Abstract: In the background of the new power system, constructing typical source-load temporal scenarios can provide scenario support for the day-ahead or intra-day electricity production organizations. To further improve the quality of temporal scenarios generated by statistical methods, and reduce the number of typical source-load temporal scenarios, a method for generating typical source-load temporal scenarios based on the modification of sampling probability interval is proposed. Firstly, considering the historical fluctuation characteristics of renewable energy, the probability interval of the prediction error temporal sampling is modified, and the temporal scenarios are sampled and generated using Latin hypercube sampling at the first sampling point and using random sampling in modified probability interval at subsequent sampling points. Secondly, typical temporal scenarios are obtained based on clustering method. Then, the typical source-load temporal scenario combinations are reduced from the perspective of system security based on the source-load “AND” operation. Finally, based on the actual data from a province in Northwest China, the source-load temporal scenarios for an IEEE 39-bus system are constructed to verify the effectiveness and feasibility of the proposed method.