The invention provides a photovoltaic power generation output joint probability distributed modeling method. The method comprises the following steps: initializing the output sample data of each photovoltaic power station; obeying Beta distribution according to the sunlight irradiance in a space where the photovoltaic power stations are positioned within a certain time interval to obtain an output probability model and a Beta distribution parameter of each photovoltaic power station; calculating a probability distribution sample space vector corresponding to each sampling point of each photovoltaic power station according to the output probability model and the Beta distribution parameter of each photovoltaic power station; selecting a multivariate standard normal distribution function for serving as a connection function of the joint probability distribution feature of each photovoltaic power station; performing parameter identification on the relevant coefficient of the multi-component standard normal distribution function based on the probability distribution sample space vectors; and substituting the output probability models of the photovoltaic power stations into the multi-component standard normal distribution function to obtain joint probability distribution models of the photovoltaic power station. Through adoption of the method, the difference among the output of each photovoltaic power station on the aspect of probability distribution can be reflected, and mutual relevance can be considered.