Abstract: Large-scale new energy grid integration has intensified the supply-demand balancing difficulty of regional power grids. Existing dispatching modes fail to coordinate inter-provincial regulating power sources, and the uneven allocation of provincial regulating resources further amplifies the risks of power supply guarantee and new energy consumption, thus necessitating the establishment of a risk-driven collaborative optimization framework. A multi-dimensional operation risk evaluation index system covering power supply guarantee, new energy consumption and regulation difficulty is constructed, and a two-stage generation allocation optimization method for regional power grids considering operation risks is proposed. Stage 1 optimizes the generation schedule of grid-dispatched power sources to minimize the overall grid risks of power supply guarantee and new energy consumption. Stage 2 achieves differentiated inter-provincial power allocation based on the regulation difficulty index. The second-order cone relaxation technique is employed to address nonlinear constraints for efficient model solving. Simulations with actual data from the East China Power Grid show that, under extreme scenarios, Stage 1 reduces the power supply guarantee risk by 286 GW·h (17% drop) and the new energy consumption risk by 573 GW·h (35% drop). Stage 2 cuts the average upward adjustment risk of provincial-dispatched power sources by 65.7% and the downward adjustment risk by 66.5% under extreme scenarios.