The plan for improving wind power forecasting by using upwind meteorological observations.

1. Reducing Wind Forecasting Uncertainty with PI and ECHO Using Real-time Offsite Meteorological Observations Marty Wilde – CEO, Principal EngineerWINData, Inc. Gregg Le Blanc – Flunky WINData, Inc. F5Direct, LP

2. PI-enabled Real Time Off-site Meteorological Observations for Improved Short-term Wind Forecasting, Operating Economics and System Reliability Marty Wilde – CEO, Principal EngineerWINData, Inc. Gregg Le Blanc – Flunky WINData, Inc. F5Direct, LP

3. Topics Brief introductions The importance of forecasting Wind power: ups and downs The impact of remote observations How WINData leverages ECHO and PI Future plans

4. WINData background Started by Marty Wilde In wind energy 1991 Wind energy specialists From raw dirt to developed wind farm Over 500 meteorological tower installs Specializing in high-end instrumentation and tall towers Goal: Advance the “state-of-the-art” by using real time measurements for model and investment validation

5. Why forecasting is important

6. The critical role of forecasting Wind penetration California: 20% renewables by 2010 33% renewables by 2020 Europe: 20% renewables by 2020 Already in place Denmark 20% of demand today 50% by 2025 Spain 20,000 MW capacity by 2010 > 15,000 MW in place today Peaked at 40% of demand (typically ~25% of demand)

7. Working together to improve forecasts Goal: Designing a program that results in better forecasts WINData &AWS Truewind Teaming up with utilities to improve system operations Better wind energy integration

8. Bonneville Power’s balancing act 1,500 MW Balancing Area 19 Wind Farms

9. BPA’s Balancing Area

10. Scheduling wind operations is complex

12. Large deviations power use of Portland, ORTheoretical Scheduled MW 730 MW 680 MW 400 MW 625 MW 0 MW

15. EPA Regulations0 MW

16. Ramp Events Not your friend.

17. Ramp events captured by ERCOT (1)

18. Ramp events captured by ERCOT (2)

19. Ramp effects at ERCOT News article from Reuters (2/27/08): Sudden 1,700 to 300 MW drop in wind power Increased demand due to colder temperatures Ordered 1,100 MW to be curtailed from interruptible customers within 10 min Other suppliers were unable to meet commitments Happy ending No other customers lost power Interruptible customers were back online 90 min later

20. Scheduling around ramp events Hour ahead schedules -forecast interpretation Forecasts are good at predicting that there WILL be a ramp event When will it arrive? Phase error How significant? Amplitude error Forecast error tends to increase around ramp events

21. WINData’s Plan Introducing WINDataNOW

22. Goal: Reduce the forecast uncertainty Idea Can short-term forecasts be improved by using actual upstream measurements from physical instruments? Requirements Off-site meteorological data from predominant wind directions Deliver measurements with enough resolution and timeliness to be useful

23. WINData’s technology Create a modern, remote “met” sensing system Combine off the shelf technologies: PI System infrastructure Sensors Logger equipment Focus on configuration vs. programming

24. WINDataNOW! – remote met observations Towers are strategically located Transmit high fidelity “line of site” data Use this data: Collaborate with forecaster Refine short-term forecasts Back-cast to develop more detailed climatology “Line of Site” Locations Wind Farm 60 – 100 minutes of prior notice Upwind Met Tower 30 – 60 miles upwind

25. Tower placement Prevailing Wind Direction

26. Tower placement Wind Speed meters / sec Wind Speed meters / sec Wind Speed meters / sec Wind Speed meters / sec

27. Anticipate changes Ramp up Ramp down Improved wind energy integration on the grid More data: smoother operations Apply “line of site” datato better understand near-term transients Power Production vs. Augmented forecast MWh Wind Speed vs. Augmented forecast meters / sec

28. Example A Site in Montana

29. Today’s data expectations (m/s) Typical industry-standard data logger output

30. The same time period, now in high fidelity Actual Wind Speed (m/s)

31. Comparison Average vs. Actual Wind Speed (m/s) Averages vs. higher fidelity data

32. Using PI’s 10m Averages

33. Using PI’s 5m Averages

34. Zooming out a little more High wind speed cut-off

35. Increased meteorological insight Re-forecast Account for actual upstream observations Decrease forecast error around ramp events Advanced warning Situational awareness Reduce costs Avoid penalties Use this Hour Ahead To reduce this

36. C-Gorge Project WINData in The Columbia Gorge

37. Map of C-Gorge vs. existing sites

38. WDN Locations vs. BPA towers Goodnoe 7 Mile Hill WDN1 WDN2 Wasco

39. How WINDataNOW! works From full-service to self-service Ways of engaging WINData Project management Deployment & management of early warning met arrays Enterprise-wide retrofit projects “WindTelligence” Analysis Development of situational awareness tools and IP Subscription service Data from reference locations Hosted central PI Server Stream data to your local system

40. Subscriber architecture * Subscriber 1 Connection OffsitePI System WINData Subscriber 2 WINData Managed Subscriber 3 WINData Managed PI SCADA Your Operations * Not to scale

42. One-time setup and validation

43. Ownership and management by customer

44. Your own IPConnection WINData Hardware Managed by you PI SCADA Your Operations

45. Combination architecture Other sites in the area Subscriber 1 Connection OffsitePI System WINData Hardware Subscriber 2 WINData Managed Subscriber 3 Managed by you PI SCADA Your Operations

46. Acquisition architecture Advantages Higher resolution data Rolls out quickly OffsitePI System Met Tower Secure Connection Solar Panel Wireless Connectivity Architecture Fits in place with existing PI infrastructure Fault tolerant Low power WINData Acquisition Hardware / Software Data Buffering Data Acquisition Battery

47. Data features Focus on data confidence Logger spools data if connectivity goes down Supports “industry standard” At least 2Hz data always streamed to PI Keep the data at its original resolution Tune the data with standard PI settings 10m averages always available via PI

48. Benefits of working on OSIsoft’s platform Component PI Server ECHO New ECHO to PI Interface Partner network Advantage Directly integrate with utilities and system operators Spool data during connectivity outages Speaks PINET over TCP/IP natively Solution works immediately with partners like Transpara’sVisualKPI

49. Demo D.I.Y. http://demo.transpara.com/

50. Future possibilities Smart Grid Provide “fuel insight” to consumers Encourage informed choices – demand side management Better investments Prospecting – High Fidelity “Bankable Data” Improved climatology for new wind energy sites Advanced wind farm design Turbulence, FFT Analysis, Matrix correlations Integrate shiny new technologies Vertical profiling SODAR, LIDAR, etc.

51. Looking forward Upstream observations can dramatically improve forecasts (Papers by Kristin Larson, 3TIER) Significant interest from utilities, operators, and forecasters in short-term ramp event prediction WINData wants to work with you Contact Marty Wilde, Dave Roberts, or Gregg

52. Special Thanks OSIsoft for its support Johnson City office and the embedded team for ECHO work Dave Roberts, Business Development, and Sales for advocacy ProSoft for its continued support Transpara for use of it’s VisualKPI software

53. Find us after the talk! Marty Wilde marty.wilde@windata-inc.com Gregg Le Blanc gregg.leblanc@windata-inc.com gregg@f5direct.com Dave Roberts – OSIsoft droberts@osisoft.com