There is proposed a new method of structural identification of interval nonlinear models in the paper, which is based on the selection of structural elements for the synthesis of the model structure using the analysis of the gradient of the objective function. The necessary and sufficient conditions for the completeness or optimality of a set of structural elements are defined and substantiated based on the analysis of the gradient of the objective function, and the basic rules for the selection of structural elements are formulated. A structural identification algorithm was implemented and a study of its convergence was carried out.