Abstract <p>The problem of structural optimization in socioeconomic systems using the example of developing a portfolio of innovative projects is examined. A key approach to solving this problem is the use of constrained multivariate optimization methods. The specific nature of the problem (nonlinear objective function, large-dimensional search space, high computational complexity) makes it advisable to use metaheuristic methods, specifically the proposed modification of the “gray wolf” algorithm, which lacks “free parameters.” To improve its efficiency, it is proposed to run several such algorithms in parallel on preferred “sections” of the feasible solution space, partitioned using fuzzy clustering.</p>