Short-term wind power forecasting based on an improved persistence approach

Wind power forecasting is of importance for power grids. It can mitigate the disadvantageous impacts of wind farms on power systems and enhance the competitiveness of wind power in electricity markets. This paper proposed an improved persistence approach based on wavelet. First, the original data of wind speeds were decomposed into high-frequency component and low-frequency component by using wavelet. Moving average method was used for predicting the high-frequency subseries, in terms of low similarity and easy fluctuation properties in high-frequency component, and the low-frequency sub-series were still predicted by the persistence method. Then, all sub-series were recomposed to form a time series of wind speeds. Wind power of a wind turbine can be further forecasted through a power curve which could transfer wind speeds data into wind power. Compared with the original method, the average relative error reduced to 11.81% from 17.10%, and the average absolute error decreased to 23.48 kW from 39.58 kW. Case study showed that the wind power forecasting accuracy was effectively improved by use of an improved persistence approach which could be put into operation in practice.