25. – 27. April 2012
Czech Technical University in Prague, Faculty of Mechanical Engineering

Neural Network Model for Prediction of NOx AT Coal-Powder Powerplant Melnik 1
Ivo BUKOVSKY, Michal KOLOVRATNIK
Abstract:
The paper presents nonconventional dynamic neural network that was designed for real time 

Keywords: dynamic neural networks; prediction of process variables; signal processing
Introduction 
 Neural networks (NN) are popular and widely studied tool for real data-driven nonlinear modeling for complicated systems where mathematical-physical analysis is unavailable for 
Data Preprocessing and Network Training
The NOx dynamics of the pulverized boiler is highly nonstationary due to varying technical conditions of the boiler, varying quality of coal powder.

Figure 1: The training data preprocessing before each reconfiguration and retraining of neural network
where U(k) is a matrix of recent history of all measured input variables as follows
The considered model inputs in U(k) are the primary, secondary, and tertiary air valves. 
Neural Network for NOx prediction
A short summarization of the paper should be given here as well as possible view for further work and some other remarks. 
Conclusions
A short summarization of the paper should be given here as well as possible view for further work and some other remarks. 
References: 
GUPTA M. M., LIANG J., HOMMA N. Static and Dynamic Neural Networks: From Fundamentals to Advanced Theory, IEEE Press and Wiley-Interscience, pub. John Wiley & Sons, Inc., 2003.
Acknowledgment
This work has been supported by grant MPO _ FR-TI1/538 and in part by grant SGS10/252/OHK2/3T/12. 
2