nerc gads benchmarking workshop 2011 (1)

Using Reliability Data to Improve Power PlantImprove Power Plant

PerformanceNERC-GADS Workshop

presented by Robert R. (Bob) Richwinepresented by Robert R. (Bob) RichwineReliability Management ConsultantRichwine Consulting Group, LLC

Oct 27, 2011

Workshop Agenda

I. Background and Case StudyI. Background and Case StudyII. Common Elements in Successful Programs

A. Awareness PhaseA. Awareness PhaseB. Identification PhaseC. Evaluation PhaseD. Implementation Phase

III. Transforming to a Market-Driven Business gEnvironment

Richwine Consulting Group, LLC 2

Background

• From a 2006 Wall Street Journal article

– Business today is awash in data and data hcrunchers

Only a few companies use data as a strategic– Only a few companies use data as a strategic weapon

– The ability to collect, analyze and act on data is the essence of a company’s competitive advantage


Survey Results in WSJ

• 450 executives; 370 companies; 35 countries; 19 industries

• Identified a strong link between extensive and sophisticated use of analytics and sustained high performancehigh performance

T f i i 5 ti• Top performing companies were 5 times more likely to single out analytics as critical to their competitive edge

Richwine Consulting Group, LLC

to their competitive edge4

A substantial gap exists between actual and potential performance

Potential Performance

Actual PerformanceActual Performance


The Worldwide Value of Closing the Gap (WEC estimate)

E i• Economic– US$80 Billion per Year

• Environmental– 1 Billion Tons of CO2 Reduction

per Year and Proportional Reductionsper Year and Proportional Reductions of Other Emissions


Source of Performance Variation

• Variation of performance due toVariation of performance due to– Technology/mode of operation = 20-25 % – Human factors/management = 75-80 %– Human factors/management = 75-80 %

C fi d b• Confirmed by– Analytical studies (Reference 2)– Practical experiences


Closing the Gap

Better Use of Reliability Data is a Key Factor in Achieving and Sustaining Top Performance


NERC-GADS

• The NERC-GADS system is the premier reliability data collection and analysis system worldwidedata collection and analysis system worldwide

• NERC has been collecting power plant reliability data in the GADS format for 30 yearsin the GADS format for 30 years

• An increasing number of international companies have begun using the NERC-GADS system to collect and analyze their plant’s performance

• The World Energy Council has adapted NERC-GADS for international usefor international use

NERC-GADS

• Some companies have used the GADS database in innovative ways to help them achieve topinnovative ways to help them achieve top performance of their generating plants

• NERC created the Generating Availability Trend Evaluation (GATE) working group to analyzeEvaluation (GATE) working group to analyze reliability trends and their causes

• A new NERC-GATE type group is now being organizedorganized

• GATE’s publications can be found on NERC’s website at www.nerc.com

• The World Energy Council has summarized many of• The World Energy Council has summarized many of the GATE studies and has published them, along with other similar studies on its website at www worldenergy orgwww.worldenergy.org

30+ WEC Published Case Studies

• Objective – Demonstrate that the valueObjective Demonstrate that the value of performance data is far greater than the cost of collecting the data plus thethe cost of collecting the data plus the risk of sharing the datawww worldenergy org• www.worldenergy.org• Click on “Performance of Generating Plant”

under “Work Programme”under Work Programme• Click on “Performance of Generating Plant” on

right side of pageright side of page• Click on “Case Studies”

WEC Case Study Topics include the Use of Data in:

• Performance ImprovementB h ki

• Maintenance Planning• Risk Management

• Benchmarking• Configuration

Optimization

• Catastrophic Event Reduction

Optimization• Generation Planning

O ti

• Life Management• Equipment Design

• Operations• Goal Optimization

• Peak Season Performance


Southern Company ExperienceMay 2004 WEC Case Study (Ref 4)

AlabamaPower

GeorgiaPower

MississippiPower

Gulf

SavannahElectric

GulfPower

Southern Company HeadquartersAtlanta, GA, USA

Southern CompanyCoal-Fired Power Stations

Equivalent Availability Factor (EAF) q y ( )Trend

9 39 5

S o u th e rn C o m p a n y W o rld

8 38 58 78 99 19 3

7 37 57 77 98 1

6 56 76 97 1

1 97 0 19 71 19 72 19 73 1 97 4 1 97 5 1 97 6 19 77 19 78 19 79 1 98 0 1 98 1 19 82 1 98 3 1 98 4 19 85 19 86 19 87 1 98 8 1 98 9 1 99 0 19 91

Reasons for Performance Decline

I bilit t id d t• Inability to provide adequate resources to power stations

• See May 2004 WEC-PGP case study and Ref 4 for detailed discussion ofand Ref 4 for detailed discussion of other reasons including:• Limited use of performance data p

Southern Company Coal-FiredPower Stations

Equivalent Availability Factor (EAF) Trend

9 5

1 0 0

S o u th e rn C o m p a n y W o rld

q y ( )

8 5

9 0

9 5

7 5

8 0

6 0

6 5

7 0

6 019 70 19 71 19 72 19 73 19 74 19 75 19 76 19 77 19 78 19 79 19 80 1 98 1 1 98 2 19 83 19 84 19 85 19 86 19 87 19 88 19 89 19 90 19 91

Reasons for Improved Performance

H i ht d f ti t• Heightened awareness of executive management of the need for performance improvementC it t f dditi l f• Commitment of additional resources for performance improvementS M 2004 WEC PGP t d d R f 4 f• See May 2004 WEC-PGP case study and Ref 4 for other reasons including:• Improved data collection analysis and application of results• Improved data collection, analysis and application of results

Performance Improvement Benefits

Savings of >US$1 billion (1991$) per year equaled:~12 percent of Annual Revenue

~100 percent of Net Income (Profit)

• Annual avoided emissions included seven million tons of CO2e per year at2 p y

NO EXTRA COST!

Availability Improvement at Other Generating Companies

PREPA – Puerto Rico +25%

NEES – USA +13%

ESB – Ireland +10%

ESKOM – South Africa +20%

Observations

E h / t f it t f• Each company/country faces its own set of challenges, constraints, and opportunitiesNo single program is optimal for e er• No single program is optimal for every company/countryThere are common elements within each• There are common elements within each successful program


Common Elements In Successful Improvement Programs

January – April 2003 WEC Case Studiesy p

Performance Improvement


Performance Improvement Process Phase 1 - Awareness

• Benchmarking

• Forecasting

• Communications


AwarenessJanuary 2003 WEC Case Study

• BenchmarkingApril 2002 WEC Case Study– April 2002 WEC Case Study

– August 2002 WEC Case Study– September 2003 WEC Case StudySeptember 2003 WEC Case Study– ASME technical papers (Ref 5,6)


Unit Level Benchmarking Unit Level Benchmarking The First Step in Improving Plant PerformanceThe First Step in Improving Plant Performance

Wh B h ki ?Why Benchmarking?• Set realistic, achievable goals

Id if b f i l i• Identify best areas for potential improvement• Give advance warning of threats

T d k l d d i ith• Trade knowledge and experience with peers• Quantify and manage performance risks

Create increased awareness of the potential for• Create increased awareness of the potential for and the value of increased plant performance

Reliability Benchmarking Process

Id tif li bilit i bl t d th• Identify reliability variables to measure and the databases required

• Select peer power plants having similar design or modeSelect peer power plants having similar design or mode of operations characteristics

• Compare the candidate power plant’s reliability against these peer plants

25Richwine Consulting Group, LLC

Reliability Benchmarking Process

Identify reliability variables to measure and the• Identify reliability variables to measure and the databases required:

Typical Reliability VariablesTypical Reliability Variables

– Equivalent Availability Factor (EAF)E i l t F d O t R t (EFOR)– Equivalent Forced Outage Rate (EFOR)

– Scheduled Outage Factor (SOF)and increasingly,and increasingly,

– Equivalent Forced Outage Rate Demand (EFORd)


Benchmarking Process

• Select peer power plants having similar design or mode of operations

h t i ticharacteristics:

Selection Procedure (NERC/Richwine developed)– Selection Procedure (NERC/Richwine developed)• Advanced statistical methodology• Has been applied numerous times over theHas been applied numerous times over the

past 20 + years at companies and countries around the world


Peer Selection Criteria

Large Population

NERC-GADS Data Base5000 + units



Exact Match

x xxx xx

xx

x xx

xxx

xx

xxx

x x xxx xNumber of Exact Matches 0Number of Exact Matches 0



Exact MatchesLarge Population

Must Balance CriteriaMust Balance Criteria30Richwine Consulting Group, LLC


Etc Etc.Etc.

Etc.VintageFiring

Fuel

ASSUME

Etc.g

DutyBoilerManufacturer

EtcAge

Manufacturer

TurbineCriticality

Etc.

Etc. SizeDraft

TurbineManufacturer



Etc Etc.Etc.

Etc.VintageFiring Fuel

ANALYSISEtc.

DutyBoilerManufacturer

EtcAge

Manufacturer

TurbineCriticality

Etc.

Etc. SizeDraft

TurbineManufacturer



Significance TestingSignificance TestingSubcritical Supercritical Cyclic DutyBaseload Duty

EFOR EFOREFOR EFOR33Richwine Consulting Group, LLC

Peer Groups Select Criteria Fossil Units

All Fossil Units

CRITICALITY

SubSuper

CRITICALITY

MODE OF OPERATIONVINTAGE

Cycling<1972 BaseloadSize

Draft TypeFuel

Boiler Mfr.Draft Type

Size

≥1972

Fuel Size


Does Peer Selection Make a Difference?

SUPERCRITICAL TECHNOLOGY

EARLY VINTAGE RECENT VINTAGE

EFOR(mean) 15.60% 9.68%EFOR(mean) 15.60% 9.68%EFOR(median) 12.17% 8.08%EFOR(best quartile) 8.14% 5.47%



EFOR - PLANT A

OLD CRITERIA NEW CRITERIA % differenceOLD CRITERIA NEW CRITERIA % difference (Coal; 100-199MW)

mean 6.47% 5.53% -14%mean 6.47% 5.53% 14%

median 4.78% 5.07% +6%

best quartile 2.65% 3.26% +23%



EFOR - PLANT B

OLD CRITERIA NEW CRITERIA % difference(Coal; 800-1300MW)

mean 5.83% 7.63% +31%

Median 4.55% 5.87% +29%

best quartile 2.70% 3.97% +47%


Reliability Benchmarking Results 30 Peer Units

P it l ti it i• Peer unit selection criteria– Subcritical

Reserve shutdown hours less than 963 hours– Reserve shutdown hours less than 963 hours per year

– Natural boiler circulationNatural boiler circulation– Primary fuel = coal– Single reheatg– Net output factor greater than 85.6%


Peer Unit EFOR Distribution

100

70

80

90

100

CEN

T

40

50

60

70

ATI

VE P

ERC

10

20

30

40

CU

MUL

A

0

10

0 1 2 3 4 5 6 7 8 9 10 11

EFOR (%)EFOR (%)


Peer Unit SOF Distribution

1 00

70

80

90

RCEN

T

40

50

60

ULA

TIVE

PER

10

20

30

CU

M

00 5 10 1 5 2 0

S O F ( % )


Peer Unit EAF Distribution

1 00

70

80

90

RCEN

T

40

50

60

ULA

TIVE

PER

10

20

30

CU

M

07 0 7 5 80 8 5 9 0 9 5 1 0 0

EA F (% )


Conclusions

• Benchmarking is helping utilities• Benchmarking is helping utilities– Set goals– Develop incentivese e op ce t es– Identify improvement opportunities– Quantify and manage risks– Create increased awareness of the potential for and

the value of increased plant performance

• Proper peer group selection is essential


Forecasting

Statistics Versus Probability

• Statistics – Yesterday’s actual resultsP b bilit T ’ di t d lt• Probability – Tomorrow’s predicted results


Forecasting Performance September 2002 WEC Case Study

• Southern Company’s historic plant EFOR’s trends were used as projections

• Comparison of actual EFORs to projected EFORs (statistical process control) showed a need for improved forecasting methodology

• New forecasting methodology was developed using multi-variable regression techniques that showed great potential


EFORACTUAL - EFORPREDICTED


Forecasting’s Basic Principle

Past Conditions Future ConditionsPast Conditions

Past Results

Future Conditions

Future ResultsPast Results Future Results


Predicting EFOR

Most Important Parameters (from statistical l i )analysis)

• Lagging Equivalent Forced Outage Hours• Lagging Service Factor• Current Year Planned Outage Hours• Lagging O&M Spending• Current Year O&M Spendingp g• Fuel Type• Other Factors


Other Factors47

EFORACTUAL - EFORPREDICTED


Other Forecasting Example

• New Technology “learning curve” – Supercritical – Ref 7

• November 2002 WEC Case Studyy

• Applies learning curve theory to historic EFORsApplies learning curve theory to historic EFORs to forecast expected improved in new technology performances


gy p

49

Outage Rates versus Yearof Initial Operation


Other Forecasting Example

• Estimating a Generating Plant’s Future• Estimating a Generating Plant’s Future Maintenance Cost

• October 2003 WEC Case Study & Ref 8October 2003 WEC Case Study & Ref 8• Uses probabilistic forecasting techniques to

estimate the range of future cost (O&M + g (Retrofit Capital, recurring and non-recurring) and its probability distribution

• Results used by many organizations including operations & maintenance, resource planning, generation planning and financegeneration planning and finance

Communications

• Communications to all stakeholders, especially employees, is vital to clearly show p y p y , ythe “GAP” between your plant’s reliability compared to the best performers in their peer group


Phase 2 - Identification February 2003 Case Study


Identification Phase – Part 2Problem Area Identification

Goal – To identify problem areas with best payback potential

• Component Benchmarking

• High Impact – Low Probability Event Reduction –February 2002 WEC Case Study & Ref 9

• Trend Analysis WEC Case Studies:– March 2002 – Peak Season Reliability (Ref 10)y ( )– June 2002 – Availability Following Planned Outages (Ref 11)– December 2002 – Reliability Versus Demand


Component Benchmarking

• Three options for peer group selection:# 1 U t i f it l l– # 1 – Use components in peer group from unit-level benchmarking peer group analysis (but this will exclude many valid peer components)

– #2 – Perform a similar analysis to that described in the unit-level benchmarking but using component design and operational data for analyzing and selecting peer

( b i d igroups (can become very expensive and time consuming)

– #3 – Use a combination of experience and engineering judgment with unit-level peer group analysis results (recommended)


Component Benchmarking

• Compare the performance of each /system/equipment to its peer distribution

• The system/equipment with the largest “percentile gap” between its performance and th “b t i l ” i it h ld bthe “best in class” in its peer group should be a high priority system to study

Component UnavailabilitySteam Turbine

S te am T u r b in e U n a va ila b i l ity

5 06 0

7 0

8 0

9 0

1 0 0

Perc

ent

0

1 0

2 0

3 0

4 0

5 0

Cum

ulat

ive

0 2 4 6 8 1 0 1 2 1 4 1 6

Un ava ilab ilit y (P e r ce n t )

Component UnavailabilityAll Boiler Tube Leaks

B o ile r T u b e L e a k U n a va ila b i l ity

5 06 0

7 0

8 0

9 0

1 0 0

Perc

ent

0

1 0

2 0

3 0

4 0

5 0

Cum

ulat

ive

0 2 4 6 8 1 0 1 2 1 4 1 6

Un ava ilab ilit y (P e r ce n t )

Component BenchmarkingBoiler Tube Leaks vs. Steam Turbine

B o ile r Tu b e s an d S tm Tu rb in e U n av a ila b ility

8 0

1 0 0

cent

2 0

4 0

6 0

Cum

ulat

ive Pe

rc

0

0 2 4 6 8 1 0 1 2 1 4 1 6

Un av a ilab ilit y (P e r ce n t )

C

S tea m Tu r b i n e B o i l e r Tu b es


C t U l d I tComponent Unplanned ImprovementUnavailability Potential Rank

BT leaks 1% #1

Turbine 0.5% #2Turbine 0.5% #2


B o ile r Tu b e s an d S tm Tu rb in e U n av a ila b ility

8 0

1 0 0

cent

2 0

4 0

6 0

Cum

ulat

ive Pe

rc

0

0 2 4 6 8 1 0 1 2 1 4 1 6

Un av a ilab ilit y (P e r ce n t )

C

S tea m Tu r b i n e B o i l e r Tu b es


Improvement Potential Ranking

Component Unplanned Improvement Peer ImprovementUnavailability Potential Rank Percentile Potential RankUnavailability Potential Rank Percentile Potential Rank

BT leaks 1% #1 50th #2

Turbine 0.5% #2 80th #1

Component Unavailabilityp yWaterwall Tube Leaks

B o ile r W a te rw a ll T u b e L e a k s

5 0

6 0

7 0

8 0

9 0

1 0 0

Per

cent

0

1 0

2 0

3 0

4 0

5 0

ACu

mul

ative

0 1 2 3 4 5 6 7 8 9 1 0 1 1 1 2

Un ava ilab lit y (P e r ce n t )

Component Benchmarking p gNuclear Reactor Only Nuclear Reactors Only Vs. Total Nuclear Unitsy

2005-2009

90100

60708090

t, %

Total Unit

20304050

Perc

ent Total Unit

Reactor Only

010

0 5 10 15 20 25 30

U l d O t F t (UOF)Unplanned Outage Factor (UOF)

Component Benchmarking p gBalance of Plant

Balance of Plant Vs. Total Nuclear Units2005-2009

90100

607080

90

t, %

Total Unit

20304050

Perc

ent Total Unit

Balance of Plant

010

20

0 5 10 15 20 25 30

Unplanned Outage Factor (UOF)

Component Benchmarking p gHydro Turbines

Hydro Turbines Only Vs Total Hydro Units (No PS)Hydro Turbines Only Vs. Total Hydro Units (No PS)2005-2009

90

100

60

70

80

90

t, %

Total Unit

20

30

40

50

Perc

ent Total Unit

Turbine only

0

10

0 10 20 30 40 50 60 70 80 90 100

Unplanned Outage Factor (UOF)Unplanned Outage Factor (UOF)

Component Benchmarking B l f Pl tBalance of Plant

Balance of Plant Only Vs. Total Hydro Units (No PS)2005-2009

90100

50607080

ent,

%

Total Unit

10203040

Perc

e Balance of Plant

010

0 10 20 30 40 50 60 70 80 90 100


Component Benchmarking p gGas Turbines in Combined Cycle Blocks

Combined Cycle Blocks Vs. CC Gas Turbinesy2005-2009

90100

506070

80

nt, % Block

1020

304050

Perc

e GT Only

010

0 5 10 15 20 25 30 35 40


Component Benchmarking p gHeat Recovery Steam Generators

Combined Cycle Blocks Vs HRSGsCombined Cycle Blocks Vs. HRSGs2005-2009

90100

607080

90

nt, % Block

20304050

Perc

en HRSG

010

0 5 10 15 20 25 30 35 40


Component Benchmarking p gCC Steam Turbines

Combined Cycle Blocks Vs. Steam Turbines200 20092005-2009

90100

50607080

ent,

%

Block

St T bi

10203040

Perc

e Steam Turbine

00 10 20 30 40


Is Your Power Plant Headed for a HILP?? How to Avoid, Detect or Mitigate High Impact – Low Probability (HILP) Events

R b t Ri h i Ri h i C ltiRobert Richwine – Richwine Consulting Michael Curley – NERCG. Scott Stallard – Black & Veatch


What is a HILP?

• High Impact – Low Probability Event• Happens infrequently but results in extended

unplanned outages• Sometimes called “First Time Event”

(at least the first time it has happened at your ( pp yplant)

• Similar to the concept described in Nassim pTaleb’s book “The Black Swan: The Impact of the Highly Improbable”


Typical HILPsTypical HILPs

• Turbine Water Induction• Boiler Explosions• Generator Winding Failures• Many, many others

HILP Reduction Programs

• Some companies have successfully reduced their HILP frequencies and/or magnitudes with a formal HILP Reduction P i NERC GADS d tProgram using NERC-GADS data.

• NERC-GADS database contains ~30 years of detailed design and reliability data from

5 100 ti it ith idover 5,100 generating units with a wide variety of technologies.


HILP Effect on Forced Outage Rate (FOR)

• FOR made up of two type of events– Routine expected events with small/medium

outage consequences– Unexpected major events with large outageUnexpected major events with large outage

consequences

• Should separate these two elements of FOR when benchmarking reliability and establishing reliability improvementestablishing reliability improvement programs


Benchmarking two units’ Forced Outage Rate (FOR) - Example

Unit A Unit B

FOR 10% 10%FOR 10% 10%Fewer, smaller

events but 1

Type of OutagesMany small

events

events but 1 major event of 3 weeks lengthType of Outages events 3 weeks length

"Normal" FOR 10% ~4%


Benchmarking two units’ Forced Outage Rate (FOR)

Implications1) The two units have had very different failure

modes2) We should adapt our benchmarking analysis

and improvements efforts to account for th diffthese differences.


HILP Reduction Program

• Step 1 – Select the best peer group for benchmarking against your unity

• Step 2 – Find the peer group’s HILP contribution to FOR or EFOR and compare to your unit’s HILP contributionSt 3 P i iti th ’ HILP bl• Step 3 – Prioritize the peer group’s HILP problem areas

• Step 4 – Review GADS root cause information• Step 5 – Assess your plant’s susceptibility to HILPs• Step 5 – Assess your plant s susceptibility to HILPs• Step 6 – Identify options to address HILPS• Step 7 – Evaluate and select HILP reduction optionsp p• Step 8 – Track results of implemented options, compare to

expectations and feedback into program to improve the processprocess

Step 1 – Select Peer Group

• It is vital to select the best peer groupIt is vital to select the best peer group• You don’t want to be comparing apples to

oranges• Actually, the best we can usually do is

compare apples to oranges – at least they are b th f itboth fruit

• If you don’t go through an valid peer selection process you might be comparing apples toprocess you might be comparing apples to zebras


Step 2 – Compare Unit to Peer Group’s HILP Contribution to FOR

• Using NERC’s pc-GAR software calculate FOR• Using NERC’s pc-GAR-MT software determine

the number of full forced outage hours with outage durations greater that the value “you”outage durations greater that the value you define as a HILP (typically greater than 1 week or longer)g )

• Using the HILP full forced outage hours calculate the FOR due to HILPs

• Compare the unit’s HILP contribution to FOR to its peer groups

f fRichwine Consulting Group, LLC

• Repeat for your company’s fleet 80

Step 3 – Prioritize the Peer Group’s HILP Problem Areas

• Using pc-GAR-MT and excluding non-HILP t ( ti f th ft ) ilevents (an option of the software) compile a

frequency chart of HILP cause codes that the peer group has experiencedthe peer group has experienced

• Use the frequency chart to focus on the most likely HILP areas for your unity y

• Consider exporting the files from pc-GAR-MT to a spreadsheet for easier manipulation and more detailed analysis as well as graphical reports


Step 4 – Review GADS Root Cause Data

• GADS input contains an optional 80 character free-format data field, often containing valuable data regarding the outages.Alth h t tl il bl i GAR• Although not currently available in pc-GAR or pc-GAR-MT, Mike Curley, Manger of NERC-GADS Services can advise you on how to retrieve thisServices, can advise you on how to retrieve this information.

• Reviewing this data for HILP events can indicate gthe root causes of events that your unit’s peer group has experienced and can point you in di ti f i it’ tibilit


directions for assessing your unit’s susceptibility to those HILPs. 82

Step 5 – Assess Your Unit’s Susceptibility to HILPs

• HILP susceptibility is usually the result of several factors occurring together

• Assessing HILP risk must rely on a structured process focusing on if these factors could exist

• Catalogue key HILP events and the circumstances th t ld i d th HILPthat could induce the HILP

• Evaluate your unit to determine if these circ mstances are present s ch as eq ipmentcircumstances are present such as equipment condition, O&M experiences & practices, QA, etc.

• Create a scorecard to quantify the level of HILPRichwine Consulting Group, LLC

• Create a scorecard to quantify the level of HILP risk 83

Step 6 – Identify Options to Address HILPs

• HILP reduction options are usually very specific to the issue

• HILP reduction options should consider ways to:– Prevent the HILP– Detect the HILP event early so as to minimize

downstream damageMitigate the impact of an undetected HILP– Mitigate the impact of an undetected HILP


Step 7 – Evaluate and Select HILP Reduction Options

• Sufficient information should be gathered to be able to forecast the effect of each option

• An economic analysis for each option should be done to: – Justify– Time

P i iti– Prioritize• Using the option evaluations and considering

the fact of limited company resources (timethe fact of limited company resources (time, money, manpower) the best set of options should be chosen for implementation


should be chosen for implementation85

Step 8 – Track Results, Compare to Expectations & Feedback

• Monitor the actual results of each implemented HILP improvement option and compare to expected results

• Compare the fleet’s FOR trend due to HILPs over time

• Feedback successes and failures into the HILP reduction program to learn from past

i d i thexperiences and improve the process


Conclusions & Recommendations

HILPs Happen!!• No power plant in immune to HILPs• While your staff must react to the “problems

of the day” some resources should be devoted to searching for cost-effective ways t t d t t iti t HILPto prevent, detect or mitigate HILPs

• Addressing HILP causes and seeking l ti “b f HILP ” isolutions “before a HILP occurs” is a proven

way to move from a fire-fighting to pro-active style of management


style of management87

Trend Analysis

• EFOR Following Planned Outages– WEC June 2002 Case Study– NERC-GATE Study

R f 11– Ref 11

EFOR Following Planned Outages

Week Following Schedule Outage


EFOR Following Planned Outage Trend

• If your plants exhibit this trend you can seek y p ycost-effective ways to reduce this unreliability

• If you cannot find ways to reduce the problem, you can incorporate this tendency p , y p yinto the dispatch optimization process (perhaps by not scheduling outages at two major units back-to-back or some other planning/scheduling method)


Trend Analysis

• Forced Outage Rate versus Demand• WEC December 2002 Case Study

Forced Outage Rate Versus Demand Trend

Forced Outage Rate Versus Demand Trend

• High Output Factor (maximum generation most of the time) units have fewer failures but take more time to repair

• Low Output Factor (often generating at minimum and load-following) units have morefailures but take less time to repair

• If your staff is aware of these trends they may have ideas to address these issues


Trend Analysis

• Peak Season ReliabilityWEC M h 2002 C St d– WEC March 2002 Case Study

– NERC-GATE Study– Based upon a study originally performed at Southern p y g y p

Company that resulted in a 1% reduction in its reserve margin criteria with no reduction in customer service reliability (a $100 million savings)

• Aging Versus Reliability– WEC July 2004 Case Study

• Aging or Vintage – which is most responsible for differences in boiler tube leak rates?– WEC November December 2003 Case StudyWEC November, December 2003 Case Study

Identification Phase - Part 2 Solution Option Identification

For Each Problem Area

• Conduct Failure Modes and Effects AnalysisD t i C• Determine Consequences

• Identify Technologically Feasible AlternativesPl t & G l Offi St ff– Plant & General Office Staff

– Successful implementations at other plants– Vendor recommendationsVendor recommendations– Research organizations recommendations (EPRI)– Other outside sources

Common Elements Phase 3 - EvaluationMarch 2003 Case Study


Retrofit Capital ProjectEvaluation Process

• Elements of an Evaluation Analysis1) IMPACT A di ti f diff i f t l t– 1) IMPACT – A prediction of difference in future plant performance if the project is implemented versus if it is not implemented (this is the primary place in an evaluation

h NERC GADS i i l bl )where NERC-GADS is invaluable).

– 2) WORTH of PERFORMANCE IMPROVEMENT – An estimate of the value to the company resulting from a change in the unit’s performance

– 3) COST – The total budget cost including equipment procurement and installation costs and should include all financing chargesfinancing charges

Capital Project Evaluation Process – Typical Impacts

• Future with/without the project (positive/negative)– Availability– Availability– Efficiency– O & M Savings (or increased cost)

A ili P R i t– Auxiliary Power Requirements– Maximum or Minimum Capacity– Environmental– Other quantifiable impacts– Intangibles


Impact Data Sources

• Knowledgeable plant and support staff• Engineering staff• Plant data !!! (NERC-GADS)• Industry data !!! (NERC-GADS)• Manufacturers and consultantsManufacturers and consultants• Other projects results• Test Results

Oth• Other


Capital Project Evaluation Process

• Justification• Justification• Timing

P i iti ti• Prioritization

Capital Project Evaluation Process Part 1

• Justification - The first (and easiest)Justification The first (and easiest) obstacle a project must pass– Addresses the question “Are the total Benefits

f thi j t t th it t t l C t ?”of this project greater than its total Costs?”– A variety of financial terms can be used

• Net Present ValueNet Present Value• Internal Rate of Return• Payback Period

Life Cycle Benefit to Cost Ratio• Life Cycle Benefit to Cost Ratio• Others


• Timing – The second obstacle a project must overcomeovercome– Addresses the question “If a project is justified, when

should it be implemented”.– Many “wear-out” project need to have this analysis

performed based on their technical risk profile – All projects should be timed based on their economicAll projects should be timed based on their economic

risk profile



• Prioritization - The third (and hardest)Prioritization The third (and hardest) obstacle a project must overcome

– Addresses the question “ If the company does not have all of the resources (money, time, manpower) necessary to implement all of the justified projects that should be done this year, which projects will hurt the least to delay?”which projects will hurt the least to delay?


Operations & MaintenanceEconomic Decision Analysis

• Company’s business economics applied to day-to-day O&M decisions

• Helps identify the best economic option for recovering from abnormal conditionsHelps identif the best economic option for• Helps identify the best economic option for establishing normal O&M programs


O&M Decision Analysis

• Problem solution steps

– Define problem and identify viable options

– Quantify technical consequences of options

– Combine technical consequences with company’s economics

– Evaluate results and incorporate into decision


O&M Decision ExampleLeaking Feedwater Heater

P blProblem– Tube failure in 7A feedwater heater

– Requires isolation of 6A & 7A heaters

– 1% efficiency loss during isolation



• Solution Options 1) Remove unit from service immediately, locate and plug leaking

tube

2) Wait until weekend to repair

3) Wait until next planned outage and imbed repair

4) Wait until next forced outage of sufficient duration to imbed4) Wait until next forced outage of sufficient duration to imbed repair



• ConsequencesConsequences– Option 1- repair immediately

• 48 hour outage during high demand periodg g g p• Overtime labor cost of $1000• Start-up cost of $20,000

T t l t $217 000• Total cost = $217,000



• ConsequencesConsequences– Option 2- repair during weekend

• 48 hour outage during lower demand period• Overtime labor cost of $1000• Start-up cost of $20,000• 1% efficiency penalty until weekend1% efficiency penalty until weekend• Total cost = $137,000



• Consequences• Consequences– Option 3 -repair during next planned outage

• 1% efficiency penalty until next planned outage1% efficiency penalty until next planned outage• No outage cost since repair will be embedded in

planned outageT t l t $202 000• Total cost = $202,000



• Consequences– Option 4- repair during next forced outage

1% ffi i lt til t 48 h t• 1% efficiency penalty until next 48 hour outage– uncertain when next 48 hour outage will occur– NERC-GADS data for your unit & peer units can help

quantify the uncertainty range

• Total cost range = $0 - $202,000Total cost range $0 $202,000– maximum cost of $202,000– “break even” point in 8 weeks



• What would YOU decide?– Option 1 (repair now) - $217,000

O ti 2 ( i d i k d) $137 000– Option 2 (repair during weekend) - $137,000– Option 3 (repair during P. O.) - $202,000

Option 4 (repair during F O ) $0 $202 000– Option 4 (repair during F.O.) - $0-$202,000



• If the same event happened at a different time the following economic results could happen:– Option 1- repair now $ 75,000– Option 2- repair during weekend $ 90,000– Option 3- repair during next PO $450,000Option 3 repair during next PO $450,000– Option 4- repair during next FO $0-$450,000

N h t ld d id ???Now what would you decide???


Operations & MaintenanceEconomic Decision Analysis

• Applications

– Maintenance• Reactive (e g planned outage extension)• Reactive (e.g. planned outage extension)• Proactive (e.g. condition directed maintenance)

– Operations• Reactive (e.g. tube leaks)• Proactive (e.g. pump operations)Proactive (e.g. pump operations)


Common Elements Phase 4 Implementation April 2003 Case Study


Implementation

• Project choice (economic plus intangibles)j ( p g )• Financing• Goal selection• Monitor actual results and compare against

expected results

Awareness


Implementation

• Goal Setting Case Studies• Goal Setting Case Studies

– May & June 2003 – Are Reliability MeasuresMay & June 2003 Are Reliability Measures Unreliable?? (Ref 12)

– July 2003 – Planned vs. Unplanned Outages -Effects on Goals (Ref 13)


Are Reliability Measures Unreliable?

• Factors – EAF, FOF, etc.Factors use the entire time period as the

denominator without regard to unit demand

EXAMPLE: Peaking Gas Turbine100 hrs/year demand00 s/yea de a d25 forced hours during demandEAF = 99.71% FOF = 00.29%


Are Reliability Measures Unreliable?

• FOR EFOR• FOR, EFORIn Example

FOR = EFOR = 25%FOR = EFOR = 25%

In reality the GT is likely to have had many moreIn reality the GT is likely to have had many more FOH reported since GADS counts all forced outage hours, not just ones during demand periods. Therefore act al EFOR statistics are m ch higherTherefore, actual EFOR statistics are much higher, often 60% +.


Equivalent Forced Outage Rate –Demand (EFORd)

• Markov equation developed in 1970’s• Markov equation developed in 1970 s• Used by the industry for many years

PJM Interconnection (20 years)– PJM Interconnection (20 years)– Similar to that used by the Canadian Electricity

Association (20 years)Association (20 years) – Being use by the New York ISO, ISO New

England, and California ISO.– Now a part of IEEE standard 762 & NERC-GADS


EFORd Equation:

EFORd= [f(FOH) + fp(EFDH)] * 100%EFORd [f(FOH) + fp(EFDH)] 100%[SH + f(FOH)]

Where: f = (1/r)+(1/T) (1/r)+(1/T)+(1/D)(1/r)+(1/T)+(1/D)

fp = SH/AHr= FOH/(# of FOH occur.)r FOH/(# of FOH occur.)T= RSH/(# of attempted Starts)D= SH/(# of actual starts)


D SH/(# of actual starts)121

EFORd Concept

• Equation is complex, but concept is simpleO “• Reported Forced Outage Hours are “reduced”

• Reduction % is the ratio of Reserve Shutdown Hours to the Service Hours in the time periodHours to the Service Hours in the time period

• This is an approximation (since actual demand hours are not collected by NERC)demand hours are not collected by NERC) that estimates the hours of forced outage during demand periods

• Advantage – We can calculate historic EFORd without collecting new data!!!

Example of EFORd

EFOR vs EFORd

30 0035.0040.0045.00

, %

EFOR range from 3 9 to 42 4%

10 0015.0020.0025.0030.00

OR

& E

FOR

d, EFOR, range from 3.9 to 42.4%

EFORd, range from 3.9 to 10.6%

0.005.00

10.00

21.7

0

21.7

0

21.7

0

21.7

0

21.7

0

21.7

0

21.7

0

21.7

0

21.7

0

21.7

0

21.7

0

21.7

0

21.7

0

21.7

0

21.7

0

21.7

0

EFO

762

712

662

612

562

512

462

412

362

312

262

212

162

112 62 12

Reserve Shutdown Hours


EFOR - demand

• For units with zero Reserve Shutdown HoursFor units with zero Reserve Shutdown Hours (RSH), EFOR and EFORd have the same value.

• For units with high RSH, EFOR will have an unrealistically high value compared to EFORd

• Because EFORd comes closer than EFOR to ti th it’ i bilit t drepresenting the unit’s inability to produce

power when needed, I recommend that EFORd be used for all generating units in your fleetbe used for all generating units in your fleet

Goal SettingPlanned versus Unplanned Outages

WEC JULY 2003 Case Study & Ref 13

F t d f l d t l d• For every extra day of planned outage, unplanned outages only were reduced by 0.6 of a day

• This suggests that planned outages should be gg p gminimized in order to maximize availability

• However, planned outages almost always occur during the non-peak season when financial consequences arethe non-peak season, when financial consequences are much lower by as much as 75% (compared to an average peak season day)Th f th t t th t ill lt i th l t t• Therefore, the strategy that will result in the lowest cost of electricity is to maximize planned outages (within reason) so as to minimize the expensive forced outages


Implementation

• Monitor Actual Results and CompareMonitor Actual Results and Compare Against Expected Results– September 2002 Case Study – Predicting Unit p y g

Reliability• Feedback Into Awareness, Identification &

Evaluation Phases


Common Elements January–April 2003 Case Studies

P f Performance Improvement


Performance Improvement Programs

• Successful Performance Improvement Programs (PIP) utilize a systemic approach to improving plant(PIP) utilize a systemic approach to improving plant performance, tailored to each company’s unique set of constraints and opportunities

• There are many diverse elements that must be integrated together

Th NERC GADS d t t i k t f• The NERC-GADS data system is a key aspect of many of these elements

GADS in Successful Performance Improvement Programs

• Awareness– Unit BenchmarkingUnit Benchmarking– Forecasting– Communications

• IdentificationIdentification– Component Benchmarking– HILP Benchmarking– Trend AnalysisTrend Analysis

• Evaluation– Proposed Project’s Impact

• ImplementationImplementation– Goal Selection– Results Monitoring and Comparison to Expectations– Feedback into ProcessFeedback into Process

The Future

Transforming to a Market-Driven Business Environment


The Future Is Not What It U d TUsed To BBe

IncreasingIncreasing CompetitionIs The FutureOf O I d tOf Our Industry

Past Business Environment

Regulated with Suppressed Competition

Basic Equation was Cost (prudent) + profit (mandated) = Price

Management Decision Focus was toAVOID RISK!!AVOID RISK!!

Evolving Business Environment

Market-Driven with Increased Competition

Basic Equation is Price (market) – Cost (total) = Profit

Management Decision Focus is toIdentify Quantify and Manage RiskIdentify, Quantify and Manage Risk

Example 1 of Reward/Risk Decisions

• You are playing a video poker “jacks or better” gamep y g p j g• You bet $5• You are dealt the 10, jack, king and ace of hearts and the

queen of spades• The reward for a straight is $20• The reward for a royal flush is $2000• Should you

1) keep the five cards you are dealt for a sure $20 payoff?– 1) keep the five cards you are dealt for a sure $20 payoff? – 2) discard the queen of spades, hoping for the queen of hearts

for a possible $2000 payoff?



Video poker example analysis

Risk = $5RewardReward

Option 1 - $20 @ 100% probability = $20.00Option 2 - $2000 @ 1/47 probability = $42.55

Reward/RiskOption 1 - $20/$5 = 4Option 2 $42 55/ $5 = 8 51 (Actually better sinceOption 2 - $42.55/ $5 = 8.51 (Actually better since other winning cards could be drawn; i.e. a different heart for a flush or a different queen for a straight)


Example 1A of Reward/Risk Decisions

• You are playing a video poker “jacks or better” gamep y g p j g• You bet $5• You are dealt the 9, 10, jack, king of hearts and the

queen of spades• The reward for a straight is $20• The reward for a straight flush is $250• Should you

1) keep the five cards you are dealt for a sure $20 payoff?– 1) keep the five cards you are dealt for a sure $20 payoff? – 2) discard the queen of spades, hoping for the queen of hearts

for a possible $250 payoff?


Example 1A of Reward/Risk Decisions

Video poker example analysis

Risk = $5Reward

$ $Option 1 - $20 @ 100% probability = $20.00Option 2 - $250 @ 1/47 probability = $5.32

Reward/RiskReward/RiskOption 1 - $20/$5 = 4Option 2 - $5.32/ $5 = 1.06 (Actually, slightly better since other winning cards could be drawn; i e asince other winning cards could be drawn; i.e. a different heart for a flush or a different queen for a straight)



• You are at a horse track with $100 in your pocket

• Your instructions are to bet $10 on one horse to win in each of 10 raceseach of 10 races

• You will be graded on how many times you cash a winning ticket (this is analogous to a regulated, risk-adverse business environment)

What is your best betting strategy??

Horse Track Reward/Risk ExampleRisk-Adverse Winning Strategy

Below are the horses in the first race and the odds that they will win the race (as set by the track handicapper)( y pp )

Horse OddsSecretariat (favorite) 2-1Whirlaway 4-1yCitation 5-1Gallant Fox 7-1Alysheba 8-1Alysheba 8 1Seattle Slew 10-1Northern Dancer 12-1Swaps 15-1Swaps 15 1War Admiral 17-1Aristides 20-1

Which horse would you bet on?Which horse would you bet on?

Horse Track Reward/Risk ExampleRisk-Adverse Winning Strategy

• As the famous sportswriter, Grantland Rice said, the race doesn’t always go to the swiftest, but that’s therace doesn t always go to the swiftest, but that s the way to bet!

• The best strategy is maximize the number of times you pick a winning horse is to “bet the favorite”!pick a winning horse is to bet the favorite !

• This virtually guarantees that you will cash a winning ticket the most times (versus other betting options)

• However you will almost certainly walk out of the track• However, you will almost certainly walk out of the track with less than $100 in your pocket (the total payoff on the times when the favorite wins is not enough to counteract the times it doesn’t win). But how muchcounteract the times it doesn t win). But how much money you end up with is not the way you will be graded, so you don’t care.

Horse Track Reward/Risk ExampleRisk-Manage Winning Strategy

• This time you will be graded on how much f 10money you have at the end of the 10 races

• In this case you must first “handicap” ( ti t th t dd ) h h ’ h(estimate the true odds) each horse’s chance of winning its race using all available and relevant data about each horse such as pastrelevant data about each horse, such as past performance, previous workout times, jockey’s success rate track conditions etcjockey s success rate, track conditions, etc.

• But you will not simply bet on the horse with the best chance of winning its race!!the best chance of winning its race!!

Horse Track Reward/Risk ExampleRisk-Manage Winning Strategy

• Instead, you will now compare each horse’s “ f “ ff“true odds” of winning against the “pay-off” as noted on the tote board just prior to the raceY h ld th b t th h ith th• You should then bet on the horse with the highest pay-off to true odds ratio – the Reward to Risk ratioto Risk ratio

Horse Track Reward/Risk ExampleRisk-Management Winning Strategy

Below are the horses in the first race and the odds that they will win the race (as set by the track handicapper) and the payoff rates (as set by the betting

public)public)Horse Odds Payoff Reward/Risk Ratio

Secretariat (favorite) 2-1 1.5/1 0.75/1Whi l 4 1 6/1 1 5/1Whirlaway 4-1 6/1 1.5/1Citation 5-1 10-1 2/1Gallant Fox 7-1 10-1 1.43/1Alysheba 8-1 4-1 0.5/1ySeattle Slew 10-1 17-1 1.7/1Northern Dancer 12-1 20-1 1.67/1Swaps 15-1 25-1 1.67/1War Admiral 17 1 30 1 1 76/1War Admiral 17-1 30-1 1.76/1Aristides 20-1 35-1 1.75/1

Which horse would you bet on?

Horse Track Reward/Risk ExampleRisk-Management Winning Strategy

• Example – You estimate that the horse with the best chance of winning g

(Secretariat) will win one out of two times (against this field of horses on this track on this night)

– However, since many people are betting on this horse to win, the payoff will only be 1.5-1 if it winsp y y

– Therefore, since the reward to risk ratio is only 0.75, this would not be a good bet

– Another horse that you have handicapped (Citation) at 5-1 has a payoff of 10-1 since not many people bet on itof 10 1 since not many people bet on it

– Therefore, its reward to risk ratio is 10/5 (or 2/1) making it an excellent bet

– You should anticipate having $200 in your pocket after 10 races if you bet on Citation ($100 profit) but only $75 if you bet on Secretariat ($25bet on Citation ($100 profit) but only $75 if you bet on Secretariat ($25 loss)

– This is actually how people make a living at a horse track – they are betting that they can make better choices of where to invest their money than the other people at the track that daymoney than the other people at the track that day

Power Plant Managementusing Reward/Risk Decision-Making

• Identify multiple solution options for any particular issueparticular issue

• Use all relevant data (especially NERC-GADS data for reliability effects) to estimate eachdata for reliability effects) to estimate each option’s impact

• Estimate each option’s reward/risk ratio p• Consider Reward/Risk Ratio in decision-

making• Compare results against expectations and

improve future decision-making

Need Better

• Key Performance Indicators (KPI)– EFOR (demand)( )– Commercial Availability

• Un-weighted• weighted


Market-Based KPIsMay & June 2003 Case Studiesy

• Demand EFOR – EFOR (demand)• Demand EFOR – EFOR (demand)– Developed for non-base loaded units– Approximates the reliability of a unit during

demand periodsdemand periods

• Commercial Availabilityy– Un-weighted – measures a unit’s availability

only during demand periods– Weighted – measures a unit’s availability onlyWeighted measures a unit s availability only

during demand periods and “weights” each hour’s impact by the unit’s net value during that hour


Commercial Availability

• Commercial Availability (CA) is a relatively new y ( ) yperformance metric that has become very important at many market-based generating companies and other companies who wish to minimize their cost of electricitycompanies who wish to minimize their cost of electricity

• CA combines the generating plant’s technical availability with the economic consequences of its availability (or unavailability)

• For more information see the Synopsis of the 2004 WEC-PGP paper “Performance of Generating Plant: NewPGP paper Performance of Generating Plant: New Realities, New Needs” published on the WEC website and the May and June 2003 WEC-PGP case studies

Why was Commercial Availability Why was Commercial Availability Created?Created?

• Traditional availability indicators such as EquivalentTraditional availability indicators such as Equivalent Availability Factor are sometimes inadequate or even misleadingA il bilit i di t d d th t id d• An availability indicator was needed that provided a direct linkage between the plant’s availability and the company’s electricity costs (or profits)p y y ( p )

• Commercial Availability (CA) was created to meet that need

Why are Traditional Availability Why are Traditional Availability Indicators Sometimes Inadequate?Indicators Sometimes Inadequate?

• There are often large differences in the economic consequences of an unexpected outage at a power plantp

• They economic consequences vary from season to season, day to day, and from hour to hour.

• This is most obvious in generators used primarily in peaking applications, but is also true for more base loaded fossil, hydro and even nuclear plants oaded oss , yd o a d e e uc ea p a ts(although to a lesser degree).

Variation in Financial Impact of Variation in Financial Impact of OutagesOutages

E lExampleA pin hole leak is detected in a boiler tube at a base-loaded unit on a weekday

Option 1 – Remove the unit from service immediately in order to minimize the damage and outage timeoutage timeOption 2 – Wait until the weekend to remove the unit (when the hourly economic consequences are less) but expect to have a longer outageless) but expect to have a longer outage

Variation in Variation in Technical Impactof Outagesof Outages

High or Low Demand Period

Option 1 1 day outage 0 27% reduction in EAFOption 1 – 1 day outage - 0.27% reduction in EAF

Option 2 – 2 day outage – 0.54% reduction in EAFp y g

Variation in Variation in Financial Financial ImpactImpactof Outagesof Outages

• Low Demand Period Increased Cost or Lost Profits• Option 1 (1 day outage during week)- $115,000 p ( y g g ) ,• Option 2 ( 2 day outage on weekend)-$184,000

• High Demand Period Increased Cost or Lost Profits• Option 1 (1 day outage during week)- $354 000• Option 1 (1 day outage during week)- $354,000

• Option 2 (2 day outage on weekend)- $265,000

Traditional IndicesTraditional Indices

• None of the traditional performance metrics are adequate to indicate or reflect the economic optimal d i idecision

• The term Commercial Availability (CA) was created to• The term Commercial Availability (CA) was created to address the problem

Commercial AvailabilityCommercial Availability

• Simple Concept

O d fi iti i th ti f t l fit (– One definition is the ratio of actual profit (gross margin) delivered by a plant relative to its potential profit if it had been able to deliver p pevery MW-HR required

– Other definitions are being used at different companies around the worldcompanies around the world

Commercial Availability Calculation Commercial Availability Calculation

• Example 1 ( 300MW unit for 10 random hours in year)

Hour G M EAF G MHour G. M. EAF G. M.potent actual CA = $72000/$73500 X 100 = 98.0%

1 $ 3000 1 $ 3000 EAF = 7/10 X 100 = 70%2 $ 0 1 $03 $ 1500 0 $0 Lost Margin = $73500-$720004 $ 6000 1 $6000 = $15005 $12000 1 $120005 $12000 1 $120006 $24000 1 $24000 EFORd = 1/7 X 100 = 14.3%7 $18000 1 $180008 $ 9000 1 $ 90009 $ 0 0 $ 010 $ 0 0 $ 0Total $73500 $72000

Commercial Availability CalculationCommercial Availability Calculation

• Example 2 (same as example except available in all hours but hour 6)

Hour G. M. EAF G. M.potent actual CA = $49500/$73500 = 67.3%

1 $ 3000 1 $ 3000 EAF = 9/10 = 90%2 $ 0 1 $ 03 $ 1500 1 $ 1500 Lost Margin = $73500-$495004 $ 6000 1 $ 6000 $ 24 0004 $ 6000 1 $ 6000 = $ 24,0005 $12000 1 $120006 $24000 0 $ 07 $18000 1 $18000 EFORd = 1/7 = 14.3%$ $8 $ 9000 1 $ 90009 $ 0 1 $ 010 $ 0 1 $ 0Total $73500 $49500Total $73500 $49500

Commercial Availability Commercial Availability vs. EAF & EFORdvs. EAF & EFORd

• EXAMPLE 1 EXAMPLE 2

Lost Margin = $1500 Lost Margin = $24000

EAF 70% EAF 90%EAF = 70% EAF = 90%

CA = 98% CA = 67.3%

EFORd = 14.3% EFORd = 14.3%

Benchmarking Commercial Benchmarking Commercial AvailabilityAvailability

• Cannot benchmark directly except against your own units and their trends

• Can benchmark indirectly using “ diti l b biliti ” l“conditional probabilities” plus yourplant’s actual economics

Conditional ProbabilityConditional Probability

When required (that’s the conditional part) what is the likelihood (that’s the probability part) that the unit will be able to generate at its rated capacity

Conditional Probability has been shown to vary depending upon the plant’s economic necessity!(WEC March 2002 case study)

(1-EFORd) is a good metric for Conditional Probability

Conditional ProbabilityConditional Probability

• Using reliability data from the GADS g ydatabase we can determine frequency distributions of Conditional Probability (C.P.) for the peer group of each individual ( ) p g punit (1-EFORd)

• There will be different probability distributions during different demanddistributions during different demand periods (peak/non-peak season, day/night, weekday/weekend day, etc.)Selecting each unit’s optimum CA goal will• Selecting each unit’s optimum CA goal will start with these C.P. distributions

Setting Commercial Availability Goals Setting Commercial Availability Goals Using Conditional ProbabilitiesUsing Conditional Probabilities

From the example of 10 random hours we might chooseFrom the example of 10 random hours we might choose Conditional Probability Goals CPG as:

Hours 1,2,3,4 = 92% (hours of mid expected value)

Hours 5 6 7 8 = 98% (hours of high expected value)Hours 5,6,7,8 = 98% (hours of high expected value)

Hours 9,10 = 90% (hours of low expected value)

These might be the best quartile reliabilities (1-EFORd) of the peer group for each demand periodof the peer group for each demand period

Setting Commercial Availability Goals Setting Commercial Availability Goals U i C di i l P b bili iU i C di i l P b bili iUsing Conditional ProbabilitiesUsing Conditional Probabilities

The gross margin goal (GMG) for each hour is calculated b lti l i h h ’ C diti l P b bilit G lby multiplying each hour’s Conditional Probability Goal (CPG) by that hour‘s Gross Margin Potential

Note: If the unit has other value to the company in addition to gross margin (such as ancillary power) that value should be incorporatedvalue should be incorporated

Setting Commercial Availability Goals Setting Commercial Availability Goals U i C di i l P b bili iU i C di i l P b bili iUsing Conditional ProbabilitiesUsing Conditional Probabilities

H GMP $ CPG GMG $Hour GMP-$ CPG GMG-$1 3000 .92 27602 0 .92 03 1500 92 13803 1500 .92 13804 6000 .92 5520 CA Goal = 71400/73500 5 12000 .98 11760 = 97.1%6 24000 .98 235206 24000 .98 235207 18000 .98 176408 9000 .98 88209 0 .90 010 0 .90 0Total 73500 71400

Commercial Availability GoalCommercial Availability Goal

• EXAMPLE 1 EXAMPLE 2EXAMPLE 1 EXAMPLE 2

Lost Margin = $1500 Lost Margin = $24000

EAF = 70% EAF = 90%

EFORd = 14.3% EFORd = 14.3%

CA = 98% CA = 67 3%CA 98% CA 67.3%

Goal CA = 97.1% Goal CA = 97.1%

Setting Commercial Availability Goals Using Conditional Probabilities

1) Id if i ’ d i & i l1) Identify your unit’s design & operational peers2) Calculate the probability distribution of these unit’s

Conditional Probabilities during their demand periods that are similar to yours.

3) Estimate your unit’s Optimum Economic Conditional Probabilities during each demand gperiod (often the top quartile or top decile C.P. of your peers).

4) Apply those Conditional Probability goals to your ) pp y y g yunit’s economics (forecast or actual) using whatever definition of Commercial Availability you choose


Setting Commercial Availability Goals Using Conditional Probabilities

Although this process might seem complicated remember the following:g

For every complex, difficult to understand, hard problem,there is a simple, easy to understand,

WRONG solution!!


Commercial Availability ImplicationsCommercial Availability Implications

• See WEC Case Studies for May & June 2003• Benchmarking• Design Changes• Goals Systems

D i i A l i ( ti & ti )• Decision Analysis (proactive & reactive)• Impact on traditional performance metrics• Perception by other stakeholders• Perception by other stakeholders

Commercial Availability Implications

– Technology tells us what can be done


– Technology tells us what can be done

WHILE

– Economics tells us what should be done


– Technology tells us what can be doneWhile

– Economics tells us what should be doneBUTBUT

– Politics tells us what will be done

Optimum Economic AvailabilityWEC July 2002 case study

• In order to maximize profitability each generating plant must seek to optimize (not simply maximize) its performance

• In addition to the WEC July 2002 case study an• In addition to the WEC July 2002 case study an article by Robert Richwine was published in Power Magazine in 2004 that more fully describes this concept (Ref 14)

Optimum Economic AvailabilityOptimum Economic Availability

Availability 100%Availability 100%


Top QuartileFrontier

A il bilit 100%Availability 100%


Total O&M Cost $ Cost ofUnavailability

$ Cost OfUnavailability

Availability 100%


Total O&M Cost +Unavailability Cost


Unavailability Cost

$ Cost OfU il bilitUnavailability

Availability 100%




y


Optimum Economic Availability

Availability 100%





Total O&M Cost TargetUnavailability

Optimum Economic Availability

Availability 100%


Top QuartileFrontier

ReactiveProactive

A il bilit 100%Availability 100%




Proactive Cost Target

Optimum Economic AvailabilityReactive

ProactiveProactive Cost Target

Reactive Cost Target

Availability 100%

Future - Market Oriented System

• Consequences– More revenue uncertainties– More cost uncertainties– More risk– More opportunities


Summary

• Change is occurring everywhere• Changes are not the same everywhere• Company specific programs should be

developed and implemented that will allow each company to anticipate and respond

i kl t it i t f hquickly to its unique set of changes• The companies that are best able to respond

t k t i d d ill b thto market-induced pressures will be the survivors

Performance Improvement

Better Use of Reliability Data will be a Key Factor in Achieving and Sustaining

Optimal Performance


Using Reliability Data to Improve Power Plant Performance

References (copies available upon request)

1) World Energy Council case studies www.worldenergy.orgwww.worldenergy.org

2) Optimizing O&M Cost to Maximize Profitability3) NERC-GATE trend studies www.nerc.com4) Performance Improvement at Southern Company5) Peer Unit Benchmarking6) Establishing Goals using Statistical Benchmarking6) Establishing Goals using Statistical Benchmarking7) Predicting Future Plant’s Reliability Using Learning

Curve Theory


References (copies available upon request)

8) Predicting Long Range Plant Maintenance Cost9) Is Your Plant Headed for a HILP?)9) Seasonal Performance Trends10) Availability Following Planned Outages12) R li bilit M U li bl It Ti f12) Reliability Measures Unreliable – Its Time for a

Change13) Planned vs. Unplanned Outages: The Economic ) p g

Impact of Selecting Availability Goals14) Maximizing Plant Availability May Not Optimize Plant

EconomicsEconomics


Presented by

Robert R. (Bob) RichwineR li bilit M t C lt tReliability Management ConsultantRichwine Consulting Group, LLC

[email protected] 678 231 3606+1-678-231-3606

Atlanta, Georgia, USA 30076


nerc gads benchmarking workshop 2011 (1)

Documents

nercgads system

gads format

data hcrunchersonly

gads database

new nercgate type group

power plant reliability

analysis system worldwide

generating plants nerc