Application of multivariate statistical process control to supervising NOx emissions from large-scale combustion systems

Abstract Computer modeling of NOx formation in combustion systems provides a tool that can be used to investigate and improve understanding of the systems. In this work the systematic method based on the data-driven model was proposed to supervise nitrogen oxides emitted from a large number of fired heaters which go through one common stack. The relationships between process variables in the fired heaters and the NOx were investigated by employing projection to least squares (PLS) regression in a hierarchical manner. Multivariate statistical contributions of fired heaters and process variables in a fired heater to NOx emissions were introduced as a key means to cope with the NOx overemissions. The proposed approach was evaluated by using data collected from an industrial combustion system.