 300 2%
50 250 • 7FA+e DLN2.6 gas 0% • Closed-loop
48 200
46
MWI
Fuel Temp. 150 turbine operating in -2% simulation of model-
44 100 combined-cycle at -4% based control
42 50 -6%
40 0 base-load algorithm (7FA+e
-20 0 20 40 60 80 100
0 200 400 600 DLN2.6, base-load,
Time [sec]
Time [sec]
ISO Day)
• ~260ºF fuel

Gas Turbine Output [%]

10 10 120
temperature

NOx [ppm@15%O2]
NOx [ppm @15%O2]

NOx
9 excursion imposed 9 Load 110 • ~10% WI change
8 (~20% MWI) over 8 100 imposed over ~30
five minutes (max seconds (rate
7 7 90
capability of fuel >18%/minute)
6 80
6 heat exchanger)
0 200 400 600 -20 0 20 40 60 80 100
Time [sec] Time [sec]
• OpFlex Wide Wobbe
• OpFlex Wide algorithm maintains
120
Amplitude [% Of Target]

120
emissions &

Amplitude [% Of Target]
Wobbe system
Combustion Dynamics

Combustion Dynamics
Frequency 1 Frequency 1
100 100
80
Frequency 2
maintains Frequency 2 dynamics levels
80
60
emissions & using fuel
60
40 dynamics levels 40
distribution only
20 using fuel 20
0 distribution only 0
0 200 400 600 -20 0 20 40 60 80 100
Time [sec] Time [sec]

26
Assessment
The Model-Based Control system provides many
advantages over competing technologies with similar
objectives:
Cost
 No additional auxiliary equipment required beyond control system sensor
redundancy. No gas analyzer required
Operability
 Negligible change in output or efficiency as a result of changing fuel
properties
 Lower combustion dynamics across the operational envelope
 Improved output & efficiency at off-design conditions
Reliability
 Increased system availability due to sensor fault detection and
accommodation
Emissions
 Tighter NOx control over a wider operational envelope

27
The Road Ahead
Advanced Controls & Modeling Will Play A Greater Role In
Thermal Sector Technology / Solutions
• Fuel Flexibility
• Integrated Gasification / Combined-Cycle
• Plant-Level Optimization
• Grid-Code Compliance
• Health Management

28
Fuel flex … expanding the envelope
Power producers seeking fuel diversification & flexibility
• Increasing fuel prices & volatility driving substitution
• Cleaner & more flexible technology … lower emissions, increased turndown,
multi-fuel, durability

Gas fuels Liquid fuels Synthetic fuels

• NG … LNG wide wobbe • Light crude … • Pet coke … refining

• High BTU … hydrogen/EOR • Heavy crude … • Coal syngas … IGCC/SNG
• Low BTU … Steel BFG/COG vanadium & sulfur • Biofuels … ethanol

29
Integrated Gasification Combined-Cycle

HRSG
ST Gen

Cooling Pump Cond

Clean-
O2 Up Syngas
Gasifier Air
Comb
Feed
Prep. Gen Comp GT
Fuel + H2O

Electricity /
Steam

as
Gasifier Syn g

Sulfur
Removal

Combined Cycle
Solid feed – Slag Sulfur Power Block
Gas/Liquid feed - Ash

30
Plant-Level Optimization
Model Predictive Controls for
Combined-Cycle Plant Start-Up Optimized
Load Profile
• Physics-based models to predict stresses
• real-time optimization to choose best loading profile
• Handles multiple ST Stress constraints simultaneously Time
• Handles multiple control actions simultaneously
• Accommodates any initial thermal state of the plant

Stress
constraints
HP & IP maximum rotor stresses
Final CC load Time

MPC Controller GT load

reference
GT, HRSG, ST models Optimize Control
HP & IP rotor stresses GT loading System
over Time
Horizon
State estimation

Measurements Measurements
Steam & metal Temperatures, Steam Pressures

31
Back-Up

32
2007
2030
Combustion Operability
Dynamics

Cold Tone Hot Tone

Window NOx

NOx
Dynamics Dynamics
Window
Limit Guarantee

Dynamics NOx
Fuel-Air Ratio
Fuel-Air Ratio

Operability
Window
Lean
Blow Out CO
Tfire (Power)

Lean Window CO Window

Blow

CO
Out Guarantee

Fuel-Air Ratio
Fuel-Air Ratio
35