Open AccessPredictive Modelling of Grate Combustion and Boiler Dynamics
Author(s) -
R.S. Jha,
Navani Niharika Jha,
Mandar M. Lele
Publication year - 2022
Publication title -
mathematical modelling and engineering problems/mathematical modelling of engineering problems
Language(s) - English
Resource type - Journals
eISSN - 2369-0747
pISSN - 2369-0739
DOI - 10.18280/mmep.090129
Subject(s) - boiler (water heating) , combustion , airflow , coal , environmental science , engineering , control theory (sociology) , computer science , waste management , mechanical engineering , chemistry , control (management) , organic chemistry , artificial intelligence
Pressure and air to fuel ratio control are extremely difficult in coal-fired grate boilers due to a significant lag in combustion. This leads to suboptimal operation of the boiler and poor efficiency of the plant. This also leads to higher level fluctuation. Fluctuation in pressure, water level and oxygen level are quite evident in the operation of coal-fired grate boilers in fluctuating load conditions. The present work aims to develop a predictive and dynamic simulation model of a coal-fired grate boiler for the prediction of pressure, and water level in fluctuating load conditions and its extension for the prediction of oxygen level. A data-driven approach has been used for the prediction of heat release, distribution of heat transfer, circulation analysis and airflow through the various dampers. This model has been integrated with the boiler dynamics model of a hybrid boiler. Errors in pressure and water level are measured for training data and the multi-objective optimisation method is used for the minimisation of errors. The Batch Gradient Descent method has been used for the minimisation of errors. The proposed integrated model is used for the estimation of heat release and the rate of combustion. Stochiometric combustion calculation is used to predict oxygen level by using the predicted value of airflow and rate of combustion. Root mean squared error is calculated for oxygen level and minimised by the Batch Gradient Descent algorithm. The model has good accuracy in the prediction of boiler pressure and water level and can be extended to improve the boiler controls of a solid fuel fired reciprocating grate boiler in extremely fluctuating load conditions.