Kozlov O. Methods and models of optimization-oriented synthesis for the improvement of intelligent control systems of nonlinear dynamic objects.

The dissertation is aimed at solving the urgent scientific and technical problem of creating and improving methods and models of optimization-oriented synthesis of fuzzy automatic control and decision support systems, used to automate complex dynamic plants and processes in various sectors of the economy. The simulation model of a combined fuzzy automatic control system is proposed, which consists of fuzzy logic output modules for: formalization of the parameters of the main feedback based on the sensor of the controlled variable; formalization of internal feedback parameters based on the built-in model of the control object; combination of the PID law and the control model of sliding modes. The multi-agent method of parametric optimization of fuzzy control and decision-making systems was further developed, in which, unlike the existing ones, an improved algorithm of "gray wolves" and an algorithm of the extended Kalman filter were introduced for implementation of parallel procedures of global search with learning and local search. The intelligent method of optimization-oriented synthesis of the rule bases of fuzzy systems with an optimal vector of conclusions and a minimum number of rules has been developed, which is based on the sequential search for optimal rule conclusions and the operations of identifying and excluding rules from the base that do not significantly affect the process of functioning of the fuzzy system, and on the ant colony algorithm with the construction of a rule base graph based on the "rule-node" formalism and a multi-criteria objective function. The criterion for quantitative assessment of the dimension of the rule bases of fuzzy systems is proposed, which is calculated as the product of the total number of rules in the rule base and the total number of possible conclusions of the rules for all output variables, and the indicator of the effectiveness of the synthesis and optimization methods of the rule bases, which is calculated as the ratio of the dimension value of the rule base to the total number of calculations of the objective function necessary to find its best solution in the process of synthesis and optimization. The method of finding optimal membership functions of linguistic terms of fuzzy systems by applying a biogeographical evolutionary algorithm and a multi-criteria objective function has been improved. The method of structural optimization of fuzzy systems by the number of linguistic terms of input and output variables has been improved by conducting a sequential search for the optimal number of linguistic terms based on a multi-criteria objective function with parallel execution of the automatic synthesis of the rule base for each solution obtained. The method of finding the optimal hierarchical structure of fuzzy decision support systems is proposed, which is based on the generation of several variants of the hierarchical structural organization followed by finding their optimal parameters and choosing the best solution based on a multi-criteria objective function. The bio-inspired method of complex structural-parametric synthesis of fuzzy control systems is proposed, which is based on procedures of determining the optimal number of linguistic terms of input and output variables, synthesis of the rule base, determination of optimal types and parameters of membership functions, as well as optimal types of operations of fuzzy inference and defuzzification according to the most rational sequence.