Download Wind Insight Product Brief
To plan for power system disturbances from fluctuating wind generation, you need the best forewarning you can get. Wind Insight™ is a wind power forecasting tool that specialises in predicting large rapid changes in wind power from one hour to several days ahead. It also provides point forecasts (expected generation), using sophisticated models to make use of multiple weather forecasts and real-time observations. Wind Insight™ is currently providing point forecasts for over ten wind farms in India.

Wind Insight™ provides:

  • Alerts with likelihoods for potential large rapid change events. For example:

    Figure A: Wind Insight™ point forecast and alerts for potential changes of > 65 MW in 30 mins from Lake Bonney 1 and Canunda. How to interpret this chart

  • Wind power field forecast animations for aggregations of wind farms in close proximity. This allows the forecast user to visualise the wind features predicted in the region of single or multiple wind farms and assess likely scenarios in wind power production from displacing these features. This allows the user a much greater appreciation for predicting large rapid changes in wind power than if they only had the point forecast. For example, the wind power field animation corresponding to Figure A is:

    Map: Region covered
    by wind power fields.

    Figure B: Wind Insight™ wind power field animation for Lake Bonney 1 and Canunda. How to interpret this chart

  • Wind power field animations for large regions with multiple wind farms. This allows predicted large wind features to be observed as they pass different wind farms, allowing their timing to be more accurately predicted at other wind farms downstream. For example, for South Australia:
    Figure C: Wind Insight™ animated field forecast for wind farm aggregates in South Australia. How to interpret this chart

    Table A gives a summary of the alerting performance* of the Wind Insight™ for large rapid changes in the total wind power generation in South Australia. The system was trained on 2011 data and the results are shown for 2012 data for forecasts between 7 and 30 hours ahead. Large rapid changes caused by high wind speed cut-outs or by horizontally propagating large weather features (such as cold fronts and low pressure systems) causing large changes in wind speed are all alerted and captured well by Wind Insight™.

    Table A: A summary of the Wind Insight™ demonstration alert performance for wind power changes > 200 MW in up to 30 minutes in total South Australian wind power production (five-minute data) over the 2012 calendar year.

    Performance measure Number of events forecast correctly (out of 62) Percentage forecast correctly Percentage of time alerted
    Wind Insight - using point forecast 35 56% 14%
    Wind Insight - using wind power fields 41 66% 13%

    The skill of Wind Insight™ is highlighted by the fact that alerts are raised only 13% of the time while 66% of the large rapid changes in wind power are alerted correctly. This, in combination with the range of likelihoods provided for each alert and the wind power field animations give the Wind Insight™ tool its value in decision-making. Using Wind Insight™'s unique wind power fields captures of 6 more events (out of 62) than from only using the point forecast. There is a trade off between the alert frequency and the number of events captured - Wind Insight™ can be tuned to meet the user's needs.

    For more information or questions on our Wind Insight™ wind power forecasting tool, please contact us.

    You can read more about how Wind Insight™ works, and the philosophy behind it in the following peer-reviewed and other publications.

    Table B: Publications on the principles behind the Wind Insight™ wind power forecasting tool.

    Cutler NJ (2013). Wind Insight - a wind power forecasting tool for power system security management. New Zealand Wind Energy Conference, Wellington, New Zealand. available
    Cutler NJ, Outhred HR and MacGill IF (2011). Final report on UNSW project for AEMO to develop a prototype wind power forecasting tool for potential large rapid changes in wind power. Sydney, Australia. pp. 107. available
    Cutler NJ, Outhred HR, MacGill IF and Kepert JD (2011). Predicting and presenting plausible future scenarios of wind power production from numerical weather prediction systems: a qualitative ex-ante evaluation for decision-making, Wind Energy 15(3), 473-488.
    Cutler NJ, Outhred HR and MacGill IF (2011). Using nacelle-based wind speed observations to improve wind power forecasts, Wind Energy 15(2), 245-258.
    Cutler NJ, Outhred HR and MacGill IF (2010). Utilising Multiple Grid Points in Numerical Weather Prediction System Forecasts to Characterise Potential Large Rapid Changes in Wind Power Generation. Modern Energy Review 2(2), 44-47. available
    Cutler NJ, Kepert JD, Outhred HR, MacGill IF and Kay MJ (2009). Characterizing future large, rapid changes in aggregated wind power using Numerical Weather Prediction spatial fields, Wind Energy 12(6), 542-555.
    Cutler NJ, Kepert JD, Outhred HR and MacGill IF (2008) Characterizing wind power forecast uncertainty with Numerical Weather Prediction spatial fields, Wind Engineering 32(6), 509-524.
    Cutler NJ, Kay MJ, Jacka K and Nielsen TS (2007). Detecting, categorizing and forecasting large ramps in wind farm power output using meteorological observations and WPPT, Wind Energy 10(5), 453-470.

    The Wind Insight™ forecasts demonstrated here use the ACCESS-A numerical weather prediction system forecasts from the Australian Bureau of Meteorology. The conversion of wind speed forecasts to available wind power was devised using publicly available historical wind power observation data from the Australian Energy Market Operator. The results shown do not currently make use of any real-time wind power observations or turbine availability data but if this were available the forecasts could be improved (to help determine what events are due to changes in wind speed, which are due to high wind speed cut-outs and which are caused by network constraints).

    * The Wind Insight™ demonstration alerting performance is based on detecting large rapid change events in five-minute historical wind power observations of the relevant wind farms' wind power production. This data was obtained from publicly available market data on the Australian Energy Market Operator website. These results are for one specific case study, and were presented at the New Zealand Wind Energy Conference (NZWEC) in March 2013 (see presentation from NZWEC 2013).