基于改进遗传算法的Claus脱硫尾气加氢反应动力学

由于Claus反应在热力学上存在反应平衡，传统的Claus装置中尾气不能满足目前的排放要求。加氢还原吸收新工艺将尾气中的SO₂、有机硫、单质硫等硫化物转化为H₂S，提浓后再返回Claus装置，从而达到较高的总硫回收率。本文提出了一种幂指数型反应动力学模型，并采用改进遗传算法对动力学模型参数进行估值。统计检验和误差分析表明此模型计算值与实验值吻合良好，所建立的反应动力学模型适宜。

Kinetic Study of Hydrogenation Reaction of Claus Tail Gas Based on Improved Genetic Algorithm

The Claus process is widely applied in sulphur-containing gas processing, but the tail gas from traditional Claus reactors cannot meet current emission requirements due to the thermodynamic reaction equilibrium of the Claus reaction. The novel hydrogenation and absorption process converts SO₂, organic sulphides and elemental sulphides in Claus tail gas into H₂S, and H₂S is concentrated and returns back to Claus reactors. Thus, the process can acquire high sulphur conversion. However, literatures related to kinetic studies of Claus tail gas hydrogenation are rather limited. An appropriate reaction kinetic model for the tail gas hydrogenation not only contributes to the scale-up of the hydrogenation reactor, but also benefits the process optimization of Claus desulphurization. A power exponent type reaction kinetic model for the hydrogenation reaction of tail gas was proposed in this work and the parameters in kinetic model were estimated using improved genetic algorithm. An approximate value near the global minimum can be obtained by the genetic algorithm, but it requires rather long time and considerable computing resources to obtain the exact numerical solution due to the intrinsic properties of the traditional algorithm. The genetic algorithm can be improved by combining with the Marquardt algorithm in which the calculation stops if a calculation fault occurs in the searching process and steps into the next set of calculations instead. The improved algorithm can obtain the optimal solution at the global range and becomes more stable and more flexible compared to the traditional genetic algorithm. The statistic test and error analysis have demonstrated that the calculation results are in consistence to the experimental data, supporting the effectiveness of the established model.