In order to improve the accuracy of wind power prediction and reduce the adverse impact of the fluctuation of output wind energy on wind power integration, a short-term wind power prediction method based on WT-IAGA-BP neural network is proposed in this paper. Firstly, data cleaning technologies, including wind speed partition, 3σ criterion and Lagrange interpolation method, are applied to remove abnormal values from the historical data of wind farm. Secondly, according to the wavelet reconstruction error, db4 wavelet transform (WT) is used to extract the different frequency characteristic signals of wind speed, wind direction and historical wind power respectively. Then, the improved adaptive genetic algorithm (IAGA) is introduced to obtain the optimized values for initial weights and thresholds of the BP neural network of each sequence, and Sigmoid function is used to adaptively change the crossover probability and mutation probability through the fitness value. Finally, the WT-IAGA-BP model of each sequence is constructed to predict the short-term wind power combination. According to the simulation analysis and comparison with other prediction models including ELM, IAGA-B, WT-ELM and WT-LSSVM, the obtained results suggest that the presented prediction model has better prediction performance and higher prediction accuracy.