slide1 n.
Download
Skip this Video
Loading SlideShow in 5 Seconds..
Modelling Through High Frequency Data Sampling and Other Advantages PowerPoint Presentation
Download Presentation
Modelling Through High Frequency Data Sampling and Other Advantages

Loading in 2 Seconds...

play fullscreen
1 / 22

Modelling Through High Frequency Data Sampling and Other Advantages - PowerPoint PPT Presentation


  • 144 Views
  • Uploaded on

Modelling Through High Frequency Data Sampling and Other Advantages. by Francis Pelletier - PhD Candidate, ETS Christian Masson – PhD Director, ETS Antoine Tahan – PhD Co-Director, ETS Martin Jetté , General Manager, OSIsoft Canada ULC. Presented at :. In collaboration with.

loader
I am the owner, or an agent authorized to act on behalf of the owner, of the copyrighted work described.
capcha
Download Presentation

PowerPoint Slideshow about 'Modelling Through High Frequency Data Sampling and Other Advantages' - cicada


Download Now An Image/Link below is provided (as is) to download presentation

Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author.While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server.


- - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - -
Presentation Transcript
slide1

Modelling Through High Frequency Data Sampling and Other Advantages

by

Francis Pelletier - PhD Candidate, ETS

Christian Masson – PhD Director, ETS

Antoine Tahan – PhD Co-Director, ETS

Martin Jetté, General Manager, OSIsoft Canada ULC

Presentedat :

In collaboration with

École de Technologie Supérieure

slide3

PhD Project: Power Performance Evaluation and Improvement of Operational Wind Power Plant

Project’s objectives

1. Improve Actual Power Performance Evaluation Techniques (modelling)

2. Improve Power Performance of Operational Wind Power Plant

An energy increase of 1% in energy output could represent a financial benefit of

3-4 Millions Euros (100MW Wind Power Plant)

/21

advanced data acquisitioning system
Advanced Data acquisitioning system

FrequencySampling = 1 Hz

Number of tags per second ≈ 100 000 tags

PI DiskSpace per year ≈ 0.6 Tbyte / year

/21

industry standard vs high frequency data sampling 10 minute average vs 1 second
Industry-Standard vs High frequency data sampling(10 minute-average VS 1 second)

POWER

1 second data output

Typical industry-standard data logger output

(10 minute-average)

TIME

Power (10 min)

Power (1 sec)

/21

slide7

Advantages related to high frequency data samplingNew Metrics

  • ACTUALLY AVAILABLE WITH 10 MINUTE-AVERAGE DATA
  • AVERAGE10min
  • STD10min
  • min10min
  • max10min
  • NEW METRICS WITH 1 SECOND DATA SAMPLING
  • Avg, Std, min, max for period different than 10 minutes
  • Statistical distribution
  • Skewness - Kurtosis
  • Derivatives
  • Filters / smoothing
  • Dynamic response
  • (V10min)2 ≠ (V2)10min
  • Etc…

/21

advantages related to high frequency data sampling increases the number of data for modelling
Advantages related to high frequency data samplingIncreases the Number of Data for Modelling

Period: 2009-02-01

00:00:00

to 2010-02-01

00:00:00

2009-02-01

00:00:00

to 2010-02-01

00:00:00

2009-02-01

00:02:00

to 2010-02-01

00:02:00

P10min = 100 kW

P10min = 500 kW

00:00:00

00:10:00

00:20:00 ...

to 2010-02-01

00:04:00

2009-02-01

00:04:00

2009-02-01

00:06:00

to 2010-02-01

00:06:00

00:05:00

00:15:00

to 2010-02-01

00:08:00

2009-02-01

00:08:00

P10min = 300 kW

/21

power curve analysis standard analysis
Power CurveAnalysis – Standard Analysis

Tower vibration

Anemometer

defect

Power [kW]

Faults and downtime

Ice accretion on blades

High wind speed cut-out

Wind Speed [m/s]

/21

Ref.: Jean-Daniel Langlois – GL : GreenPowerConferences – Houston09

slide10

Problems faced by O&M personel

Power Curve

STD>50kW

Power [kW]

Wind Speed [m/s]

/21

slide11

Review of literature – Power curve modelling

  • Discrete models (Bin)
  • P=f(WS@nacelle)
  • IEC61400-12-1
    • Air density correction
    • TI correction
    • Multivariate analysis
  • IEC61400-12-2
  • Parametric models
  • Polynomial functions
  • Logistic function with 4 parameters (G.A.)
  • Power curve partitions (3 regions)
  • Stochastic models
  • Markov Chain
  • Non-parametric models
  • K-NN
  • SVM
  • Boosting Tree
  • ANN
  • etc...
  • Data Reduction Techniques
  • PCA
  • Self Organizing Map (SOM)

It has been found that the power curve modeling’s precision is difficult to improve because of: - Non-linearities

- Interaction between variables

/21

slide12

Some Results with ANN modeling

  • P = f (WS & Air Density)

Air Density

1.22 kg/m^3

to

1.3 kg/m^3

POWER [kW]

Wind Speed [m/s]

slide13

Some Results with ANN modeling

  • P = f (WS & Turbulence Intensity)

Turbulence Intensity (TI)

7.5%

to

15%

POWER [kW]

Wind Speed [m/s]

slide14

Some Results with ANN modeling

  • P = f (WS & Wind Shear)

Wind Shear

(V80/V40)

1.15

to

1.5

POWER [kW]

Wind Speed [m/s]

slide15

Some Results with ANN modeling

  • P = f (WS & WSskew)

Wind Speed Skewness

-0.5

to

+0.25

POWER [kW]

Wind Speed [m/s]

control validation cut in cut out validation
Control validation (Cut-in & Cut-out validation)

Wind Speed (1sec)

Cut-out

( Lowwind speed)

Cut-in

Power (1sec)

168.75

137,5

75

43.75

12.5

Cut-in Wind Speed

(3.5 m/s)

-18.75

Enables: - Yaw validation & Yawerrordetection - Diverter position

- Pitch & RPM validation - Faults investigation

- Availability validation - etc...

/21

enhanced trouble shooting possibilities
Enhanced Trouble Shooting Possibilities

INSTANTANEOUS OVER POWER

Power (1sec) – Normal Behaviour

Power (1sec) – Over Power

Maximum Instantaneous Power

/21

advantages related to high frequency data sampling iii trouble shooting possibilities
Advantages related to high frequency data samplingIII- Trouble shooting possibilities

POWER

T X RPMgenerator

ENCODER MALFUNCTION

(1 second Data)

/21

slide20

Other opportunities related to high frequency data sampling

  • New faults and underperformance investigations techniques
  • Availability validation
  • Support for Insurance or legal claims
    • Ex.: Gear box, blade damage etc…
  • Advance power curve modelling (Markov chain)
  • Condition monitoring (ex.: FFT analysis…)
  • Forecasting Improvement (Ramp)
  • Improve quality control of data (Outlier’s identification)
  • Wake’s validation and investigation
  • Control validation and optimization
  • Etc…

/21

conclusion
Conclusion
  • High frequency data sampling improves modelling possibilities
  • Several other advantages related to high frequency data sampling have been found.
  • Future investigations will certainly demonstrate new advantages related to high frequency data sampling of wind power plants

/21

slide22

Merci!

Thankyou!