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: 10 th 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? … 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 Maintenance Reactive 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 Solved via Monte Carlo Sim. Wind Regime, Turbine Power Curve, Turbine and Market Economics Evaluated via Wind Turbine performance Power Performance Evaluation Model Yield, cost, revenue, spares, operational life & maintenance objectives Tested asset management 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 D C2 U C3 U C4 U 22 C1 D C2 DER C3 U C4 U 11 C1 DER C2 U C3 D C4 U 1 C1 U C2 U C3 U C4 U 9 C1 U C2 U C3 D C4 U 15 C1 DER C2 DER C3 D C4 U 12 C1 U C2 DER C3 D C4 U 3 C1 U C2 DER C4 U C3 U 25 C1 U C2 D C3 U C4 U 26 C1 DER C2 D C3 U C4 U 8 C1 DER C2 DER C4 DER C3 U 6 C1 U C2 DER C4 DER C3 U 5 C1 U C2 U C4 DER C3 U 17 C1 U C2 U C4 D C3 U 27 C1 U C2 D C4 DER C3 U 19 C1 U C2 DER C4 D C3 U 28 C1 DER C2 D C4 DER C3 U 24 C1 D C2 DER C4 DER C3 U 20 C1 DER C2 DER C4 D C3 U 4 C1 DER C2 DER C3 U C4 U 7 C1 DER C2 U C4 DER C3 U 2 C1 DER C2 U C3 U C4 U 18 C1 DER C2 U C4 D C3 U 23 C1 D C2 U C4 DER C3 U 13 C1 DER C2 U C4 DER C3 D 10 C1 U C2 U C4 DER C3 D 14 C1 U C2 DER C4 DER C3 D 16 C1 DER C2 DER C4 DER C3 D StatesComp #State # Normal/ DeratedAll1to8 ElectronicsC39to16 BladeC417to20 GearboxC121to24 GeneratorC225to28
The Power Performance Evaluation Model Yield, cost, revenue, spares, operational life & maintenance objectives Tested asset management policies Asset Management & High-Level Objectives Model Turbine condition states, Turbine component reliability Quantified: Availability, Reliability & Condition Condition Model Markov Chain Solved via Monte Carlo Sim. Wind Regime, Turbine Power Curve, Turbine and Market Economics Evaluated via Wind Turbine performance Power Performance Evaluation Model Modelling Approach Summary
Wind Turbine Yield Model 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 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 asset management policies Asset Management & High-Level Objectives Model Turbine condition states, Turbine component reliability Quantified: Availability, Reliability & Condition Condition Model Markov Chain Solved via Monte Carlo Sim. Wind Regime, Turbine Power Curve, Turbine and Market Economics Evaluated via Wind Turbine performance Power Performance Evaluation Model 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 30No access to site 20No climbing turbines 18No opening roof doors fully 15No working on roof of nacelle 12No going into hub 10No lifting roof of nacelle 7No blade removal 5No climbing met masts
CBM Maintenance Regime Condition-Based decision model: couple condition & maintenance Use concept of operating risk as decision-metric: Im(event) = Economic term(s), currently replacement cost of component
Risk Magnitude Informing Maintenance Intervals 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 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. Ault 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.