Parameter identification for time-varying systems by evolutionary neural network

Elman, which is one of the well-known recurrent neural networks, has been improved to easily apply in parameter identification of time-varying systems during the past decade. In this paper, a learning algorithm for Elman neural networks (ENN) based on improved particle swarm optimization (IPSO), which is a swarm intelligent algorithm, is presented. IPSO and Elman are hybridized to form IPSO-ENN evolutionary algorithm, which is employed to parameter estimation. Simulation experiments show that IPSO-ENN is a more effective swarm intelligent algorithm, which results in an identifier with the best trained model. Time-varying system of the IPSO-ENN is obtained.