Adaptive nonlinear system identification using Comprehensive Learning PSO

In this paper we introduce the Comprehensive Learning Particle Swarm Optimization (CLEPSO) technique for identification of nonlinear systems. System identification in noisy environment has been a matter of concern for researchers in control theory for nonlinear analysis and optimization. In the recent past the Least Mean Square Algorithm (LMS), Genetic Algorithm (GA), Particle Swarm Optimization (PSO) etc. have been employed for developing mathematical archetype of an anonymous system. LMS performs inversely with nonlinearity. Although PSO performs better than GA in terms of convergence rate, it suffers from premature convergence. To alleviate the problem we propose a novel CLEPSO technique for updating the parameters of the Functional Link Artificial Neural Network (FLANN) model. The CLEPSO is a variant of PSO which ascertains the convergence of the model parameters to the global optimum with a faster speed and better accuracy. Comprehensive computer simulations corroborate that CLEPSO is a better parameter updating algorithm than PSO even in noisy conditions, both in terms of accuracy and convergence speed.