A Neural Network Approach for Identification and Modeling of Delayed Cocking Plan

In this study, an artificial neural network (ANN) modeling of a delayed cocking unit (DCU) is proposed. Different data from various DCUs have been collected. Feed API and cat cracker (CCR) weight percent have been considered as network inputs. Coke, output CCR, light gases, gasoline, gas-oil and C5+ weight percents are the network outputs. 70 percent of the data have been used for training of ANNs. Among the multi layer feed forward architectures a network with 31 hidden layer neurons has been implemented. Radial basis function (RBF) also has been implemented for identification of the unit. An RBF network with 20 spread was found as best estimator of the DCU in this paper. Best RBF network and best feed forward network performance in prediction for 30 percent of unseen data were compared finally. RBF method had the best generalization capability and was used for DCU modeling.