Predictive control of CO2 emissions from a grate boiler based on fuel nature structures using intelligent neural network and Box-Behnken design

Abstract

Abstract The aim of this research was to predictive control of CO2 emissions by modelling the correlations between fuel nature structure (elementary composition) and CO2 emissions from a grate boiler. Back Propagation Neural Network (BPNN) coupled with Genetic Algorithms (GA), which facilitates the learning algorithms to figure out the local minimum deviation, is employed to map the highly nonlinear relationships between elements such as C, H and O in fuels and final CO2 emission. A total of 15,000 training and testing data come from the recordings of a grate boiler within six months. And the predicted CO2 emissions based on fuel nature structure matched the measured data with fairly good agreement. Finally, the Box-Behnken experimental design methodology was used to extract the mathematical expression between elements in fuels and CO2 emission. Consequently, by knowing the C, H and O composition in fuels, the CO2 emission can be well forecasted, in such way, it is sensible to optimize the future fuel nature structure in order to achieve clean carbon footprint and control the CO2 emissions.