The ever-increasing demand for electricity and the need for conventional sources to cooperate with renewable ones generates the need to increase the efficiency and safety of its generation sources. Therefore, it is necessary to find a way to operate existing facilities more efficiently with full detection of emerging faults. In order to improve the operation of steam power plants of electricity generating plants, thermal-fluid diagnostics have been traditionally used, and in this paper a three-body steam turbine, having a high-pressure, a medium-pressure and a low-pressure part, has been selected for analysis. The turbine class is of the order of 200 MW electric power. Genetic algorithms (GA) were used in the process of creating a diagnostic model. So far, they have been used for diagnostic purposes in gas turbines, and no paper has been found in the literature using GA for the diagnostic process of such complex objects as steam turbines located in professional industrial facilities. The use of genetic algorithms have allowed rapid acquisition of global extremes, e.g. efficiency and power of the unit. The result of the work carried out is the possibility to perform a full diagnostic process, i.e. detection, localisation and identification of single and double degradations. In this way 100 % of the main faults are found, but there are sometimes additional ones and these are not perfectly identified especially for single time detection. Thus, the results showing a very high success rate of the investigations to the simulated damages of the steam turbine geometrical components is discovered.