Pultrusion is a technological process, where fibers impregnated with resin move through the heated die and solidify into composite profile with a constant cross section as in the metallic die. To improve an effectiveness of conventional pultrusion processes, preserving the quality of pultruded profiles, new optimization methodology is developed and applied for a production of thin-walled profiles. Due to the large dimension of the numerical problem to be solved and multiple iterations applied for the solution of governing equations, an optimization methodology is developed, employing the method of experimental design and response surface technique. In each point of the plan of experiments the coupled thermo-chemical problem is solved iteratively by the mixed time integration scheme and nodal control volumes method. Different order polynomial functions are evaluated for a development of suitable approximation of simulation results by using a conventional un-weighted least square estimation. To demonstrate an application of the developed non-direct optimization methodology, the real industrial process, producing thin-walled angle profiles made of glass fibers TEX4800 and epoxy resin RESOLTECH 1401+1407+AC140 was chosen. In the present study the objective function is formulated as the minimum energy applied per one meter of pultruded profile. The design variables describe important parameters of the technological process (pull speed, control temperature on electrical heaters, position of electrical heaters) and ambient room temperature. Constraints are introduced into optimization procedure with the aim to provide a qualitative profile production, when the resin is fully cured and no overheated during the pultrusion process. The developed optimization approach has allowed to estimate effectiveness and productivity of the industrial pultrusion process with the temperature control executed by the heaters switch-on and -off strategy. By this way more realistic process optimization was achieved and more accurate values of electrical energy spent for a die heating were obtained.