Impact of Forecast Uncertainty on Wind Farm Profitability

摘要

The case in which a WF participating on the day-ahead energy market is obliged to submit in advance the production schedule and it is considered financially responsible for its observance has been investigated. The potential revenue losses due to the forecast errors that could affect the WF have been calculated based on a U.S. case study provided by NREL. The case study features both actual production and forecast at different lead times, even though the forecast data are not complete (some lead times are not pro-vided). The forecast errors have been statistically analyzed to reconstruct the error PDFs for several lead times and forecasted power levels. The PDFs for the missing lead times have beenestimated by interpolating those already provided. By using a Gaussian copula, from the error marginal distributions and from an error correlation matrix retrieved in the literature, forecast error scenarios have been generated. Each of these scenarios is a realis-tic forecast, which can be thought as the schedule submitted to the market, in absence of other decision tools. To reduce the complex-ity of the analysis, the year has been divided into representative units (clusters). Based on the annual revenue, nine clusters have been found to be suitable to describe the whole year. For each of these "typical days," the revenue losses due to the forecast errors have been calculated. In comparison with the perfect forecast case, low production days have been found to be very sensitive to forecast errors and to the price that the correction might have in the case the imbalance is traded into balancing markets. On an annual basis, the profit reduction ranged from 5 to 45%, according to the balancing and day-ahead market prices. Finally, the effec-tiveness of two techniques to reduce losses has been estimated. Curtailment is inexpensive but can be used only in those cases inwhich the negative prices make uneconomical the excess produc-tion trading. It has been found that the curtailment could reduce the losses up to 16%, bringing them to be roughly comprised between 20 and 35%.