Abstract. The neuroevolutionary approach is proposed to construct the mathematical models of nonlinear nonstationary objects in the presence of non-Gaussian noise. The general structure of an evolutionary feed-forward neural network is considered. The modeling of the proposed approach for various cases of nonstationarity is carried out, which proved the efficiency of the approach. Figs: 9. Tabl.: 2. Refs: 51 titles.

Keywords: identification, robustness, nonlinear nonstationary object, artificial neural network, evolutionary calculation, genetic algorithm.