JP2001067103A - Method and device for controlling process - Google Patents

Abstract

PROBLEM TO BE SOLVED: To prevent the increase of a calculation load in optimizing calculation and to continue model prediction control by monitoring the calculation load by means of the optimizing processing of an evaluating function and mitigating a restricting condition in accordance with the calculation load. SOLUTION: A restricting condition setting means 14 sets the searching range of a manipulated variable updating value in model prediction control. A manipulated variable optimizing means 15 receives an inequality concerning the evaluating function and the restricting condition and decides the optimum manipulated variable by control period through the use of a numeral optimizing algorithm. The means 15 monitors the calculation load concerning solution searching in the numeral optimizing algorithm. When the calculation load becomes a certain prescribed value, a trigger S156 is outputted. Besides, the means 15 judges whether an optimum solution exists or not. When a restricting condition mitigating means 16 receives the trigger S156, it executes a mitigating processing with respect to restricting condition mitigating data. The mitigating processing corresponds to the extension of the searching range of the optimum solution in the manipulated variable updating value.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a process control method and apparatus for optimally controlling a controlled object while predicting a future behavior by using a mathematical model expressing a dynamic characteristic of the controlled object. In particular, the present invention relates to a process control method and apparatus for accurately and stably controlling a plant that needs to consider a constraint condition related to an operation management constraint.

[0002]

2. Description of the Related Art In recent years, in the field of process control such as petrochemicals, a control method called model predictive control has been actively applied in order to realize optimal control while satisfying restrictions on operation control of a controlled object. It is.

In this control method, a future behavior of a controlled object is predicted using a prediction formula based on a dynamic characteristic model of a plant, and a deviation between the prediction result and a target value and an evaluation function relating to a future change of a manipulated variable are calculated. Then, an optimal operation amount for optimizing the evaluation function is calculated sequentially for each control cycle. This control method is described in IEICE Transactions C, Vol. 116, No. 10,
(1996), pp. 1089-1096.

When this control method is used, an operation is performed to determine the operation amount at the current time so that the aforementioned control amount future value approaches the target value under a constant operation amount as much as possible.

[0005] The above-mentioned prediction formula is expressed by a mathematical formula relating to the past control amount and operation amount. The prediction calculation is executed each time based on the past control amount and operation amount so that the future control amount value approaches the target value.

Subsequently, an evaluation function for model predictive control is set based on the prediction calculation result. In many cases, this evaluation function is expressed in the form of the sum of squares or the sum of absolute values of “deviation between control amount predicted value and target value” and “operated amount update value”. For the evaluation function, an optimal solution is obtained by applying a numerical optimization algorithm. This optimal solution corresponds to the manipulated variable update value, and the optimal manipulated variable is determined.

[0007] The above-described constraint condition relating to the control target is used as a search range for an optimum solution in the optimization calculation. In addition, as the optimization algorithm, a linear programming method, a quadratic programming method, or the like is known, and by applying these algorithms, an optimal operation amount update value can be derived.

[0008]

As described above, model predictive control can be said to be a suitable control method when a chemical process or the like is to be controlled.

However, many of the control targets in the process control are multivariable processes, and the control system becomes large-scale. Therefore, the above-described optimization problem including the combination of the evaluation function and the constraint condition has a very large scale, and it is expected that the calculation load of the optimization calculation will increase accordingly.

In the process control, a target value and a manipulated variable are often set at or near the above-mentioned constraint conditions in order to improve the operation efficiency. In such a situation, it is expected that the above-mentioned constraint condition will be temporarily deviated due to the influence of measurement noise. Accordingly, it is expected that the calculation load of the optimization calculation will increase.

On the other hand, the operation of the controller in the process control requires real-time performance, and an algorithm that sufficiently considers the calculation speed and the calculation load is required for the above-described optimization calculation.

The above prior art does not consider such calculation speed or calculation load, and has been a problem when applying model predictive control to an actual plant.

An object of the present invention is to apply the above-described model predictive control to an actual plant, since the control target is large-scale. An object of the present invention is to provide a process control method and apparatus capable of preventing an increase in the calculation load of the optimization calculation, which is expected by deviating from the above-described constraints, and continuing model prediction control.

[0014]

In order to achieve the above object, according to the first aspect of the present invention, a control object is optimized while predicting a future behavior by using a mathematical model expressing a dynamic characteristic of the control object. In the process control method, the future behavior of the control amount is predicted using the model, the control target satisfies the constraint conditions, and the operation amount for optimizing the evaluation function is derived. It is characterized in that it comprises means for monitoring the calculation load due to optimization processing, and means for relaxing the constraint conditions according to the calculation load.

According to the first aspect of the present invention, even if the calculation load of the above-described optimization calculation changes due to the operating condition of the process to be controlled, the change is monitored, and immediately after the increase in the calculation load is confirmed, Constraint conditions can be relaxed, leading to optimal solution derivation.

In order to achieve the above object, according to the second aspect of the present invention, the number of repetition operations in the optimization processing of the evaluation function is monitored as the above-mentioned calculation load, and the number of repetition operations exceeds a predetermined value. The present invention is characterized in that a means for requesting a process of relaxing the constraint condition is provided.

According to the second aspect of the present invention, by monitoring the number of repetition operations in the optimization calculation, it is possible to prevent the termination of the operation due to failure to satisfy the stop condition of the repetition operation during the search for the optimum solution. Can be.

In order to achieve the above object, according to the invention of claim 3, the computer occupancy of the optimization processing of the evaluation function is monitored as a calculation load, and when the occupancy exceeds a predetermined value, It is characterized by having a means for requesting a process of relaxing the constraint condition.

According to the third aspect of the invention, by monitoring the computer occupancy in the optimization calculation, it is possible to prevent the calculation efficiency of other control logic from being reduced by the optimization calculation.

In order to achieve the above object, in the invention according to the fourth aspect, at the time of receiving the mitigation request, a means for selecting a specific constraint condition from the constraint conditions on the control variables and the operation variables of the controlled object. It is characterized by having.

According to the fourth aspect of the present invention, by selecting a specific constraint condition, the number of constraint conditions is reduced, and the search area for the optimal solution is expanded, so that the optimal solution can be obtained without increasing the calculation load. Can be derived.

In order to achieve the above object, according to the invention of claim 5, at the time of receiving the mitigation request, means for extending a specific constraint range from constraints on control variables and operation variables of the controlled object. It is characterized by having.

According to the fifth aspect of the present invention, by expanding the specific constraint range, the search area for the optimum solution is expanded, and the optimum solution can be derived without increasing the calculation load.

In order to achieve the above object, according to the present invention, the above-mentioned monitoring of the calculation load and the existence determination of the optimal solution are executed, and when it is determined that the optimal solution does not exist, the constraint is set. It is characterized in that it has means for requesting a condition relaxation process.

According to the sixth aspect of the present invention, by monitoring the number of repetitive operations in the optimization calculation, it is possible to prevent a situation in which the stop of the iterative operation is not satisfied and the operation cannot be ended during the search for the optimum solution. Can be.

[0026]

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is a diagram showing an entire process control system to which a process control device according to an embodiment of the present invention is applied.

The process control system shown in FIG. 1 is divided into a process control device 1 of the present invention and a process 2 to be controlled.

The process control apparatus 1 of the present invention includes a model database 11A, a controller database 11B, a constraint condition database 11C, a prediction calculation unit 12, an evaluation function setting unit 13, a constraint condition setting unit 14, an operation amount optimizing unit 15, a constraint The condition relaxing means 16 is a basic component. Each database has a data setting unit 3A, 3B,
3C are set correspondingly.

The model database 11A stores model data including a mathematical model representing the dynamic characteristics of the process 2 to be controlled.

The controller database 11B stores controller data including adjustment parameters for model predictive control.

The constraint condition database 11C stores, for operation control, constraint values imposed on the control target for each process variable.

The predicted value calculation means 12 reads the process data of the process 2 to be controlled, calculates future changes of the process 2 to be controlled using the model data, and obtains a prediction result.

The evaluation function setting means 13 derives an evaluation function relating to the optimal operation amount in the model predictive control based on the above-mentioned prediction result.

The constraint condition setting means 14 compares the constraint condition with the constraint condition based on the prediction result, and determines the search range of the operation amount update value derived by optimizing the evaluation function.

The manipulated variable optimizing means 15 compares the data determined by the evaluation function setting means 13 with the constraint condition setting means 1.
With the data determined in step 4, data received via the constraint condition mitigation means 16 is received, and the optimal operation amount is determined by numerical optimization. In the case of FIG. 1, the opening degree of the adjustment valve 23 corresponds to the above-described operation amount.

The constraint condition easing means 16 receives the data S 156 indicating the calculation load of the operation amount optimizing means 15, performs relaxation or selection processing on the constraint data, and sends the data to the operation amount optimizing means 16. It is set again as data S165.

The process 2 to be controlled has a chemical process device 21, a process measuring device 22, a control valve 23, and a process measuring device 24 as basic components.

The process measuring instrument 22 is used for measuring the temperature inside the chemical process equipment or measuring the liquid level. On the other hand, the process measuring instrument 24 installed on the pipe is used for flow rate measurement or temperature measurement in the pipe. Control valve 2
Reference numeral 3 has a function of adjusting the flow rate of the piping, and is used, for example, for adjusting the raw material.

Although the process control system of the present invention has a small-scale configuration such as two process measurement points and one operating end, the same control system configuration can be used even if the process to be controlled is a large-scale process. Take.

Next, the prediction calculating means 12 of the process control device 1 of the present invention will be described with reference to FIGS.

FIG. 2 shows the trend of one control variable at the top,
Below is the trend of one manipulated variable. In this trend, the vertical axis indicates the magnitude of the process amount, and the horizontal axis indicates time. On the time axis, the time at which the manipulated variable calculation is being performed is defined as the current time, and the trend to the right from this indicates the predicted trend of the process. Dotted lines in the graph represent the upper limit and lower limit of the operation constraint condition of the control variables and operation variables of the process.

The prediction calculation means 12 of FIG. 1 predicts and derives a future change of the controlled variable of the process 2 using a mathematical model stored in the model database 11A and representing a dynamic characteristic of the process to be controlled.

This mathematical model expresses the correlation of the temporal change between the manipulated variables and disturbance variables of the process and the control variables, and generally corresponds to a time series model or the like.

In the prediction calculation of the change in the control amount, the measurement data S211 and S241 of the process 2 shown in FIG.
The past process data is substituted into the above-described mathematical model, and a change in the control amount from the measured time is obtained.

Since the manipulated variable optimizing means 15 determines the future change of the manipulated variable, the prediction calculating means 12 assumes that the future manipulated variable will maintain a constant value from the current time, and temporarily changes the controlled variable. Predict. This prediction result is called a steady response prediction trend of the process 2, and corresponds to the control amount change 200 in FIG.

The steady-state response prediction trend 200 is a prediction result of a control amount change when the operation amount is kept constant from the current time. The steady response prediction trend 200 is output from the prediction calculation unit 12 to the evaluation function setting unit 13 and the constraint condition setting unit 14 in FIG.

Next, the evaluation function setting means 13 of the process control device 1 of the present invention will be described with reference to FIGS.

The evaluation function setting means 13 derives an evaluation function for determining an operation amount update value in the model predictive control. The evaluation function can be defined in various ways, but in general,
As variables to optimize the manipulated variable update value for the controlled process,

[0050]

## EQU1 ## In many cases, (predicted control amount-target value) ＾ 2 + (operating amount update value) ＾ 2. Here, ＾ 2 means a square operation.

First, the first term will be described. The evaluation function setting means 13 shown in FIG.
, A control variable future change 201 with respect to the manipulated variable future change 202 ahead of the current time is derived using the mathematical model. The control amount future change 201 is compared with a control amount target value over a prediction evaluation section. This deviation is called a prediction deviation, and in the model prediction control, an operation amount update value that minimizes the prediction deviation is obtained. Here, the prediction evaluation section is a section in which the prediction deviation is evaluated in one operation amount calculation.

Next, the second term will be described. As a feature of the process control, it is desirable that the change in the operation amount be as small as possible for the process to be controlled. In FIG. 2, a section in which an update value is determined in one operation amount calculation is set as a prediction control section.

Since the evaluation function takes a quadratic form, it corresponds to the area sum of the above-described predicted deviation and the manipulated variable update value corresponding to each section. In FIG. 2, an area 201 in the control variable trend and an area 202 in the operation variable trend correspond to each other. These areas use the manipulated variable update value as a variable as described above.

The above-described evaluation function can be expressed by the following expressions.

[0055]

| Control amount predicted value−Target value | + | Operation amount update value | Here, | · | means an absolute value. Again,
It can be said that this corresponds to the above-mentioned area sum.

Next, the constraint condition setting means 14 of the process control apparatus 1 according to the present invention will be described with reference to FIGS.

The constraint condition setting means 14 sets a search range of the manipulated variable update value in the model predictive control. In process control, generally, an upper limit value and a lower limit value are often set for a control variable, an operation variable, and an operation amount update value, respectively, as shown in FIG.

In the constraint condition setting means 14 of FIG. 1, based on the steady-state response prediction trend 200 of the process, similarly to the evaluation function setting means 13, the future change of the control amount and the operation amount is updated, and the operation amount update value is updated. Derived as unknown variables. this,
Comparing the mathematical expression with the manipulated variable update value as an unknown variable and the constraint condition data, it can be described by the inequality condition as shown below.

[0059]

(Constraint lower limit value) <(control amount predicted value) <(constraint upper limit value) (constraint lower limit value) <(operation amount future value) <(constraint upper limit value) (updated lower limit value) <(operation amount) (Updated value) <(upper limit of updated value) These inequalities are used by the manipulated variable optimizing means 15 as a solution search range.

Next, the operation amount optimizing means 15 of the process control device 1 of the present invention will be described with reference to FIGS.

The operation amount optimizing means 15 receives the above-mentioned evaluation function and the inequalities relating to the constraint conditions, and determines an optimum operation amount for each control cycle by using a numerical optimization algorithm. As the numerical optimization algorithm, a linear programming, a quadratic programming, or the like can be applied according to the mathematical expression of the evaluation function.

In the search for a solution, for example, an operation amount update value for minimizing the area sum shown in FIG. 2 is derived.

The manipulated variable optimizing means 15 monitors the calculation load related to the solution search by the numerical optimization algorithm.
As the calculation load, an index that can be easily monitored, such as the number of repetitive operations related to the above-described solution search and the occupancy of a computer, is used.

When the calculation load reaches a certain prescribed value, a trigger S156 is output.

Further, the manipulated variable optimizing means 15 monitors the above-mentioned calculation load and determines whether or not an optimal solution exists. The determination of the existence of the optimal solution is generally performed by using the above-described linear programming algorithm and the quadratic programming algorithm, which also have a determination parameter.
156 applies.

Next, the constraint condition relaxing means 16 of the process control apparatus 1 of the present invention will be described with reference to FIG.

Upon receiving the above-described trigger S156, the constraint condition mitigation means 16 executes a relaxation process on the constraint data. Here, the relaxation processing corresponds to expanding the search range of the optimal solution of the operation amount update value. As a method of expanding the search range, one method is to re-select a constraint condition item in the constraint data described above,
Reset as constraint data of a specific control variable or manipulated variable. In this method, the search range is expanded by reducing the number of constraints.

Another method is to change the constraint value itself. That is, the upper limit value and the lower limit value are reset so as to increase the absolute value. In addition, a method of expanding the search range may be applied.

The process control device 1 of the present invention described above
In FIG. 3, a flowchart summarizing the arithmetic processing for each control cycle is shown.

FIG. 3 shows the calculation processing at the k-th time, which corresponds to the current time 0:00 in FIG. That is, the arithmetic processing shown in FIG. 3 is repeated with the progress of time.

The flowchart shown in FIG. 3 will be described below.

First, the above-described process control device 1 executes the reading of plant data 302, and derives the above-described steady-state response prediction trend 200 by the predicted value calculation 303.

Subsequently, an evaluation function setting 304 and a constraint condition setting 305 are executed to derive an evaluation function and a constraint condition relating to the model predictive control, and the execution of an optimum solution search execution 306 is started.

When the solution search is executed, the calculation load monitoring 307 is executed.
Execute As described above, the calculation load monitoring 307 uses the number of repetitive operations for solution search and the computer occupancy as monitoring items. Subsequently, it is determined that the search for the optimal solution has been completed, and if the search needs to be repeated, the process returns to the optimal solution search execution 306 described above. These steps 306 to 308 are realized as the operation amount optimizing unit 15.

When the calculation load monitoring 307 determines that the calculation load has reached the limit or that there is no optimal solution, the constraint condition relaxation 309 is executed, and the data processing of the constraint condition data is performed. Then, the operation returns to the optimal solution search 306, and the operation amount optimization calculation continues. At this time, since the search range of the solution is expanded, the calculation load is reduced as compared with the situation before the constraint condition is relaxed.

In the process control device 1 of the present invention, the operation of the process control system in the case where the above-described constraint mitigation 309 shown in FIG. 3 is executed will be described with reference to FIGS.

FIG. 4 shows a case where the target value has been changed by the current time, but the target value is substantially the same as the upper limit value of the control variable.

In this case, if the ability to follow the target value is emphasized, the control amount prediction trend 401 exceeds the upper limit of the constraint. Therefore, the process control device 2 requires time to reach the target value. The result is that the quantity update value 402 is output. In addition, since there is no margin in the solution search range, the calculation load increases, and it is considered that processing such as aborting the operation amount optimization in the middle is required.

Next, FIG. 5 shows the operation of the process control apparatus 1 of the present invention when the above-described constraint mitigation 309 is executed.

In FIG. 5, as in the case of FIG. 4 described above, since the upper limit value setting of the control amount constraint and the target value are close to each other, it is assumed that the above-mentioned calculation load, especially the number of repetitive calculations, has increased.
The above-described constraint mitigation 309 is executed. Alternatively, it is determined that the optimal solution does not exist, and the above-described constraint condition relaxation 3
09 is executed. As a result, in FIG. 5, although the control amount prediction trend 501 partially deviates from the upper limit value of the constraint condition due to the relaxation of the constraint condition, it is predicted that the control amount prediction trend 501 quickly follows the target value. The results shown in FIG. 5 are the operation results per control cycle, and do not mean that the control variables actually deviated from the constraint conditions.

That is, when the control variable comes close to the target value, the next control operation functions, a new operation amount update value is added, and the stability of the controlled object can be maintained.

[0082]

According to the present invention, in a process control method and apparatus for optimally controlling a controlled object while predicting future behavior using a mathematical model expressing the dynamic characteristics of the controlled object, By monitoring the calculation load due to the optimization processing and providing means for relaxing the constraint conditions used for the operation amount optimization according to the calculation load, it is possible to prevent the calculation load from increasing, and furthermore, the Control with set values can be executed. As a result, it becomes possible to control a large-scale plant as a control target, and there is an effect of reducing operating costs due to marginal operation near the constraint conditions.

[Brief description of the drawings]

FIG. 1 is a diagram schematically illustrating a process control device and a control target process according to an embodiment of the present invention.

FIG. 2 illustrates a process control method according to an embodiment of the present invention.
The figure which shows a part of operation | movement of a control object.

FIG. 3 is a diagram schematically showing a flowchart relating to a process control method according to an embodiment of the present invention.

FIG. 4 is a diagram showing a part of the operation of the control target when the constraint condition mitigation processing does not operate in the process control method according to one embodiment of the present invention.

FIG. 5 is a diagram showing a part of the operation of the control target when the constraint condition mitigation process is operated in the process control method according to one embodiment of the present invention.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 ... Process control apparatus, 2 ... Control target process, 3A,
3B, 3C: Data setting unit, 11A: Model database, 11B: Controller database, 11C: Constraint condition database, 12: Prediction value calculation means, 13: Evaluation function setting means, 14: Constraint condition setting means, 15: Operation amount optimization Conversion means, 16 ... constraint condition relaxation means.

 ──────────────────────────────────────────────────続 き Continued on the front page F term (reference) 5B045 AA05 GG04 JJ02 5H004 GA13 GA18 GA40 JA12 JA17 JB14 KA54 KC09 KC10 KC12 KC13 KC27 LB10 MA36

Claims (6)

[Claims] 1. A process control method for optimally controlling a controlled object while predicting a future behavior by using a mathematical model expressing a dynamic characteristic of the controlled object. And the control target satisfies the constraint conditions, and derives an operation amount for optimizing the evaluation function, and, based on the calculation load, a unit that monitors a calculation load by the optimization processing of the evaluation function, A process control method and apparatus, comprising: means for relaxing the constraint condition. 2. The method according to claim 1, wherein the number of repetition operations in the optimization processing of the evaluation function is monitored as a calculation load, and when the number of repetition operations exceeds a predetermined value, a request for relaxing the constraint condition is made. The method and apparatus of claim 1, further comprising: 3. A computer occupancy ratio according to claim 1, wherein a computer occupancy of the optimization process of the evaluation function is monitored as a calculation load, and when the occupancy exceeds a predetermined value, a request for relaxing the constraint condition is requested. The process control method and apparatus of claim 1, comprising means. 4. The process according to claim 1, further comprising means for selecting a specific constraint condition from the control variable and the control variable of the controlled object at the time of receiving the mitigation request. Control method and device. 5. The process according to claim 1, further comprising: means for expanding a specific constraint range from constraints on the control variables and manipulated variables of the controlled object when the mitigation request is received. Control method and device. 6. The method according to claim 1, further comprising means for monitoring the calculation load and performing the existence determination of the optimal solution, and requesting a process of relaxing the constraint condition when it is determined that the optimal solution does not exist. Item 6. The process control method and apparatus according to Item 1.