In order to improve the prediction accuracy of wind farm output power, wavelet analysis(WD) and cuckoo optimization support vector machine(CS-SVM) algorithm are used to predict wind power in ultra-short term, which is more direct and accurate than indirect wind power obtained by predicting wind speed. Firstly, the wind power model is decomposed into approximate sequence and detail sequence by using wavelet decomposition and reconstruction. Then, the support vector machine optimized by cuckoo algorithm is used to predict each sequence, and the prediction results of each sequence are obtained. Finally, the prediction results of each sequence are superimposed to form the final prediction value of wind power. The results of numerical examples show that the prediction results have high accuracy, and the method used in this paper is more accurate, superior and practical than the support vector machine and other methods optimized support vector machine prediction results.