С использованием данных испытаний турбогенератора типа ТГВ-300 мощностью 300 МВт на Трипольской ТЭС проведены численные исследования температурных полей ротора в различных эксплуатационных режимах. Создана база данных для определения максимальных температур обмотки возбуждения в зависимости от степени развития в ней термодефектов и активной и реактивной нагрузки ТГ. Построены статистические модели зависимости температуры от технологических параметров и степени ухудшения условий охлаждения витков обмотки для диагностирования дефектов в системе охлаждения обмотки ротора. The technique of numerical calculation heating of a rotor powerful turbogenerator is developed. The technique takes into account real geometry of object, presence of defects cooling, a condition thermolysis from a surface and in ventilating channels of a winding excitation, parameters of a loading mode.The technique can be applied for research of temperature fields of a rotor turbogenerator in asynchronous, asymmetrical, dynamic and other modes. The analysis of a technical condition turbogenerators on thermal power stations of Ukraine is carried out. The main malfunctions of TG influencing reliability of work are defined. On the basis of the given tests of a turbogenerator such as ТGV-300 by capacity 300 МWt on Тriрolskaya power station numerical researches of temperature fields of a rotor in various operational modes are carried out. The database for definition of the maximal temperatures of a excitation winding depending on a degree of development in her thermo defects and active and jet loading TG is created. Comparison of settlement and experimental values on heatings is carried out under nominal conditionsof cooling of a rotor winding. Statistical models of temperature dependence from technological parameters and degrees of deterioration of cooling conditions of winding coils are developed. It enables to diagnose many defects in system of cooling of a rotor winding under operating conditions generators at constantly changing loadings and technological parameters. The received multiple-factor regression dependences are based on a method of the smallest squares. They provide high reliability of models. For all dependences confidential intervals are defined. It is shown, that the data are in significant correlation with a technical condition of a turbogenerator.