JP2575818B2 - Plant control device - Google Patents

Description

Description: BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a plant control device, and more particularly to a plant control device suitable for controlling a system having a time delay that cannot be ignored from generation of a state quantity to detection thereof. It concerns the device.

[Conventional technology]

As a control method for a large-scale system, a blocking non-interference optimal control method that divides the system into several sub-blocks and controls while avoiding interference between the blocks has been proposed.

When this blocking non-interference optimal control method is applied to each stand in order to decoupling and optimally control each stand of the tandem rolling mill, there is a distance from the rolling position to the sheet thickness detection position. There is a time delay from the occurrence of the state quantity (sheet thickness) to detection, and the sheet thickness immediately below the rolling rolls cannot be obtained. The dynamic characteristics of the controlled object changes due to acceleration / deceleration. However, there is a problem that the speed is no longer optimal at the changed speed.

This problem is not limited to tandem rolling mills, but is a problem common to systems in which there is a time delay from the occurrence of a state quantity to detection, and the dynamic characteristics of a controlled object fluctuate.

Conventionally, this problem has been addressed by an observer control method or a model reference control method, but it has not been a fundamental solution.

Incidentally, as an example showing this kind of conventional technology, Japanese Patent Application Laid-Open No.
-131103 and the like.

[Problems to be solved by the invention]

FIG. 4 shows an example of a basic configuration of a conventional plant control device. FIG. 4A shows an observer control method, and FIG. 4B shows a model scale control method. In these figures, X is a state quantity vector actually obtained from the control target 2, is a state estimation quantity vector obtained from the control target model 10, and Y is a detection quantity vector of a detection target parameter obtained by conversion from the state quantity , Are detection estimation amount vectors obtained from the control target model 10, and C are detection matrices 4, 12 used for the conversion.

Conventionally, the deviation −Y between the estimated value obtained by the model 10 and the detected value Y is input to only the control target model 10 in the case of observer control, or the control target 2 in the case of model reference control.
Then, the dynamic characteristics of the controlled object 2 and the dynamic characteristics of the controlled object model 10 are intended to be matched with each other by adding the input to the input of the controlled object model 10.

In the case of observer control, the state estimator vector is X
However, since the feedback control is performed without considering the change in the dynamic characteristic of the control target 2, there is a problem that the feedback control is not optimal, and the control system may be collapsed.

On the other hand, in the case of the model reference control, a control output is output to the control target so that the dynamic characteristic of the control target 2 matches the dynamic characteristic of the control target model 10. Therefore, when the controlled object model 10 deviates greatly from the actual state, a large control output may be generated. In many cases, a limiter or the like is applied to the control output. It is difficult to match the dynamic characteristics of 10
It did not provide a solution to the problem.

An object of the present invention is to keep the estimation error of the state quantity to a minimum even when the dynamic characteristic of the controlled object changes, and further, in accordance with the change in the dynamic characteristic of the controlled object, perform feedback control or feedforward control in addition thereto. An object of the present invention is to provide a plant control device capable of maintaining an optimum gain.

[Means for solving the problem]

 The above object is achieved by the following means.

A state quantity including a time delay before detection is estimated by the control target model using each control output state quantity of the control target, and feedback control is executed.

Using the deviation between the estimated value and the detected value considering the time delay, the influence coefficient of the controlled model is adaptively corrected, and the feedback control gain calculated using the influence coefficient and the feedforward control gain for the observable disturbance are calculated. , The dynamic characteristics of the control target model and the control target are made to coincide with each other, and the control system is also corrected so that the control gain is optimized.

If a steady-state error exists in the model to be controlled, a control error occurs. In order to quickly remove the control error, a steady-state error compensator using a detected value is provided.

That is, in order to achieve the above object, the present invention replaces one state quantity (Δh) of a control target having a detection time delay with another state quantity (ΔS, ΔP, ΔH,...) Related to the state quantity. Control target model simulated by a linear function as a variable (120)
In the plant control apparatus, an estimated value (Δh *) of one state quantity having no detection time delay is obtained, and the control value is controlled as a feedback value to reduce a deviation from a control target value. An estimated value ((Δh *) ₀ ) obtained by delaying the estimated value (Δh *) of the one state quantity by a time corresponding to the detection time delay, and a detected value of the one state quantity actually detected with a time delay A model identifier () that corrects the coefficient of the linear function of the controlled object model so as to reduce the deviation from ((Δh) ₀ ), and corrects the gain related to the feedback value based on the corrected coefficient. A plant control device characterized by providing 110) is proposed.

Further, in the above plant control device, a feedforward control circuit (220) for controlling the control target in response to a disturbance which can be detected in advance among the disturbances to the control target (220)
And correcting the gain of the feedforward control based on the coefficient corrected by the model identifier (110).

Further, in any one of the above plant control devices, a steady-state deviation compensator for correcting the control target value by a value obtained by integrating at least a detection value ((Δh) ₀ ) of the one state quantity detected with the time delay. 200).

By using the above means, it is possible to perform control using the estimated value while adaptively modifying the control target model and the control system, and even when the dynamic characteristics of the control target change, it is stable against disturbance and still has a steady-state deviation. An optimal control system that does not remain can be configured.

[Action]

FIG. 1 is a diagram showing the basic principle and configuration of the present invention.

The present invention estimates a state estimation amount for a detection amount Y including a time delay before detection using a control target model. Since the state estimator includes X * that does not include a time delay before detection and * that includes a time delay before detection, = (X *, *) is set. At this time, the detection matrix C is determined so that = C = C (X *, *) (1).

Is delayed in consideration of the time delay, and used as a detection estimator.
Compared with the actual detection value Y, by a method such as model identification,
Adaptively correct the influence coefficient of the model itself.

At this time, if a constant term is added to the model, the steady-state deviation of the model can be removed.

The control gain of the feedback control using the state feedback and the feedforward control for the observable disturbance is determined using the influence coefficient of the model, but the control gain is adaptively corrected according to the correction of the influence coefficient. The system can be optimized.

It is the model estimator that performs the above functions. In the model estimator, the model is modified so that -Y → 0, that is, → Y. Further, the feedback control is executed such that Y → 0. Therefore, if the model includes a steady-state deviation and Y has a deviation,
It takes time to converge to the deviation 0. Therefore, the deviation of the detected value Y is compensated to be 0 using a steady deviation compensator. As a result, the time required for the adaptive correction of the model and the control deviation to be zero can be minimized.

〔Example〕

As one embodiment of the present invention, a case where the present invention is applied to automatic thickness control of a tandem rolling mill will be described.

In each stand of the tandem rolling mill, since there is a distance from immediately below the roll where the rolling is actually performed to the entrance or exit thickness gauge, which is a thickness detector, for example, the plate immediately below the roll is the exit thickness There is a time delay before reaching the total. Rolling, after passing the plate between the rolls at a low speed (10 to 20 mpm), accelerate to a high speed (about 1000 mpm), maintain a constant speed, finally decelerate, pull out the plate at a low speed and finish I do. When the rolling speed changes, the coefficient of friction changes, and the dynamic characteristics of the system change. For this reason, the control gains of the feedback control and the feedforward control determined by the dynamic characteristics of the system deviate from the optimal values.
Control performance deteriorates. The present invention is used to correct this deviation and maintain an optimal control gain.

FIG. 3 shows an outline of rolling at each stand. Rolling is performed by passing a material to be rolled between the rolls. In Figure, H is thickness at entrance side, h is the thickness at delivery side, P is the rolling load, S is the roll gap, T _f is the exit side tension, T _b represents the entry side tension. In the following description, a deviation from the set value is represented by a symbol Δ.

FIG. 2 is a block diagram of each stand control device to which the present invention is applied. By modeling the exit side thickness deviation Δh,
An estimated value Δh * is calculated by the control target model (state estimator) 120 according to the equation (1).

Δh * = a ₁ ΔS + a ₂ ΔP + a ₃ ΔH + a ₄ ΔT _b + a ₅ ΔT _f + e (2) where ΔS: roll gap deviation ΔP: rolling load deviation ΔH: entry side thickness deviation ΔT _b : entry side tension deviation ΔT _f: exit side tension deviation a ₁ ~a _5: influence coefficient e: using state deviation term estimated thickness deviation Delta] h *, the output U _fb of the feedback _{control, U fb = {(M +} Q) / M} Δh * ... (3) Calculate by and perform feedback control. Here, M is a mill constant, and Q is a plastic constant.

The difference between the estimated thickness deviation Δh * and the estimated thickness (Δh *) ₀ and the detected value (Δh) ₀ according to the moving speed of the plate Δh = (Δh *) ₀ − (Δh) ₀ = Δh * e ^{− TS -Δhe -TS} ... and (4), (Δh *) 0 ΔS that was used to calculate the, ΔP, ΔH,
Calculation is performed using ΔT _b and ΔT _f , the control target model 120 is identified by the model identifier 110, and a _{1 to} a ₅ and e are adaptively corrected. As a method of model identification, for example, RLS (recursi
Although there are various methods such as ve least squares, there is a large number of documents concerning these methods, and thus description thereof is omitted here.

In the model identifier 110 is a ₁ ~a ₅ and e is adapted modifications, in the rolling mill, Δh = {(M / ( M + Q)} ΔS ... (5) Δh = {(Q / (M + Q) ｝ ΔH (6), and a ₁ = M / (M + Q) (7) a ₃ = Q / (M + Q) (8)

Therefore, the control gain in the feedback control is
From the equation (3), 1 / a ₁ is obtained, so if the calculation is performed by U _fb = (1 / a ₁ ) Δh * (9), the adaptive correction of the control gain becomes possible. Even when the dynamic characteristic of the control target 2 changes, the optimum gain can be maintained.

Feed-forward control is possible by detecting the entry-side sheet thickness deviation ΔH with the entry-side thickness gauge and tracking the sheet thickness just below the rolling mill. In this case, the control output U _ff is U _ff = (Q / M) ΔH. (10) is calculated by Therefore, U _ff = (a ₃ / a ₁ ) ΔH (11) enables adaptive control of the feedforward control gain.

The estimated value Δh * may include a steady-state deviation. For example, if there is an offset error of S ₀ to _{ΔS, Δh * = Δh * 0} + a 1 S 0 ... (12) However, Delta] h _{* 0} is the true value of thickness deviation. Here, in order to cancel a ₁ S ₀ , the model 120 includes an e term, and this e is modified so that a ₁ S ₀ + e = 0 (13). The feedback control is performed using Δh *, and the modification of the model 120 is performed by Δh * → Δ
h, so that Δh → 0 finally. However, when Δh ≠ 0, the time until Δh = 0 is determined by the convergence time of model identification. Therefore, in order to remove the steady-state error as quickly as possible, the steady-state error compensator 200 using an integrator is used to perform control so that (steady-state error) → 0.

The example of the case where the present invention is applied to the automatic thickness control of the rolling mill has been described above. Here, a control system with one input and one output is considered as the control, but the same method can be applied to the case of performing multivariable optimal control with multiple inputs and multiple outputs. In the case of conventional multivariable optimal control, the control gain is calculated using a model.If the dynamic characteristics of the model do not match the dynamic characteristics of the control target, the desired control performance cannot be obtained. By adaptively controlling the dynamic characteristics of the model by using the control characteristics, the control gain is optimized, and even when the dynamic characteristics of the control target change, optimal control is always performed.

Applicable objects of the present invention include not only tandem rolling mills but also all systems in which there is a time delay from the occurrence of a state quantity to detection thereof, and the dynamic characteristics of the controlled object change.

〔The invention's effect〕

According to the present invention, since there is a time delay from the occurrence of a state quantity to the detection thereof, and the optimality of the control gain of the control system with respect to the system in which the dynamic characteristic of the controlled object changes can be always maintained, the control deviation can be reduced. A stable control system can be constructed.

[Brief description of the drawings]

FIG. 1 is a diagram showing the basic principle and configuration of the present invention, FIG. 2 is a block diagram of a plant control device according to the present invention, FIG. 3 is a diagram schematically showing a rolling mill, and FIG. FIG. 9 is a diagram illustrating an example of a system configuration of a conventional observer type control device and a model reference type control device. 2 ... Control target, 4,12 ... Detection matrix, 8,14 ... Delay time calculator, 10 ... Conventional control target model, 16,20 ... Adder, 18 ... Feedback control circuit, 22 ... ... feedforward control circuit, 100 ... model estimator, 110 ... model identifier, 120 ... control target model (state estimator) of the present invention, 180 ... gain variable feedback control circuit, 200
…… Steady deviation compensator, 220 …… Gain variable feedforward control circuit.

 ──────────────────────────────────────────────────続 き Continuation of the front page (56) References JP-A-63-128401 (JP, A) JP-A-59-30104 (JP, A) JP-A-54-17480 (JP, A) 18517 (JP, B2) Automatic Control Handbook (Basic), pp. 58.10.30, PP. 701-702, Ohmsha

Claims (3)

(57) [Claims] 1. A state quantity (.DELTA.h) of a control object having a detection time delay is replaced with another state quantity (.DELTA.S, .DELTA.P,
.DELTA.H,...) Are used as variables to obtain an estimated value (.DELTA.h *) of one state quantity which does not have the detection time delay and is used as a feedback value. In the plant control device that controls the control target so as to reduce a deviation from a control target value, an estimated value ((h *) of the one state quantity is delayed by a time corresponding to the detection time delay (( Δh *) ₀ ),
The coefficient of the linear function of the controlled object model is modified so as to reduce the deviation from the detected value ((Δh) ₀ ) of one state quantity actually detected with a time delay, and the modified coefficient is A plant control device, comprising: a model identifier (110) for correcting a gain related to the feedback value based on the feedback. 2. The plant control apparatus according to claim 1, further comprising a feedforward control circuit (220) for controlling the controlled object in response to a disturbance that can be detected in advance among the disturbances on the controlled object, A plant control device, wherein a gain of forward control is corrected based on a coefficient corrected by the model identifier (110). 3. The plant control device according to claim 1, wherein the control target value is determined by at least integrating a detection value ((Δh) ₀ ) of the one state quantity detected with a time delay. A plant control device comprising a steady-state deviation compensator (200) for correction.