ECN publication
Title:
Neuro-fuzzy control in a steam boiler at Hoogovens
 
Author(s):
 
Published by: Publication date:
ECN 1998
 
ECN report number: Document type:
ECN-I--98-009 Other
 
Number of pages: Full text:
44 Download PDF  

Abstract:
A neuro-fuzzy controller has been designed for a burner system in a gasfuelled steam boiler at Hoogovens Staal in IJmuiden, Netherlands. In comparison to the existing conventional controller, the intelligent controller will yield a tighter NOx emission control and enhance energy efficiency. The controller maximises energy efficiency under the constraint of a maximum NOx emission level, and was designed with a fuzzy network structure. This structure makes optimal usage of the fruitful cross-fertilisation of two advanced control techniques: fuzzy logic, providing transparency and interpretability, and neural networks, providing learning ability. The controller has been trained on real historic process data to perform optinially. The underlying assumption is that the non-linear static map generated by the neural network can adequately represent the system's inverse dynamic behaviour in the ranges of interest. This design procedure is less time-consuming and more cost-effective than the conventional design method of analytic modelling, frequency response analysis, and manual fine-tuning. In addition to these advantages in its design phase, the neuro-fuzzy controller can show several important advantages in its application phase. First, the costs of practical implementation will be low since standard fuzzy logic controllers can be used, and additional sensors are not needed. Second, the neuro-fuzzy controller will show an excellent performance. This is due to its non-linearity, which makes it very flexible: its performance is optimal under a larger range of process conditions instead of only in one working point. Also, the controller can detect sudden process perturbations, such as strong load changes or large variations in fuel gas composition, and has learned how to compensate for these accordingly before they would affect the NOx emission level. The improved controllability of the NOx emission level allows the steam boiler to be operated closer to its NOx emission constraint and thus closer to its maximum efficiency. Moreover, energy efficiency will increase by minimization of the electrical energy consumption of the recycle flow ventilator. Furthermore, a tighter fuel/air ratio control and a more stable boiler operation are expected. Superior performance over the existing controller can be proven by on-line application. 25 refs.


Back to List