The thesis emphasizes on the development ofpractical daily generator maintenance scheduling methodfor a large-scale power system. Several constraints,e.g. crew, reserve, multiple times of maintenanceconstraints, etc., are taken into account includingseveral energy limited units. With the proposed method,an appropriate generator maintenance plan can beobtained. The proposed method first employs the totalproduction cost to decreases the size of problem and toobtain initial solution. Then, the levelized Loss ofLoad Probability (LOLP) method taking into accountreliability index as an objective function is used tosearch for a generator maintenance plan. The effect ofenergy limited units which are included in thereliability indices and probabilistic production costcalculation are solved by the peak shaving method. Based on the proposed method, a computer programhas been developed and tested with the IEEE-RTS (IEEEreliability test system) and the EGAT (ElectricityGenerating Authority of Thailand) system of the fiscalyear 1998 and 1999. The results show that thecomputation time by the proposed method is less thanthat of conventional methods. The reliability indicesand total production costs of the plan obtained fromthe method are close to or slightly better than theindices and costs of the plans obtained from theconventional methods. Finally, the results demonstratethat the proposed method can be practically used in theactual operation of a large power system.