Thermoelectric power plants have critical units, such as the boiler and the turbine-generator, which are complex multivariate systems. These units exhibit non-stationary behavior and multiple operational modes that imply constant changes of set points of key performance variables. A methodology based on MSPC (Multivariate Statistical Process Control) techniques and PCA (Principal Component Analysis) is presented with an adaptive mean estimator that deals with frequent changes of set points, both for design and just in time monitoring. The proposed methodology is implemented in a thermoelectric power plant using a commercial PIMS (Process Information Management System) software suite. Experimental results illustrate and validate the proposition, its just-in-time implementation and usage.

Multivariate Statistical Process Control (MSPC) techniques require process stationarity as a main rule for design and monitoring. However, a critical process such as Boiler and Turbine-Generator units of Thermoelectric Power Plants, which is a multivariable complex system, features different types of non-stationary behavior and operational modes with constant changes of set points of key performance variables. A methodology based on MSPC and Principal Component Analysis (PCA) is presented with an adaptive mean estimator that deals with frequent changes of set points, both for design and just in time monitoring. Experimental results, based on data from a power plant, illustrate the application and use of the methodology.