Towards Quantification of Condition Monitoring Benefit for Wind Turbine Generators Mr. David...

Towards Quantification of Condition Monitoring Benefit for Wind Turbine Generators

Mr. David McMillan (Presenting)and Dr. Graham W. Ault

Evaluate and quantify condition monitoring system benefit for wind turbines via probabilistic simulation

Summary:

10th May 2007, EWEC Milan

Overview of this Morning’s Presentation

• Trends in Wind Capacity: UK focus

• Barriers to employment of condition based maintenance (CBM) for Wind Turbine Generators (WTGs)

• Forming an Economic case for Wind Turbine CM systems

• Modelling and Analysis to Quantify WTG CM Benefit

• Results, Conclusions and Discussion

Trends in Wind Capacity and Condition Monitoring

Renewables Obligation fuelling rapid build of Wind Capacity in the UK

Over 10GW (>10%) of capacity in the UK planning system alone

Turbines increasingly installed with a Condition Monitoring (CM) system

… However, utilities are reluctant to move towards condition - based maintenance of wind farms.

WHY?

Level of Renewables Obligation

02468

101214161820

2006

2007

2008

2009

2010

2011

2012

2013

2014

2015

2016

Lev

el o

f R

O %

… Until 2027

Possible Reasons Against Employment of CBM

The low technical impact of a single wind turbine outage

The low economic yield of WT– relative to fossil fueled plant

An assumption that the simplest methods will serve best

- Hence use of scheduled (periodic) maintenance

Time

Periodic MaintenanceReactive Maintenance

Well-known practical difficulties with CM

- Transducers (Mounting, Spurious Signals, Reliability)

- Interpretation: Pay a team of experts or rely on automation

Motivation: The Economic Case for CM

Fact: Wind farm operators will only move towards widespread use of CM systems if the economic case for their use is clear.

… So answers to the following questions would be useful:

1. What is the value of a wind turbine condition monitoring system?

2. Are CM systems currently cost-effective?

3. What are the necessary conditions for cost-effective WTG CM?

Build a set of models to answer these questions.

The ‘Component’ Model

Turbine condition states, Turbine component reliability

Quantified: Availability, Reliability & Condition

Condition Model Markov Chain Solvedvia Monte Carlo Sim.

Wind Regime, Turbine Power Curve, Turbine and Market Economics

Evaluated via Wind Turbine performance

Power PerformanceEvaluation Model

0

0.5

1

1.5

2

1 4 7 10 13 16 19 22 25

Yield, cost, revenue, spares, operational life & maintenance objectives

Tested assetmanagement policies

Asset Management & High-Level Objectives Model

Modelling Approach Summary

Modelled Components

Two sources of information to decide which components to model:

1. Wind Turbine Sub-Component Reliability Data

2. Wind Farm Operational Experience

Sub-Component Reliability Data: Failures

[1] G.J.W. van Bussel et al, Reliability, Availability and Maintenance aspects of large-scale offshore wind farms, MAREC01, [2] P.J. Tavner et al, Machine and Converter Reliabilities in Wind Turbines, PEMD 06, [3] H. Braam et al, Models to Analyse Operation and Maintenance Aspects of Offshore Wind Farms, ECN

Operational Experience

The published data focuses primarily on annual failure rate

- What about the severity of the failure: cost, downtime etc?

Dialogue with WT operators reveals the most severe failures:

Gearbox Generator

- High Capital Cost

- Long Lead Time

- In-Situ Repair Difficult

- Large Size/ Weight

- Position: in Nacelle, at the top of the tower

Model Development: Selected Components and Monitoring

Blade

Optical Strain

Generator

Vibration

Lube Oil Analysis

Temperature

Gearbox

Vibration

Lube Oil Analysis

Temperature

Power Electronics

No Monitoring

Input Values

Annual Turbine Failure Rate = 2.1

Gearbox 1.25

Generator 0.28

Blade 0.1

Electronics 0.46

WT Component Modelling Technique

Method: Discrete-time Markov chain solved via Monte Carlo simulation

Useful Features of this Modelling Framework

• Flexible approach: easily add new features

• Able to model Condition Monitoring (knowledge of states)

• Probabilistic nature can take account of future uncertainties

Model wind turbine as a deteriorating system of sub-components

• A recognised method for equipment degradation modelling

• Easily interfaced with other models

• Multi-stage model can capture time dependence

State Space of 4-Component Model

21 C1 DC2 UC3 UC4 U

22 C1 D

C2 DER

C3 U

C4 U

11 C1 DER

C2 UC3 DC4 U

1 C1 U

C2 UC3 U

C4 U

9 C1 U

C2 UC3 DC4 U

15 C1 DER

C2 DERC3 DC4 U

12 C1 U

C2 DERC3 DC4 U

3 C1 UC2 DER

C4 U

C3 U

25 C1 UC2 DC3 UC4 U

26 C1 DERC2 DC3 UC4 U

8 C1 DER

C2 DER

C4 DERC3 U

6 C1 UC2 DER

C4 DERC3 U

5 C1 UC2 U

C4 DERC3 U

17 C1 U

C2 U

C4 DC3 U

27 C1 U

C2 D

C4 DERC3 U

19 C1 U

C2 DER

C4 DC3 U

28 C1 DER

C2 D

C4 DERC3 U

24 C1 D

C2 DER

C4 DERC3 U

20 C1 DERC2 DER

C4 DC3 U

4 C1 DERC2 DER

C3 UC4 U

7 C1 DER

C2 U

C4 DERC3 U

2 C1 DER

C2 UC3 U

C4 U

18 C1 DER

C2 U

C4 DC3 U

23 C1 D

C2 U

C4 DERC3 U

13 C1 DER

C2 U

C4 DERC3 D

10 C1 UC2 U

C4 DERC3 D

14 C1 UC2 DER

C4 DERC3 D

16 C1 DER

C2 DER

C4 DERC3 D

States Comp # State #

Normal/ Derated All 1 to 8

Electronics C3 9 to 16

Blade C4 17 to 20

Gearbox C1 21 to 24

Generator C2 25 to 28

The Power Performance Evaluation Model

Yield, cost, revenue, spares, operational life & maintenance objectives

Tested assetmanagement policies

Asset Management & High-Level Objectives Model

Turbine condition states, Turbine component reliability

Quantified: Availability, Reliability & Condition

Condition Model Markov Chain Solvedvia Monte Carlo Sim.

Wind Regime, Turbine Power Curve, Turbine and Market Economics

Evaluated via Wind Turbine performance

Power PerformanceEvaluation Model

0

0.5

1

1.5

2

1 4 7 10 13 16 19 22 25

Modelling Approach Summary

Wind Turbine Yield Model

Wind Speed Distribution

0

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

0.0

1.4

2.7

4.1

5.4

6.8

8.1

9.5

10.8

12.2

13.5

14.9

16.2

17.6

18.9

20.3

21.6

23.0

24.3

25.7

27.0

28.4

29.7

31.1

32.4

Wind Speed m/s

Pro

babi

lity

Yearly site wind speed data used to form probability distribution

Wind speed is generated by randomly sampling this distribution

This is fed into the turbine curve model

Wind Speed Model Turbine Curve Model

Sampled from manufacturers data sheet

Trials

TROCElecT MPMPMWhvenue

1Re

Revenue Calculation

Market Price Electricity = £36/MWh Market Price ROCs = £40/MWh

The Asset Management Model

Yield, cost, revenue, spares, operational life & maintenance objectives

Tested assetmanagement policies

Asset Management & High-Level Objectives Model

Turbine condition states, Turbine component reliability

Quantified: Availability, Reliability & Condition

Condition Model Markov Chain Solvedvia Monte Carlo Sim.

Wind Regime, Turbine Power Curve, Turbine and Market Economics

Evaluated via Wind Turbine performance

Power PerformanceEvaluation Model

0

0.5

1

1.5

2

1 4 7 10 13 16 19 22 25

Modelling Approach Summary

Maintenance Models

1. Scheduled 6-Month Maintenance

Maintain at set intervals (current practice)

2. Condition Based Maintenance

Maintain at intervals informed via condition information

Replacement and Repair Costs

These costs are subtracted from the turbine revenue stream

Maintenance Frequency

Weather Constraints

Other conditions such as lightning or snowfall also prohibitive.

Wind speed (M/S)

Restrictions

30 No access to site

20 No climbing turbines

18 No opening roof doors fully

15 No working on roof of nacelle

12 No going into hub

10 No lifting roof of nacelle

7 No blade removal

5 No climbing met masts

CBM Maintenance Regime

Condition-Based decision model: couple condition & maintenance

Use concept of operating risk as decision-metric:

)Im()Pr()( eventeventstateRiskNN

Im(event) = Economic term(s), currently replacement cost of component

Risk Magnitude Informing Maintenance Intervals

0

200

400

600

800

1000

1200

1 2 3 4 5 6 7 8

MODEL STATE

RIS

K

Risk associated with each model state

The risk magnitude sets the level of maintenance urgency via delay time

Condition and Maintenance are now Linked

Simulation Study for CM Evaluation

These concepts were coded in Fortran 95

Multiple runs of the program were conducted: Average values and confidence limits calculated:

The following metrics were produced for a periodic maintenance policy:

N

SDevZL

@ 95% degree of confidence,

Z score =1.65

Gearbox 1.31

Generator 0.21

Electronics 0.43

Blade 0.07

Maintenance Comparison: Periodic Vs. CBM

• Simulation resolution is 1 day

• Simulation run 30 times, for 14,000 trials

Conclusions

• It appears possible to quantify the benefits of WTG CM via modelling

• Less assumptions, better characterisation of actual processes

• Test model for different conditions: wind regime, turbine ratings etc.

• Model reliability of condition monitoring system itself

Implications for Future Research

• CBM annual value of £2,000: borderline cost-effective

… But how is value of CM affected by different operating conditions?

• The model outputs can provide ballpark figures but the assumptions must be valid!

Towards Quantification of Condition Monitoring Benefit for Wind Turbine Generators

Mr David McMillan and Dr. Graham W. AultE-mail: dmcmillan@eee.strath.ac.uk

Many thanks to Yusuf Patel at Scottish Power and Peter Diver at ITI Energy.

This research was conducted under the PROSEN project, EPSRC grant number EP/C014790/1.