A wind power forecasting method and its confidence interval estimation

This paper describes a wind power forecasting method and its confidence interval estimation. Recently, flat control of wind power generators using various batteries has been required. In flat control, accurate wind power forecasts and their error confidence intervals are needed. In this paper, wind speed forecasts are calculated by regression models using Grid Point Value (GPV) weather forecasts. The forecasts are adjusted by fuzzy inference using the latest errors. The wind power forecasts are translated from the wind speed forecasts using two power curves. The power curves are selected or combined by fuzzy inference depending on the wind direction. The error confidence interval models are generated for each forecasting target time. Each confidence interval is combined by the other fuzzy inference. The proposed methods were applied to actual wind power generators, and it was found that the forecasting errors were smaller than in the conventional methods. Almost all of the forecasts can be within the error confidence intervals estimated by the proposed methods. The results show the effectiveness of the proposed methods.