Wind power prediction plays a very important role in the dispatching and operation to maintain power system safety and stability. Due to the limitation of single prediction model with large errors at some individual points

a novel twostep dynamic optimal combination model of wind power prediction based on variable adaptive vector forgetting factor is proposed

which takes the numerical weather prediction (NWP) data as inputs. Three single prediction models are adopted considering advantages of physical and statistical models. Recursive least square (RLS)

covariance optimization combination and multilayer perceptron (MLP) network are proposed to calculate the weights of each single prediction model. Effect of new observations on parameter estimation is further evaluated based on Cook’s distance

and the weight coefficients of three combined prediction models are calculated through adaptive exponential dynamic combination model by use of variable vector forgetting factor. The proposed model is used to combine the results from the first step to obtain the optimal model in the adaptive process. Case studies show that the optimal forecasts can be obtained in ultrashort term wind power prediction using the proposed model

which improves the accuracy of prediction effectively. This work is supported by National Natural Science Foundation of China (No. 51477174

No. 51077126).