Diese Präsentation wurde erfolgreich gemeldet.
Die SlideShare-Präsentation wird heruntergeladen. ×

Delle monache luca

Anzeige
Anzeige
Anzeige
Anzeige
Anzeige
Anzeige
Anzeige
Anzeige
Anzeige
Anzeige
Anzeige
Anzeige
Wird geladen in …3
×

Hier ansehen

1 von 46 Anzeige
Anzeige

Weitere Verwandte Inhalte

Ähnlich wie Delle monache luca (20)

Anzeige

Aktuellste (20)

Delle monache luca

  1. 1.   NCAR’s  Wind  Forecas2ng  System       Luca  Delle  Monache  et  al.   (lucadm@ucar.edu)   Na2onal  Center  for  Atmospheric  Research  (NCAR)  −  Boulder,  CO,  USA         Winterwind  –  Interna2onal  Wind  Energy  Conference   12-­‐13  February,  2013,  Östersund,  Sweden  
  2. 2. Outline   •  The  U.S.  Na2onal  Center  for  Atmospheric  Research  (NCAR)   •  Renewable  energy  forecas2ng  research  and  development   •  NCAR-­‐Xcel  energy  project   •  Probabilis2c  power  predic2ons   o  The  Analog  Ensemble  (AnEn)   o  Test  cases   •  Summary   2
  3. 3. What  is  the  US  Na2onal  Center  for   Atmospheric  Research  (NCAR)?   •  NCAR  is  a  Federally  funded  research  and   development  center  sponsored  by  the  U.S.   Na2onal  Science  Founda2on   •  NCAR  is  operated  by  the  University   Corpora2on  for  Atmospheric  Research   (UCAR),  a  non-­‐profit  corpora2on.   •  UCAR  has  1400  employees  and  ~$250M   budget.   •  Research  is  conducted  on  climate  and   weather  modeling,  air  chemistry,   NCAR, Boulder, CO thunderstorms,  hurricanes,  icing,  turbulence,   societal  impacts  of  weather,  energy,  solar   physics,  etc.  
  4. 4. Power  Predic2on   Goal:   Accurate  power  forecasts  and  reliable  quan2fica2on  of  forecast   uncertainty       Mo+va+on:   •  Wind  power  forecas2ng  is  necessary  for  effec2ve  grid  integra2on   o  Day-­‐ahead  forecas2ng  –  energy  trading   o  Short-­‐term  forecas2ng  –  grid  integra2on  &  stabiliza2on   •  Thus,  an  effec2ve  forecas2ng  system  should  include  components  for   both    
  5. 5. Renewable  Energy  Forecas2ng  
  6. 6. Xcel  Energy  Service  Areas   Wind Farms (50+) 3585 Turbines (growing) 4842 MW+ (wind) ~10% Wind (highest in continental US) 3.4 million customers (electric) Annual revenue $11B Provides  good  geographical  diversity  for  research  and  tes2ng  
  7. 7. NCAR’s  Wind  Energy  Predic2on     System  for  Xcel  Energy   CSV  Data   NCEP Data Wind Farm Data Operator GUI NAM Nacelle wind speed GFS Generator power RUC Node power GEM (Canada) Met tower Availability Statistical WRF RTFDDA Verification System Dynamic, Meteorologist Integrated Wind to Energy GUI WRF+MM5 Forecast Conversion Ensemble System System Subsystem (DICast®) WRF  Model  Output   Supplemental VDRAS Wind Farm Data (nowcasting) Met towers Wind profiler Surface Stations Expert System Windcube Lidar (nowcasting)
  8. 8. WRF  RTFDDA  Model  Domains   Determinis2c  System   Ensemble  System  (30  members)   D3 = 3.3 km D2 = 10 km D1 = 30 km 0-72 hrs D1 = 30 km 0-48 hrs D2 = 10 km 0-72 hrs D2 = 10 km 0-48 hrs D3 = 3.3 km 0-24 hrs
  9. 9. NCAR-Xcel Energy Project Accurate prediction economical benefits ~$1.9M  per  each   percent     improvement  
  10. 10. NCAR-Xcel Energy Project CO2 reduction due to accurate predictions “The avoided generation occurred when Xcel cycled offline baseload thermal units (coal or natural gas combined cycle) due to extended periods of forecasted low loads and high winds.” AVOIDED EMISSIONS DUE TO IMPROVED PREDICTIONS: 238,136 TONS OF CO2 MWh’s of avoided generation in 2011 Arapahoe 3 = 317 Arapahoe 4 = 6,941 Cherokee 1 = 11,606 Cherokee 2 = 13,772 Valmont 5 = 10,061 FSV CC = 93,626 RMEC CC = 308,989 10
  11. 11. Probabilis2c  Power  Predic2on   Goal:   Accurate  power  forecasts  and  reliable  quan2fica2on  of  forecast   uncertainty       Mo+va+on:   •  Wind  power  forecas2ng  is  necessary  for  effec2ve  grid  integra2on   o  Day-­‐ahead  forecas2ng  –  energy  trading   o  Short-­‐term  forecas2ng  –  grid  integra2on  &  stabiliza2on   •  Thus,  an  effec2ve  forecas2ng  system  should  include  components  for   both    
  12. 12. Ensemble  (En)  Predic2on   fi The  single  determinis+c  forecast  f0   fails  to  predict  the  TRUE     f0 Wind  Speed   The  ini2al  probability  density  func2on   PDF(0)  represents  the  ini2al   uncertain2es       PDF(t) TRUE An  ensemble  of  perturbed  forecasts   fi,  star2ng  from  perturbed  ini2al   condi2ons  designed  to  sample  the   ini2al  uncertain2es  can  be  used  to   PDF(0) es2mate  the  probability  of  future   states  PDF(t)   Forecast  +me   12
  13. 13. Weather  analogs:  basic  idea   13
  14. 14. Weather  analogs:  basic  idea   Today
  15. 15. Weather  analogs:  basic  idea   Today One week ago? 15
  16. 16. Weather  analogs:  basic  idea   Today One week ago? 5 years ago?!? 16
  17. 17. Weather  analogs:  basic  idea   Today One week ago? 5 years ago?!? 17
  18. 18. Weather  analogs:  basic  idea   Today One week ago? 5 years ago?!? 18
  19. 19. Weather  analogs:  basic  idea   Today One week ago? 5 years ago?!? 19
  20. 20. Weather  analogs:  basic  idea   Today   Can  we  use  this  informa2on   (i.e.,  both  obs  and  re-­‐analysis),     to  improve  forecasts  or  resource  es2mates?   One week ago?   5 years ago?!? 20
  21. 21. Analog  Ensemble  (AnEn)   Mon Tue Wed Thu Fri Sat Sun t=0 Time   Analog search as in Delle Monache et al. (MWR 2011) 21
  22. 22. Analog  Ensemble  (AnEn)   PRED Mon Tue Wed Thu Fri Sat Sun t=0 Time   Analog search as in Delle Monache et al. (MWR 2011) 22
  23. 23. Analog  Ensemble  (AnEn)   PRED OBS Mon Tue Wed Thu Fri Sat Sun t=0 Time   Analog search as in Delle Monache et al. (MWR 2011) 23
  24. 24. Analog  Ensemble  (AnEn)   PRED OBS Mon Tue Wed Thu Fri Sat Sun t=0 Time   Analog search as in Delle Monache et al. (MWR 2011) 24
  25. 25. Analog  Ensemble  (AnEn)   PRED OBS Mon Tue Wed Thu Fri Sat Sun t=0 Time   Analog search as in Delle Monache et al. (MWR 2011) 25
  26. 26. Analog  Ensemble  (AnEn)   PRED OBS Mon Tue Wed Thu Fri Sat Sun t=0 Time   Analog search as in Delle Monache et al. (MWR 2011) 26
  27. 27. Analog  Ensemble  (AnEn)   PRED OBS Mon Tue Wed Thu Fri Sat Sun t=0 Time   Wed Fri Sat Tue Sun Mon Thu farthest   closest   analog   analog   “Analog”  Space   Analog search as in Delle Monache et al. (MWR 2011) 27
  28. 28. Analog  Ensemble  (AnEn)   PRED OBS Mon Tue Wed Thu Fri Sat Sun t=0 Time   Wed Fri Sat Tue Sun Mon Thu farthest   closest   analog   analog   “Analog”  Space   Analog search as in Delle Monache et al. (MWR 2011) 28
  29. 29. Analog  Ensemble  (AnEn)   PRED OBS Mon Tue Wed Thu Fri Sat Sun t=0 Time   2-­‐member  AnEn   Wed Fri Sat Tue Sun Mon Thu farthest   closest   analog   analog   “Analog”  Space   Analog search as in Delle Monache et al. (MWR 2011) 29
  30. 30. How skillful is AnEn? •  AnEn generated with Environment Canada GEM (15 km), 0-48 hours •  Comparison with: o  Environment Canada Regional Ensemble Prediction System (REPS, next slide) o  Logistic Regression (LR) out of 15-km GEM o  LR our of REPS, i.e., Ensemble Model Output Statistics (EMOS) •  Period of 15 months (verification over the last 3 months) •  10-m wind speed •  550 surface stations over CONUS (in two slides) •  Probabilistic prediction attributes: statistical consistency, reliability, sharpness, resolution, spread-error consistency 30
  31. 31. Regional  Ensemble  Predic2on  System  (REPS) •  Model: GEM 4.2.0 (vertical staggering) •  20 members + 1 control run •  72 hours forecast lead time •  Resolution: ~33 km with 28 levels •  Initial conditions (i.e., cold start) and 3-hourly boundary condition updates from GEPS (EnKF + multi-physics) •  Physics: o  Kain et Fritsch (1993) for deep convection o  Li et Barker (2005) for the radiation o  ISBA scheme (Noilhan et Planton, 1989) for surface •  Stochastic Physics: Markov Chains on physical tendencies 31
  32. 32. Ground  truth  dataset   •  550 hourly METAR Surface Observations •  1 May 2010 – 31 July 2011, for a total of 457 days •  10-m wind speed 32
  33. 33. Probabilis2c  forecast  ahributes:  Reliability   Example: ①  An event (e.g., wind speed > 5 m/s) is predicted to happen with a 30% probability ②  We collect the observations that verified every time we made the prediction in 1 ③  If the frequency of the event in the observation collected is 30%, then the forecast is perfectly RELIABLE
  34. 34. Analysis  of  reliability  &  sharpness   Reliability and sharpness diagram: 10-m wind speed > 5 m s-1, 9-h fcst 1 1 (a) (b) Observed Relative Frequency REPS EMOS 0.8 0.8 Forecast Probability 0.6 0.6 0.4 0.4 0.2 0.2 0 0 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 1 1 (c) (d) Observed Relative Frequency LR AnEn 0.8 0.8 Forecast Probability 0.6 0.6 0.4 0.4 0.2 0.2 0 0 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 Forecast Probability Forecast Probability 34
  35. 35. Probabilis2c  forecast  ahributes:  Sharpness   Sharpness refers to the degree of concentration of a forecast PDF’s probability density, and is a property of the forecasts only. Ideally, we want the forecast system, while mainly reliable, with as many forecasts as possible close to 0% and 100%, corresponding to a perfect deterministic forecast system. However, an improvement in sharpness does not necessarily mean that the forecast system has improved. 0 5 10 15 20 T 0 5 10 15 20 T Sharper Forecast Less Sharp Forecast
  36. 36. Analysis  of  reliability  &  sharpness   Reliability and sharpness diagram: 10-m wind speed > 5 m s-1, 9-h fcst 1 1 (a) (b) Observed Relative Frequency REPS EMOS 0.8 0.8 Forecast Probability 0.6 0.6 0.4 0.4 0.2 0.2 0 0 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 1 1 (c) (d) Observed Relative Frequency LR AnEn 0.8 0.8 Forecast Probability 0.6 0.6 0.4 0.4 0.2 0.2 0 0 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 Forecast Probability Forecast Probability 36
  37. 37. ATTRIBUTES OF FORECASTSahributes:   Probabilis2c  forecast   atic error RESOLUTION – Different forecasts Resolu2on   observed events precede different stConsider different classes of classes of fcst events Consider different forecast events. bsIf all observed classes corresponds to different by If all observed classes are preceded forecast classes, ons => probabilisticdifferent forecasts => RESOLUTION. then the distinctly forecast has perfect Y PERFECT RESOLUTION orrected Resolution CANNOT BE statistically 37
  38. 38. Analysis  of  Resolu2on  (1)   Rela2ve  Opera2ng  Characteris2cs  skill  score,  10-­‐m  wind  speed  ≥  5,  10  m  s-­‐1   WSPD  >  5  m  s-­‐1   WSPD  >  10  m  s-­‐1   0.8 1 beNer   AnEn EMOS REPS LR (a) (b) 0.7 0.9 0.6 ROCSS 0.8 0.5 0.7 0.4 worse   0.6 0 6 12 18 24 30 36 42 48 0 6 12 18 24 30 36 42 48 Forecast Lead Time (hours) Forecast Lead Time (hours) 38
  39. 39. AnEn  sensi2vity   Rela2ve  Opera2ng  Characteris2cs  skill  score,  10-­‐m  wind  speed  ≥  5  m  s-­‐1   AnEn  with  a  shorter     AnEn  built  with  a  coarser   training  data  set  (15  à  9  months)   dynamical  model  (15  à  33  km)   beNer   0.8 0.8 0.75 0.7 ROCSS ROCSS 0.7 0.6 0.65 AnEn AnEn LR worse   LR Fine Resolution (15−km GEM) Full Training Short Training Coarse Resolution (33−km REPS mem. # 20) 0.6 0.5 0 6 12 18 24 30 36 42 48 0 6 12 18 24 30 36 42 48 Forecast Lead Time (hours) Forecast Lead Time (hours) 39
  40. 40. Power predictions: Experiment design •  Test site: Wind farm in northern Sicily − 9 turbines, 850 kW Nominal Power (NP) •  Training period: November 2010 - October 2012 •  Verification period: November 2011 – October 2012 •  Probabilistic prediction systems: ECMWF EPS, COSMO LEPS, AnEn
  41. 41. Power  predic2ons   Normalized Power Forecast Lead Time 41
  42. 42. RMSE  of  ensemble  means  
  43. 43. Probabilis2c  forecast  ahributes:     Sta2s2cal  and  spread-­‐error  consistency   ①  The ensemble spread tell us how uncertain a forecast is. Ideally, large spread should be associate with larger uncertainties, low spread should indicate higher accuracy ②  If an ensemble is perfect, than the observations are indistinguishable from the ensemble members
  44. 44. Spread-­‐skill  relathionship   ECMWF  EPS  
  45. 45. Summary   •  NCAR’s  wind  energy  research  and  development   o  Icing,  fine-­‐scale  &  boundary  layer  meteorology  research   o  Data  assimila2on,  sta2s2cal  learning   o  Wind  &  power  predic2ons,  wind  resource  assessment   •  The  NCAR-­‐Xcel  Project,  a  successful  story   •  The  analog  ensemble  provides  accurate  predic2ons/es2mates  and   reliable  uncertainty  quan2fica2on  (at  a  lower  computa2onal  cost)     •  The  analog  ensemble  can  be  used  for  dynamical  downscaling  and  wind   resource  assessment       45
  46. 46. THANKS!   (lucadm@ucar.edu)   Collaborators  include:   Bill  Mahoney,  Sue  Haupt,  Greg  Thompson,  Gerry  Wiener,  Bill  Myers,  David  Johnson,   Yubao  Liu,  Jenny  Sun,  Tom  Hopson,  Branko  Kosovic,  Julie  Lundquist,  Stefano   Alessandrini,  Seth  Linden,  Julia  Pearson,  Frank  McDonough,  …   References   •  Delle  Monache  et  al.,  2011:  Kalman  filter  and  analog  schemes  to  postprocess  numerical  weather  predic2ons.  Monthly   Weather  Review,  139,  3554−3570.   •  Delle  Monache  et  al.,  2013:  Probabilis2c  weather  predic2ons  with  an  analog  ensemble.  Mon.  Weather  Review,  sub.     •  Vanvyve  and  Delle  Monache,  2013:  Wind  resource  assessment  with  an  analog  ensemble.  Journal  of  Applied   Meteorology,  in  prepara2on.   46

×