Wind speed spatio-temporal forecasting of wind farms based on universal kriging and Bayesian dynamic model

The safe and stable operation of power grid has long been challenged due to the instability of connected wind power. Current research work rarely considers the temporal and spatial correlation among different positions in wind farms. In this paper, a wind speed statistical model based on multiple spatial and temporal scales is presented. First, spatio-temporal wind speed data is analyzed using empirical variogram, to provide basis for the analysis of the spatial correlation of different locations in wind farm. By using geographic information and the spatial covariance matrix, the spatial structure matrix is constructed, showing the spatial correlation between wind turbines. Then universal kriging and a Bayesian dynamic model are used for modeling. Gibbs sampling method is utilized to analysis the model and a prediction is developed based on it. Forecasting results using actual data of a wind farm confirm the effectiveness of the methods in the paper.