Control method of air separator

【0001】

[Technical field to which the invention belongs]

The present invention relates to a method for controlling an air separation device that collects a product of oxygen, nitrogen, and argon, or a combination thereof, as a product.

【0002】

[Conventional technology]

Conventionally, in industries that consume a large amount of industrial gases such as oxygen, nitrogen, and argon, gas holders and liquefied gases are used in air separation devices as a countermeasure against frequent and large fluctuations in demand that exceed the capacity of supply equipment from gas users. A method equipped with a vaporizing backup device is adopted. For example, in response to a large increase request, product gas is replenished from a gas holder or backup device, and conversely, when demand decreases, product gas is released.

【0003】

However, in the gas holder method, in order to secure the supply pressure of the product when the demand increases, the discharge pressure of the compressor for supplying the product must be set higher than necessary for operation, resulting in a large waste of power. There is a problem in terms of energy saving. On the other hand, in the case of the method using a backup device, the liquefied gas is produced, taken out from the storage tank in which the liquefied gas is stored, vaporized and pressure-fed, and supplied. Since a vaporizer or the like is required, capital investment costs are high.

【0004】

Therefore, for these, a method of responding to fluctuations in the demand amount of product gas by operating the increase / decrease amount of the air separation device itself has been proposed. The increase / decrease operation operation in the air separation device is substantially the same as the increase / decrease operation in the distillation column. In this case, there is almost no delay in the change in the flow rate of the rising gas in the distillation column, whereas the flow rate of the falling liquid is almost the same. Changes with a time delay. Therefore, when the load is changed, the fluctuation of the product concentration can be suppressed by operating so that the amount of falling liquid changes faster than the amount of rising gas. Regarding this, the applicant has proposed a control method for an air separation device that enables rapid increase / decrease operation without using additional equipment such as a gas holder or a backup device (see, for example, Patent Document 1).

【0005】

However, in the above method, after a rapid increase / decrease operation operation, it takes a considerably long time for the air separation device to reach a steady state due to the unique characteristics of the air separation device. Therefore, when the increase / decrease operation is repeated in a short time, The product concentration may not be stable and the product concentration may exceed the set spec value. In order to stabilize the air separation device in the shortest possible time, it is effective to perform feedback control on the product concentration and the intermediate product concentration. Commonly used control methods include (1) PID control and (2) model predictive control (Mode1 Predictive Contro1, abbreviated as "MPC"). In PID control, it is difficult to obtain good control results when the quick response of the controlled object is poor (such as when the dead time is long) or when mutual interference between control variables is strong. Also, in model prediction control, it is not easy to obtain desirable control performance for all loops because mutual interference between control loops having different time constants reduces the prediction accuracy.

【0006】

By the way, the applicant forms a direct control loop for the argon concentration in the feed argon stream and the oxygen concentration in the crude argon stream for the air separation device whose product is oxygen, nitrogen, and argon, and the response is slow. We have proposed a control method that stabilizes the product concentration when the amount of sudden increase or decrease is achieved by configuring it with two 1 input / output control systems that ignore the influence of interference from the loop to the fast loop (Japanese Patent Application No. 2001-143820). The effectiveness of the proposed control method was confirmed by the simulation results using the simulator of the air separation device. However, when control that enables repeated operation changes in a short time (for example, repeated increase / decrease operation in a 2-hour cycle) is required, this control method may not be able to handle it.

【0007】

In addition, an example of using the general model predictive control (abbreviated as "GPC") method for controlling the product purity of an air separation device has already been described by Shigeyuki Tani et al., "Measurement and Control No. 39". It is reported in the literature of Vol. 5, pp. 343-345 (2000), and in the literature of "Automatica Vol. 23, pp. 2, pp. 137-160 (1987)" by DW Clark et al. The outline will be described below.

【0008】

In the generalized model control (GPC) theory in these documents, the dynamic characteristics of the plant are expressed by the following CARIMA (Controlled Auto-Regressive and Integrated Moving-Average) model. A (z-¹) y (t) = B (z-¹) z-^Lu (t-1) + d (t) (1) Here, A (z-¹) Is the target set operating state, B (z-¹) Is the current operating state, u (t-1) is the manipulated variable, y (t) is the controlled variable, d (t) is the noise, and L is the wasted time.

【0009】

Consider the following evaluation function (2) for this plant.

[Number 1]<img file="JP2004163003A_D0001.tif" />Where N₁Is the minimum rating horizon, N₂Is the highest rating horizon, N₃Is a control horizon.

【0010】

In generalized model predictive control (GPC), the manipulated variable u that minimizes this objective function J is determined based on the predicted value of y. The general theory for this has been studied for decades as described above, and general-purpose soft packages are also commercially available.

【0011】

In the literature by Shigeyuki Tani et al., Who applied this GPC to an air separation device, the manipulated variable u is: 1 Raw material air flow rate 2 Expansion turbine flow rate 3 Product oxygen extraction flow rate 4 Liquid nitrogen recirculation flow rate ( Valve opening) 5 Crude argon tower differential pressure 6 Control amount y using crude argon extraction flow rate 1 Product oxygen concentration 2 Product nitrogen concentration 3 Crude argon concentration 4 Product oxygen Purge amount 5 Expansion turbine / raw material air amount ratio 6 Argon yield. In this example, a certain effect is obtained in stabilizing the product purity during steady operation, but the control algorithm is complicated due to the large number of instrumental variables, and many control loops with different time constants are controlled in the same manner. Since the amount of operation is determined by the instrument, when the load changes significantly, mutual interference between the loops tends to adversely affect the control accuracy.

【0012】

[Patent Document 1]

Japanese Patent No. 3027368 [0013]

[Problems to be Solved by the Invention]

However, when repeating the increase / decrease operation of the air separation device in a short time, if the next increase / decrease operation is performed before the product concentration becomes stable, the purity of the product cannot be maintained, so the product concentration is stabilized in the shortest possible time. However, it is difficult to obtain good control with conventional PID control. Moreover, even if model predictive control is applied, the control calculation becomes complicated because there are many instrumental variables. That is, since the operation amount of many control loops having different time constants is determined by the same controller, sufficient control accuracy cannot be obtained when the load changes significantly. Moreover, it is not easy to obtain the desired overall control performance for each loop. In addition, the argon concentration in the feed argon flow and the oxygen concentration in the coarse argon flow shall be configured by two 1 input / output control systems that directly configure the control loop and ignore the interference effect from the slow response loop to the fast loop. Therefore, even in the case of a control method that stabilizes the product concentration when the amount is suddenly increased or decreased, if control that enables repeated operation changes in a short time is required, this control method may not be able to handle it. There were various problems.

【0014】

The present invention has been made in view of such a situation, and a control method capable of a large and frequent increase / decrease operation operation in an air separation device without using additional equipment such as a gas holder type or backup equipment. It is what makes it possible to provide.

【0015】

[Means for solving problems]

In order to solve the above problem, as an invention according to claim 1, in an air separation device for producing oxygen, nitrogen, and argon products by separating and collecting air components from raw material air by low temperature rectification, all the products It is possible to control the purity of all products by optimizing only the argon concentration in feed argon, the oxygen concentration in crude argon, and the product oxygen concentration based on the predicted values without using the concentration control loop directly. This is a characteristic control method for the air separation device. As an invention according to claim 2, the optimization control based on the predicted value is a prediction operation based on an approximate model that approximates the dynamic characteristic of the concentration control loop as a first-order lag or second-order lag characteristic having a waste time and its prediction. The control method for the air separation device according to claim 1, wherein the air separation device is configured by an optimization calculation based on a value. As an invention according to claim 3, the air separation device includes at least one of a crude argon column and a deoxidizing column, has a concentration control loop of 3 inputs and 3 outputs or more, and controls each by a multivariate prediction algorithm. For a control loop in which the output prediction of the loop is not performed at the same time and the time constant is small and the dead time is short, the output prediction value of the control loop is obtained in consideration of the influence on the prediction value of other control loop operations. This is the control method for the air separation device according to claim 1. As an invention according to claim 4, the air separation device includes at least one of a crude argon column and a deoxidizing column, has a concentration control loop of 3 inputs and 3 outputs or more, and controls each by a multivariate prediction algorithm. The first aspect of claim 1, wherein a control loop in which the output prediction of the loop is not performed at the same time and the time constant is large and the dead time is long is treated as a disturbance by treating the influence of the settling operation of the other control loop on the predicted value. This is a control method for the air separation device. The invention according to claim 5 is the control method for an air separation device according to any one of claims 1 to 4, wherein the invention is applied to a process in which argon is not collected. As an invention according to claim 6, the air separation device includes a distillation column. The method for controlling the air separation device according to any one of claims 1 to 4, further comprising a distillation selection and having a step of separating nitrogen as a product. The air separation according to claim 7, wherein the air separation device further includes a distillation column or a distillation section, and has a step of separating oxygen as a product. It is a control method for the device.

【0016】

BEST MODE FOR CARRYING OUT THE INVENTION

The present invention is composed of two parts as follows. (i) The operation of the air separation device shall be performed by a basic control loop such as flow rate control, pressure control, and liquid level control . (ii) On top of that, it further has the following product concentration control loop. (B) Low-dimensional mathematics in which the argon concentration in feed argon, the oxygen concentration in crude argon, and the product oxygen concentration are output, and the oxygen flow rate, crude argon flow rate, and reflux liquid nitrogen flow rate are control inputs for concentration stabilization control. Extracting a model. (B) Ignoring the influence of product oxygen concentration adjustment, and having a control rule for the argon concentration in feed argon based on model prediction control in which the interference term of oxygen concentration adjustment in crude argon is a disturbance. (C) Have a model prediction type oxygen concentration control rule in crude argon that ignores the effects of argon concentration adjustment in feed argon and product oxygen concentration. (D) Have a model prediction type product oxygen concentration control rule that ignores the effects of argon concentration adjustment in feed argon and oxygen concentration adjustment in crude argon.

【0017】

Hereinafter, an embodiment of an argon extraction plant to which the present invention is applied will be described. FIG. 1 shows a configuration example of an embodiment of an argon extraction plant to which the present invention is applied. As shown in the figure, in the present embodiment, the compressed raw material air AIR from which water and carbon dioxide gas have been removed enters the main heat exchanger 3 from the pipeline 1 through the control valve 2, and is a low-temperature fluid. It is cooled by and supplied to the bottom of the lower tower 4. The raw air AIR is distilled here and separated into oxygen-rich liquid air at the bottom of the tower and high-purity nitrogen at the top of the tower. A part of the high-purity nitrogen extracted from the conduit 5 at the top of the lower tower 4 is liquefied in the main condenser 7 at the bottom of the upper tower 9 and then led out to the conduit 8, and a part is taken out. Returning to the lower tower 4, the remaining liquid nitrogen is cooled by the supercooler 13 via the conduit 12, then expanded by the control valve 14 and supplied to the top of the upper tower 9. The remaining medium-pressure nitrogen gas is divided into two through the pipeline 10, one of which is heated by the main heat exchanger 3 and the cold box is led out through the control valve 11 as the medium-pressure product nitrogen gas MGN2. The other medium-pressure nitrogen gas is heated by the main heat exchanger 3 through the pipeline 20, further heated by the turbine heat exchanger 21, pressurized by the turbine blower 22, and then cooled by the cooler 23 and the turbine heat exchanger. After being cooled by 21, it expands by the expansion turbine 24, merges with the waste nitrogen gas from the upper tower 9, is heated by the main heat exchanger 3 through the pipeline 29, and becomes the waste nitrogen gas RN2, which is a cold box. Is derived via the control valve 30. The liquid air extracted from the bottom of the lower tower 4 is cooled by the supercooler 13 and then divided into two parts, the liquid air LAIR1 and the liquid air LAIR2 for the argon condenser. One of the liquid air LAIR1 is the conduit 15. It is expanded by the control valve 16 via the control valve 16 and supplied to the intermediate portion of the upper tower 9. The other liquid air for argon condenser LAIR2 is expanded by the control valve 18 through the conduit 17, heated by the argon condenser 19, and introduced into the portion above the position of the outlet pipe 35 of the feed argon stage of the upper tower 9. To.

【0018】

The upper tower 9 is a tower in which the main condenser 7 is arranged at the bottom of the tower, and is composed of the reflux liquid nitrogen RLN2 supplied from the lower tower 4 via the conduit 12 and the liquid air LAIR 1 branched into the conduit 15. , The liquid air LAIR2 for the argon condenser that has flowed through the argon condenser 19 through the conduit 17 is supplied to the upper tower 9 and rectified here. As a result, low-pressure product nitrogen gas GN2, waste nitrogen gas RN2 from pipe 28, product oxygen gas GO2 from pipe 31, and liquid oxygen LO2 from pipe 33 are separated and collected from the pipe 26 at the top of the tower. To. The low-pressure product nitrogen gas GN2 derived from the pipeline 26 and the waste nitrogen gas RN2 derived from the pipeline 28 are heated to room temperature by the supercooler 13 and the main heat exchanger 3 and exit the cold box. The product oxygen gas GO2 led out from the pipeline 31 of the upper tower 9 is heated to room temperature by the main heat exchanger 3 and led out from the cold box via the control valve 32. Further, the liquid oxygen LO2 led out from the bottom of the upper tower 9 via the conduit 33 is cooled by the supercooler 13 and sent to the liquid oxygen tank (not shown) via the valve 34.

【0019】

The crude argon column 36 is a column in which an argon condenser 19 is arranged at the top of the column, and the feed argon gas FAr supplied from the upper column 9 via the conduit 35 is enriched with argon in the crude argon column 36. Separated into crude argon gas RAr. A part of the crude argon gas RR led out from the top of the crude argon column 36 in the conduit 38 is branched into the conduit 40 and heated by the main heat exchanger 3, and then the cold box is led out through the valve 43. .. The remaining crude argon gas RAr is liquefied in the crude argon column condenser 19 via the conduit 39 and returned so as to be supplied to the top of the crude argon column 36. The oxygen-enriched liquid (liquid argon) LRAr containing argon from the bottom of the crude argon column 36 is returned to the upper column 9 by the conduit 37.

【0020】

In FIG. 1, reference numeral 45 is a raw material air flow rate controller, 46 is an argon gas flow rate controller, 47 is a product oxygen flow rate controller, 48 is a product nitrogen gas flow rate controller, and 49 is a medium pressure product nitrogen gas flow rate controller. , 50 is the pressure regulator of the upper tower 9, 51 is the liquid nitrogen flow regulator returning to the upper tower 9, 52 is the liquid level regulator of the lower tower 4, 53 is the liquid air flow controller, 54 is the upper tower 9. The liquid level controller, 55 is a liquid air flow controller for an argon condenser, and 56 is a liquid oxygen flow controller. And such an air separation device is generally operated at a pressure of 120 to 160 kPa for the upper tower 9 and 450 to 600 kPa for the lower tower 4, but at a pressure higher or lower than these pressures. It is also possible to operate. Further, in the case of load fluctuation, in order to balance the substance of the entire air separation device, the set value of each corresponding flow rate controller is optimized and operated.

【0021】

Next, the control method proposed by the present invention will be described. In order to simplify and enable the control algorithm, the present invention does not form a feedback loop for all control variables, but feeds the control algorithm as being most important for plant operation and capable of continuous measurement. The argon concentration in argon, the oxygen concentration in crude argon, and the product oxygen concentration are selected as control quantities, and the product oxygen flow rate, crude argon flow rate, and reflux liquid nitrogen flow rate are selected as operation quantities.

【0022】

In addition, it is difficult to describe mathematically accurately and completely because the air separation device has a long waste time and each element influences each other in a complicated manner. Here, we focus only on the main dynamic characteristics of the plant. However, the mathematical model is made into a simple structure by expressing the dynamic characteristics of the plant as the first-order lag and second-order lag characteristics with wasted time, and considering the effects of the ignored elements as disturbances.

【0023】

Therefore, Fig. 2 shows the model structure of the controlled object by this mathematical model display. The symbols and formulas in Fig. 2 indicate the following. U₁(t): Product oxygen flow rate / u₂(t): Crude argon flow rate u₃(t): Reflux liquid nitrogen flow rate y_l(t): Argon concentration in feed argon y₂(t): Oxygen concentration in crude argon y₃(t): Product oxygen concentration d₁(t): Disturbance that affects the feed argon concentration d₂(t): Disturbance that affects the oxygen concentration in crude argon d₃(t): Disturbances that affect the product oxygen concentration u₁₂(t) = A₁₂-¹(z-¹) u₂(t) u₁₃(t) = A₁₃-¹(z-¹) u₃(t) u₂₁(t) = A₂₁-¹(z-¹) u₁(t) u₂₃(t) = A₂₃-¹(z-¹) u₃(t) u₃₁(t) = A₃₁-¹(z-¹) u₁(t) u₃₂(t) = A₃₂-¹(z-¹) u₂(t) [B₁₁(z-¹) / A₁₁(z-¹)] z-^L11: u₁(t) to y₁Transfer function to (t) · [B₁₂(z-¹) / A₁₂(z-¹)] z-^L12: u₂(t) to y₁Transfer function to (t) · [B₁₃(z-¹) / A₁₃(z-¹)] z-^L13: u₃(t) to y₁Transfer function to (t) · [B₂₁(z-¹) / A₂₁(z-¹)] z-^L21: u₁(t) to y₂Transfer function to (t) · [B₂₂(z-¹) / A₂₂(z-¹)] z-^L22: u₂(t) to y₂Transfer function to (t) · [B₂₃(z-¹) / A₂₃(z-¹)] z-^L23: u₃(t) to y₂Transfer function to (t) · [B₃₁(z-¹) / A₃₁(z-¹)] z-^L31: u₁(t) to y₃Transfer function to (t) · [B₃₂(z-¹) / A₃₂(z-¹)] z-^L32: u₂(t) to y₃Transfer function to (t) · [B₃₃(z-¹) / A₃₃(z-¹)] z-^L33: u₃(t) to y₃Transfer function to (t) [0024]

Next, the step response of the plant was calculated for each manipulated variable using a dynamic simulator, and the model of the controlled object (transfer function of each channel) was identified. Model prediction control was performed using the obtained identification model.

【0025】

From this simulation, the above-mentioned mathematical model is a 3-input, 3-output system. Based on this 3-input, 3-output mathematical model, the desired optimal control law can be derived by applying a multivariable model predictive control algorithm, but the time constants of each loop are significantly different, so the loops are inter-loop. It is generally difficult to obtain good control performance by mutual interference. For example, the dead time and time constant of the oxygen concentration control loop in the crude argon flow are significantly longer and larger than those of the argon concentration control loop in the field argon flow, so when both are handled at the same time, the latter settling time Even after the end, the former operation amount interferes, and a control deviation occurs in the argon concentration in the feed argon flow. Therefore, here, the interference term is treated as a disturbance input in each loop, and the single loop control side for each loop is designed.

【0026】

First, the product oxygen flow rate u₁Input, feed argon flow argon concentration y₁Is derived from the single-loop control law that outputs. Equation (1) can be expressed as shown in Fig. 2, so the output y of this loop₁Is given as in the following equation (3). A₁₁(z-¹) y₁(t) = B₁₁(z-¹) z-^L11u₁(t) + A₁₁(z-¹) B₁₂(z-¹) z-^L12u₁₂(t) + A₁₁(z-¹) B₁₃(z-¹) z-^L13u₁₃(t) + d₁(t) (3) However, u₁₂(t) = A₁₂-¹(z-¹) u₂(t) u₁₃(t) = A₁₃-¹(z-¹) u₃(t) represents the interference due to the crude argon flow rate and the interference due to the reflux liquid nitrogen flow rate, which are the manipulated variables, respectively.

【0027】

Where u₃Has a long settling time, so u₃Manipulation amount of y₁If you incorporate it into the prediction algorithm of₁There is a risk of lengthening the settling time and deteriorating the control accuracy. Therefore, here u₃Ignore the information about and consider the effects of this interference as a generalized disturbance. y₁Is given as the following equations (4) and (5). A₁₁(z-¹) y₁(t) = B₁₁(z-¹) z-^L11u₁(t) + A₁₁(z-¹) B₁₂(z-¹) z-^L12u₁₂(t) + d₁'(t) (4) Here, d₁'(t) = A₁₁(z-¹) B₁₃(z-¹) z-^L13u₁₃(t) + d₁(t).

【0028】

The optimum predicted value of y1 is given by the following equation (5).

[Number 2]<img file="JP2004163003A_D0002.tif" />Based on this predicted value, the control rule that minimizes the evaluation function (3) is as shown in the following equation (6). u₁= (R₁G₁^TG₁+ Q₁I)-¹G₁^TR₁(w₁-f₁) ・・・(6)【0029】

Next, the crude argon flow rate u₂Oxygen concentration in the crude argon stream from y₂Consider the loop of. From Fig. 2, y₂Is given as in the following equation (7). A₂₂(z-¹) y₂(t) = B₂₂(z-¹) z-^L22u₂(t) + A₂₂(z-¹) B₂₁(z-¹) z-^L21u₂₁(t) + A₂₂(z-¹) B₂₃(z-¹) z-^L23u₂₃(t) + d₂(t) (7) However, u₂₁(t) = A₂₁-¹(z-¹) u₁(t) u₂₃(t) = A₂₃-¹(z-¹) u₃(t) represents the interference due to the product oxygen flow rate and the interference due to the reflux liquid nitrogen flow rate, which are the manipulated variables, respectively.

【0030】

Where u₁And u₃Because the settling time is long, u₁And u₃Manipulation amount of y₂If you incorporate it into the prediction algorithm of₂There is a risk of lengthening the settling time and deteriorating the control accuracy. Therefore, here u₁, U₃Ignore the information about and consider the effects of this interference as a generalized disturbance. That is, y₂Is given by the following equation (8). A₂₂(z-¹) y₂(t) = B₂₂(z-¹) z-^L22u₂(t) + d₂'(t) (8) Here, d₂'(t) = A₂₂(z-¹) B₂₁(z-¹) z-^L21u₂₁(t) + A₂₂(z-¹) B₂₃(z-¹) z-^L23u₂₃(t) + d₂(t).

【0031】

y₂The optimum predicted value of is calculated by the following equation (9).

[Number 3]<img file="JP2004163003A_D0003.tif" />When the predictive control algorithm is applied, the control rule that minimizes the evaluation function (3) is as shown in the following equation (10). u₂= (R₂G₂^TG₂+ Q₂I)-¹G₂^TR₂(w₂-f₂) ・・・(10)【0032】

Finally, the reflux liquid nitrogen flow rate u₃From product oxygen concentration y₃Consider the loop of. From Fig. 2, y₃Is given as in the following equation (11). A₃₃(z-¹) y₃(t) = B₃₃(z-¹) z-^L33u₃(t) + A₃₃(z-¹) B₃₁z-^L31(z-¹) u₃₁(t) + A₃₃(z-¹) B₃₂(z-¹) z-^L32u₃₂(t) + d₃(t) (11) However, u₃₁(t) = A₃₁-¹(z-¹) u₁(t) u₃₂(t) = A₃₂-¹(z-¹) u₂(t) represents the interference due to the product oxygen flow rate and the interference due to the crude argon flow rate, which are the manipulated variables, respectively. Where u₁And u₂Because the settling time is long, u₁And u₂Manipulation amount of y₃If you incorporate it into the prediction algorithm of₃There is a risk of lengthening the settling time and deteriorating the control accuracy. Therefore, here u₁, U₂Ignore the information about and consider the effects of this interference as a generalized disturbance. That is, y₃Is given as in the following equation (12). A₃₃(z-¹) y₃(t) = B₃₃(z-¹) z-^L33u₃(t-1) + d₃'(t) (12) Here, d₃'(t) = A₃₃(z-¹) B₃₁z-^L31(z-¹) u₃₁(t) + A₃₃(z-¹) B₃₂(z-¹) z-^L32u₃₂(t) + d₃(t).

【0033】

y₃The optimum predicted value of is as shown in the following equation (13).

[Number 4]<img file="JP2004163003A_D0004.tif" />When the predictive control algorithm is applied, the control rule that minimizes the evaluation function (3) is obtained as in the following equation (14). u₃= (R₃G₃^TG₃+ Q₃I)-¹G₃^TR₃(w₃-f₃) ・・・(14)【0034】

Further, in order to confirm the effectiveness of the control method of the present invention, a dynamic simulation was carried out at a speed of 3% / sec (% / min) to repeatedly increase / decrease the amount of operation in a 2-hour cycle with a width of 30%. The change in the raw material air flow rate is shown by the solid line in Fig. 3. The change in argon concentration in feed argon is shown by the solid line in Fig. 4. The change in oxygen concentration in crude argon is shown by the solid line in Fig. 5. The change in nitrogen concentration in crude argon is shown by the solid line in Fig. 6. The change in product oxygen concentration is shown by the solid line in Fig. 7. The broken line in the figure shows a comparative example by the conventional control method. From the results of the simulation, according to the control method of the present invention, the increase in the product oxygen concentration and the nitrogen concentration in the crude argon is suppressed as compared with the conventional control method (Japanese Patent Application No. 2001-143820), and relatively good control performance is obtained. It was confirmed that it could be obtained.

【0035】

[Effect of the invention]

The control method of the air separation device of the present invention is carried out in the above-described embodiment, and has the following effects. That is, when operating the air separation device, it is based on the basic control loop of flow rate, pressure, liquid level, etc., and further model predictive control type product purity control is added, and in particular, it represents the stability of product purity. By directly feeding back the argon concentration in feed argon and the oxygen concentration in crude argon and performing the optimum control operation based on the predicted value, it is possible to quickly follow the load fluctuation and keep the product concentration within the allowable range. It is something that can be done.

【0036】

In addition, it has the following effects. (I) Compared with the control method based only on basic control such as flow rate control, pressure control, and liquid level control, it is possible to follow a large and sudden load change operation well, and a large and sudden load fluctuation is possible. (II) It can follow the repetitive fluctuation of the load in a short time, and the repetitive change of the load in a short time is possible. (III) During the load change and load change operation in (I) and (II) above, the fluctuation of the operating state of the device can be suppressed within the permissible range. (IV) Fluctuations in product concentration can be well suppressed even for unpredictable disturbance factors. (V) Energy saving can be expected. (VI) Since the buffer tank for product oxygen is not required, the capital investment cost can be reduced.

[Simple explanation of drawings]

FIG. 1 is a process system diagram of an air separation device.

FIG. 2 is a model structure diagram of a controlled object.

FIG. 3 is a graph of changes in raw material air flow rate.

FIG. 4 is a graph of changes in argon concentration during feed argon.

FIG. 5 is a graph of changes in oxygen concentration in crude argon.

FIG. 6 is a graph of changes in nitrogen concentration in crude argon.

FIG. 7 is a graph of changes in product oxygen concentration.

[Explanation of symbols]

1,5,6,8,10,12,15,17,20,25,26,28,29,31,33,35,37,38,39,40 Pipeline, 2,11,14 , 16,18,27,30,32,34,43 Control valve, 3 Main heat exchanger, 4 Lower tower, 7 Main condenser, 9 Upper tower, 13 ... supercooler, 19 ... argon condenser, 21 ... turbine heat exchanger, 22 ... turbine blower, 23 ... cooler, 24 ... expansion turbine, 36 ... coarse Argon tower, 45 ... Raw material air flow rate controller, 46 ... Coarse argon gas flow rate controller, 47 ... Product oxygen gas flow rate controller, 48 ... Low pressure product Nitrogen gas flow rate controller, 49 ... Medium pressure product Nitrogen gas flow regulator, 50 Pressure regulator of upper tower 9, 51 Liquid nitrogen flow regulator recirculated to upper tower 9, 52 Liquid level regulator of lower tower 4 , 53 Liquid air flow controller, 54 Liquid level controller of upper tower 9, 55 Liquid air flow controller for argon condenser, 56 Liquid oxygen flow controller, AIR Raw material air, GO2 Product oxygen gas, GN2 Low pressure product Nitrogen gas, RN2 Waste nitrogen gas, MGN2 Medium pressure product Nitrogen gas, LAIR1 Liquid air, LAIR2 Liquid air for argon condenser, ETN2 Nitrogen gas for expansion turbine, RLN2 Refluxing liquid nitrogen, FAr Feed argon, LO2 Liquid oxygen, RAr Crude argon gas, LRAr Oxygen-enriched liquid containing trace amounts of argon