Maximization of Combustion Efficiency: A Data Mining Approach
Maximizing combustion efficiency with minimizing emissions is of importance to electric power industry. In this research, the impact of data transformation on boiler efficiency is investigated. The study showed that the data transformed with wavelet algorithms (Haar and Daubechies) provided better cross-validation accuracy, while moving average and wavelets resulted in similar prediction accuracy. The relationship between the length of the control horizon and prediction accuracy is studied. The study shows that the control horizon of a 3-hour to a half-week long provided acceptable prediction accuracy. An ensemble predictive model of the two control horizons is proposed to increase prediction accuracy. The research findings have established foundation for maximizing combustion efficiency by introduction of meta-controllers based on data mining algorithms.
KeywordsPrediction Accuracy Move Average Haar Wavelet Decision Category Data Mining Approach
Unable to display preview. Download preview PDF.