The problem of multi-objective short term unit commitment in power systems containing a large number of thermal units is considered in this thesis. The total production cost and the produced emissions are considered as the multiple objectives of the problem to be minimized simultaneously in the programming period. To solve the problem a method is suggested based on the genetic algorithm to handle the large dimensionality of the problem and fuzzy modeling and fuzzy optimization are used to find the compromise fitness values of the chromosomes according to multiple objectives. By the proposed method, elementary generations of chromosomes are firstly produced that is composed of feasible solutions of the problem. Using genetic operators like cross-over and mutation, new generations of children are produced with better compromised fitness values. To prevent the algorithm to stop at local optimum solutions and to increase the convergence speed, some modification such as a dynamic variation of mutation rate and producing apart of chromosomes by a priority list are considered. When the on units in each hour of the program are known, their optimal share to maintain the forecast load of the hour should be determined due to multiple objective functions. This is also done by a fuzzy linear programming method. Various case studies are considered by running program in a small 10 unit test system to verify and evaluate the proposed algorithm. Results obtained using a test system containing 100 thermal units are also analyzed and show the ability and effectiveness of the algorithm to overcome difficulties of multi-purpose short term unit commitment in real large scale power systems.