Novel system for predicting slagging behaviour of fuel blends in large scale utility boilers. Tan, C.K.; Kakietek, S.; Wilcox, S.J.; Ward, J.; Golec, T.; Szymczak, J. Journal of the Energy Institute, Volume 79, Number 4, December 2006 , pp. 251-256

The work described in this paper aims to address the limitations of conventional techniques for predicting the risk of slagging in pulverised coal fired boilers by developing a set of laboratory based slagging parameters and a predictive model which can be easily scaled up for full scale boiler application. The predictive model is based on artificial neural networks (ANNs) which are known for their robustness and the ability to learn complex data relationships. In the current study, the ANN models were further optimised by using a genetic algorithm (GA) which searches for an optimum set of training data so as to minimise overfitting of the data. This GA based approach was also compared with several other more conventional methods of selecting data for ANN training and the results suggested that GA based data selection outperformed all other methods in this exercise. In the final part of the paper, the ANN based predictive model was coupled with a simple heat transfer calculation and implemented on a full scale boiler. The results from this validation exercise suggest that the approach is useful and offers plant operators a tool to quickly assess the risk of slagging.