Power heterogeneous communication networks plays an important role in the future smart grid. Due to its complex structure, power heterogeneous communication networks can be easily paralyzed in large-area network. To solve this problem, according to a complex network of self-organized criticality theory, the power of heterogeneous networks using the theory of triangulation is divided into some sub-regions, and according to the each node failures associated probability as metrics, an improved genetic algorithm to search for key network node failures is proposed, and the critical failure network nodes are maintained or cut off, so that the power of heterogeneous communication network is in self-organized and keeps away from critical state, thus avoiding a large area network paralysis event. Simulation results show that the algorithm is of effectiveness and robustness in variety of networks.