Overview of statistical methods of process monitoring

Abstract

At present, process monitoring by comparing the current parameters against a specified setpoint is widespread at thermal power plants. This approach does not allow diagnosing the emergence of a trend leading to an emergency mode at early stages. On the other hand, the analysis of time series of parameters by means of methods of statistical process monitoring allows detecting deviations from normal mode of equipment operation before emergency conditions appear. The purpose of this work is to analyze existing works in the field of application of statistical methods of monitoring. A systematic literature review (SLR) method was used in the analysis. The research process included clarifying research questions, searching for articles in databases, and forming article evaluation criteria. The specifics of the process at TPPs, which impose restrictions on the application of individual methods of statistical monitoring, were formulated. The study resulted in 64 publications, which were filtered and classified into rating groups. The highest ranked articles were used to determine the most effective statistical monitoring methods to be used on thermal power equipment. The answer to the questions posed in the study also reveals the existing problems of applying statistical methods of monitoring. The results of the study will help to identify the most applicable statistical methods for thermal power industry. It is concluded that the most suitable methods for application on thermal power equipment are methods of control charts (both Shewhart charts and Hotelling charts) based on either process parameters or generalized variance, as well as the use of autocorrelation models.