A short-term wind power combination forecasting model has been proposed based on deep belief network (DBN) and Kalman algorithm combined with the error correction algorithm (EC). Firstly, the empirical modal algorithm (EMD) is to be used to decompose the original wind speed sequence, thus extracting its main features and reducing the mutation of the wind speed sequence. Then, by adopting DBN method, two kinds of input and output matrices can be constructed so as to obtain the predicted powers of pro_1 and pro_2, thus obtaining the two kinds of prediction errors. Next, with pro_1 as the measurement value, and bias_1 as the measurement error, the introduction into Kalman model of pro_2 as the observation value and bias_2 as the process error helps to obtain the prediction result pro and prediction error bias, followed by a modification of pro and bias based on EC algorithm. The simulation results show that DBN-kalman-EC combined prediction model can effectively neutralize bias_1 and bias_2 errors, reduce prediction errors, correct predicted values, and effectively improve the accuracy of the short-term wind power prediction.