Ethyl acetate-normal propyl alcohol-water pressure swing distillation process based on neural network control

Abstract

The invention discloses an ethyl acetate-normal propyl alcohol-water pressure swing distillation process based on neural network control, and provides a heat integration three-tower pressure swing distillation process aiming at the problems that a ternary azeotropic system is difficult to separate, high in energy consumption and poor in disturbance rejection capacity in the prior art. The azeotropic composition is changed by adjusting the operating pressure of the high-pressure tower T1, the medium-pressure tower T2 and the low-pressure tower T3, and dynamic optimization control is achieved in combination with a feedforward neural network. In the process, a neural network prediction module (NNPM) acquires parameters such as feed flow, composition and product purity in real time, predicts and optimizes operation variables such as reflux ratio, pressure and thermal load of each tower, and combines proportional-integral-differential (PID) control to form an optimal control system, so that + / -20% of feed disturbance can be stably treated, and the quality of products is improved. The purity of the finally separated ethyl acetate, n-propyl alcohol and water reaches 99.9% or above, and effective energy saving is achieved.

Description

Technical Field

[0001] The present invention belongs to the field of chemical separation and purification, and particularly relates to a pressure-swing distillation process for ethyl acetate - n-propanol - water based on neural network control.

Background Art

[0002] Ethyl acetate and n-propanol are key raw materials in chemical production. Due to their excellent solubility and reactivity, they are widely used in the national defense, electronics, textile, and rubber industries. In recent years, with the rapid growth of the demand for high-purity solvents in fields such as packaging and printing, the importance of their separation and purification technology has become increasingly prominent. In the ternary system of ethyl acetate - n-propanol - water, there are three binary azeotropes: under standard atmospheric pressure, the azeotropic point of ethyl acetate - water is the lowest (70.16 °C), followed by ethyl acetate - n-propanol (77.18 °C), and the azeotropic point of n-propanol - water is the highest (80.16 °C). Since the formation of these azeotropes hinders the complete separation of components, it is impossible to directly obtain high-purity ethyl acetate and n-propanol using ordinary distillation methods, so special distillation methods need to be used. Pressure-swing distillation adjusts the relative volatility of each component in the mixture by changing the operating pressure, and utilizes the difference in separation difficulty at different pressures to achieve the efficient separation of azeotropic or near-boiling mixtures in stages. By adjusting the operating pressure, the azeotropic composition of these three binary azeotropic systems all changes significantly (the change range exceeds 5%), so the pressure-swing distillation method can be used. At the same time, in terms of process optimization, two core problems need to be solved: First, the cascade utilization of energy - by constructing a reboiler heat integration system, double optimization of waste heat recovery and steam consumption is achieved; Second, the economic balance of equipment investment - based on the systematic optimization design of distillation column parameters, the optimal solution of minimizing equipment investment and maximizing economic benefits is sought under the premise of ensuring separation purity.

[0003] The dynamic control system is an automation technology based on real-time monitoring and feedback regulation. By continuously collecting the operating parameters of the system, it realizes the high-precision, strong robustness and adaptive optimization operation of complex industrial processes, and is widely used in fields such as chemical production, energy management, and intelligent manufacturing that require real-time regulation. Therefore, implementing a dynamic control system in the rectification process is of great significance for improving product quality and production efficiency, reducing production costs and energy consumption. In the field of rectification process control, although the traditional proportional-integral-derivative control method has been widely used, due to factors such as non-linear dynamic response characteristics, strong multi-variable coupling, hardware execution delay, real-time optimization lag, and amplification of thermal integration disturbances, it is prone to comprehensive defects such as control instability, energy efficiency fluctuations, and decline in equipment economy. In view of the energy-saving control problem of pressure swing distillation (PSD), the present invention designs a real-time intelligent optimization control model, uses the simulated annealing algorithm for optimization, obtains the optimal operation parameter combination and constructs a training data set. The variables related to the initial feed parameters are used as the input of the FNN, and the operation parameters are used as the output of the training to establish an FNN prediction model with dynamic response ability. Finally, through the collaborative application of the dynamic model and thermal integration technology, not only the energy utilization efficiency is improved, but also the performance of FNN-assisted optimization control in the thermal integration process is verified.

[0004] In view of the energy-saving control of the pressure swing rectification process, the present invention realizes the efficient energy-saving separation of the ethyl acetate - n-propanol - water azeotrope, and the purities of the separated ethyl acetate and n-propanol can both reach 99.9%. In view of the increase in production energy consumption caused by the change of the feed parameters in the actual process production, a real-time optimization control model is designed, and finally a real-time optimization and improvement control process that combines thermal integration technology with the heterogeneous separation process is formed. By optimizing the operation parameters in real time during the disturbance process to reach the optimal value, not only can the energy utilization efficiency of the rectification process be effectively improved, but also the performance of feedforward neural network-assisted optimization control in the thermal integration process is verified.

Summary of the Invention

[0005] [Technical Problems to be Solved]

[0006] The purpose of the present invention is to provide a pressure swing rectification process for ethyl acetate - n-propanol - water based on neural network control, which realizes the optimized control of the energy-saving thermal integration process while obtaining high-purity products.

[0007] [Technical Solutions]

[0008] The present invention proposes a pressure-swing distillation process for ethyl acetate - n-propanol - water based on neural network control. By utilizing the characteristic that the azeotropic composition of the azeotrope changes with pressure, the pressure-swing distillation method is adopted to achieve the efficient separation of three binary azeotropes. In order to reduce the energy consumption during the disturbance process, the neural network is combined with dynamic control to predict the optimal operating parameters during the disturbance process in real time and effectively control the process. The proposed control structure can achieve the robust control of the process, effectively reducing the economic cost and energy consumption.

[0009] The operating pressure of the distillation column affects the azeotropic composition of the azeotrope, thus affecting the separation effect of the products. In the pressure-swing distillation process for ethyl acetate - n-propanol - water based on neural network control proposed by the present invention, the neural network prediction model is trained with the optimal process data within ±20% of the original feed composition. The optimization algorithm adopts the simulated annealing optimization algorithm, and the neural network adopts a feedforward neural network, including input, hidden layer, output layer and output. During the distillation process, the process parameters correspond to a certain feed flow rate, feed composition and product purity. The learning process takes these as input variables, and the reflux ratio of the high-pressure column T1, the heat load of the high-pressure column T1, the pressure of the high-pressure column T1, the reflux ratio of the medium-pressure column T2 and the pressure of the low-pressure column T3 are used as output variables for training. After training, there is a certain functional relationship between the input variables and the output variables, and a double-layer feedforward neural network prediction model is constructed. During the simulation operation process, the feed flow rate, feed composition and product purity are transmitted as input signals to the neural network prediction module NNPM. The neural network prediction module NNPM analyzes the collected data to predict the reflux ratio of the high-pressure column T1, the heat load of the high-pressure column T1, the pressure of the high-pressure column T1, the reflux ratio of the medium-pressure column T2 and the pressure of the low-pressure column T3 in the process, and inputs the analyzed data into the corresponding controller to control the dynamic process, and then the process operation data is cyclically input into the neural network prediction module NNPM.

[0010] The pressure-swing distillation process for ethyl acetate - n-propanol - water based on neural network control proposed by the present invention is characterized in that the process for separating the ethyl acetate - n-propanol - water azeotropic system is a heat-integrated three-column pressure-swing distillation process, and the dynamic control scheme is a real-time optimization and improvement control scheme combining traditional proportional-integral-derivative control and neural network prediction control. The process device includes the following parts:

[0011] High-pressure tower T1, medium-pressure tower T2, low-pressure tower T3, condenser C1, condenser C2, condenser C3, reboiler R1, reboiler R2, reboiler R3, valve V1, valve V2, valve V3, valve V4, valve V5, valve V6, valve V7, level controller LC1, level controller LC2, level controller LC3, level controller LC4, level controller LC5, level controller LC6, pressure controller PC1, pressure controller PC2, pressure controller PC3, temperature controller TC1, temperature controller TC2, flow controller FC, reflux ratio controller RR1, reflux ratio controller RR2, reflux ratio controller RR3, composition controller CC, reflux drum M1, reflux drum M2, reflux drum M3, pressure pump P1, pressure pump P2, pressure pump P3, pressure pump P4, pressure pump P5, pressure pump P6, pressure pump P7, neural network predictive control module NNPM;

[0012] The pressure-swing distillation process of ethyl acetate - n-propanol - water based on neural network control mainly includes the following steps:

[0013] The mixture of ethyl acetate - n-propanol - water and the stream from the top of the low-pressure tower T3 enter the high-pressure tower T1. The reboiler R1 heats the bottom liquid of the high-pressure tower T1. Part of the stream returns to the bottom of the high-pressure tower T1, and the other part of the stream is withdrawn as high-purity ethyl acetate through the pressure pump P3 and the valve V3. The overhead distillate of the high-pressure tower T1 enters the reflux drum M1 after being condensed by the condenser C1. Part of the stream is refluxed to the high-pressure tower T1, and the other part of the stream enters the medium-pressure tower T2 through the pressure pump P2 and the valve V2; The bottom stream of the medium-pressure tower T2 is heated by the reboiler R2. Part of it returns to the bottom of the medium-pressure tower T2, and the other part is withdrawn as high-purity n-propanol through the pressure pump P5 and the valve V5. The overhead vapor enters the hot stream inlet of the reboiler R3 from the overhead gas outlet of the tower and exchanges heat with the bottom liquid phase of the low-pressure tower T3. The exchanged stream enters the condenser C2, and after being condensed by the condenser C2, it enters the reflux drum M2. Part of the material in the reflux drum M2 is refluxed into the medium-pressure tower T2, and the other part is transported through the pressure pump P4 and the valve V4 and enters the low-pressure tower T3; After the bottom liquid mixture of the low-pressure tower T3 exchanges heat with the overhead gas of the medium-pressure tower T2 through the reboiler R3, part of the stream returns to the bottom of the low-pressure tower T3, and the other part of the stream is withdrawn as high-purity water through the pressure pump P7 and the valve V7. The overhead distillate enters the reflux drum M3 after being condensed by the condenser C3. Part of it is refluxed to the low-pressure tower T3, and the other part enters the high-pressure tower T1 through the pressure pump P6 and the valve V6;

[0014] The improved control structure for real-time optimization based on the above steady-state process includes the following steps:

[0015] The flow controller FC controls the feed rate of the high-pressure tower T1 by adjusting the opening degree of the valve V1, and the flow controller FC is of reverse action; the top pressure controller PC1 of the high-pressure tower T1 controls the top pressure by controlling the input heat load of the condenser C1, and the pressure controller PC1 is of reverse action; the top pressure controller PC2 of the medium-pressure tower T2 controls the top pressure by controlling the input heat load of the condenser C2, and the pressure controller PC2 is of reverse action; the top pressure controller PC3 of the low-pressure tower T3 controls the top pressure by controlling the input heat load of the condenser C3, and the pressure controller PC3 is of reverse action; the level controllers LC1 of the reflux drum M1 and the level controller LC2 of the bottom of the high-pressure tower T1 maintain the level stability by adjusting the opening degrees of the valves V2 and V3 respectively, the level controllers LC3 of the reflux drum M2 and the level controller LC4 of the bottom of the medium-pressure tower T2 maintain the level stability by adjusting the opening degrees of the valves V4 and V5 respectively, the level controllers LC5 of the reflux drum M3 and the level controller LC6 of the bottom of the low-pressure tower T3 maintain the level stability by adjusting the opening degrees of the valves V6 and V7 respectively, and the level controllers LC1, LC2, LC3, LC4, LC5 and LC6 are all of forward action; the reflux ratio controller RR1 controls the ratio between the reflux flow and the top product withdrawal of the high-pressure tower T1, the reflux ratio controller RR2 controls the ratio between the reflux flow and the top product withdrawal of the medium-pressure tower T2, and the reflux ratio controller RR3 controls the ratio between the reflux flow and the top product withdrawal of the low-pressure tower T3; the temperature controller TC1 controls the sensitive plate temperature of the medium-pressure tower T2 by adjusting the reboiler heat load of the medium-pressure tower T2; the composition controller CC is connected in series with the temperature controller TC2 and then in series with the reflux ratio controller RR3. By detecting the purity of water in the bottom stream, the signal input to the reflux ratio controller RR3 is manipulated to control the sensitive plate temperature of the low-pressure tower T3 so that the purity of the product water meets the standard; the neural network predictive control module NNPM predicts the heat load of the high-pressure tower T1, the reflux ratio of the high-pressure tower T1, the pressure of the high-pressure tower T1, the reflux ratio of the medium-pressure tower T2 and the pressure of the low-pressure tower T3 according to the input signals of the feed flow rate F, the ethyl acetate feed composition Fx EA , the n-propanol feed composition Fx NP , the water feed composition Fx H2O , the ethyl acetate product purity x EA , the n-propanol product purity x NP and the water product purity x H2O and respectively inputs them into the reboiler R1, the reflux ratio controller RR1, the pressure controller PC1, the reflux ratio controller RR2 and the pressure controller PC3.

[0016] The pressure swing distillation process of ethyl acetate - n - propanol - water based on neural network control proposed by the present invention realizes process intensification by constructing a "data acquisition - prediction optimization - dynamic regulation" closed - loop system: dynamically operating data such as the temperature, pressure, and component concentration of the distillation column are collected in real - time and input into the neural network prediction module (NNPM) for multi - objective parameter optimization to generate the best combination of operating parameters, ensuring that the process operates under the best operating conditions to achieve real - time optimization and control of the process.

[0017] According to another preferred embodiment of the present invention, it is characterized in that: the reflux ratio of the high - pressure tower T1 is 0.8 - 1.2, the reflux ratio of the medium - pressure tower T2 is 1.3 - 1.7, and the reflux ratio of the low - pressure tower T3 is 0.7 - 1.1. The pressure of the high - pressure tower T1 is 10.8 - 11.2 atm, the pressure of the medium - pressure tower T2 is 2 atm, and the pressure of the low - pressure tower T3 is 0.1 - 0.3 atm.

[0018] According to another preferred embodiment of the present invention, it is characterized in that: the number of trays of the high - pressure tower T1 is 70, the feed location is the 40th tray, and the sensitive tray location is the 16th tray; the number of trays of the medium - pressure tower T2 is 70, the feed location is the 40th tray, and the sensitive tray location is the 65th tray; the number of trays of the low - pressure tower T3 is 70, the feed location is the 35th tray, and the sensitive tray location is the 36th tray; the location of the top stream of the low - pressure tower T3 returning to the feed tray of the high - pressure tower T1 is the 15th tray.

[0019] According to another preferred embodiment of the present invention, it is characterized in that: the mass fraction of ethyl acetate after separation is greater than 99.9%; the mass fraction of n - propanol after separation is greater than 99.9%; the mass fraction of water after separation is greater than 99.9%.

[0020] According to another preferred embodiment of the present invention, it is characterized in that: after the top steam of the medium - pressure tower T2 enters the hot stream inlet of the reboiler R3 from the top gas phase outlet and exchanges heat with the bottom liquid phase of the low - pressure tower T3, the heat load saved by the reboiler R3 of the low - pressure tower T3 is not less than 2000 kW.

[0021] [Advantageous Effects]

[0022] Compared with the prior art, the present invention has the following beneficial effects:

[0023] (1) Optimizing and controlling the heat - integrated pressure swing distillation process of ethyl acetate - n - propanol - water based on a feed - forward neural network realizes effective regulation of the energy - saving distillation process.

[0024] (2) The control method involved in the present invention can robustly control the feed component disturbance within 20%, and the purity of the separated product is above 99.9%. [Description of the Drawings]

[0025] Figure 1 It is a pressure-swing distillation process for ethyl acetate - n-propanol - water based on neural network control.

[0026] High-pressure tower T1, medium-pressure tower T2, low-pressure tower T3, condenser C1, condenser C2, condenser C3, reboiler R1, reboiler R2, reboiler R3, valve V1, valve V2, valve V3, valve V4, valve V5, valve V6, valve V7, liquid level controller LC1, liquid level controller LC2, liquid level controller LC3, liquid level controller LC4, position controller LC5, liquid level controller LC6, pressure controller PC1, pressure controller PC2, pressure controller PC3, temperature controller TC1, temperature controller TC2, flow controller FC, reflux ratio controller RR1, reflux ratio controller RR2, reflux ratio controller RR3, composition controller CC, reflux drum M1, reflux drum M2, reflux drum M3, pressure pump P1, pressure pump P2, pressure pump P3, pressure pump P4, pressure pump P5, pressure pump P6, pressure pump P7, neural network predictive control module NNPM.

Detailed implementation manner

[0027] The following is further described with reference to the drawings, which does not limit the scope involved in the present invention.

[0028] Embodiment 1:

[0029] Utilize Figure 1The shown flow chart, with a feed flow rate of 6840 kg / h, a feed temperature of 25 °C, and a feed composition of 77% ethyl acetate, 17% n-propanol, and 6% water. Among them, the number of theoretical plates of the high-pressure tower T1 is 70, the operating pressure under steady state is 11 atm, the feed is from the 40th plate, the location of the sensitive plate is the 16th plate, the position where the overhead stream of the low-pressure tower T3 returns to the feed plate of the high-pressure tower T1 is the 15th tray. The mixture feed liquid is pressurized by the pressure pump P1 and then enters the high-pressure tower T1 through the valve V1. Under steady state, the reflux ratio is 1. The overhead distillate enters the reflux drum M1 after being condensed by the condenser C1. Part of the stream is refluxed to the high-pressure tower T1, and the other part of the stream enters the medium-pressure tower T2 through the pressure pump P2 and the valve V2. High-purity ethyl acetate is withdrawn from the bottom through the pressure pump P3 and the valve V3; the operating pressure of the medium-pressure tower T2 is 2 atm, the number of theoretical plates is 70, the feed location is the 40th plate, the location of the sensitive plate is the 65th plate, under steady state, the reflux ratio is 1.5. Part of the bottom stream of the medium-pressure tower T2 returns to the bottom of the medium-pressure tower T2, and the other part is withdrawn as high-purity n-propanol through the pressure pump P5 and the valve V5. The overhead steam enters the hot stream inlet of the reboiler R3 from the overhead gas outlet of the tower and exchanges heat with the bottom liquid phase of the low-pressure tower T3. Part of the heat-exchanged stream is refluxed into the medium-pressure tower T2, and the other part is transported through the pressure pump P4 and the valve V4 and enters the low-pressure tower T3; the operating pressure of the low-pressure tower T3 under steady state is 0.2 atm, the number of theoretical plates is 70, the feed location is the 35th plate, the location of the sensitive plate is the 36th plate, under steady state, the reflux ratio is 0.9. The overhead distillate of the low-pressure tower T3 is condensed by the condenser C3, and part of it is refluxed to the low-pressure tower T3, and part of it enters the high-pressure tower T1 through the pressure pump P6 and the valve V6. After the bottom liquid mixture of the low-pressure tower T3 exchanges heat with the overhead gas phase of the medium-pressure tower T2 through the reboiler R3, part of it returns to the bottom of the low-pressure tower T3, and the other part of the stream is withdrawn as high-purity water through the pressure pump P7 and the valve V7; after stabilizing for two hours, the feed composition of ethyl acetate is increased to 92%, the feed composition of n-propanol is 4%, and the feed composition of water is 4%. After the disturbance returns to stability, the pressure of the high-pressure tower T1 is 11.2 atm, the reflux ratio of the high-pressure tower T1 is 1.2, the pressure of the medium-pressure tower T2 is 2 atm, the reflux ratio of the medium-pressure tower T2 is 1.7, the pressure of the low-pressure tower T3 is 0.3 atm, the reflux ratio of the low-pressure tower T3 is 1.1. After separation, the mass fraction of the ethyl acetate product is greater than 99.9%, the mass fraction of the n-propanol product is greater than 99.9%, the mass fraction of the water product is greater than 99.9%, and the heat load saved by the reboiler R3 of the low-pressure tower T3 is 2000 kW.

[0030] Example 2:

[0031] Utilize Figure 1The shown flow chart: The feed flow rate is 8,000 kg / h, the feed temperature is 25 °C, and the feed composition is 77% ethyl acetate, 17% n-propanol, and 6% water. Among them, the number of theoretical plates of the high-pressure tower T1 is 70, the operating pressure under steady state is 11 atm, the feed is from the 40th plate, and the location of the sensitive plate is the 16th plate. The top stream of the low-pressure tower T3 returns to the feed plate position of the high-pressure tower T1, which is the 15th plate. The mixture feed liquid is pressurized by the pressure pump P1 and then enters the high-pressure tower T1 through the valve V1. Under steady state, the reflux ratio is 1. The overhead distillate enters the reflux drum M1 after being condensed by the condenser C1. Part of the stream is refluxed to the high-pressure tower T1, and the other part of the stream passes through the pressure pump P2 and the valve V2 and enters the medium-pressure tower T2. The high-purity ethyl acetate is withdrawn from the bottom through the pressure pump P3 and the valve V3; the operating pressure of the medium-pressure tower T2 is 2 atm, the number of theoretical plates is 70, the feed position is the 40th plate, the location of the sensitive plate is the 65th plate, and the reflux ratio under steady state is 1.5. Part of the bottom stream of the medium-pressure tower T2 returns to the bottom of the medium-pressure tower T2, and the other part is withdrawn as high-purity n-propanol through the pressure pump P5 and the valve V5. The top steam enters the hot stream inlet of the reboiler R3 from the top gas outlet and exchanges heat with the bottom liquid phase of the low-pressure tower T3. After heat exchange, part of the stream is refluxed into the medium-pressure tower T2, and the other part is transported through the pressure pump P4 and the valve V4 and enters the low-pressure tower T3; the operating pressure of the low-pressure tower T3 under steady state is 0.2 atm, the number of theoretical plates is 70, the feed position is the 35th plate, the location of the sensitive plate is the 36th plate, and the reflux ratio under steady state is 0.9. The overhead distillate of the low-pressure tower T3 is condensed by the condenser C3, and part of it is refluxed to the low-pressure tower T3, and part of it passes through the pressure pump P6 and the valve V6 and enters the high-pressure tower T1. After the bottom liquid mixture of the low-pressure tower T3 exchanges heat with the top gas phase of the medium-pressure tower T2 through the reboiler R3, part of it returns to the bottom of the low-pressure tower T3, and the other part of the stream is withdrawn as high-purity water through the pressure pump P7 and the valve V7; After stabilizing for two hours, the feed composition of ethyl acetate is reduced to 62%, the feed composition of n-propanol is 19%, and the feed composition of water is 19%. After the disturbance returns to steady state, the pressure of the high-pressure tower T1 is 10.8 atm, the reflux ratio of the high-pressure tower T1 is 0.8, the pressure of the medium-pressure tower T2 is 2 atm, the reflux ratio of the medium-pressure tower T2 is 1.3, the pressure of the low-pressure tower T3 is 0.1 atm, the reflux ratio of the low-pressure tower T2 is 0.7. After separation, the mass fraction of the ethyl acetate product is greater than 99.9%, the mass fraction of the n-propanol product is greater than 99.9%, the mass fraction of the water product is greater than 99.9%, and the heat load saved by the reboiler R3 of the low-pressure tower T3 is 2,340 kW.

[0032] Example 3:

[0033] Utilize Figure 1The shown flow chart has a feed flow rate of 8500 kg / h, a feed temperature of 25 °C, and a feed composition of 77% ethyl acetate, 17% n-propanol, and 6% water. Among them, the high-pressure tower T1 has 70 theoretical plates, an operating pressure of 11 atm under steady state, feeds from the 40th plate, and the sensitive plate is at the 16th plate. The top stream of the low-pressure tower T3 returns to the feed plate position of the high-pressure tower T1, which is the 15th plate. The mixture feed liquid is pressurized by the pressure pump P1 and enters the high-pressure tower T1 through the valve V1. Under steady state, the reflux ratio is 1. The overhead distillate enters the reflux drum M1 after being condensed by the condenser C1. Part of the stream is refluxed to the high-pressure tower T1, and the other part passes through the pressure pump P2 and the valve V2 and enters the medium-pressure tower T2. High-purity ethyl acetate is withdrawn from the bottom through the pressure pump P3 and the valve V3; the medium-pressure tower T2 has an operating pressure of 2 atm, 70 theoretical plates, a feed position of the 40th plate, and a sensitive plate at the 65th plate. Under steady state, the reflux ratio is 1.5. Part of the bottom stream of the medium-pressure tower T2 returns to the bottom of the medium-pressure tower T2, and the other part withdraws high-purity n-propanol through the pressure pump P5 and the valve V5. The top steam enters the hot stream inlet of the reboiler R3 from the top gas outlet of the tower and exchanges heat with the bottom liquid phase of the low-pressure tower T3. After heat exchange, part of the stream is refluxed into the medium-pressure tower T2, and the other part is transported through the pressure pump P4 and the valve V4 and enters the low-pressure tower T3; the low-pressure tower T3 has an operating pressure of 0.2 atm under steady state, 70 theoretical plates, a feed position of the 35th plate, and a sensitive plate at the 36th plate. Under steady state, the reflux ratio is 0.9. The overhead distillate of the low-pressure tower T3 enters the reflux drum M1 after being condensed by the condenser C3. Part of the stream is refluxed to the low-pressure tower T3, and part passes through the pressure pump P6 and the valve V6 and enters the high-pressure tower T1. The bottom liquid phase mixture of the low-pressure tower T3 exchanges heat with the top gas phase of the medium-pressure tower T2 through the reboiler R3. Part of it returns to the bottom of the low-pressure tower T3, and the other part of the stream withdraws high-purity water through the pressure pump P7 and the valve V7; after two hours of stabilization, the feed composition of ethyl acetate is increased to 72%, the feed composition of n-propanol is 14%, and the feed composition of water is 14%. After the disturbance returns to stability, the pressure of the high-pressure tower T1 is 10.93 atm, the reflux ratio of the high-pressure tower T1 is 0.93, the pressure of the medium-pressure tower T2 is 2 atm, the reflux ratio of the medium-pressure tower T2 is 1.43, the pressure of the low-pressure tower T3 is 0.16 atm, and the reflux ratio of the low-pressure tower T3 is 0.83. After separation, the mass fraction of the ethyl acetate product is greater than 99.9%, the mass fraction of the n-propanol product is greater than 99.9%, the mass fraction of the water product is greater than 99.9%, and the heat load saved by the reboiler R3 of the low-pressure tower T3 is 2485 kW.

Claims

1. A pressure-swing distillation process for ethyl acetate - n-propanol - water based on neural network control, characterized in that The device for implementing this process mainly includes the following parts: High-pressure tower T1, medium-pressure tower T2, low-pressure tower T3, condenser C1, condenser C2, condenser C3, reboiler R1, reboiler R2, reboiler R3, valves V1, V2, V3, V4, V5, V6, V7, level controllers LC1, LC2, LC3, LC4, LC5, LC6, pressure controllers PC1, PC2, PC3, temperature controllers TC1, TC2, flow controller FC, reflux ratio controllers RR1, RR2, RR3, composition controller CC, reflux drums M1, M2, M3, pressure pumps P1, P2, P3, P4, P5, P6, P7, neural network predictive control module NNPM; Using the above device for the pressure swing distillation process of ethyl acetate - n-propanol - water based on neural network control, its steady-state process includes the following steps: The mixture of ethyl acetate - n-propanol - water and the stream from the top of the low-pressure tower T3 enter the high-pressure tower T1. The reboiler R1 heats the bottom liquid of the high-pressure tower T1. Part of the stream returns to the bottom of the high-pressure tower T1, and another part of the stream is withdrawn as high-purity ethyl acetate through the pressure pump P3 and valve V3. The overhead distillate of the high-pressure tower T1 enters the reflux drum M1 after being condensed by the condenser C1. Part of the stream refluxes to the high-pressure tower T1, and another part of the stream enters the medium-pressure tower T2 through the pressure pump P2 and valve V2; The bottom stream of the medium-pressure tower T2 is heated by the reboiler R2. Part of it returns to the bottom of the medium-pressure tower T2, and another part is withdrawn as high-purity n-propanol through the pressure pump P5 and valve V5. The overhead vapor of the medium-pressure tower T2 enters the hot stream inlet of the reboiler R3 from the overhead gas outlet and exchanges heat with the bottom liquid phase of the low-pressure tower T3. The exchanged stream enters the condenser C2 and then enters the reflux drum M2 after being condensed by the condenser C2. Part of the material in the reflux drum M2 refluxes into the medium-pressure tower T2, and another part is transported through the pressure pump P4 and valve V4 into the low-pressure tower T3; After the bottom liquid mixture of the low-pressure tower T3 exchanges heat with the overhead gas of the medium-pressure tower T2 through the reboiler R3, part of the stream returns to the bottom of the low-pressure tower T3, and another part of the stream is withdrawn as high-purity water through the pressure pump P7 and valve V7. The overhead distillate enters the reflux drum M3 after being condensed by the condenser C3. Part of it refluxes to the low-pressure tower T3, and another part enters the high-pressure tower T1 through the pressure pump P6 and valve V6; The process of realizing neural network control based on the above steady-state process includes the following steps: The flow controller FC controls the feed rate of the high-pressure tower T1 by adjusting the opening degree of the valve V1, and the flow controller FC is of reverse action; the top pressure controller PC1 of the high-pressure tower T1 controls the top pressure by controlling the input heat load of the condenser C1, and the pressure controller PC1 is of reverse action; the top pressure controller PC2 of the medium-pressure tower T2 controls the top pressure by controlling the input heat load of the condenser C2, and the pressure controller PC2 is of reverse action; the top pressure controller PC3 of the low-pressure tower T3 controls the top pressure by controlling the input heat load of the condenser C3, and the pressure controller PC3 is of reverse action; the level controllers LC1 of the reflux drum M1 and the level controller LC2 of the reboiler of the high-pressure tower T1 maintain the stability of the liquid level by adjusting the opening degrees of the valves V2 and V3 respectively, the level controllers LC3 of the reflux drum M2 and the level controller LC4 of the reboiler of the medium-pressure tower T2 maintain the stability of the liquid level by adjusting the opening degrees of the valves V4 and V5 respectively, the level controllers LC5 of the reflux drum M3 and the level controller LC6 of the reboiler of the low-pressure tower T3 maintain the stability of the liquid level by adjusting the opening degrees of the valves V6 and V7 respectively, and the level controllers LC1, LC2, LC3, LC4, LC5 and LC6 are all of forward action; the reflux ratio controller RR1 controls the ratio between the reflux rate and the top product withdrawal rate of the high-pressure tower T1, the reflux ratio controller RR2 controls the ratio between the reflux rate and the top product withdrawal rate of the medium-pressure tower T2, and the reflux ratio controller RR3 controls the ratio between the reflux rate and the top product withdrawal rate of the low-pressure tower T3; the temperature controller TC1 controls the sensitive plate temperature of the medium-pressure tower T2 by adjusting the reboiler heat load of the medium-pressure tower T2; the composition controller CC is connected in series with the temperature controller TC2 and then in series with the reflux ratio controller RR3. By detecting the purity of water in the bottom stream, the signal input to the reflux ratio controller RR3 is manipulated to control the sensitive plate temperature of the low-pressure tower T3 so that the purity of the product water meets the standard; the neural network predictive control module NNPM predicts the heat load of the high-pressure tower T1, the reflux ratio of the high-pressure tower T1, the pressure of the high-pressure tower T1, the reflux ratio of the medium-pressure tower T2 and the pressure of the low-pressure tower T3 according to the input signals of the feed flow rate F, the ethyl acetate feed composition Fx EA , the n-propanol feed composition Fx NP , the water feed composition Fx H2O , the ethyl acetate product purity x EA , the n-propanol product purity x NP and the water product purity x H2O and respectively inputs them into the reboiler R1, the reflux ratio controller RR1, the pressure controller PC1, the reflux ratio controller RR2 and the pressure controller PC3.

2. The pressure-swing distillation process of ethyl acetate - n-propanol - water based on neural network control according to claim 1, characterized in that: The reflux ratio of the high-pressure tower T1 is 0.8 - 1.2, the reflux ratio of the medium-pressure tower T2 is 1.3 - 1.7, and the reflux ratio of the low-pressure tower T3 is 0.7 - 1.

1. The pressure of the high-pressure tower T1 is 10.8 - 11.2 atm, the pressure of the medium-pressure tower T2 is 2 atm, and the pressure of the low-pressure tower T3 is 0.1 - 0.3 atm.

3. The pressure-swing distillation process of ethyl acetate - n-propanol - water based on neural network control according to claim 1, wherein: The number of trays of high-pressure tower T1 is 70, the feed location is the 40th tray, and the sensitive tray location is the 16th tray; the number of trays of medium-pressure tower T2 is 70, the feed location is the 40th tray, and the sensitive tray location is the 65th tray; the number of trays of low-pressure tower T3 is 70, the feed location is the 35th tray, and the sensitive tray location is the 36th tray; the return location of the overhead stream of low-pressure tower T3 to high-pressure tower T1 is the 15th tray of the feed plate location.

4. The pressure-swing distillation process of ethyl acetate - n-propanol - water based on neural network control according to claim 1, characterized in that: The mass fraction of ethyl acetate after separation is greater than 99.9%; the mass fraction of n-propanol after separation is greater than 99.9%; the mass fraction of water after separation is greater than 99.9%.

5. The pressure-swing distillation process of ethyl acetate - n-propanol - water based on neural network control according to claim 1, characterized in that: After the overhead steam of medium-pressure tower T2 enters the hot stream inlet of reboiler R3 from the overhead gas outlet and exchanges heat with the bottom liquid phase of low-pressure tower T3, the heat load saved by reboiler R3 of low-pressure tower T3 is not less than 2000 kW.

Images