Wind Speed Forecasting for Power System Operation

In order to support large-scale integration of wind power into current electric energy system, accurate wind speed forecasting is essential, because the high variation and limited predictability of wind pose profound challenges to the power system operation in terms of the efficiency of the system. The goal of this dissertation is to develop advanced statistical wind speed predictive models to reduce the uncertainties in wind, especially the short-term future wind speed. Moreover, a criterion is proposed to evaluate the performance of models. Cost reduction in power system operation, as proposed, is more realistic than prevalent criteria, such as, root mean square error (RMSE) and absolute mean error (MAE).

Two advanced space-time statistical models are introduced for short-term wind speed forecasting. One is a modified regime-switching, space-time wind speed fore- casting model, which allows the forecast regimes to vary according to the dominant wind direction and seasons. Thus, it avoids a subjective choice of regimes. The other one is a novel model that incorporates a new variable, geostrophic wind, which has strong influence on the surface wind, into one of the advanced space-time statistical forecasting models. This model is motivated by the lack of improvement in forecast accuracy when using air pressure and temperature directly. Using geostrophic wind in the model is not only critical, it also has a meaningful geophysical interpretation.

The importance of model evaluation is emphasized in the dissertation as well. Rather than using RMSE or MAE, the performance of both wind forecasting models mentioned above are assessed by economic benefits with real wind farm data from Pacific Northwest of the U.S and West Texas. Wind forecasts are incorporated into power system economic dispatch models, and the power system operation cost is used as a loss measure for the performance of the forecasting models. From another perspective, the new criterion leads to cost-effective scheduling of system-wide wind generation with potential economic benefits arising from the system-wide generation of cost savings and ancillary services cost savings.

As an illustration, the integrated forecasts and economic dispatch framework are applied to the Electric Reliability Council of Texas (ERCOT) equivalent 24- bus system. Compared with persistence and autoregressive models, the first model suggests that cost savings from integration of wind power could be on the scale of tens of millions of dollars. For the second model, numerical simulations suggest that the overall generation cost can be reduced by up to 6.6% using look-ahead dispatch coupled with spatio-temporal wind forecast as compared with dispatch with persistent wind forecast model.