Infinitive Impulse Response system identification using an improved particle swarm optimization algorithm

This paper proposes an improved particle swarm optimization (IPSO) algorithm for IIR system identification problem. IPSO adopts three novel steps as follows: The population initialization step is based on golden ratio, which is beneficial for improving the quality of candidate solutions. In velocity updating step, all particles use different inertia weights, which is helpful to preserve the balance between global search and local search. Moreover, the proposed algorithm incorporates normal distribution to disturb the global best particle, which enhances its capacity of escaping from the local optimums. These properties of IPSO make it better suited for IIR system identification problems. IPSO is applied to twelve instances, and experimental results demonstrate that IPSO is capable of obtaining the best objective function values in all cases. Compared with the other four PSO approaches, it has better convergence and reliability which clearly points out its better performance in search accuracy.