Method of Parametric Identification for Interval Discrete Dynamic Models and the Computational Scheme of Its Implementation

Abstract

A method of parametric identification of interval discrete dynamic models is considered. In the case of an interval data set, finding estimations for parameters of such models requires solving an interval system of nonlinear algebraic equations for some known vector of basic functions. The solution of these equations forms a non-convex area in the parameter space which can consist of several unconnected subareas. For solving this parametric identification problem, methods of random search are widely used including that based on the procedure of the Rastrigin’s director cone having high time complexity. Therefore, the detailed analysis of this parametric identification method was carried out in this work to reduce the time complexity. A new improved scheme of computational implementation of the method is proposed which takes into account areas of permissible values of the modeled characteristic. Results of the comparative efficiency analysis of implementation scheme of the proposed method and the known one are presented demonstrating that the time complexity of the improved scheme of the method is at least twice less compared to the known implementation scheme.