Abstract: Monthly scenario generation is the basis for monthly time series production simulation, power planning and thinning scenario analysis. Aiming at the problems of difficulty in fitting multivariate high-dimensional variables and intensified source-load uncertainty that monthly source and load scenarios modelling usually faces, we put forward a new method for monthly source and load scenario generation based on progressive growing of generative adversarial nets. The monthly time series are split into mean values to obtain the power series on multi-time scale, and the source and load series are spliced vertically to design the corresponding two-dimensional convolution structure in combination with multi-time scale properties. A smooth progressive growth method and some special training strategies are used to gradually generate the source and load scenarios with multi-granularity. The generative network deconstructs the low-dimensional and high-dimensional features, firstly learns the daily scale characteristics under the month and week, and then gradually fits the high-dimensional nonlinear features to generate the 720 monthly scenarios at the hourly level. Finally, an example analysis is carried out based on the actual Wind-Photovoltaic-Load dataset. The simulation results verify the effectiveness of the proposed algorithm in monthly source and load scenario generation, which can be used as a reference for monthly dispatching and planning.