A new optimization model is established in this paper to build power system emergency load shedding scheme adaptive to multi operation modes. Particle swarm optimization (PSO) is used to solve the model heuristically. To improve performance of PSO, the concept of belief space in cultural algorithm is introduced to form the cultural particle swarm optimization (CPSO). A new solution is generated in the belief space to replace the inferior solution in CPSO, and motion parameters of CPSO are adjusted by crowding distance to make the algorithm escape from local optimum. The proposed model and algorithm are validated with a provincial power grid model.