Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
5000264 | Control Engineering Practice | 2017 | 15 Pages |
Abstract
New emission regulations will increase the need for inexpensive NOx emission monitoring solutions also in smaller power plants. The objective in this study is to find easily maintainable and transparent but still valid models to predict NOx emissions in natural gas fired hot water boilers utilizing existing process instrumentation. With a focus on long-term applicability in practical installations, the performance of linear regression is compared in two municipal 43 MW boilers with three widely used nonlinear methods: multilayer perceptron, support vector regression, and fuzzy inference system. The linear models were the most applicable providing the best estimation results (relative error of <3% in all cases), generalizability and simplicity. Therefore, the approach fulfils the requirements of the Industrial Emission Directive and is valid to be applied as a soft sensor in PEMS1applications in practise. However, each boiler model should be identified individually.
Related Topics
Physical Sciences and Engineering
Engineering
Aerospace Engineering
Authors
Timo Korpela, Pekka Kumpulainen, Yrjö Majanne, Anna Häyrinen, Pentti Lautala,