The invention relates to the field of wind power generation, and in particular relates to a multiple-wind-farm output correlation modeling method by using a Copula-garch function. The method comprisesthe steps that the active power distribution data of wind farms are collected; ark effect testing is carried out on the wind farms, and further a Garch (P, Q) model is established to fit the edge distribution function of the wind farms; the combined output distribution of multiple wind farms is split into the form of multiplication of the edge distribution function and Copula functions accordingto the SKlar theorem; four indexes corresponding to five types of Copula functions are acquired; the weight of each index in each function is determined; an attribute measurement value is calculated based on each weight; the rating standard corresponding to each kind of copula function is determined according to an attribute measurement value; and the copula function corresponding to the optimal level is selected as the best copula function of the wind farms. The research on the uncertainty of wind power output is transformed into deterministic research, and the problem that choosing the optimal correlation function is not consistent with the actual situation is solved.