The chapter proposes an adaptive algorithm for selecting and calibrating internal parameters of parallel population algorithms used to search for extremes of nonlinear multimodal functions with a system of constraints on the final solution. The results of computational experiments on the search for the best solution of some mathematical functions of two-dimensional space (Rastrigin, Rosenbrock and Levy functions) with various stochastic parameters of a parallel population algorithm are presented. A comparative analysis of the impact on the quality of the final solution of each of the stochastic parameters of the main population algorithm is carried out and presented. The proposed adaptive population algorithm is able to store the history of the selected parameters of the main algorithm, as well as use it as a template when calibrating parameters to find the best solution to a new function.