The invention relates to an electric power system risk scheduling method and system. The method includes the following steps: acquiring the architecture data of an electric power system and the load profile data of a new task; based on the architecture data and the load profile data of the new task, using the bacterial foraging algorithm to iterate a preset initial knowledge matrix to obtain a corresponding risk scheduling object function value and a updated knowledge matrix; based on the updated knowledge matrix corresponding to the minimum risk scheduling object function value, online optimizing the new task to obtain the result of the risk scheduling optimization and outputting the result; the optimal knowledge matrix of a source task serving as the initial matrix of the new task for conducting knowledge transfer, and using the bacterial foraging algorithm of the knowledge transfer to conduct online optimization on the new task. Transfer learning greatly increases the speed of online learning, achieves online dynamic optimization of risk scheduling, can guarantee rapid resolution when a problem develops, and can adapt to rapid optimization of large-scale and complex risk scheduling.