JPS5925242B2 - Simple predictive control device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a simple predictive control device that uses a mathematical model to improve prediction accuracy and effectively utilizes information in the mathematical model to cope with disturbances applied to a controlled object. .

FIG. 1 is a block diagram showing the configuration of a conventionally known simple predictive control device.

In FIG. 1, A is a controlled object, and B is a proportional-integral operation controller (hereinafter referred to as a PI controller). Disturbance U2(t) and operation signal u from PI controller B are applied to this controlled object A.
1(t) to obtain the control amount x(t), and this control amount x(t) is added to the predictor C to obtain the predicted value 'x(t+lΔt) of the control amount. This predicted value ′x(
t+lΔt) is added to the code converter D and after code conversion,
It is designed to be added to adder E. The adder E is also configured to add a target value γ(t+1JT) lΔT ahead from the target value setter F. As a result, the adder E outputs the output of the code converter D and the target value γ(t+lΔT) ahead of lΔT.
are added to obtain the control deviation '7(t+11JT), and this is added to the PI controller B to obtain the operation signal u1(t).
I'm trying to get it. In this way, the predicted value 'x(t+lΔT) is obtained only from the information of the control amount x(t), this predicted value 'x(t+1ΔT) is compared with the target value lΔT ahead, and the predicted value is calculated via the PI controller B. It is designed to obtain an operation signal u1(t).

Therefore, since the prediction is made as an extension of the history up to the present time, if the manipulated variable or disturbance suddenly changes midway, the accuracy of the prediction will be extremely degraded. Furthermore, even if the manipulated variable and disturbance do not change during the process, if lΔT is large, accuracy will be poor. Therefore, with the conventional simple predictive control device shown in FIG. 1, it is not possible to expect a significant improvement in control performance compared to conventional PI control. This invention has been made in consideration of the above points, and sets the controlled variable of the controlled object and the target value of each of the above-mentioned systems, which are changed by inputting the manipulated variable signals and disturbances of at least two or more systems to the controlled object. A proportional-integral-derivative control system that obtains manipulated variable signals for each system by subtracting the control deviation from the target value set by the controller into a proportional-integral-derivative controller for each system, and the target value for each system. It is composed of elements that do not include the dead time characteristic of the first approximate transfer characteristic that represents the amount of variation in the control amount with respect to the amount of variation in the disturbance with the output of the value setter being constant, and the disturbance is input. a disturbance fluctuation predictor for each system that outputs a predicted disturbance fluctuation value and its time differential value, and a fluctuation amount of the output of the target value setter of a system other than the predetermined system in the proportional-integral-derivative control system. The second approximate transfer characteristic representing the amount of variation in the controlled variable of the predetermined system corresponding to the second approximate transfer characteristic does not include the dead time characteristic, and another loop target fluctuation predictor for each system that receives the output and outputs a predicted target fluctuation value and a time derivative fluctuation value of a system other than the predetermined system in the proportional-integral-derivative control system; A third parameter representing the amount of variation in the controlled variable of the predetermined system corresponding to the amount of variation in the output of the target value setter of the system.
The target value fluctuation for each system is composed of elements that do not include dead time characteristics among the approximate transfer characteristics of a predictor, and each system that adds the time differential value of the disturbance fluctuation predicted value outputted from each of the disturbance fluctuation predictors and the time differential value of the target fluctuation predicted value outputted from each of the other rube target fluctuation predictors. a first adder for each;
First means for each system that multiplies the result of addition of the output of each first adder and the time differential value of the target value fluctuation predicted value outputted from the target value fluctuation predictor by a predetermined coefficient and outputs the result. and a second adder for each system that adds the target variation predicted value output from each of the other loop target variation predictors and the target variation predicted value output from each of the target value variation predictors, a second means for multiplying by a predetermined coefficient the addition result of the output of each of the second adders and the disturbance fluctuation predicted value outputted from each of the disturbance fluctuation predictors and outputting the result, and a first means for each of the systems;
The result of adding the outputs of the means and the second means is code-converted, one is outputted as the intermediate variable to the target value predictor for each system, and the other is set for the target value provided for each system. a third means for outputting to a dead time generator that generates dead time for correcting the output of the device, and adding the output of the target value setter for each system and the output of the dead time generator. By comprising a third adder that outputs the addition result to the above-mentioned subtracter and other loop target fluctuation predictors, prediction accuracy is improved, high control performance is obtained, and the mathematical model The purpose of the present invention is to provide a simple predictive control device that can obtain a value equivalent to a differential value from a variable in a variable without differentiation, and has remarkable control performance, especially for thermal processes with large dead times. .

Embodiments of the simple predictive control device of the present invention will be described below with reference to the drawings.

FIG. 2 is a block diagram showing the configuration of one embodiment.
In FIG. 2, the part surrounded by a broken line 100 shows a conventionally known proportional-integral-derivative control system (hereinafter referred to as a PID control system). I will briefly explain this part. Starting from the upper part of this PID control system 100,
The target value set by the target value setter 1 is subtracted from the control amount 4 from the controlled object 3 by the subtracter 2, so that a control deviation 5 is obtained. This control deviation 5 is determined by the PID controller 6 (
It is designed to be input to a controller that performs proportional, integral, and differential operations. When the control deviation 5 is input to the PID regulator 6, the PID regulator 6 outputs a manipulated variable signal 7 for operating the controlled object 3. Furthermore, in the controlled object 3, the controlled variable 4 is in a relationship such that it fluctuates in response to the manipulated variable signal 7 and the disturbance 8. Such a PID control system 10
0, the present invention provides 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 target value set by the target value setter 1. This is an attempt to correct the conventional target value.

In addition, in FIG. 2, the lower loop of the PID control system 100 is the same as the upper loop described above, so parts corresponding to the upper loop are given the same reference numerals as each part of the upper loop. Unless there is a particular change, the explanation will be omitted. Next, the part of the predictive control circuit described above will be explained. The signals 10 and 10' calculated by this predictive control circuit are cut, and the disturbance
= constant, the PID control system 10 surrounded by a broken line
The output of the target value setter 1 at 0, that is, the target value γ
Control amount 4(y) for the step change (variation amount Δγ)
Examine the response of the variation Δy of
e-1t0).

Here, Fρ(F'1D) has a form that does not include dead time characteristics, and D means a differential operator -.

Dt where t is time.

Next, the target value γ set by the target value setter 1 is kept constant (signals 10 and 10' are cut off), and the disturbance 8 (w
) Variation amount Δy of control amount 4(y) with respect to step change
Similarly, the response of is examined to determine the approximate transfer characteristic F3 old e-L3.

Similarly, for the lower loop, F1 recommendation-L'
1 bottom, FP)e-L2D and F3O))e-L3'.
Let's find out. Based on the above preparations, the configuration of the predictive control circuit will be explained below. First, the information of the disturbance 8 is expressed by the approximate transfer characteristic F3O))e-L3 of (F3O))e
-L3 special) F,? , (F!(self)) section is applied to the disturbance fluctuation predictor 11,1', which is composed of only the (F!(self)) section, and the output 22(F,)
,23(fu,),~
S: (Produces output 27 (3 Y and output 23''(Y3'').

::Here, Y3 is the time differential value of Y3, and as will be described next, the structure of the predictor is such that it can be extracted from the middle of the circuit of the disturbance fluctuation predictor 11.

The structure of the predictor is a rational function, and the relationship m<n is established between the order m of the numerator D (differential operator) and the order n of the denominator D, and the differential value of the output of the predictor is predicted. It can be expressed using variables inside the container. For example, as a simple example, what is the first-order lag characteristic? Since the order of the numerator is m-0 and the order of the denominator is n1+τD 1, the above condition of m<n is satisfied, and the numerical model in this case is (X: input Y: output).

In addition, the specific calculation formula is transformed into Y + τDY-X, and further expressed as Z-DY, it becomes Z--(X-Y) τ, and Y can be calculated by integrating Z in the above formula. Desired.

Therefore, if the intermediate variable z is output, it is the differential value of the output Y. Next, the other loop target fluctuation predictors 14, 14' configured only with the FP (self) part of the aforementioned approximate transfer characteristic F2D)e-L2D(F2'Cf))e-L2'D) Adder 9' of each other loop
, 9 output ~ 8-
is input, output 25 (Y2) and output 16 (Y2) (output 15''(F,'') and output 16''(J2'') are output.

Here, like Y3, Y2 is also a time differential value of Y2, and the predictor is configured so that it can be extracted from the middle of the circuit of the other loop target fluctuation predictors 14, 14'.

In addition, the approximate value transfer characteristic FlD)e−Ll value (F/
D) Intermediate variables (Δγ*) 20, 20″ (Δγ″*) of the corrected target value are added to the target value fluctuation predictor 17, 17′ which is configured only from the FlI)) (F1]) part of D) e-L/0). and output 18 (V1) and output mode.
゜, 8shi1
9 (V1), (output 18'' (y1) and output 19'' (y1)
”).

Here, like V3, i1 is also a time differential value of F, and the predictor is configured such that it can be extracted from the middle of the circuit of the target value fluctuation predictor 17. 21 indicates an adder.

The adder 21 is configured to add the output 12 (F) of the disturbance fluctuation predictor 11 and the output 22 of the adder 23. The adder 23 receives the output 15 (su,
) and the output 18 (?1) of the target value fluctuation predictor 17 are added. The same applies to adders 2V and 23'. Further, the adder 24 adds the output 13 (Y3) of the disturbance fluctuation predictor 11 and the output 16 (?2) of the other loop target fluctuation predictor 14, and sends the output 25 to the adder 26. ing.

This adder 26 adds the output 25 of the adder 24 and the output 19 (?1) of the target value fluctuation predictor 17. The same applies to adders 24' and 26'. The output of the adder 21 is sent to a coefficient multiplier 27.

When the output of the adder 21 is introduced into the coefficient multiplier 27, it is multiplied by a coefficient and sent to the adder 28. Further, the output of the adder 26 is introduced into a coefficient unit 29, where it is multiplied by a coefficient and then added to the adder 28. Therefore, the adder 28 adds the outputs of the coefficient units 27 and 29 and transmits the addition result to the code converter 30. This code converter 3
One of the zero outputs (intermediate variables of the corrected target value) is transmitted to the target value fluctuation predictor 17, and the other is configured to be transmitted to the dead time generator 31. This dead time generator 31 outputs a signal 10 and sends it to the adder 9. As a result, the adder 9 becomes the target value setter 1.
The output of the dead time generator 31 is added to the output 10 of the dead time generator 31, and the result of the addition is sent to the loop target fluctuation predictor 14' and the subtracter 2 in the lower loop. The set value L of the dead time generator 31 is (σ3L3+σ2L2-L1), and if L is zero or negative, the dead time set value is set to zero or bypassed.

Here, σ2 and σ3 are weights that satisfy σ2+σ31 and are both non-negative. In order to prevent the set value L from becoming negative, it may be better to take this into consideration at the approximate transfer characteristic stage, but in that case it is not necessary to bypass the dead time generator 31. The same applies to the coefficient generators 2r and 29'', the adder 28'', the code converter 30', and the dead time generator 31'.

Next, the operation of the simple predictive control device of the present invention configured as described above will be explained.

First, assuming that the disturbance 8 increases, the control amount 4,
4'' deviates from the target value setters 1 and 1'', but in order to minimize the deviation, the present invention operates as follows. First, input the value of the disturbance 8 to the disturbance fluctuation predictors 11 and 11',
The disturbance fluctuation predictors 11, 11' output outputs 12, 12'' and outputs 13, 13''. Next, target value setter 1
, 1' and the output 10 of the dead time generator 31, 3'.
, 10' are added by adders 9, 9'. Of these, the output of the adder 9 is input to another loop target fluctuation predictor 14''.When this other loop target fluctuation predictor 14' receives the output of the adder 9, it outputs outputs 15'' and 16'. do. Further, the output of the adder 9'' is input to another loop target fluctuation predictor 14, and this other loop target fluctuation predictor 14 outputs 1
Outputs 5 and 16. Therefore, the disturbance fluctuation predictor 11,1
17 outputs 12, 12' and other loop target variation predictor 1
Outputs 15, 15' of 4, 14' pass through adders 21, 211, coefficient multipliers 27, 27'', adders 28, 28', code converters 30, 30', and one side passes through the dead time generator 3.
1, 31', and the other is applied to the target value variation predictor 17, 17''.

When the target value fluctuation predictors 17, 17' receive the outputs from the code converters 30, 30', they output outputs 18, 18'' and 19, 19'.
3,23'. The adders 23, 23' receive the outputs 15, 15' of the other loop target fluctuation predictors 14, 14''.
Therefore, the adders 23, 23'' add these two outputs 18, 18'' and 15, 15', and add the outputs 22, 22' to the adders 21, 21'. Adder 2
1 and 2' add this output and the outputs 12 and 12' of the disturbance fluctuation predictors 11 and 1', and the addition result is sent to the coefficient unit 27.
, 27'' and then added to adders 28, 28''. Further, the adders 24 and 24' are connected to the disturbance fluctuation predictors 11 and 1.
1' output 13, 13! and loop target variation predictor 14,
The outputs 16 and 16' of 14' are added, and outputs 25 and 25' resulting from the addition are sent to adders 26 and 26'. Adders 26, 26″ are the output 1 of target value fluctuation predictor 17.
9, 19' and the outputs 25, 25' of the adders 24, 24' are added, and the output is sent to the coefficient units 29, 29''.The coefficient units 29, 29' multiply this output by a coefficient, and then add the outputs 25, 25'. vessel 28
, 28'. Adders 28, 28″ are coefficient units 27,
Adding the outputs of 27″ and 29,29″, code converter 3
0,30'. That is, the disturbance fluctuation predictor 1
After adding the outputs 12, 12'' of 1,1'', the outputs 15, 15'' of the other loop target variation predictors 14, 14'', and the outputs 18, 18'' of the target value variation predictors 17, 17'', A system that multiplies the coefficient and feeds it back to the target value fluctuation predictors 17, 17', and an output 13 of the disturbance fluctuation predictors 11, 1'.
13'', the outputs 16, 16' of the other loop target fluctuation predictors 14, 147, and the outputs 19, 19' of the target value fluctuation predictors 17, 17'' are added, and then the coefficients are multiplied by the coefficient units 29, 29'. The output of the adders 21, 2' and the adders 26, 26' is made to approach zero.

When one of the outputs of the aforementioned code converters 30, 30'' is applied to the dead time generators 31, 3f, the dead time generator 31
, 3' outputs a signal 10,10'', and this signal 10
, 10' are added to adders 9, 9'.

These signals 10, 10' are added to the outputs of the target value setters 1, 1'' by adders 9, 9', and the outputs of the target value setters 1, 1'' are added to the original target value setters 1, ``.
Correct the output of Here, the dead time generators 31, 3'
The set values L and L' of the dead time are σ3L3+σ2L2-L
l, σ3″L《+σ6L4−Ll, and when L and L′ are negative, the set value of the dead time is set to zero or bypassed. The output is transmitted to the subtractors 2, 2' and compared with the controlled variables 4, 4', and when the controlled variables 4, 4' disturbed by the disturbance 8 are smaller than the target value, the inputs of the PID regulators 6, 6' are changed. Increase, P
The outputs of the ID controllers 6, 6', that is, the manipulated variable signals 7,
7'' increases, the control amounts 4 and 4'' are increased, and the control deviations 5 and 5' are gradually brought to zero.

In this way, the control performance is greatly improved, which will now be explained in more detail with reference to FIG.

In this case as well, the process is performed for the upper loop, and the same applies to the lower loop, so the explanation will be omitted. First, in the conventional PID control system 100 in which the signal 10 of the corrected target value Δγ is not input to the adder 9, the disturbance 8
Let us consider a case in which the control amount 4 changes as shown in Y3 when .

The predictive control circuit (the part excluding the PID control system 100) receives changes in the disturbance 8 through a disturbance fluctuation predictor 11, and as shown in FIG. Outputs the value of and the value of V3. However, No.●●
●'Zj''Jj''j-- In Figure 3, Y3 and Y2 and yl, which will be described later, are not specified.

Next, in the lower loop, the fluctuation of the target value or the modified target value Δ
The influence on the control amount 4 due to the change in γ'' is expressed as Y2, and when the disturbance 8 and the interference from the lower loop to the upper loop are added, Y2+Y3 is obtained.

And from the other loop target fluctuation predictor 14, 72 and ma,
is output. Therefore, (Ma, +V3) and (Ma, +
Y3) is transmitted to the adder 28 and code converter 30 through the coefficient unit 27 and the coefficient unit 29, and is applied to the target value fluctuation predictor 17. the result,? 1 and 71 are outputted from the target value variation predictor 17 and are added to (Y2+Y3) and (Y2+Y3) and multiplied by a coefficient.

Feedback is provided to the target value fluctuation predictor 17. This feedback value is shown in FIG. 3 as Δγ*. The value obtained by further passing this value Δr* through the dead time generator 31 is the corrected target value, which is shown as Δγ by the broken line in FIG. 3E.

The conventional PID control system 100 receives this corrected target value Δγ and exhibits the behavior of y1 as a response to the target value fluctuation,
The control amount 4(y) appears as a sum of Y3 due to the previous disturbance and Y2 due to other loop target fluctuations. This value of y corresponds to the control performance of the conventional PID control system 100.
Very small compared to 3. Although the above description has focused on a two-input control system, it can be similarly extended to a multivariable system.

As described above, in this invention, the multivariable feedback system consisting of the conventional PID controller is left as is,
And the target value setter 1 of the conventional system? The value of ``is to be manipulated by the predictive control circuit of this invention,
This predictive control circuit consists of a simple mathematical model that approximates the behavior of the controlled object and a feedback loop for the variables of that mathematical model, and is able to achieve high control performance that could not be achieved with conventional predictive control. .

Therefore, although the present invention is applicable to control objects in general, it is particularly effective for processes with large delays. As detailed above, according to the simple predictive control device of the present invention, the controlled variable of the controlled object that changes by inputting the manipulated variable signals and disturbances of at least two or more systems to the controlled object, A proportional-integral-derivative control system that obtains a manipulated variable signal for each system by subtracting the control deviation obtained by subtracting the target value set by the target value setter into a proportional-integral-derivative controller for each system, and consisting of elements that do not include dead time characteristics of a first approximate transfer characteristic representing the amount of variation in the control amount with respect to the amount of variation in the disturbance with the output of the target value setter of each system being held constant; a disturbance fluctuation predictor for each system that receives the disturbance and outputs a predicted disturbance fluctuation value and its time derivative, and a target value setter for other systems other than the predetermined system in the proportional-integral-derivative control system. The second approximate transfer characteristic representing the amount of variation in the control amount of the predetermined system corresponding to the amount of variation in the output, which does not include the dead time characteristic, Another loop target fluctuation predictor for each system that receives the output of the value setter and outputs the predicted target fluctuation value and its time derivative fluctuation value for systems other than the predetermined system in the proportional-integral-derivative control system. and a third approximate transfer characteristic representing the amount of variation in the controlled variable of the predetermined system corresponding to the amount of variation in the output of the target value setter of the predetermined system, which does not include the dead time characteristic. and a target value fluctuation predictor for each system that outputs the target fluctuation predicted value and its time derivative value for each system, into which the intermediate variable of the corrected target value is input, and the disturbance output from each of the disturbance fluctuation predictors. a first adder for each system that adds the time differential value of the predicted variation value and the time differential value of the target variation predicted value output from each of the other loop target variation predictors;
a first means for each system that multiplies the addition result of the output of the adder and the time differential value of the target value fluctuation predicted value output from the target value fluctuation predictor by a predetermined coefficient and outputs the result; a second adder for each system that adds the target fluctuation predicted values output from the other loop target fluctuation predictors and the target fluctuation predicted values output from each of the target value fluctuation predictors; a second means for multiplying by a predetermined coefficient the addition result of the output of the adder and the disturbance fluctuation predicted value outputted from each of the disturbance fluctuation predictors, and outputting the result by multiplying the result by a predetermined coefficient; The results of adding the outputs of the means 2 are converted into codes, one of which is output as the intermediate variable to the target value predictor for each system, and the other is the output of the target value setter provided for each system. A third means for outputting an output to a dead time generator that generates dead time for correction, and adding the output of the target value setter for each system and the output of the dead time generator to obtain the addition result. The gist of this is that it consists of the above subtracter and a third adder that outputs to other loop target fluctuation predictors, improving prediction accuracy and obtaining high control performance. A value corresponding to a differential value can be obtained from a variable, and the control performance is particularly effective for thermal processes with large dead time.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing the configuration of a conventional simple predictive control device, FIG. 2 is a block diagram showing the configuration of an embodiment of the simple predictive control device of the present invention, and FIG. 3 explains the operation of the same embodiment. This is a time chart for 1,1''...Target value setter, 2,2''...
...Subtractor, 3,3'...Controlled object, 6,6ξ
...PID controller, 8...Disturbance, 9,2
1, 23, 24, 26, 28, 97, 21″, 23′,
24″, 26″, 28′... Adder, 11, 1
``...disturbance fluctuation predictor, 14, 14'...
...Other loop target variation predictors, 17, 17'...
・・Target value fluctuation predictor, 27, 29, 27″, 29′・
...Coefficient unit, 30, 30'... Code converter, 31, 31'... Dead time generator.

Claims (1)

[Claims] 1. Using a subtractor, subtract the controlled variable of the controlled object that changes by inputting the manipulated variable signals and disturbances of at least two or more systems into the controlled object and the target value set by the target value setter of each of the above systems. A proportional-integral-derivative control system that inputs the control deviation into a proportional-integral-derivative controller for each system to obtain manipulated variable signals for each system, and a proportional-integral-derivative control system that obtains manipulated variable signals for each system. It is composed of elements that do not include the dead time characteristic of the first approximate transfer characteristic representing the amount of variation of the control amount with respect to the amount of variation of the disturbance, and when the disturbance is input, the disturbance variation predicted value and its time differential value are output. the disturbance fluctuation predictor for each system, and the amount of variation in the controlled variable of the predetermined system corresponding to the amount of variation in the output of the target value setter of other systems other than the predetermined system in the proportional-integral-derivative control system. The output of the target value setter of a system other than the predetermined system is inputted to the second approximate transfer characteristic representing the second approximate transfer characteristic, which does not include the dead time characteristic. Another loop target fluctuation predictor for each system that outputs the predicted target fluctuation value and its time differential fluctuation value for systems other than the predetermined system, and the amount of variation in the output of the target value setter for the predetermined system. The third approximate transfer characteristic representing the amount of variation in the controlled variable of the predetermined system corresponding to A target value fluctuation predictor for each system that outputs a fluctuation predicted value and its time derivative, and a time derivative of the disturbance fluctuation predicted value output from each of the disturbance fluctuation predictors and each other loop target fluctuation predictor described above. a first adder for each system that adds the time differential value of the predicted target fluctuation value output from the target value fluctuation predictor; a first means for each system that multiplies the addition result of the predicted value with the time differential value by a predetermined coefficient and outputs the result, and a target fluctuation predicted value output from each of the other loop target fluctuation predictors and each of the above targets. a second adder for each system that adds the target fluctuation predicted value output from the fluctuation predictor; and a disturbance fluctuation predicted value output from the output of each second adder and each disturbance fluctuation predictor. a second means for multiplying the addition result by a predetermined coefficient and outputting the result, and a second means for adding the outputs of the first means and the second means for each of the above-mentioned systems, converting the sign, and converting one of the results for each of the above-mentioned systems. third means for outputting the intermediate variable as the intermediate variable to the target value predictor of the target value setter, and outputting the other to a dead time generator provided for each system that generates a dead time for correcting the output of the target value setter. and a third adder that adds the output of the target value setter for each system and the output of the dead time generator and outputs the addition result to the subtracter and other loop target fluctuation predictors. A simple predictive control device consisting of: