The use of real-time off-site observations as a methodology for increasing forecast skill in prediction of large wind power ramps one or more hours ahead of their impact on a wind plant.

[摘要] ABSTRACT Application of Real-Time Offsite Measurements in Improved Short-Term Wind Ramp Prediction Skill Improved forecasting performance immediately preceding wind ramp events is of preeminent concern to most wind energy companies, system operators, and balancing authorities. The value of near real-time hub height-level wind data and more general meteorological measurements to short-term wind power forecasting is well understood. For some sites, access to onsite measured wind data - even historical - can reduce forecast error in the short-range to medium-range horizons by as much as 50%. Unfortunately, valuable free-stream wind measurements at tall tower are not typically available at most wind plants, thereby forcing wind forecasters to rely upon wind measurements below hub height and/or turbine nacelle anemometry. Free-stream measurements can be appropriately scaled to hub-height levels, using existing empirically-derived relationships that account for surface roughness and turbulence. But there is large uncertainty in these relationships for a given time of day and state of the boundary layer. Alternatively, forecasts can rely entirely on turbine anemometry measurements, though such measurements are themselves subject to wake effects that are not stationary. The void in free-stream hub-height level measurements of wind can be filled by remote sensing (e.g., sodar, lidar, and radar). However, the expense of such equipment may not be sustainable. There is a growing market for traditional anemometry on tall tower networks, maintained by third parties to the forecasting process (i.e., independent of forecasters and the forecast users). This study examines the value of offsite tall-tower data from the WINDataNOW Technology network for short-horizon wind power predictions at a wind farm in northern Montana. The presentation shall describe successful physical and statistical techniques for its application and the practicality of its application in an operational setting. It shall be demonstrated that when used properly, the real-time offsite measurements materially improve wind ramp capture and prediction statistics, when compared to traditional wind forecasting techniques and to a simple persistence model.

[发布日期] 2012-12-31 [发布机构] WINData, LLC

[关键词] 17 Wind Energy Real-Time, Off-Site, Observations, Methodology, Increasing, Forecast Skill, Prediction, Wind Ramps, Hours Ahead, Wind Plant, Operations, Scheduling, Marty Wilde, Gregg Leblanc, Osisoft, Naturener, Craig Collier, Patrick Shaw, Garrad Hassan America, Eric Grimit, 3Tier, Usdoe, Project, Transpara Kpi, Windata, Wind Resource, Met Towers, Pressure Gradients, Windata 60 Meter Met Tower, Realtime Logger, East Glacier Mt, Sensor, Pi Data Base, Pi Processbook, Upstream Pressure, Pressure Wave Propagation, Ramp Event, Bias, Mean Absolute Error, Mae, Forecast Horizon, Nudged Forecast, Clustering, Training, Algorithm, Forecast, Probability Of Detection, Pod, False Alarm Ratio, Far, Critical Success Index, Csi, Metric, Persistence, K-Means Clustering, Glacier Wind, Montana, United States Department Of Energy, American Recovery And Reinvestment Act Of 2009, High Fidelity, Meteorological, Sensor Data, Short Term Wind Forecasting, Situational Awareness, System Operators, Power Forecasters, Sensor Net;Real-Time, Off-Site, Observations, Methodology, Increasing, Forecast Skill, Prediction, Wind Ramps, Hours Ahead, Wind Plant, Operations, Scheduling, Marty Wilde, Gregg Leblanc, Osisoft, Naturener, Craig Collier, Patrick Shaw, Garrad Hassan America, Eric Grimit, 3Tier, Usdoe, Project, Transpara Kpi, Windata, Wind Resource, Met Towers, Pressure Gradients, Windata 60 Meter Met Tower, Realtime Logger, East Glacier Mt, Sensor, Pi Data Base, Pi Processbook, Upstream Pressure, Pressure Wave Propagation, Ramp Event, Bias, Mean Absolute Error, Mae, Forecast Horizon, Nudged Forecast, Clustering, Training, Algorithm, Forecast, Probability Of Detection, Pod, False Alarm Ratio, Far, Critical Success Index, Csi, Metric, Persistence, K-Means Clustering, Glacier Wind, Montana, United States Department Of Energy, American Recovery And Reinvestment Act Of 2009, High Fidelity, Meteorological, Sensor Data, Short Term Wind Forecasting, Situational Awareness, System Operators, Power Forecasters, Sensor Net