Download - A self-guided presentation on Wind Insight Dr Nicholas Cutler February 2012 [email protected]
A self-guided presentation on Wind Insight
Dr Nicholas CutlerFebruary 2012
[email protected] www.roamconsulting.com.au
Overview
• This presentation is intended as a guide to interpreting and using ROAM’s wind power forecasting tool, Wind Insight
• Presentation structure:– Overview and advantages of Wind Insight– Performance statistics of Wind Insight– Interactive examples– Subscription options
Overview of Wind Insight
• Wind Insight provides:– Point forecasts of wind power (“expected generation”)
Overview of Wind Insight
• Wind Insight provides:– Point forecasts of wind power (“expected generation”)– Alerts of potential large rapid changes in wind power, with
likelihoods
Overview of Wind Insight (cont’d)
• Wind power field forecast animations provide additional insight into potential scenarios during uncertain periods
Overview of Wind Insight (cont’d)
• Can be used for single wind farms or the aggregated wind power from groups of wind farms
• Based on novel approach using numerical weather prediction systems, developed by Dr Nick Cutler during his research at the University of New South Wales and in consultation with the Australian Energy Market Operator
Advantages of Wind Insight- compared with other wind power forecasting systems such as the Australian
Wind Energy Forecasting System (AWEFS)
• By considering that significant wind features may deviate from their forecast positions, Wind Insight provides a superior assessment of possible wind generation scenarios compared to most wind power forecasting tools.– Due to wind feature forecast displacement, point forecasts (such as those
that AWEFS provides) may not capture possible large rapid changes
• Wind Insight provides alerts with likelihoods of possible large rapid changes, covering high wind speed cut-outs, and rapid changes in wind speed caused by large synoptic systems as well as periods of localised high variability
• Wind power fields: graphical animations of the underlying wind patterns allow forecast users to visualise potential scenarios of wind power and respond accordingly
Wind power fields
Estimated speed and direction of the most prominent moving wind features
Australian coastlineForecast hub height wind directions
Wind farm location
Wind speed forecasts are transformed
• Local elevation and surface roughness affect local wind speeds, making displacing wind features not a simple task
• Thus, the wind power fields use transformed wind speed forecasts where local modelled terrain effects are made ‘equivalent’ to the terrain of the wind farm site
• This allows feature displacements to be used and visualised directly
Site-equivalent wind speedsRaw wind speed forecasts
Wind speeds over the ocean are reduced
Wind speed transformation over land is more complex
Other wind power field display formats (1)
• 3D Wind power field can be displayed in 2D format
• Press “Page Up” and “Page Down” to play again.
2D wind power formatStandard 3D field format
Other wind power field display formats (2)
• ‘Coloured changes’ format colours according to the severity of large rapid changes
• This example also includes locations of high wind speed alerts
• Click to start animation
Coloured changes format2D wind power format
Alerting results from demo Wind Insight forecasting tool
• Canunda and Lake Bonney 1 wind power changes by > 65 MW in 30 minutes in 12-month period using ACCESS-A (BoM)
Performance measure Change in wind speed events
High wind speed cut-out events
All events
Number of events 58 15 73
Number of events correctly alerted
39 12 51
Percentage of events correctly alerted
67% 80% 70%
Percentage of time-steps alerted 5% 6% 10%
126 MW
(These results are based only on AEMO wind power dispatch data – results would improve if turbine availability or wind speed information were available)
Alerting results from Wind Insight prototypedeveloped for AEMO in 2010
• Total SA wind power changes by > 200 MW in 30 minutes in 18-month period (548 days) using reduced alerts tuning 868 MW
Performance measure Change in wind speed events
High wind speed cut-out events
All events
Number of events 15 6 21
Number of events correctly alerted
11 6 17
Percentage of events correctly alerted
73% 100% 81%
Percentage of time-steps alerted 7% 3% 10%
(These results are based only on AEMO wind power dispatch data – results would improve if turbine availability or wind speed information were available)
How the likelihoods work• Likelihood values apply to the
hour time window• E.g. The change in wind speed
alert here gives:– 20% likelihood of event occurring in
the period 1:30 to 2:30– 20% likelihood of event occurring in
the period 2:30 to 3:30– 4% likelihood for 3:30 to 4:30 and
4:30 to 5:30 – 41% of event occurring in the full
alerted period from 1:30 to 5:30
= 1 – likelihood no event occurs in this
period = 1 – 0.8*0.8*0.96*0.96
= 41%
Interactive examples for using Wind InsightExample 1: point forecast and alerts
• Imagine the time is midnight on 12th October 2010
Interactive examples for using Wind InsightExample 1: point forecast and alerts
• Imagine the time is midnight on 12th October 2010
• Wind power production is around 270 MW
• Point forecast shows rapid decrease at around 4:00
Interactive examples for using Wind InsightExample 1: point forecast and alerts
• Imagine the time is midnight on 12th October 2010
• Wind power production is around 270 MW
• Point forecast shows rapid decrease at around 4:00
• High wind speed alert raised with 10% likelihood at midnight to 1:00
• Change in wind speed alert raised for period 2:00 to 5:00 with 40% likelihood around 2-3:00
• Press “Page Up” and “Page Down” to play again.
• What kind of wind power changes do you think are most likely to occur in the alerted period? (See next slide for what happened)
Example 1: animations
Example 1: with observations
• A rapid decrease in wind power occurred at 2:00 on 12 October 2010
• This was 1-2 hours earlier and more rapid than the point forecast suggested
• However the alerts and wind power fields did suggest this possibility
• During the 10% likelihood high wind speed cut-out alert, an actual event did not occur this time
Interactive examples for using Wind InsightExample 2: point forecast and alerts
• Imagine the time is 9:45am on 19th December 2010
• Wind power production has been around 200-250 MW for the past 4 hours
• Point forecast is rated power for next 12 hours followed by rapid decrease
• High wind speed and change in wind speed alerts raised over next 12 hours
Example 2: animations
• Press “Page Up” and “Page Down” to play again.
• What kind of wind power changes do you think are most likely to occur in the alerted period? (See next slide for what happened)
Example 2: with observations
• Two high wind speed cut-out events occurred around 11:00 and 13:30.
• A large decrease in wind power occurred at 17:00.
• None of these events were suggested in the point forecast
Wind power fields for South Australia
• Wind power fields in any format can be displayed for multiple wind farms sites in large regions such as South Australia – click to start animation
Four regions covering SA wind power in 2D format
Wind speed forecast contours outside wind farm regions (ms-1 at hub height)
Cathedral Rocks and Mt Millar: 136 MW
Mid-north region: 321 MW
Wattle Point and Starfish Hill: 125 MW SESA: 325 MW
Subscription Options
• Wind Insight forecasts can be provided up to 48 hours ahead for – single wind farm sites– aggregations of groups of wind farms, such as total wind generation in South
Australia.
• The alerts can be tuned to the user’s needs– User can specify the size and duration for a large rapid change as well as the
threshold for alerts
• Wind power fields can be provided in different formats• The point forecast can be provided for
– any time resolution desired (eg. 30 minutes or 5 minutes)– tailored for its use (i.e. minimised root mean square error, or optimised for wind
fluctuations)
• Other specific features of Wind Insight are also possible• Any further questions or enquiries, please contact Nick Cutler on
+61 7 3112 6018, or [email protected]