Hybrid optimization algorithm based parallel processing method

The invention relates to a hybrid optimization algorithm based parallel processing method. The method comprises that a universal principal and subordinate parallel computing frame is designed and achieved, principal processes are in charge of new point generation and sending, adaptive value receiving, sorting and convergence analysis, and subordinate processes only perform adaptive value sending and new point receiving and forward modeling processes. In the new point generation process of the principal processes, a particle swarm optimization and genetic algorithm characteristic combined hybrid algorithm is designed, and rapid convergence of the particle swarm optimization and the global convergence capacity of the genetic algorithm are provided. On the basis, a multi-communication-subspace concept is introduced, that is, the whole communication space is portioned into a plurality of sub communication spaces, and the process forward modeling process is completed in parallel by a plurality of processes in resp