Abstract: Aiming at the local convergence problem of particle swa rm optimization algorithm, we proposed a particle swarm optimization algorithm b ased on the inertia weight adjustment and group best position variation. In this algorithm, the state information of individual particles was introduced into th e inertia weight strategy. The inertia weight of each particle was adjusted inde pendently, which reflected the difference of individual particles to the weight demand. In the mutation strategy of the best position, the classification idea w as used. According to the searching state of the particle swarm, the correspondi ng extreme mutation mode was selected, which made the mutation operation more ta rgeted. The experimental results indicate that the new algorithm shows good opti mization performance for several test functions, which can effectively avoid local convergence problem and improve the global sea rch ability of the particle swarm.

Key words: particle swarm, optimization algorithm, classification mutation, independent inertia weight