在我国石油、化工等工业生产中，精馏塔一直是最重要的设备之一，整个石油工厂30%以上的能量消耗均是来自于精馏塔，而温度是实现产品质量控制主要的间接参数。精馏塔不但机理复杂，变量之间相互关联，而且温度控制惯性大，传统的精馏塔温度控制方法控制精度低，因此研究更优的精馏塔温度控制方法具有重要意义。

本文详细介绍了精馏塔的结构组成以及分馏原理，并对精馏的生产工艺流程、温度控制系统进行了分析。

针对精馏塔塔顶和塔釜温度存在严重的耦合关系，本文分别采用前馈补偿解耦和神经网络解耦。解耦后的两个回路互相独立，以基于改进人工蜂群算法的PID控制器进行控制，根据被控对象当前特征，实时调整PID控制器的参数，提高被控温度的稳定性和准确性。基于改进人工蜂群算法的PID控制器与神经网络解耦结合就构成了GABC-PID-神经网络解耦控制，不仅解决了只采用PID控制时，被控对象耦合严重，控制效果不佳的问题；又解决了只采用解耦算法时，常规PID受各种扰动后，原有控制器参数不能很好适应变化后对象的问题。最后，分别以常规不解耦、前馈补偿解耦、GABC-PID-神经网络解耦控制方案对精馏塔温度控制系统进行仿真实验。对比仿真结果，GABC-PID-神经网络解耦控制在温度最大动态偏差、调节时间指标方面具有更好的控制效果。

通过对精馏塔塔顶温度和塔底温度的解耦控制，有效减小了塔顶温度和塔底温度动态过程中的最大动态偏差和调节时间，提高了温度控制的精度和稳定性，实现了塔顶馏出物（塔顶产品）和塔底产品（残液）的“卡边”控制，对解决实际生产过程中的产品纯度控制不精准和能耗大的问题具有一定理论意义。

In China's petroleum, chemical and other industrial production, the rectification tower has always been one of the most important equipment. More than 30% of the energy consumption of the entire petroleum plant comes from the rectification tower, and temperature is the main indirect parameter to achieve product quality control. Not only is the mechanism of the rectification tower complex, the variables are interrelated, but also the temperature control inertia is large. The traditional rectification tower temperature control method has low control accuracy, so it is of great significance to study a better rectification tower temperature control method.In this paper, the structure composition and fractionation principle of distillation column are described in detail, and the production process flow of distillation column, influencing factors of product quality and relevant model calculation are described and analyzed.

This article introduces the structure and fractionation principle of the distillation column in detail, and analyzes the production process flow and temperature control system of the distillation.

In view of the serious coupling relation between the temperature of the top of the rectifying tower and the temperature of the tower kettle, feed-forward compensation decoupling and neural network decoupling are used respectively in this paper. The decoupled two loops are independent of each other, and the PID controller based on the improved artificial bee colony algorithm is used for control. According to the current characteristics of the controlled object, the parameters of the PID controller are adjusted in real time to improve the stability and accuracy of the controlled temperature. The combination of PID controller based on improved artificial bee colony algorithm and neural network decoupling constitutes GABC-PID -neural network decoupling control, which not only solves the problem of serious coupling of controlled objects and poor control effect when only PID control is used. In addition, it solves the problem that the original controller parameters cannot adapt well to the changed object when the conventional PID is disturbed by various disturbances when the decoupling algorithm is used only. Finally, the conventional uncoupling, feedforward compensation decoupling and GABC-PID-neural network decoupling control schemes are used to simulate the temperature control system of the distillation column. Compared with the simulation results, GABC-PID -neural network decoupling control has better control effect in terms of temperature maximum deviation range and adjustment time index.

Through to the rectification tower top temperature and bottom temperature of decoupling control, effectively reduced the tower top temperature and bottom temperature dynamic process and the maximum deviation of amplitude adjustment time, implements the tower distillate (top) and bottom products (" edge "of the residual liquid) control, to solve the actual production in the process of the problem of large energy consumption of the product purity control is not accurate and has a certain theoretical significance.