Smart assessment of hidden energy losses at a thermal power plant

In search of accurate and effective ways to quickly assess the hidden energy losses of power plants during the supply of electricity to the grid, this study examines the information technology of intellectual reasoning based on the precedents of stationary generation. A steam-gas electric power plant was chosen as the object of research. The input characteristics are the ambient temperature, relative humidity and ambient pressure, which are the main factors for gas turbines, as well as the exhaust vacuum measured in the steam turbine. The study used methods of clustering precedents, the technology of finding the nearest neighbors in machine learning and the usual linear multifactor regression. It was found that even with the help of such simple tools as linear local multifactor regression, it is possible to adequately estimate the hidden energy losses caused by uncontrolled factors in the operation of a steam-gas power plant. Using more sophisticated tools and proper pre-processing can significantly increase the reliability of the assessment.