In this paper a radiator heating system of a building is considered. For the purpose of the heating system optimization, a mathematical model of the system is developed. The linear quadratic algorithm with integral action is proposed and analyzed. This solution has proven to be expensive. Further analysis of the model is done and a reduction of the order of the system is proposed. An inverse-based controller design approach for minimum-phase first order system is used to provide realizable controller in the form of proportional integral controller. Optimal parameters of the control algorithm parameters have been chosen by integral of time absolute error criterion, and also by metaheuristic optimization. According to the real heating demand, a simulation of the plant is performed. Proposed controllers were tested by numerical simulation for a typical winter day for geographical region of the building. It is shown that advanced performance can be achieved with optimized control systems, and that by controller optimization a significant reduction of the energy consumption is obtained without losing the in-door comfort. This has also proved to be more economical solution. [Project of the Serbian Ministry of Education, Science and Technological Development, Grant no. TR 33047, Grant no. TR35005 and Grant no. TR 35016]