Ali Shaeri
Detailed operating condition were gathered from a live Delay Coker unit (DCU) (Lab. And DCS) for two years. The most significant parameters were selected by both simulation and experience. Coke, output CCR, light gases, gasoline, gas-oil and C5+ weight percent are the network outputs. The R2 and MSE of the proposed model were 0.92583 and 0.1424, respectively. It was found in the optimum operation conditions. Then, by considering all operational constraints, the results confirmed that the decision variables generated by the optimization approach can enhance the gross profit of the hydrocracking process to more than $0.51 million annually, which is significant for the economy of the target refinery. Among the Multi Layer Perceptron (MLP) architectures a network with 31 hidden neurons has been found as best MLP predictor. 80 percent of the data have been used for training of ANN. Radial Basis Function (RBF) also has been implemented for identification of the plant. Best RBF network and best MLP network performance in prediction of 25 percent of unseen data were compared. It was found that RBF method has the best generalization capability and was used in DCU modeling.
The delayed coking process is used to convert heavy oils into more lucrative light liquid products while producing less valuable gas and solid coke byproducts. Although the first delayed coking plant was established in 1930, the delayed coking process has been evolving for 78 years. In recent years, changes in feed stock have had a significant impact on the design and operation of delayed coking plants
