JPS5926042B2 - Predictive control device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION As shown in FIG. 1, the conventional predictive control method uses a predictor to calculate the predicted value of the controlled variable for a lead time lΔT based only on the information on the controlled variable x. +l△T) is calculated, and the predicted value x (+l△T) is compared with the target value (+l△T) ahead of △T. From the deviation 7 (+l△T) obtained, the proportional-integral operation is calculated. The operating signal ul() is generated via the PI controller.

The drawback of such a method is that prediction is performed on the extended J-length line of the history up to the present time, so if the manipulated variable or disturbance suddenly changes midway, the accuracy of the prediction will be extremely reduced.

Furthermore, even if the operation amount and disturbance do not change during the process, if the lead time lΔT is large, the accuracy will deteriorate. Therefore, the predictive control/control performance of this method cannot be expected to be significantly improved compared to conventional PI control. In view of the above circumstances, the present invention leaves the conventional feedback system consisting of a PID controller as it is, and changes the value of the target value setter of the conventional system to a simple mathematical formula that approximates the behavior of the object to be controlled. By operating a predictive control circuit consisting of a model section and a feedback loop of the variables of the mathematical model,
The objective is to provide a predictive control device that provides high control performance that could not be obtained with conventional predictive control, and is particularly effective for processes with large delays. Approximate characteristics F2(D)e-L2D (where D is a differential operator) that express the behavior of the controlled variable when a disturbance is input to the value are terms F, (D
) as a transfer function and receives a disturbance input to the controlled object and outputs a predicted disturbance variation value y2 and its time differential value y2; Approximate transfer characteristic F that expresses the behavior of the controlled variable when the value is input, (D) does not include dead time in e-LID ■ 5th term F1 (D) is the transfer function and intermediate variable of the corrected target value described later △ * is input to the target value fluctuation predicted value y,
and a target value fluctuation predictor that outputs the time differential value yl, and a dead time generator that outputs the dead time L2-L1, and a coefficient unit that adds y1 and Y2 and multiplies the value by a coefficient; and a coefficient unit that adds and multiplies the value by a coefficient, and an intermediate variable △r* of the corrected target value obtained by adding the outputs of both coefficient units and converting the sign. It is characterized in that the output △r obtained by inputting it to the dead moment generator is added to the output of the target value setter of the PID value control system. ) and F1(D) are rational functions, with a relationship of m<n between the order m of D in the numerator and the order n of D in the denominator, and the outputs of F2(D) and F, (D) The differential value of F2(D) and F,(D) are made to be able to be expressed without being expressed by internal variables.

−
1 For example, as a simple example, the first-order lag characteristic is +70
Since the order of the numerator is m-01 and the order of the denominator is n-1,
The above condition of m<n is satisfied, and the mathematical model in this case is as follows: If expressed as Z=DY, then Y can be found by integrating Z in the above equation, and therefore, if intermediate variable Z is output, it is the differential value of output Y.

An embodiment of the present invention will be described with reference to FIG. 2. The part surrounded by a broken line shows a conventional system consisting of a controlled object and a PTD regulator, and the signal output from the target value setter 1 is transmitted to the subtracter 2. is subtracted from the control amount 4 output by the controlled object 3, resulting in a control deviation 5.

However, the adder 9 located between the target value setter 1 and the subtracter 2
Since this is a part of the predictive control circuit according to the present invention, this will be described later. Next, the PID controller 6 which inputs the control deviation 5 outputs a manipulated variable signal 7 for the controlled object 3. Also,
The controlled object 3 is in such a relationship that its output control amount 4 fluctuates in response to the operation signal 7 and the disturbance 8. The present invention provides such a conventional control system with an adder 9 between the target value setter 1 and the subtracter 2, and adds the signal 10 calculated by the predictive control circuit described later and the signal from the target value setter 1. By this,
This is an attempt to correct the target value of the conventional PID value control system.

That is, first, the signal 10 is turned off and the disturbance 8 is kept constant, and the output of the target value setter 1 (
Control amount 4 for the variation amount △r due to step change in r)
The approximate transfer characteristic F1 (
D). - Determine LID. Here, F1(D) does not include the dead time characteristic, D means a differential operator, and t is time. Next, the signal 10 is turned off, the output (r) of the target value setter 1 is kept constant, and the disturbance 8
Similarly, the response of the variation Δy of the controlled variable 4(y) to the step change of (w) is investigated, and the approximate transfer characteristic F2(D)e
'-L2D is determined.

Based on the above preparations, the predictive control circuit will be explained below. Information on disturbance 8 is transferred using the above-mentioned approximate transmission characteristic T2(D)e-L2D
The signal is applied to the disturbance fluctuation predictor 11 consisting of only the F2 (D) section, and outputs 12 (Y2) and 13 (?) are output.

Next, the above-mentioned approximate transfer characteristic F1(D)e−Ll of F1(
The intermediate variable 15 (△r) of the corrected target value is applied to the target value fluctuation predictor 14, which is comprised only of part D), and outputs 16 (skin) and 17 (?) are output. The adder 18 outputs 12 (Y2
) and the output 16 (y1), multiplied by the coefficient in the coefficient unit 19, and the adder 20 adds the output of the coefficient unit 19 and the output 16 (y1), which will be described later.
Add the outputs of 2. The adder 21 adds the output 13 (s) and the output 17 (y1) and outputs the result to the coefficient multiplier 22. Further, the output of the adder 20 is transmitted to a code converter 23, one of the outputs of the code converter 23 is transmitted to the target value variation predictor 14, and the other to the dead time generator 24. The dead time generator 24 outputs a signal 10, which is added to the output of the target setter 1 in the adder 9. The set value of the dead time generator 24 is (L2-L,), and if L2≦L, the set value of the dead time is set to zero or bypassed. Assuming that such a device fluctuates due to the disturbance 8, the control amount 4 will deviate from the value of the target value setter 1. In order to minimize the deviation, the device of the present invention operates as follows. First, the value of disturbance 8 is input to the disturbance fluctuation predictor 11, and output 12 (Y2) and output 13 (y:2) are output. Output 12
(Y2) passes through the adder 18 and is applied from the coefficient unit 19 and code converter 23, one of which is added to the target value fluctuation predictor 14,
The other is added to the dead time generator 24. Therefore, the target value fluctuation predictor 14 receives the output of the code converter 23 and outputs an output 16 (y1) and an output 17 (y:1). Adder 18
Then output 12 (Y2) and output 16 (y1) are added,
Also, in the adder 21, output 13 (Y2) and output 17 (y1
) is added. Then, the output of the adder 21 is the coefficient unit 2
2, is added to the output of the coefficient multiplier 19 by an adder 20, and is transmitted to a code converter 23. That is, output 12
(Y2) and output 16 (y1) are added, then multiplied by a coefficient, and then fed back to the target value fluctuation predictor 14. After adding output 13 (9) and output 17 (6), the system is multiplied by a coefficient and fed back to the target value fluctuation predictor 14. It consists of a system that feeds back to the target value fluctuation predictor 14, and has the function of bringing the sum of output 12 (Y2) and output 16 (y1) and the sum of output 13 (Y2) and output 17 (Y1) close to zero. do. A signal is extracted from the middle of such a circuit as the output of the code converter 23 and is sent to the dead time generator 2.
4 to the adder 9, the original output of the target value setter 1 is corrected. Here, the dead time generator 24
The dead time set value is (L2-L,), and when L2≦L1, the dead time set value is set to zero or bypassed. Note that the corrected target value, that is, the output of the adder 9, is transmitted to the subtracter 2 and compared with the control amount 4, and when the control amount 4 disturbed by the disturbance 8 is smaller than the target value, the PID
The input to the regulator 6 increases, the output of the PID regulator 6, that is, the manipulated variable signal 7 increases, the control amount 4 increases, and the control deviation 5 is gradually brought to zero. The control performance is greatly improved by the above-mentioned action, and this point will be explained with reference to FIG. 3. First, the revised target value △
Conventional PI where the signal 10 of r is not connected to the adder 9
Consider a case where when the disturbance 8 changes in the D control system, the control amount 4 changes like Y2.

The predictive control circuit receives changes in the disturbance 8 with a fluctuation predictor 11 and outputs a value of 0 and a value of 0.

However, in FIG. 3, Zeng and C, which will be described later, are omitted. ?
and C pass through the coefficient unit 22 and the coefficient unit 19 to the adder 2
0 and the code converter 23, and the target value fluctuation predictor 1
4. the result,? and y′1 are output from the target value fluctuation predictor 14, and the wage and ? are added and multiplied by a coefficient, respectively, and fed back to the target value fluctuation predictor 14. The feedback values are shown in Figure 3.
Indicated by r*. The value obtained by passing this Δr* further through the dead time generator 24 is the corrected target value and is indicated by Δr. The PID control system surrounded by the broken line in Fig. 2 receives △r and shows the behavior of y1 as a response due to the target value fluctuation, and the fluctuation of the controlled variable 4(y) is added to Y2 due to the previous disturbance. appear. According to the device of the present invention, the value of y is the PID of FIG.
This is very small compared to the control amount x(t) in the control system. Compared to conventional predictive control devices, the device of the present invention uses a mathematical model, so prediction accuracy is much better and high control performance can be obtained.

Incidentally, since conventional predictors differentiate the control amount, if noise is superimposed on the control amount, the predicted value will be greatly disturbed. Therefore, the control performance of the apparatus of the present invention is significantly improved, especially for thermal processes with large dead times.

[Brief explanation of the drawing]

Fig. 1 is a block diagram showing a control method in a known predictive control device, Fig. 2 is a block diagram showing an embodiment of the predictive control device of the present invention, and Fig. 3 is a diagram showing the input/output relationship of Fig. 2. It is. 1...Typical value setter, 2...Subtractor, 3
...Controlled object, 4...Controlled amount, 5...
... Control deviation, 6 ... PID controller, 7.
...Manipulated amount signal, 8...Disturbance, 9...
... Adder, 10 ... Signal calculated by the prediction control circuit, 11 ... Disturbance fluctuation predictor, 12 (arm) ... Disturbance fluctuation predictor 11 Output, 13(y:
2)... Differential value of the output (Y2) of the disturbance fluctuation predictor, 14... Target value variation predictor, 15...
・Intermediate variable of corrected target value △R8, 16 (armor)...
...Output of target value fluctuation predictor 14, 17 (six)...
... Differential value of output 11 of target value fluctuation predictor 14, 18.
...Adder, 19...Coefficient unit, 20...
...Adder, 21...Adder, 22...
...Coefficient unit, 23... Code converter, 24...
...Dead time generator.

Claims (1)

[Claims] 1 Approximate characteristics that represent the behavior of the controlled variable when a disturbance is input to a PID value control system under a constant target value ▲ There are mathematical formulas, chemical formulas, tables, etc. ▼ (where D is a differential operator), the dead time It has a term F_2(D) that does not include as a transfer function, and the disturbance input to the controlled object is input and the disturbance fluctuation predicted value ■_
2 and its time differential value■ A disturbance fluctuation predictor that outputs _2, and an approximate transfer characteristic representing the behavior of the controlled variable when a target value is input to the PID control system under a constant disturbance ▲Mathematical formula, chemical formula, table The term F_1 (D) that does not include dead time in ▼ is used as a transfer function, and the intermediate variable △_r^* of the corrected target value described later is input, and the predicted target value fluctuation value ■_1 and its time differential value ■_1 are input. It is equipped with a target value fluctuation predictor that outputs, a dead time generator that outputs dead time L_2-L_1, a coefficient unit that adds ■_1 and ■_2 and multiplies the value by a coefficient, and ■_1 and ■_2. On the one hand, when the intermediate variable △_r^* of the corrected target value obtained by adding the outputs of the coefficient machine and the coefficient machine which adds the value and multiplies the value by a coefficient and converts the sign is fed back to the input side of the target value fluctuation predictor. The output △_r obtained by inputting the other dead time generator to the PI
A predictive control device characterized in that it is added to the output of a target value setter of a D-value control system.