Multivariate optimization to reduce emissions of environmental pollutants – nitrogen oxides from combustion of solid fuels
Öberg, T., Bergström, J.
Lecture at the Second Scandinavian Symposium on Chemometrics in Bergen, Norway, May 28-31, 1991.

Abstract
Nitrogen oxides are environmental pollutants of significance. In Sweden it has therefore been decided to set "environmental fees" on these emissions from large boilers. The most cost-effective measure to reduce emissions is process optimization.

We will describe a multivariate approach to reduce emissions of nitrogen oxides from the combustion of various fuels. Initial test series were carried out with fractional factorial designs. Design parameters and other process parameters of interest, 30 variables, were measured continuously and recorded as 5-minute mean values. Each test series consisted of eight 4-hour tests (about 10,000 measurement values). Data were analyzed by principal component analysis, multiple linear regression and partial least squares regression. The response functions obtained were used for optimization by the "steepest ascent" method. Response surfaces were calculated from PLS-regression on further experiments carried out in the optimal region.

It was possible to reduce the emission of nitrogen oxides with at least 50%, without causing disturbances in the normal operation in these boilers.

Full text as PDF file, 295 kb