论文部分内容阅读
吸收稳定系统是多变量、强耦合系统,难于实现基于机理模型的操作优化。提出了用带搜索约束的神经网络(ANN),结合遗传算法(GA),实现吸收稳定系统操作优化的新方法。第1步,选用富气压缩机出口压力、补充吸收剂流量和吸收剂温度作为调优变量,以产品产量、压缩机功耗、解吸塔和稳定塔再沸负荷及主要冷却负荷为输出变量,建立具有最佳初始权值和阈值的吸收稳定系统反向传播型(BP)神经网络模型,以预测系统的操作;第2步,用GA引导BP模型实施以系统综合效益最大为目标函数的优化计算,以确定系统的最佳操作变量值。鉴于ANN“黑箱”特征,不掌握调优变量之间的关系,GA寻优可能陷入无效区域,从吸收稳定工艺出发,提出用吸收塔回流板水力学特征约束GA次代的产生,以保证寻优过程有效进行。某1.2 Mt/a重油催化裂化装置吸收稳定系统应用表明,该方法是有效的,经过12代GA运算成功找到最佳操作点,相较于现有操作,综合效益增加252万RMB/a。
Absorption stability system is a multivariable, strongly coupled system, it is difficult to achieve the operation optimization based on mechanism model. A new neural network with search constraint (ANN) and a genetic algorithm (GA) are proposed to optimize the operation of the absorption stable system. Step 1, the outlet pressure of rich compressor is selected, the flow of absorbent and the temperature of absorbent are added as the tuning variable, and the output of the product, the power consumption of the compressor, the reboiling load of the desorption tower and the stabilizing tower and the main cooling load are taken as output variables, In order to predict the operation of the system, the BP neural network model with the optimal initial weight and threshold is established to predict the operation of the system. The second step is to optimize the objective function with GA-guided BP model Calculate to determine the best operating system variable values. In view of ANN “black box ” characteristics, GA does not grasp the relationship between tuning variables, and GA optimization may fall into an ineffective area. Based on the absorption stabilization process, it is proposed that the generation of GA secondary generation should be restrained by the hydraulic characteristics of the absorption tower recirculation plate Optimization process effectively carried out. The application of an absorption stability system of a 1.2 Mt / a heavy oil catalytic cracking unit shows that this method is effective. After 12 generations of GA operations, the optimal operating point has been successfully found. Compared with the existing operation, the comprehensive benefit increased by 2.52 million RMB / a.