| Popis: |
Hydrotreating of gasoil is one of the most important processes in petroleum refining; it helps to improve the characteristics of diesel fuel to make it meet the required specifications and pollution standards. Modelling of this process would, among other things, allow to predict the product quality as a function of different process variables (temperature, pressure, space velocity). There are several modelling techniques, among them the Black Box type modelling by Artificial Neural Networks (ANNs), which is used to model the hydrotreating process at pilot scale. The approach used in this work is an artificial intelligence approach. The model developed is a neural network Multi Layer Perceptron (MLP) type. Network learning (NL) is carried out according to the Backpropagation Gradient algorithm with momentum; The Early Stopping (ES) technique has been used to prevent the effect of overfitting and thereby ensure a good generalization of the model. Experimental and predicted results show good agreement with an error not exceeding 4%. |
