A load variation on power plant causes thermal stress in power plant facilities. Thermal stress not only shorten life time of facilities, but also cause unforseen accidents. Especially, thermal stresses developed in turbine rotors are most serious. So it is important to limit the thermal stress developed in turbine rotors when operating fossil power plants. In this paper, we designed a fuzzy load scheduler with thermal stress constraints in turbine rotors. A fuzzy load scheduler designed here determines load variation rate in accordance with predicted thermal stress. It has two inference engines, one is for an expert's knowledge and the other is for a supervisor's knowledge. A supervisory inference engine operates and generates a new variation rate, when the inference result based on expert's knowledge is not in the valid region confirmed by the supervisor. It also modifies expert's knowledge each time it expires. A fuzzy load scheduler uses a dynamic model to predict the boiler outlet steam conditions which are used as the basis for calculating thermal stress.
