This paper presents a new approach based on genetic algorithms to solve the thermal unit commitment problem of electric power systems. Genetic algorithms (GAs) are general search techniques based on the biological metaphor and are very suitable for solving combinatorial optimization problems. Because of its nonconvex and combinatorial nature, the unit commitment problem is difficult to solve by conventional programming methods. However, it is well suited for the application of the GAs. A key to the success of the implementation of the proposed algorithm is a newly developed knowledge-augmented mutation-like operator, named here the forced mutation. It was found to improve, significantly, the efficiency of the GA in solving the unit commitment problem. Two different coding schemes were devised and tested. In addition, the effects of GAs' control variables on convergence were extensively studied. The approach was tested on a ten-unit system. Test results clearly reveal the robustness and promise of the proposed approach.