Estimation of Calorific Value of Lignite Field in Kütahya-Gürağaç (Turkey) by means of Artificial Neural Network

Abstract

Artificial neural networks are generally information processing systems that mimic the working principles of the human brain or central nervous system. Artificial neural networks are a method that gives successful results in solving many daily life problems such as classification, modeling and prediction. Artificial neural networks accomplish this by adjusting the connection weights between neurons. It can solve prediction and classification problems with back propagation algorithm, which is widely used in artificial neural networks with multilayer perceptron. In this study, unknown calorific values were tried to be estimated by using the analysis values (depth, ash, moisture, sulfur, calorific value) of the drillings realized in the Kütahya -Gürağaç lignite field. An artificial neural network was created for this purpose. First, 8 neurons were used in the hidden layer of the network, and 10 neurons were used secondarily. In the artificial neural network, the learning function is sigma, the learning rate is 95%, and the network is trained using Levenberg-Marquardt as the training algorithm. The network with 10 neurons converged at the desired margin of error (1e-07) and was completed after 271 iterations. The relationship between actual calorie values and predicted calorie values with network training reached a high ratio of R2=0.97. After the training of the network is completed, the network is simulated for the estimation of seams with unknown caloric values. As a result, caloric values were determined with an average of 97% confidence interval for the unknown coal seams of the field.