The invention discloses a wind electricity power combined predication method based on wind farm data pre-processing. The method includes the following steps: using the principle of wavelet denoising to process historical wind velocity data and historical power data of a wind farm to separately obtain a smooth wind velocity sequence and a power sequence; respectively conducting frequent item set analysis and association rule mining on the wind velocity sequence and the power sequence to obtain a wind velocity strong association rule and a power strong association rule, and using the strong association rules to find a historical forecast that id similar to the current forecast which is to be corrected; and using the similar historical forecast to conduct error correction on the current forecast which is to be corrected; using statistics method to train the corrected current forecast of the wind velocity and establish a prediction model; using the prediction model to predict power, and using the similar historical forecast to predict for a second time the predicted power to eventually obtain a final combined prediction result.