A novel GMFO-based identification method for MIMO Hammerstein model with heavy-tailed noise

This paper deals with the identification of multi-input multi-output (MIMO) Hammerstein system with combined nonlinearities under the heavy-tailed noise. Considering the outliers in the noises may lead to the unsatisfactory identification results using analytical method, this paper proposes a novel identification scheme combining the advantages of Radial Basis Function Neural Network (RBFNN) and a recently proposed nature-inspired algorithm called Moth-Flame Optimization (MFO). We use RBFNN to construct the static nonlinear block. The identification problem could be converted to an optimization problem, and the parameters of the linear part and nonlinear part are updated simultaneously. To improve its performance for identification, a novel version of MFO based on Gaussian-mixture distribution, which is named Gaussian-mixture Moth-Flame Optimization (GMFO), is proposed. The main innovation is the discrete population initialization and the individual position adjustment using Gaussian-mixture distribution, which is conducive to jumping out of local optima caused by outliers. The simulation results illustrate the proposed method is effective and outperforms other common evolutionary algorithms.