Verification of hourly forecasts of wind turbine power output
A verification of hourly average wind speed forecasts in terms of hourly average power output of a MOD-2 was performed for four sites. Site-specific probabilistic transformation models were developed to transform the forecast and observed hourly average speeds to the percent probability of exceedance of an hourly average power output. (This transformation model also appears to have value in predicting annual energy production for use in wind energy feasibility studies.) The transformed forecasts were verified in a deterministic sense (i.e., as continuous values) and in a probabilistic sense (based upon the probability of power output falling in a specified category). Since the smoothing effects of time averaging are very pronounced, the 90% probability of exceedance was built into the transformation models. Semiobjective and objective (model output statistics) forecasts were made compared for the four sites. The verification results indicate that the correct category can be forecast an average of 75% of the time over a 24-hour period. Accuracy generally decreases with projection time out to approx. 18 hours and then may increase due to the fairly regular diurnal wind patterns that occur at many sites. The ability to forecast the correct power output category increases with increasing power output because occurrences of high hourly average power output (near rated) are relatively rare and are generally not forecast. The semiobjective forecasts proved superior to model output statistics in forecasting high values of power output and in the shorter time frames (1 to 6 hours). However, model output statistics were slightly more accurate at other power output levels and times. Noticeable differences were observed between deterministic and probabilistic (categorical) forecast verification results.
- Research Organization:
- Pacific Northwest National Lab. (PNNL), Richland, WA (United States)
- DOE Contract Number:
- AC06-76RL01830
- OSTI ID:
- 6545007
- Report Number(s):
- PNL-4900; ON: DE84017176
- Resource Relation:
- Other Information: Portions are illegible in microfiche products. Original copy available until stock is exhausted
- Country of Publication:
- United States
- Language:
- English
Similar Records
Time Series Models for Simulating Hourly Wind Power.
A Public-Private-Acadmic Partnership to Advance Solar Power Forecasting