Prediction of Calorific Value of Indian Coals by Artificial Neural Network

The experimental determination of calorific value of solid fuels is a cost intensive process, as it requires spatial instrumentation and highly trained analyst to perform the experiments, whereas proximate analysis data can be obtained easily using an ordinary muffle furnace compared to calorific value. Regression analysis and artificial neural network analysis methods have been introduced to simplify the task and also reduce the cost of analysis. An endeavor has been made in this present study to access the applicability of these correlation and artificial neural network with a spatial emphasize on Indian coals. Correlation have been created using simple linear regression and multivariable linear Regression analysis based on proximate analysis of data sets. Artificial neural network model is also designed to predict the gross calorific value of coals belonging to different Indian coal fields. 59 samples were collected from different coal fields of India including the South Eastern Coalfields (SECL), Singareni Collieries Company Limited (SCCL), Central Coalfields limited (CCL), Mahanadi Coalfield Ltd. (MCL), Eastern Coalfields Limited (ECL), North Eastern Coalfield Limited (NECL), Jindal Steel and Power Limited. The intrinsic properties were determined by carrying out proximate analysis and gross calorific value (GCV) by using bomb calorimetry. It was observed that all the three models predict the calorific value fairly accurately. However, the ANN model gives a better prediction than the other methods. Therefore, prediction of gross calorific value by ANN model could be a viable option than experimentation in the laboratory. The ANN model considers the intrinsic properties determined by proximate analysis as input parameters, which is a routine task in the field as these are required to determine the grade of coals and hardly demand any costly experimental setup