JP3350841B2 - Plant control device, manipulated variable determining device, plant control method, and manipulated variable determining method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to a plant control device,
More specifically, the present invention relates to a technique for generating operation knowledge and performing automatic operation based on the generated operation knowledge.

[0002]

2. Description of the Related Art In order to control a chemical plant or the like, a plant control device is used. FIG. 15 schematically shows the configuration of a monitoring man-machine device 2 which is a conventional plant control device disclosed in Japanese Patent Application Laid-Open No. 5-127705. In the conventional monitoring man-machine device 2, an operator performs an operation in a normal state of the plant P via the input / output device 4, the switch 6, and the controller 8.

[0003] The monitoring man machine device 2 stores a plant value (control amount) of the plant P and an operation amount given by an operator corresponding to the plant value in an operation history database 10. The conversion means 12 converts the data stored in the operation history database 10 into a rule-type operation knowledge base 14 having a plant value as a condition part (IF part) and an operation amount as a conclusion part (THEN part), and stores the same. .

In an unsteady state, the plant value collecting means 1
6 receives a plant value via the controller 8. The inference unit 18 obtains the acquired plant value and the operation knowledge base 1
4 is compared with the condition part of each rule stored in
Select a rule that has a condition part that conflicts with the captured plant value. The monitoring man machine device 2 executes the conclusion part of the selected rule. This makes it possible to automate the operation in an unsteady state using the operation history data.

[0005]

However, such a conventional monitoring man-machine device 2 has the following problems. In the inference unit 18, when comparing the captured plant value with the condition part of each rule stored in the operation knowledge base 14, when there is a rule that conflicts with the captured plant value, the selected plant value is selected. By executing the rules, an appropriate operation can be performed.

However, when there is no conflicting rule, proper operation cannot be performed, and the reliability of automatic operation of the plant is low.

In this case, by storing a large number of rules as operation knowledge, the appearance rate of competing rules can be increased. However, in this method, the storage capacity of the operation knowledge base 14 must be increased, and the utilization efficiency of the hardware resources deteriorates.

In the monitoring man-machine device 2,
Only the plant value is used as a criterion for applying the rule.
Therefore, under different situations having the same plant value (for example, when the plant value is the same and the derivative value of the plant value is different), an optimum operation depending on the situation may not be performed. For this reason, reliability when performing automatic operation of a plant is low.

In order to avoid such a situation, for example, a rule can be created in consideration of the differential value of the plant value. However, this requires an enormous number of rules, and the storage capacity of the operation knowledge base 14 must be further increased. For this reason, the utilization efficiency of the hardware resources is further deteriorated.

Furthermore, in the conventional monitoring man-machine device 2, even when noise is included in the fetched plant value or the like, rules are created without considering noise. Therefore, the reliability of the rule itself is low. As a result, the reliability of the automatic operation of the plant based on this rule decreases.

The present invention solves the problems of the plant control device such as the conventional monitoring man-machine device 2 and can perform highly reliable automatic operation with less operation knowledge (rules). , An operation amount determination device, a plant control method, and an operation amount determination method.

[0012]

[Means for Solving the Problems]

[0013]

Technical idea devised to solve the problem: A plant control device, an operation amount determining device, a plant control method and an operation amount determining method capable of performing highly reliable automatic operation with a small amount of operation knowledge (rules). To provide
Operation knowledge is generated for each model division region, and the operation amount is output based on the operation knowledge of the model division region closest to the given control amount.

That is, as shown in FIG. 1 showing the configuration of the invention described in the claims, the plant control apparatus according to the first aspect of the present invention provides the plant P in a manual mode in which the plant P is manually operated. History data constructing means 22 for constructing manual history data by collecting in a time series a set of data having a control amount indicating the operation amount and an operation amount given to the plant by the operator M, and determining an operation amount to be given to the plant P In the operation knowledge generation mode for generating operation knowledge for performing the operation, assuming a control amount space having at least one control amount on a coordinate axis, and constructing manual history data among divided regions obtained by dividing the control amount space. For the model divided region including a point in the control amount space corresponding to the control amount constituting any set of data, based on the manual history data,
An operation knowledge generating means for generating operation knowledge for each model divided region, in an automatic mode for automatically operating the plant P, selecting a model divided region closest to a point in a control amount space corresponding to a given control amount; And inference means 26 for outputting an operation amount based on the operation knowledge of the selected model divided area.

According to a second aspect of the present invention, there is provided a plant control apparatus.
In the plant control device of FIG.
Is characterized in that noise is removed from the manual history data constructed by the manual history data construction means 22 and operation knowledge is generated based on the manual history data from which the noise has been removed.

According to a third aspect of the present invention, there is provided a plant control apparatus.
Alternatively, in the plant control apparatus according to claim 2, the control amount space is configured such that at least one control amount and a time change rate of the control amount are both coordinate axes.

The operation amount determining apparatus according to claim 4 is a plant P
The manual history data in which the set data having the control amount indicating the state of the plant P and the operation amount given by the operator M to the plant P in the manual mode in which the manual operation is performed is taken in time series and given to the plant P In an operation knowledge generation mode for generating operation knowledge for determining an operation amount to be determined, a control amount space having at least one control amount on a coordinate axis is assumed, and a manual operation is performed among divided regions obtained by dividing the control amount space. Operation knowledge for generating operation knowledge for each model divided region based on manual history data for a model divided region including a point in a control amount space corresponding to a control amount constituting any of the set data constituting the history data Generation means 2
4. In an automatic mode in which the plant P is automatically operated, a model divided region closest to a point in a control amount space corresponding to a given control amount is selected, and an operation amount is determined based on operation knowledge of the selected model divided region. Output means 2 for outputting
6 is provided.

According to a fifth aspect of the present invention, in the manual mode in which the plant P is manually operated, the plant P
Operation for constructing manual history data by collecting, in a time series, a set of data having a control amount indicating the state of operation and an operation amount given to the plant P by the operator, and determining an operation amount to be given to the plant P In the operation knowledge generation mode for generating knowledge, assuming a control amount space having at least one control amount on a coordinate axis, any of the divided regions obtained by dividing the control amount space to construct manual history data In the automatic mode in which the operation knowledge is generated for each model divided region based on the manual history data for the model divided region including the point in the control amount space corresponding to the control amount constituting the data, and the plant P is automatically operated. , Selects the closest model divided region from a point in the control amount space corresponding to the given control amount, and calculates the operation amount based on the operation knowledge of the selected model divided region. To force, characterized by.

The plant control method according to claim 6 is directed to claim 5
The plant control method according to the above, characterized in that noise is removed from the constructed manual history data, and operation knowledge is generated based on the manual history data from which the noise has been removed.

According to a seventh aspect of the present invention, there is provided a plant control method.
Alternatively, in the plant control method according to claim 6, the control amount space is configured such that at least one control amount and a time change rate of the control amount are both coordinate axes.

[0021] The operation amount determining method according to claim 8 is the method of
The manual history data in which the set data having the control amount indicating the state of the plant P and the operation amount given by the operator M to the plant P in the manual mode in which the manual operation is performed is taken in time series and given to the plant P In an operation knowledge generation mode for generating operation knowledge for determining an operation amount to be determined, a control amount space having at least one control amount on a coordinate axis is assumed, and a manual operation is performed among divided regions obtained by dividing the control amount space. Based on manual history data, for a model divided region including a point in the control amount space corresponding to a control amount constituting one of the set data constituting the history data, operation knowledge is generated for each model divided region, and the plant In the automatic mode in which the automatic operation of P is performed, the model divided region closest to a point in the control amount space corresponding to the given control amount is selected and selected. Outputting an operation amount based on a working knowledge of Del divided area, characterized by.

According to a ninth aspect of the present invention, there is provided a computer-readable storage medium according to any one of the first to third aspects, the plant control apparatus according to any one of the first to third aspects, the operation amount determining apparatus according to the fourth aspect, or the fifth to fifth aspects. A program for realizing, by using a computer, the plant control method described in any one of claims 7 or the manipulated variable determination method described in claim 8 is stored.

[0023]

[Definition of terms] The concept of terms in the claims for expressing the technical idea devised to solve the problem is defined as follows, and the relationship between the terms and the embodiments is described.

"Manual history data": In a manual mode in which the plant P is manually operated, a set of data having a control amount indicating the state of the plant P and an operation amount given to the plant by the operator M is collected in time series. Means the data obtained by In the embodiment, the contents in the [DATA] column in FIG. 4 correspond.

"Operating knowledge": knowledge including a control amount indicating the state of the plant and an operation amount to be given to the plant in accordance with the control amount. In the embodiment, the contents in the [KNOWLEDGE] column in FIG. 13 correspond.

"Control amount space": A space having a control amount and the like on a coordinate axis. Therefore, when the control amount and the time change rate of the control amount are respectively provided on the coordinate axes, the space becomes a two-dimensional space.
In the embodiment, this corresponds to the control amount space shown in FIG.

"Division area": A space obtained by dividing the control amount space. Therefore, when the controlled variable space is a two-dimensional space, the divided region is also a two-dimensional space, and when the controlled variable space is an n-dimensional space, the divided region is also an n-dimensional space. In the embodiment, the divided area BA shown in FIG.

"Model division region": A region including points in the control amount space corresponding to the control amounts constituting any of the set data for constructing the manual history data among the division regions.
In the embodiment, the model division area MA (i) shown in FIG.
Is applicable.

"Noise": A sudden change component in a time waveform such as a control amount or an operation amount.

[0030]

According to a first aspect of the present invention, in the operation knowledge generating mode, the plant control device, the manipulated variable determining device according to the fourth aspect, the plant control method according to the fifth aspect, and the manipulated variable determining method according to the eighth aspect,
Generates operation knowledge for each model divided region based on the manual history data, selects a model divided region closest to a point in a control amount space corresponding to a given control amount in an automatic mode, and selects the selected model divided region. The operation amount is output on the basis of the operation knowledge.

Therefore, when a point in the control amount space corresponding to the given control amount is included in the model space, the operation amount can be output based on the operation knowledge of the model space. On the other hand, when a point in the control amount space corresponding to the given control amount is not included in the model space, the operation amount can be output based on the operation knowledge of the closest model division region. Therefore, even if the number of operation knowledge is small, highly reliable automatic operation can be performed.

Further, the coordinate axes of the controlled variable space can be determined according to the characteristics of the plant to be controlled. For this reason, it is possible to set a control amount space that expresses the characteristics of the plant without excess or shortage.

The plant control apparatus according to the second aspect and the plant control method according to the sixth aspect further include removing noise from the constructed manual history data and generating operation knowledge based on the manual history data from which the noise has been removed. Features.
Therefore, it is possible to eliminate automatic driving based on operation knowledge including noise.

The plant control device according to the third aspect and the plant control method according to the seventh aspect are further characterized in that the control amount space is configured such that at least one control amount and a time change rate of the control amount are both coordinate axes. Features.

Therefore, even in a plant in which not only the control amount but also the time rate of change of the control amount must be taken into account in the control, the control amount space according to the characteristics of the plant can be set.

According to a ninth aspect of the present invention, there is provided a computer-readable storage medium storing a program for realizing a plant control device or the like according to any one of the first to eighth aspects using a computer. Features. Therefore, a plant control device or the like can be realized by causing a computer to read the data.

[0037]

FIG. 2 shows a hardware configuration in which each function of the plant control device shown in FIG. 1 is realized by using a CPU. The plant control device 30 includes a CPU 32, a memory 34, an FD drive 36, a CRT 38, a keyboard 4
0, an I / O port 42 is provided.

The I / O port 42 takes in data such as a control amount indicating the state of the plant P and an operation amount given to the plant by the operator via the control panel 44 and transfers the data to the CPU 32. I
The / O port 42 transfers the manipulated variable output from the CPU 32 to the plant P. The FD drive 36 reads the content stored on a flexible disk, which is a computer-readable storage medium. The memory 34 stores a program that is the read storage content.

The CPU 32 processes the data taken in by the I / O port 42 in accordance with the program stored in the memory 34. The CPU 32 also outputs an operation amount as a processing result. Information on the state of the plant P and the processing by the CPU 32 is displayed on the CRT 38. The keyboard 40 is used as input means for commands and the like. Here, the CPU 32 corresponds to the manual history data construction unit 22, the operation knowledge generation unit 24, and the inference unit 26 in FIG.

Next, FIG. 3 shows a block diagram of the processing by the plant control device 30 shown in FIG. The processing includes a manual history data construction processing 50, a preprocessing 52, an operation knowledge generation processing 54, and an inference processing 56.

Here, the manual history data construction processing 50
This corresponds to the manual history data construction means 22 in FIG.
The pre-processing 52 and the operation knowledge generation processing 54 correspond to the operation knowledge generation means 24 in FIG. The inference processing 56
This corresponds to the inference means 26 in FIG.

Referring to FIG. 3, the flow of processing by the plant control device 30 will be described with reference to the hardware configuration of FIG. First, the CPU 32 performs a manual history data construction process 50. The manual history data construction process 50 is performed in the manual mode for performing the manual operation of the plant P according to the following procedure.

The signal flow in the manual mode is shown by a broken line in FIG. The control amount x1 representing the state of the plant P is
Displayed on the control panel 44. The operator M determines an operation amount x2 to be given to the plant P from the control amount x1 or the like displayed on the control panel 44 and inputs the operation amount x2 from the control panel 44. The input operation amount x2 is determined by the I / O port 42 (see FIG. 2).
To the plant P via

The CPU 32 uses the control amount x1 and the operation amount x2 in the manual mode as a set of data,
2, and is stored as time-series manual history data in a manual history data file 58 set in the memory 34.

An example of the manual history data file 58 is shown in FIG.
Shown in [DAT] of the manual history data file 58 in FIG.
[A], manual history data is stored. That is, in the [DATA] column, the set data having the control amount x1 and the operation amount x2 is a time series (TIME = 0.00, 1.00, 2.
00 ...).

In this embodiment, the existing DC
The configuration is such that the manual history data construction processing 50 is performed using a part (simple DCS) of the function of S (Distributed Control System).

Next, the operation knowledge generation mode will be described. The signal flow in the operation knowledge generation mode is shown by a two-dot chain line in FIG. CPU 3 in operation knowledge generation mode
2 first performs pre-processing 52 on the data constructed in the manual history data construction processing 50. FIG. 5 shows the specific processing contents of the pre-processing 52. The preprocessing 52 includes a control amount x
It is composed of an IIR filter process 52a for removing noise from 1 and the operation amount x2, and a least square method process 52b for obtaining the gradient (time rate of change) of the control amount x1.

The CPU 32 first performs IIR filter processing 52a on the control amount x1 and the operation amount x2.
A calculation formula of the IIR filter processing 52a regarding the control amount x1 and the operation amount x2 is shown.

Y1 (k) = ((Σ _{i = 0, lb} (i) · x1
(Ki))-(Σ _{j = 1, l} a (j) · y1 (k−
j))) ... (k ≧ l) y1 (k) = x1 (k) ... (k <l) y2 (k) = ((Σ _{i = 0, lb} (i) · x2 (k −i)) −
(Σ _{j = 1, l} a (j) · y2 (k−j))) ... (k ≧ l) y2 (k) = x2 (k) ... (k <l).

In the above equation, y1 (k): input data (control amount after noise removal) y2 (k): output data (operation amount after noise removal) x1 (k): control amount x2 (k): operation Amount k: sampling frequency l: IIR filter order (IIR_DIMENSION) a (j): IIR filter coefficient (IIR_COEFFICIENT) b (i): IIR filter coefficient (IIR_COEFFICIENT)

When l <1, the IIR process is not performed. Further, in this embodiment,
_{i = i1, i2c} (i) is given by: Σi _{= i1, i2c} (i) = c (i1) + c (i1 + 1) + ... + c
(I2).

Preprocessing definition file 6 used for preprocessing 52
FIG. 6 shows a part of the contents of 0. FIG. 6A shows the contents of the definition file used in the IIR filter processing 52a for the control amount x1 and the operation amount x2. In addition,
The pre-processing definition file 60 is set in the memory 34.

As shown in FIG. 5, the CPU 32 next obtains the gradient x1 'of the control amount by performing the least square method process 52b on the control amount x1. The calculation formula of the least squares method process 52b relating to the control amount x1 is shown.

X1 ′ (k) = (1 / n · Σ _{i = 0, n−1} t (k
−i) · y1 (ki) −1 / n · Σ _{i = 0, n−1} t (k−
i) · 1 / n · Σ _{i = 0, n-1} y1 (ki)) / (1 / n ·
_{{I = 0, n-1} t (ki) .t (ki) -1 / n.}
_{i = 0, n-1} t (ki) /1/n.Σ _{i = 0, n-1} t (k-
i)) ... (n≥2) x1 '(k) = y1 (k) -y1 (k-1) ... (n <2) x1' (k) = 0 ... (k <n X1 '(k): slope of control amount y1 (k): input data (control amount after noise removal) t (k): time k: number of samplings n: number of sample data (LSM_DIMENSION) .

It should be noted that the least squares method processing 52 relating to the control amount is performed.
The contents of the definition file used for b are shown in FIG.

As shown in FIG. 5, the CPU 32 next executes the IIR filter processing 5 on the gradient x1 'of the control amount.
By performing 2a, the slope y1 'of the input data is obtained. Least-squares method processing 52 related to gradient x1 'of control amount
The formula for calculating b is shown.

Y′1 (k) = ((Σ _{i = 0,} mb (i) · x1
'(Ki))-(Σ _{j = 1,} ma (j) · y1' (k-
j))) ... (k≥m) y1 '(k) = x1' (k) ... (k <m).

In the above equation, y1 '(k): slope of input data (slope of control amount after noise removal) x1' (k): slope of control amount k: number of samplings m: degree of IIR filter (IIR_DIMENSION) a (j): Coefficient of the IIR filter (IIR_COEFFICIENT) b (i): Coefficient of the IIR filter (IIR_COEFFICIENT)

When m <1, IIR processing is not performed. FIG. 6B shows the contents of the definition file used for the IIR filter processing 52a relating to the gradient x1 'of the control amount.

As described above, the noise is removed from the control amount x1 and the operation amount x2 by the preprocessing 52, and the data is converted into the input data y1 and the output data y2. Further, the slope y1 'of the input data is calculated. These preprocessed data 52 are stored in the input / output data file 62 set in the memory 34. FIG. 7 shows an example of the input / output data file 62.

Next, the CPU 32 performs an operation knowledge generating process 54 on the data which has been subjected to the preprocessing 52. FIG. 8 shows specific processing contents of the operation knowledge generation processing 54. The operation knowledge generation processing 54 includes an operator check processing 54a in which the operator M determines the suitability of the data based on the input data y1 and the output data y2, and a knowledge to extract the operation knowledge yw from the input data y1 and the output data y2. Extraction processing 5
4b.

First, an operator check process 54a is performed. The waveform of the input data y1 (see FIG. 9) and the waveform of the output data y2 stored in the input / output data file 62 are displayed on the CRT 38. The operator M determines whether the manual operation is appropriate based on the displayed waveform. As a result of the judgment,
The operation knowledge yw is not generated from the input data y1 and the output data y2 relating to the manual operation determined to be inappropriate.

As a result of the determination, the operation knowledge extracting process 54b is performed on the input data y1 and the output data y2 relating to the manual operation determined to be appropriate. Operation knowledge extraction processing 54b
Is performed by giving one control knowledge yw to one divided region for a plurality of divided regions obtained by dividing a control amount space described later.

In this embodiment, a two-dimensional space having the input data y1 and the gradient y1 'of the input data as coordinate axes is set as the control amount space (see FIG. 11). The divided area BA is obtained by dividing the two-dimensional space. The two-dimensional space is divided by input data y1
The axis is equally divided so that the unit division amount is d1 (see FIG. 9), and the inclination y1 'axis of the input data is equally divided so that the unit division amount is d1' (see FIG. 10). Do. The formula for calculating the unit division amounts d1 and d1 'is shown below.

D1 = (y1MX-y1MN) / nd1 d1 '= (y1'MX-y1'MN) / nd1'

In the above formula, d1: unit division amount of input data y1 axis y1MX: maximum setting value of input data y1 y1MN: minimum setting value of input data y1 nd1: number of coordinate divisions of input data y1 axis d1 ': input data The unit division amount of the y1 'axis y1'MX: Input data inclination y1' axis setting maximum value y1'MN: Input data inclination y1 'axis setting minimum value nd1': Input data inclination y1 'axis This is the number of coordinate divisions.

In the above equation, the set maximum value y1MX of the input data y1, the set minimum value y1MN of the input data y1, the set maximum value y1'MX of the input data slope y1 'axis, and the set minimum value of the input data slope y1' axis y1'MN is stored in the [SIGNAL] column of the input / output data file 62 shown in FIG. Also, the input data y1 axis coordinate division number nd1,
The coordinate division number nd1 'of the input data inclination y1' axis is shown in FIG.
[DIVIS of the operation knowledge generation definition file 64 shown in FIG.
ION] column.

In this way, as shown in FIG. 11, a two-dimensional control amount space having the input data y1 and the gradient y1 'of the input data as the coordinate axes is represented by d1 × 2.
It is equally divided into rectangular divided areas BA such as d1 '.

As described above, the operation knowledge extraction processing 54b is performed on the divided area BA including the input data y1 which is determined to be appropriate as a result of the operator check processing 54a among the divided areas BA. The operation knowledge extraction processing 54b is performed, and the divided area BA in which the operation knowledge yw is generated becomes the model divided area MA (i). A plurality of rectangular model division areas MA (i) are represented in the control amount space shown in FIG.

The operation knowledge extraction processing 54b in each model divided area MA (i) is performed as follows. First, the following data is obtained for the input data y1.

Data when entering the model divided area MA (i) of interest: y1FS (i) y1FS (i) = y1 (k) Data when leaving the model divided area: y1FN (i) y1FN (i) = y1 (k + n) the minimum value in the model divided _{region: y1 min (i) y1 min} (i) = mIN (y1 (k), ... y1 (k +
n)) Maximum value in the model division area: y1 _max (i) y1 _max (i) = MAX (y1 (k),... y1 (k +
n)) Average value in the model divided area: y1 _avg (i) y1 _avg (i) = 1 / (n + 1) Σ _{j = k, k + n} y1 (j) where the model divided area MA (i) Are sampling numbers k to k + n.

The same data is obtained for the slope y1 'of the input data. Data when entering the model division area MA (i) of interest: y1'FS (i) y1'FS (i) = y1' ( k) Data when leaving the model division area: y1'FN (i) y1'FN (i) = y1 '(k + n) Minimum value in the model division area: y1' _min (i) y1 ' _min (i) = MIN (y1 '(k), ... y1' (k
+ N)) Maximum value in the model division area: y1 ' _max (i) y1' _max (i) = MAX (y1 '(k), ... y1' (k
+ N)) Average value in the model divided area: y1 ' _avg (i) y1' _avg (i) = 1 / (n + 1) Σ _{j = k, k + n} y1 '
(J).

Next, a model operation amount Y2F (i), which is an operation amount representative of the model divided area MA (i), is obtained. In this embodiment, the model operation amount Y2F (i)
Is the output data y2 (k) when entering the model divided area MA (i) of interest. Y2F (i) = y2 (k).

The same operation is repeated for each model division area MA (i). The operation knowledge yw (i) thus obtained by the operation knowledge extraction processing 54b is stored in the operation knowledge file 66, and the operation knowledge generation mode ends. FIG. 13 shows the contents of the operation knowledge file 66.

Next, the automatic mode will be described. The signal flow in the automatic mode is shown by a solid line in FIG. In the automatic mode, the CPU 32 first controls the control amount x1 from the plant P via the I / O port 42 (see FIG. 2).
Take in. After arranging the acquired control amount x1 in a time series (simple DCS 50), the CPU 32 performs preprocessing 52 similarly to the manual history data (see FIG. 5). The noise is removed by the preprocessing 52, and the control amount x1 is converted into input data y1. Further, the slope y1 'of the input data is calculated.

Next, the CPU 32 executes these preprocessing 52
An inference process 56 is performed according to the operation knowledge yw obtained by the operation knowledge extraction process 54b on the basis of the completed data. The inference processing is performed according to the following procedure. CPU3
2 is an area distance which is a distance from a point Q (y1, y1 ') in the control amount space corresponding to the preprocessed data y1, y1' to a boundary of each model divided area MA (i). L
a (i) is calculated (see FIG. 11, part (a)). The formula for calculating the region distance La (i) is shown.

[0077] La (i) = (Ay1 ( i)) 2 + (Ay1' (i)) 2 However, Ay1 (i) = 0 ... (y1 min (i) ≦ y1 ≦ y1
_max (i)) Ay1 (i) = y1 _min (i) -y1 ... (y1 <y1
_{min (i)) Ay1 (i} ) = y1-y1 max (i) ... (y1> y1
_{max (i)) Ay1' (i} ) = 0 ... (y1' min (i) ≦ y1' ≦ y1'
_max (i)) Ay1 '(i) = y1' _min (i) -y1 '... (y1'<y
_{1'min (i)) Ay1' (} i) = y1'-y1' max (i) ... (y1'> y
1 ′ _max (i)).

Next, the CPU 32 calculates the area distance La
Find the smallest minimum area distance La _min of (i), La _min = MIN (La (i)).

Next, the CPU 32 determines the point Q (y in the control amount space corresponding to the data y1, y1 'for which the pre-processing 52 has been completed.
1, y1 ') and the center-of-gravity distance Lg (i), which is the distance between the center of gravity of each model divided area MA (i), is calculated (FIG. 11,
(B) section). The formula for calculating the center-of-gravity distance Lg (i) is shown.

Lg (i) = (Gy1 (i)) ² + (Gy1 ′ (i)) ² where Gy1 (i) = y1−y1 ′ _avg (i) Gy1 ′ (i) = y1′−y1 ′ _avg (i).

Next, the CPU 32 calculates the center-of-gravity distance Lg.
Find the smallest minimum center-of-gravity distance Lg _min of (i), Lg _min = MIN (Lg (i)).

Next, the CPU 32 outputs the selected model operation amount y2F (s) according to the procedure shown in FIG. CP
U32 first checks whether or not the model division area MA (i) where the area distance La (i) = 0 is unique (step S2).

When the model distance MA (i) for which the region distance La (i) = 0 is unique, the model operation amount Y2F of the operation knowledge yw (i) corresponding to the model distance MA (i). (I) (see FIG. 13) is output as the selected model operation amount y2F (s) (step S8).

When there is no model division area MA (i) where the area distance La (i) = 0, or the area distance La (i)
If there are a plurality of model division areas MA (i) where (i) = 0, it is checked whether the model division area MA (i) where the center of gravity distance Lg (i) = Lg _min is unique (step S4). ).

When the model division area MA (i) where the center of gravity distance Lg (i) = Lg _min is unique, the model operation amount of the operation knowledge yw (i) corresponding to the model division area MA (i) Y2F (i) is selected model manipulated variable y2F
(S) is output (step S8).

When there are a plurality of model division areas MA (i) where the center of gravity distance Lg (i) = Lg _min , the model division area MA (i) where the area distance La (i) = La _min
Is checked (step S6).

If the model division area MA (i) where the area distance La (i) = La _min is unique, the model operation amount of the operation knowledge yw (i) corresponding to the model division area MA (i) Y2F (i) is selected model manipulated variable y2F
(S) is output (step S8).

When there are a plurality of model division areas MA (i) where the area distance La (i) = La _min , the model operation amount Y2F of the operation knowledge yw (i) corresponding to those model division areas MA (i). In (i), the minimum model operation amount Y
2F (i) is output as the selected model operation amount y2F (s) (step S10).

As shown in FIG. 3, the output of the selected model manipulated variable y2F (s) is provided to the plant P via the simplified DCS 50. By repeating such processing, the CPU 32 performs the automatic operation of the plant P.

In the above-described embodiment, the output of the selected model operation amount y2F (s) is performed according to the procedure shown in FIG. 14, but the selection model operation amount y2F (s) is output.
The procedure for outputting (s) is not limited to the procedure shown in FIG. For example, a configuration may be adopted in which the selected model operation amount y2F (s) is selected using only the center of gravity distance Lg (i) without using the region distance La (i).

Although the configuration shown in FIG. 13 has been exemplified as the operation knowledge, the operation knowledge may have a configuration other than the configuration shown in FIG.

In the above embodiment, the IIR filter processing 52a is performed to remove noise. However, a method other than the IIR filter processing 52a may be used as a noise removal method. Further, the configuration may be such that the operation knowledge generation 54 and the inference processing 56 are performed without removing noise.

Although the least square method processing 52b is performed to obtain the control amount gradient x1 ', the control amount gradient x1' may be obtained by a method other than the least square method processing 52b. it can.

The input data y1 is used as the control amount space.
And a two-dimensional space having the coordinate y1 ′ of the input data as a coordinate axis is set, but a two-dimensional space having a coordinate axis other than the input data y1 and the gradient y1 ′ of the input data is set as a control amount space. You can also. In addition, a one-dimensional space or a three-dimensional or more space other than the two-dimensional space can be set as the control amount space. Further, the time axis may be one of the coordinate axes of the control amount space. Further, the operation amount is one, but the operation amount may be two or more.

Although the divided area is divided equally, the divided area may be divided into different sizes.

Although a flexible disk is used as the computer-readable storage medium, a hard disk, an optical disk, or the like may be used as the computer-readable storage medium.

In the above embodiment, the CPU 3
2, the functions of the plant control device shown in FIG. 1 have been described as examples. However, some or all of the functions may be realized by hardware logic.

[0098]

According to the first aspect of the present invention, there is provided a plant control apparatus.
The operation amount determination device, the plant control method according to claim 5, and the operation amount determination method according to claim 8 generate operation knowledge for each model division region based on manual history data in the operation knowledge generation mode, and And selecting the closest model division region from a point in the control amount space corresponding to the given control amount, and outputting the operation amount based on the operation knowledge of the selected model division region.

Therefore, when a point in the control amount space corresponding to the given control amount is included in the model space, the operation amount can be output based on the operation knowledge of the model space. On the other hand, when a point in the control amount space corresponding to the given control amount is not included in the model space, the operation amount can be output based on the operation knowledge of the closest model division region. Therefore, even if the number of operation knowledge is small, highly reliable automatic operation can be performed.

The coordinate axes of the controlled variable space can be determined according to the characteristics of the plant to be controlled. For this reason, it is possible to set a control amount space that expresses the characteristics of the plant without excess or shortage. That is, highly reliable automatic operation can be performed with a small amount of operation knowledge (rules).

The plant control apparatus according to the second aspect and the plant control method according to the sixth aspect further include removing noise from the constructed manual history data and generating operation knowledge based on the manual history data from which the noise has been removed. Features.

Therefore, automatic driving based on operation knowledge including noise can be eliminated. That is, more reliable automatic operation can be performed with less operation knowledge.

The plant control device according to the third aspect and the plant control method according to the seventh aspect are further characterized in that the controlled variable space is configured such that at least one controlled variable and a time change rate of the controlled variable are both coordinate axes. Features.

Therefore, even in a plant in which not only the control amount but also the time rate of change of the control amount must be taken into account in the control, the control amount space can be set according to the characteristics of the plant. In other words, with little operation knowledge,
Highly reliable automatic operation can be performed.

According to a ninth aspect of the present invention, there is provided a computer-readable storage medium storing a program for realizing a plant control device or the like according to any one of the first to eighth aspects using a computer. Features.

Therefore, a plant control device or the like can be realized by causing a computer to read the data. In other words, more easily, with less operation knowledge,
Highly reliable automatic operation can be performed.

[Brief description of the drawings]

FIG. 1 is a drawing showing a configuration of a plant control device described in claims.

FIG. 2 is a diagram illustrating an example of a hardware configuration in which each function of a plant control device according to an embodiment of the present invention is implemented using a CPU.

FIG. 3 is a block diagram showing a flow of processing by a plant control device according to an embodiment of the present invention.

FIG. 4 is a diagram showing contents of a manual history data file in the plant control device according to one embodiment of the present invention.

FIG. 5 is a block diagram showing details of preprocessing in the plant control device according to one embodiment of the present invention.

FIG. 6 is a diagram showing contents of a pre-processing definition file in the plant control device according to one embodiment of the present invention.

FIG. 7 is a diagram showing contents of an input / output data file in the plant control device according to one embodiment of the present invention.

FIG. 8 is a block diagram showing details of an operation knowledge generation process in the plant control device according to one embodiment of the present invention.

FIG. 9 is a diagram showing a time waveform of input data in the plant control device according to one embodiment of the present invention.

FIG. 10 is a diagram showing a time waveform of a slope of input data in the plant control device according to one embodiment of the present invention.

FIG. 11 is a drawing showing a controlled variable space in the plant control device according to one embodiment of the present invention.

FIG. 12 is a diagram showing contents of an operation knowledge generation definition file in the plant control device according to one embodiment of the present invention.

FIG. 13 is a diagram showing contents of an operation knowledge file in the plant control device according to one embodiment of the present invention.

FIG. 14 is a flowchart showing a procedure for determining a selected model operation amount in the plant control device according to one embodiment of the present invention.

FIG. 15 is a drawing showing an overall configuration of a conventional monitoring man-machine device.

[Explanation of symbols]

 22 Manual history data construction means 24 Operation knowledge generation means 26 Inference means

────────────────────────────────────────────────── ─── Continuation of the front page (56) References JP-A-5-127705 (JP, A) JP-A-6-281302 (JP, A) JP-A-4-233002 (JP, A) 132108 (JP, A) (58) Field surveyed (Int. Cl. ⁷ , DB name) G05B 13/02-13/04 G05B 7/02 G05B 23/02

Claims (9)

(57) [Claims] 1. In a manual mode in which a plant is manually operated, manual history data is constructed by collecting a set of data having a control amount indicating a plant state and an operation amount given to the plant by an operator in a time series. In the operation knowledge generation mode for generating operation knowledge for determining the operation amount to be given to the plant, dividing the control amount space by assuming a control amount space having at least one control amount on the coordinate axis Based on the manual history data, for the model divided region including a point in the control amount space corresponding to the control amount constituting one of the sets of data constituting the manual history data, An operation knowledge generating means for generating operation knowledge for each area, in an automatic mode for automatic operation of a plant, corresponding to a given control amount; Controlled variable selects the closest model divided region from a point in space, a plant control system which comprising the inference means for outputting an operation amount based on the operating knowledge of a model divided area selected. 2. The plant control apparatus according to claim 1, wherein the operation knowledge generating means removes noise from the manual history data constructed by the manual history data construction means, and operates based on the manual history data from which the noise has been removed. A plant control device, characterized in that it is configured to generate 3. The plant control device according to claim 1, wherein the control amount space is configured such that at least one control amount and a time rate of change of the control amount are both coordinate axes. Plant control equipment. 4. A manual history data in which group data having a control amount indicating a plant state in a manual mode for performing a manual operation of a plant and an operation amount given to the plant by an operator are arranged in time series, and In an operation knowledge generation mode for generating operation knowledge for determining an operation amount to be given to a control amount space having at least one control amount on a coordinate axis, and dividing a control amount space into Generating operation knowledge for each model divided region based on the manual history data for a model divided region including a point in the control amount space corresponding to the control amount constituting any of the set data for constructing the manual history data Operation knowledge generating means, in an automatic mode for automatic plant operation, a model closest to a point in a control amount space corresponding to a given control amount. And an inference means for selecting an operation amount of the selected model division region and outputting an operation amount based on operation knowledge of the selected model division region. 5. In a manual mode in which a plant is manually operated, manual history data is constructed by collecting a set of data having a control amount indicating a plant state and an operation amount given to the plant by an operator in a time series. In the operation knowledge generation mode for generating operation knowledge for determining the operation amount to be given to the plant, a division obtained by dividing the control amount space by assuming a control amount space having at least one control amount on a coordinate axis. Based on the manual history data, the operation knowledge is obtained for the model divided region including the points in the control amount space corresponding to the control amount constituting one of the sets of data that constructs the manual history data. In the automatic mode in which the plant is automatically operated, the model division area closest to the point in the control amount space corresponding to the given control amount is generated. Selected plant control method characterized by, that outputs an operation amount based on the operating knowledge of a model divided regions were selected. 6. The plant control method according to claim 5, wherein noise is removed from the constructed manual history data, and operation knowledge is generated based on the manual history data from which the noise has been removed. . 7. The plant control method according to claim 5, wherein the control amount space is configured such that at least one control amount and a time change rate of the control amount are both coordinate axes. Plant control method. 8. A manual history data in which a set data having a control amount indicating a plant state in a manual mode for performing a manual operation of a plant and an operation amount given to the plant by an operator is chronologically arranged, and In an operation knowledge generation mode for generating operation knowledge for determining an operation amount to be given to a control amount space having at least one control amount on a coordinate axis, and dividing a control amount space into On the basis of the manual history data, the operation knowledge is generated for each of the model divided regions including the points in the control amount space corresponding to the control amounts constituting any of the sets of data that constructs the manual history data. In the automatic mode for automatic operation of the plant, select the model divided area closest to the point in the control amount space corresponding to the given control amount. Outputting an operation amount based on operation knowledge of the selected model divided region. 9. The plant control device according to any one of claims 1 to 3, the operation amount determination device according to claim 4, or the control device according to any one of claims 5 to 7. A computer-readable storage medium storing a program for implementing the described plant control method or the manipulated variable determination method according to claim 8 using a computer.