A method for the optimal synthesis of prestressed steel frame structures is developed. The search for a solution is being carried out by applying discrete sets of variable parameters, including cross-sections of structural rods and cross-sections of prestressing tie-rods. It is possible to vary the location of the prestressing system. To optimize the cost of the objects under consideration, an improved meth-od of evolutionary simulation has been used, including a combined scheme for constraints set and an unusual scheme for the formation of populations. Strength and stiffness considerations for a number of objects within a population are not strict, and for variants with breached constraints, a penalty function has been applied. In forming the current population used to modify the individuals, multipoint genetic operators, random generation, and a strategy of elitism have been applied. To vary the location of the prestressing system and take into account the multivariance of loads, parallel evolving populations have been introduced, between which a limited exchange of individuals is allowed. Examples of optimization of prestressed frames with girders and trusses were considered.