Performance Optimization of Steam Generator by Reduction of Unburned Carbon

Abstract

In present work optimization of various input parameter to minimize the unburned carbon in fly ash and bottom and CO in flue gas. Data collection has been taken from the maximum distributed control system (DCS) in which current and historical data has been available. These DCS data are taken from instrument placed on location using card and transmitter. Based on these data various test has been done by changing single variable to know the pattern how it impacts on unburned carbon. Based on single test observation multi input variable are used to know the impact on unburned carbon. 15 tests have been done with multi input to find their impact on unburned carbon and based on these tests optimize input parameter are taken which give minimum unburned carbon. Machine learning tools with linear regression model (LRM) are used to predict unburned carbon based on data learns from the past. Test results are used to train the machine learning software and LRM. Java based machine learning tools are used to predict the unburned carbon in fly ash and bottom ash which give interesting result with more than 90 % accuracy. The present work in provide practical solution for the 110 MW power plant which facing problem of high unburned carbon in fly ash and bottom ash and CO in flue gas. Based on final 15 test result various optimize input parameter has been found and by using these results substantial decrease in unburned carbon which lead to increase steam generator efficiency. The result found in machine learning linear regression model give good prediction based on historical trained data. It will help plant operator to know the approximate unburned carbon losses based on running input parameters.