Meredith Sperling 1,2 Alexandra St. Pé 3, Aditya Choukulkar 4, Cristina L. Archer 5, Ruben Delgado 2,6

Transcription

1 Evaluating Wind Power Prediction Uncertainty Using Scanning Doppler Wind Lidar AMS 2018: Ninth Conference on Weather, Climate, and the New Energy Economy Meredith Sperling 1,2 Alexandra St. Pé 3, Aditya Choukulkar 4, Cristina L. Archer 5, Ruben Delgado 2,6 1. UMBC Departments of Mechanical Engineering and Mathematics 2. NOAA ESSRST 3. UMBC Department of Geography and Environmental Systems 4. NOAA CIRES 5. University of Delaware 6. UMBC Joint Center for Earth Systems Technology 1

2 Background: The Power Equation & Manufacturer s Power Curve Power Estimate = C p 0. 5 ρ A U 3 Efficiency term Available Power ρ = air density, A = rotor area C p = power coefficient - C p (U) U = wind speed MPC Plots Power as a Function of U U (m/s) 2

3 Lewes, Delaware September-October MW Coastal Turbine (Gamesa) Data Collection: The VERTEX Campaign Remote Sensing: Wind: Scanning Doppler wind lidar Windcube 200s (~3km) Temperature: Microwave Radiometer 3

4 Data Collection: Wind Profile Reconstruction Plan Position Indicator (PPI) Scans: 200s Lidar PPI scan Elevation Angle 1 PPI scan Elevation Angle 2 PPI scan Elevation Angle 3 PPI scan Elevation Angle 4 OI Reconstructed Profile * Choukulkar, A., Calhoun, R., Billings, B., and Doyle, J. D. A modified optimal interpolation technique for vector retrieval for coherent doppler lidar. IEEE Geoscience and Remote Sensing Letters, 9(6): ,

5 Motivation: Accurate and Precise Power Curves OBSERVATION: Hub-height wind speed alone does not accurately and precisely predict power ഥU hub (m/s) Accuracy how close output values are to the predicted value Precision how close output values are to each other 5

6 Method: The Uncertainty Trifecta Wind Resource Uncertainty Turbine Efficiency Uncertainty Turbine Power Prediction Uncertainty Available Power Uncertainty Power Estimate = C p 0. 5 ρ A U 3 6

7 Height (m) Previous Work: Addressing Available Power Uncertainty ഥU 7 Power Estimate = C p 0. 5 ρ A U 3 ഥU 6 ഥU 5 U hub height # UREWS = 3 1 N U i 3 A i ഥU 3 ഥU 2 ഥU 1 A i=1 Wind Speed (m/s) Included IEC standard for available power assessment * * IEC. Power performance measurements of electricity producing wind turbines edition 2, committee draft 2. Technical Report IEC , International Electrotechnical Committee; 2017 Wagner R, Antoniou I, Pedersen SM, Courtney MS, Jørgensen HE. The influence of the wind speed profile on wind turbine performance measurements. Wind Energy. 2009;12(4):

8 Height (m) Previous Work: Addressing Wind Resource Uncertainty Vertical Wind Profile Types Power Law Linear Strong Inflection Inverted Average Wind Speed (m/s) *St Pé A, Sperling M, Brodie JF, Delgado R. Classifying rotor-layer wind to reduce offshore available power uncertainty. Wind Energy. 2018; accepted 8

9 Height Research Questions: Quantifying and Comparing Uncertainty What is power prediction uncertainty when using hub-height wind speed? What is the relative power prediction uncertainty reduction achieved by accounting for wind speeds throughout the rotor layer via Wagner REWS? Does the uncertainty of hub-height wind speed and REWS predictions vary during different classified profile types? If so, how? Wind Speed 9

10 10

11 Results: The Value of REWS ഥU Hub & ഥU REWS Prediction Uncertainties 19.1% 17.2% 0.15 MW 0.14 MW MW 0.033MW REWS reduces all 3 errors measured 11

12 Height (m) Results: Hub-Height Wind Speed Prediction Uncertainty by Type Profile Type Power-Law Unclassified Strong Inflections Inverted Linear FVU 7.9% 10.3% 16.3% 31.9% 34.4% P Bias (MW) P Scatter (MW) Average Wind Speed (m/s) Increasing Error and Uncertainty 12

13 Results: Relative Value of REWS by Type 55.8% 13.3% 12.1% 9.7% 9.0% 3.3% 1.4% -0.1% 1.5% -2.9% 13

14 Height Conclusions: ഥU REWS vs. ഥU hub Overall, ഥU REWS reduces power prediction uncertainty by 1.9% ഥU hub (m/s) ഥU hub uncertainty by classified profile type Standard power-law profiles have the lowest ഥU hub prediction uncertainty (7.9%) Value of ഥU REWS by classified profile type Strong inflection types have the largest uncertainty reduction achieved by ഥU REWS (3.3%) Wind Speed 14

15 Future Work: Current Research Questions Determining the effect of atmospheric stability on power prediction uncertainty Establishing a correlation between profile types and atmospheric stability Improved Wind Power Predictions Using machine learning to individualize power prediction models in given local conditions 15

16 Acknowledgements NOAA Office of Education: Educational Partnership Program NOAA Cooperative Science Center for Earth System Sciences and Remote Sensing Technologies (NOAA-CESSRST) under the Cooperative Agreement Grant # NA16SEC Maryland Energy Administration (contract number U00P ) National Science Foundation (award number ) The statements contained within the manuscript/research article are not the opinions of the funding agency or the U.S. government, but reflect the author s opinions. 16

17 Thank you for listening For more information: Contact information: 17