Abstract: The scenario generation method based on deep learning can adaptively capture the high-dimensional nonlinear features in historical data and has been widely used in the uncertainty modeling of wind power and photovoltaic output. However, most scenario-generation methods based on deep learning are black-box models, which have problems such as poor interpretability and uncontrollable generation. Therefore, a scenario set constructing method of wind power and photovoltaic output based on improved information maximization generation adversarial network (InfoGAN) is proposed. In this method, the mutual information regularization term is added to the objective function to maximize the mutual information between the control coding and the generation scenario, unsupervised learning controls the mapping relationship between the coding and the generation scenario statistical features, and Gumbel-Softmax distribution is introduced to improve the quality of the generation scenario. The results show that the proposed method can not only accurately describe the uncertainty of wind power and photovoltaic output but also be interpretable and generate the specified wind power and photovoltaic output scenario.