Investigation on a new methodology for thermal power plant assessment through live diagnosis monitoring of selected process parameters; application to a case study

This paper proposes a new methodology for short- and long-term assessment of the operation of a thermal power plant to promote preventive maintenance as well as failure analysis while ensuring a degree of compliance with statutory regulations. This study aims to identify and monitorize the quasi-steady states associated with measurements in time series of power plant process. It was focused on a new procedure for detection over different thermodynamic variables involved, being multivariate and automatic. For each one of these states, statistical calculations are carried out to configure a data pre-processing level. The data settings now available can be used as assessment criteria based on detected deviations from a reference system that has been updated during plant-performance tests, addressing long term variations suggested variables. Although the most important outcome is the highly precise and valuable information that will be obtained on the live operating mode, leading to further improvements of the cycle efficiency and achievements in the overall control system of the thermal plant, the main goal is to detect any anomaly, reacting as quickly as possible to return the plant to a normal operation mode. This study also provides a comprehensive practical application, monitoring a power plant of reference. ► We model a new method for short- and-long-term assessment in thermal power plants. ► It is based on automatic multivariate detection of different steady states. ► We supply improved deviations from a reference system of operation's performance. ► We provide a comprehensive application to highlight the achievements mentioned. ► The study deals with the long term mode, associating deviations to possible causes.