A Wind Energy Ramp Tool and Metric for Measuring the Skill of Numerical Weather Prediction Models

Abstract

A wind energy Ramp Tool and Metric (RT&M) has been developed out of recognition that during significant ramp events (large changes in wind power over short periods of time ) it is more difficult to balance the electric load with power production than during quiescent periods between ramp events. A ramp-specific metric is needed because standard metrics do not give special consideration to ramp events and hence may not provide an appropriate measure of model skill or skill improvement. This RT&M has three components. The first identifies ramp events in the power time series. The second matches in time forecast and observed ramps. The third determines a skill score of the forecast model. This is calculated from a utility operator’s perspective, incorporates phase and duration errors in time as well as power amplitude errors, and recognizes that up and down ramps have different impacts on grid operation. The RT&M integrates skill over a matrix of ramp events of varying amplitudes and durations.

Measuring the impact of additional instrumentation on the skill of numerical weather prediction models at forecasting wind ramp events during the first Wind Forecast Improvement Project (WFIP)

“…For each of these 20 ramp definitions, the RT&M follows three basic steps: First, it identifies ramp events in the time series of observed and modeled power data. Three different identification methods are used, the “Fixed Time Interval Method,” the “Min‐Max Method,” and the “Explicit Derivative Method.” While the “Fixed Time Interval Method” only measures the difference in power over a determined time window, the “Min‐Max Method” takes into account the maximum amplitude change in power within that window, and the “Explicit Derivative Method” analyzes the value of a smoothed time derivative of the power over that time window.…”

“…The first method allows for forecast skill to be derived as a function of forecast hour, and it assumes that no artificial ramps are generated by concatenating different model runs. The second method avoids the possibility of generating artificial ramps through combining different model runs but has the disadvantage that for each model forecast the beginning and ending forecast hours will suffer from truncated ramps that potentially begin before the start or end of the forecast cycle . This disadvantage is reduced for the “stitching method” as the concatenated time series will be much longer and the impact of having truncated ramps at the beginning and end will be negligible.…”

“…Three different identification methods are used, the “Fixed Time Interval Method,” the “Min‐Max Method,” and the “Explicit Derivative Method.” While the “Fixed Time Interval Method” only measures the difference in power over a determined time window, the “Min‐Max Method” takes into account the maximum amplitude change in power within that window, and the “Explicit Derivative Method” analyzes the value of a smoothed time derivative of the power over that time window. A detailed description of these methods is presented in Bianco et al Second, the RT&M matches in time the observed ramp events with those predicted by the forecast model. Finally, it scores the ability of the model to forecast ramp events. The scoring metric accounts for both amplitude and timing (center point and duration) errors in the forecast, allowing one to evaluate up‐ramp and down‐ramp events separately.…”

“…A description of the RT&M can be found in Bianco et al, in which the tool was tested on data collected over a period of 9 days from a set of four 80‐meter‐tall towers and forecasts at the same locations from the 13‐km grid spacing National Oceanic and Atmospheric Administration/Earth System Research Laboratory (NOAA/ESRL) Rapid Refresh Numerical (RAP) model, during the first WFIP. WFIP took place in the United States Great Plains from September 2011 to September 2012 .…”

Wind Power Ramp Events Prediction with Hybrid Machine Learning Regression Techniques and Reanalysis Data

“…To begin the discussion, it is convenient to start with the recent work [54], which is an illustrative review of state-of-the-art physical models to predict WPREs. In particular, it proposes novel tools and metrics to evaluate and compare different NWP models.…”