Abstract: The randomness and uncertainty of wind power fluctuation is one of the main reasons limiting the consumption of wind energy, and the fluctuation of wind power affects the accurate prediction of wind power. How to quantitatively characterize the fluctuation of wind power has become the key to solve these problems. Therefore, a mixed t-t Location-scale distribution is proposed by combining the advantages of t-distribution and t-location-scale distribution from the characteristic of "sharp peaks and thick tails" of wind power fluctuation. In addition, in order to avoid the problem of insufficient parameter estimation accuracy in previous methods, an intelligent optimization algorithm, the sine cosine optimization algorithm, is introduced for the parameter estimation of the distribution model. The proposed model is compared with five models, including normal distribution and mixed t Location-scale distribution, and the applicability of the model is verified from three sliding average time period lengths of 15 min, 30 min and 45 min. Finally, the simulation test is conducted by the historical data of three wind farms in Northwest China, comparing the three evaluation indexes, and it is found that the proposed model has a very good fitting performance for the fluctuation of wind power.