The Improved Group Search Optimization Algorithm And Its Application On Ethylene Cracking Furnace

Ethylene production is the main pillar of the whole petrochemical industry, and ethylene cracker is the key equipment in the ethylene plant. To operate the ethylene plant in a safe, efficient mode, has become more and more important. However, in the practical production, the cracker usually can’t be operated in an optimal state. The optimal operation of the cracker only depends on the experience of the operator. The operational condition can’t be adjusted reasonably, especially in the process of changing the cracked feed, the scheduling and strategy of the production are not able to be changed, which leads to low production yield and high energy consumption.Two optimal problems need to be solved during the operation of ethylene cracking furnace.One is the time when the feed changes, the operational factors can not be adjusted reasonably. To solve this problem, the model of yield of ethylene and propylene is established, which depends on the neural net which uses input and output from the data of the plant. The improved group search optimization(GSO) algorithm is proposed to find the optimum value of the model. In the respect of algorithm, the differential evolution group search optimization algorithm and the quantum group search optimization are proposed based on the GSO, which promotes the performance of searching the optimum.The other is the process of changing the feedstock type during one operation cycle, the switching moment of the feedstock and the related operational variables are not able to be controlled reasonably. To solve this problem, the model is built up based on CoilsimlD software which predicts the operational variable of the whole cycle of the cracker. To solve the problem of nonlinear constraint, some improved strategies are put forward based on the penalty function and the method of double values. The improved algorithm has better performance in robustness and convergence through some numerical simulations of benchmark functions. Compared with other classical algorithms, the proposed algorithm has better ability in optimal performance.Finally, the proposed model and algorithm are used in the actual situation of the cracker’s operation. The operational condition can be optimized when the feedstock type changed. During the process of scheduling the feedstock, the better scheduling strategy can be acquired. Therefore, the yield of production has improved obviously, so as the productivity effect. The optimal results can provide the decision support of the switching feedstock moment for the operation staff.