JPS6314202A - Plant control method - Google Patents

Abstract

PURPOSE:To perform high-accuracy control by fixing parameters of a controlled system against plural inputs and outputs, and making the best design by using them and varying and determining the parameters on an on-line basis. CONSTITUTION:An observable state quantity (y) and an input (u) to the controlled system 1 are inputted to a data gathering control mechanism 6, which adjusts the timing of data gathering from the state quantity (y) and input (u) and the timing of transmission of data (q) to an identifying mechanism 3 with a control signal 8 of a timing control mechanism 7. The mechanism 3 outputs data (v) which is an identification result to a design computing mechanism 4 with a control signal 10 of the mechanism 7 and the mechanism 4 receives a control signal 12 of the mechanism 7 and the data (v) and determines the constitution and parameters of a controller according to design logic, thereby outputting them as output data (w) to a control and storage mechanism 13. Then mechanism 13 receives the data (w) and varies the parameters of a controller 5 according to a control signal 14 of the mechanism 7 and the relation between the signals (u) and (y), and the controller 5 uses a reference signal (r), the state quantity (y), and a control signal 16 from the mechanism 7 to determine the input (u) to the controlled system 1.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to the control of plants such as steel rolling mills, and more particularly to a control method capable of identifying and controlling sudden changes in parameters of a controlled object.

[Conventional technology]

``Automatic Control Handbook Basic Edition'' as a conventional control method.
The self-tuning regulator (STR) discussed in pages 702 to 703 and pages 723 to 730 of the Japan Society of Automatic Control Engineers (1983)
It is abbreviated as. ) has been proposed. An embodiment of the present invention is shown in FIG. 1, and in order to clarify the differences from the conventional control method, the conventional technique will be explained below. FIG. 2 shows an example of a self-tuning regulator. In this method, an output 2 and an input from a controlled object 1 are input to an identification mechanism 3. The output of the identification mechanism 3 is input to the design calculation mechanism 4, and the calculation result of the design calculation mechanism 4 becomes a parameter change command for the control device 5°.

The target value γ of the output 2 and the output 2 are input as feedback amounts to the control device 5, and an input signal of the controlled object 1 that brings the output 2 of the controlled object 1 closer to the target value is inputted to the control device 5. Device 5 outputs.

With this configuration, it is possible to control a single-power output system such as voltage control of a turbo generator, but it is possible to control a device that has multiple inputs such as a rolling device input and a speed command input, such as a steel rolling mill. This could not be applied to multi-input/multi-output systems where multiple devices are connected in series and the entire device is controlled to obtain, for example, a constant plate thickness6.Furthermore, if calculations are performed for each sampling, the processing capacity will be reduced. For this reason, there is a problem that the number of parameters to be controlled that can be identified is small, making it impossible to support a control system with many parameters such as the aforementioned plural series rolling mills.

Furthermore, in the case of a rolling mill, the parameters of the controlled object change suddenly when the welding point of the rolled steel material passes.In the above configuration method, the parameters are identified based on the input/output relationship to the controlled object 1 a certain time ago from the current time. Therefore, there was a problem that it could not be followed.

[Problem that the invention seeks to solve]

The above-mentioned conventional technology does not take into consideration the multi-input multi-output system, the relationship between sampling time and the controlled object, and sudden changes in parameters, and has the problem of increased output deviation.

An object of the present invention is to provide a control method that solves the above problems.

[Means to solve the problem]

The above objective is achieved by identifying parameters of a controlled object from a plurality of inputs and outputs, performing optimal design, and realizing a control device online.

[Effect]

In a control system that changes parameters based on the relationship between multiple inputs and multiple outputs, pre-stored parameters are used for rapid parameter changes, and the latest identification results 7 are used for parameters that change slowly. Since parameters can be used, highly accurate control is possible.

〔Example〕

An embodiment of the present invention will be described below with reference to FIG. The variable am state quantity y (vector of mX1) and the input u (vector of Q×1) with respect to the controlled object 1 are input to the data collection control mechanism 6, and in the data collection mechanism 6, the control signal 8 of the timing control mechanism 7 The available wA measurable state quantity y and the input U
Data collection timing from and data q to identification mechanism 3
The sending timing of the identification mechanism 3 is adjusted, and the identification mechanism 3 outputs data V, which is the identification result, to the design calculation mechanism 4 according to the control signal 10 of the timing control mechanism 7, and the design calculation mechanism 4 controls the timing control mechanism 7. Upon receiving the signal 12 and the data, the configuration and parameters of the control device are determined according to the design theory, and the parameter control and parameters are determined as the output data υ.
The data W is output to the storage mechanism 13, and the parameter control/storage section changes the parameters of the control device 5 based on the relationship between the data W, the control signal 14 of the timing control mechanism section, the input signal U, and the observable signal y. , the control device 15 uses the reference input signal r, the observable state y, and the control signal 16 from the timing control mechanism 7 to control the control target 1.
Determine the input U to .

The role of the data collection control mechanism 6 is that in the conventional STR system, various calculations shown in FIG. Controlled object 1
The sampling period is short in order to improve control performance, and since it is difficult to calculate parameters for each sampling period, the input U to the controlled object 1 and the observable state y are calculated for each sampling. , collects and accumulates these data, and outputs data q when the identification mechanism 9 is activated.

Note that the data collection mechanism 6 can also apply statistical processing to the collected data to compress the information amount of the data q.

The operation of the identification mechanism 3 will be explained next. Although the controlled object 1 is assumed to be a discrete value system for convenience of explanation, it can be similarly applied to a continuous system.

The model of the controlled object is x (k + 1) = Φ (k) x (k) + T' (k) u (k)
・・・・・・(1) y(k + 1)=C(k)x(k)+D(k)u(
k)......(2) Represented by:

For simplicity, the coefficient matrices Φ(k), r(k).

C(k) and D(k) are assumed to be constant coefficient systems. (1), (2)
The formula is y(k)=Cx(k)+Du(k) −・=(
4).

First, suppose that C is completely observable and C=I.

When D=O, equation (3) multiplies both sides by (x”(k)u(k)) and adds from =1 to N, then Σx (k + 1) CxT(k)uT(k)
)k=1.

If C is not completely observable, then Sagara and Akizuki.

As stated on pages 222 to 236 of ``System Identification'' by Nakamizo 9 Katayama, 2, published by the Institute of Instrumentation and Control Engineers (published in 1981), when the rank of C is known, n −= n −ran k (C)+1, the following equation holds true.

S[γn*(k+1), γ. Fist (k + 2), ・ ,
γll*(k+n+γ.umbrella)]...(6) It is expressed as follows.

The sum of equation (6) from =1 to N becomes 3 and equation (7).

k=1 k=1...
(7) However.

・・・・・・(8) k=1 So [Φ, R) is determined arbitrarily.

Next, the procedure for determining the order n and selection matrix is shown in FIG.

That is, step 101 is a process for finding 100-I, step 102 is a process for finding the minimum dimension of BNψ, and step 1
03 is a process to obtain the selection matrix S, and step 104 is [Φ
, R), step 104 is the step of determining [Φ,
R) is a process for calculating r.

On the other hand, if Φ is found, C can be found from equation (5).

The controllable standard form of the object is shown in Figure 3.

Input using equations (3) to (10) above.

The function of the identification mechanism is to obtain each parameter of equation (1) shown in FIG. 3 in the observable state y.

To explain the design calculation mechanism 11 using the optimal regulator problem as an example, the evaluation function of the optimal regulator is... (11
) think of. This is set externally to the design calculation mechanism 11 or set as a constant of the design computer 11, where Q and R are positive definite symmetric matrices.

In this case, Riccati's algebraic equation % has a unique positive definite symmetric matrix solution P, and the evaluation function J (u)
The optimal human power uo(k) that minimizes is uo(k)=Kx(k) −・=(1
3) However, K=-(rTPT'+R)-"rPo ---
--・(14).

Riccati's solution method for algebraic equations is by Ito and Kimura.

It can be obtained by implementing the method described in Hosoe, "Design Theory of Linear Control Systems", Institute of Instrumentation and Control Engineers Series (1978), pages 97 to 103.

Using the feedback amount K obtained by the design calculation mechanism 11,
FIG. 4 shows the configuration of an integral type optimal regulator.

That is, when the integral type optimal regulator is selected as the design method, the design calculation mechanism 12 calculates a group of parameters by combining the parameters shown in FIG. 4 with Kf.

The determined parameter W is stored together with the input U and the output y in the parameter control storage mechanism 13.

The configuration of the parameter control/storage mechanism 13 is shown in FIG.

Signal 16 from timing control signal 7, parameter W,
The input signal U and the output signal y are input to the control mechanism 20 of the parameter control/storage mechanism 13.

The control mechanism 20 uses the signal 16 to determine whether to store the parameter W in the parameter storage section 21 or to store the parameters stored in the storage section 21 in the control device 15 to configure a control system.

In the parameters stored in the parameter storage unit 21,
Kl is output from the control mechanism 21 to the control device 15 via the output control section 22, and the usage frequency section 23 of the parameter used at this time is updated.

The usage frequency section 23 is provided with a mechanism for monitoring the usage status of the parameter storage section 21.

The control mechanism 2o of the storage mechanism 13 calculates KI from the combination of the input U and output y of the controlled object 1 to the control parameters shown in FIG. However, the identification result does not necessarily match the combination of input U and output y. Therefore,
If they do not match, for example, use the least squares method to calculate Σ(u−u,)z+Σ(y−yt)”=min −(
15) Parameter group K that minimizes the deviation between input and output.

Select Kr. In this case, exact parameters cannot be selected, but optimal control can compensate for parameter errors and
Get the desired results.

Based on the configuration described above, the rolling mill shown in FIG. 6 will be considered as an object to be controlled.

The rolls 50 and 51 roll the input plate thickness H+ΔH to obtain the output plate thickness h+Δh. At this time, the gap between the rolls 50 and 51 is controlled by a rolling down device 52, and the rolling down device 52 is
It operates according to the interval command Sp+ΔSp. Roll 50.
51 is rotated by an electric motor 53, the speed of the electric motor 53 is controlled by a speed control device 54, and the speed control device 54
operates according to the speed command vp+Δvp.

FIG. 7 shows a minute fluctuation model created considering a steady state.

The equation of state in Figure 7 is ・・・・・・(16) shall be.

Here, for simplicity, if we discretize, assume that the coefficient of each matrix is 1, and remove the influence term of ΔH... (1
8).

Initial value of the system x(0) = (OO'3", input series = o, 1, 2..., 5...... (
20) Then, the output becomes.

......(23) Substituting the above equation into equation (3, 2)......(24) If we rearrange [Φ, r] and perform the quasi-conversion, we get the following. As a result, the parameters of the system are determined.

That is, even if the gain and time constant of the transfer function change due to changes in the spring constant due to changes in the type of steel input to the rolling mill, etc., the parameters can be determined by following the above procedure.

Furthermore, the rolling theoretical formula is nonlinear, and the parameters vary widely when the operating point changes. Conventionally, the control gain was set thick to absorb parameter fluctuations, but this had the disadvantage of reducing the margin for fluctuations.However, since the parameters can be selected accurately, increasing the gain increases the margin and increases robustness. Become.

After identifying the control system parameters in this way,
Although the feedback constants are determined, detailed procedures are omitted as they are detailed in the above-mentioned document ``Design Theory of Linear Control Systems'', but feedback coefficients corresponding to the above-mentioned parameters are determined. Now, when the relationship between input and output is given, the corresponding feedback coefficient may not be found. In this case, if there is a similar input/output relationship, it is unlikely that the parameter will deviate significantly from the corresponding parameter, so it is possible to utilize the parameter with the closest input/output relationship.

With such a configuration, the parameters of the controlled object 1 are identified via the data collection mechanism 6 and the identification mechanism 9 in response to parameter fluctuations of the controlled object 1 (for example, due to passing of a welding point of the rolled material in a rolling machine). The parameters calculated by the design calculation mechanism 11 are stored in the parameter control storage mechanism.

The time intervals for these processes are set so that the required accuracy can be obtained and from the processing time capability of the computer.

On the other hand, in order to accurately control the controlled object 1, the control device 15 is preferably controlled at as short time intervals as possible.

From these, identification control is performed using a time sequence with a long period for identification and a time sequence with a short period for control.

In addition, in response to parameter fluctuations of the controlled object 1, the parameters of the control device 15 can be retrieved from the parameter control/storage mechanism 13 based on the relationship between the input U and the output y. Since the processing has been carried out, it is possible to discard parameters that are no longer needed.

As a result, optimal control is possible even if the controlled object changes due to various factors.

〔Effect of the invention〕

According to the present invention, parameters can be identified online and control constants can be changed for a controlled object having multiple inputs and multiple outputs, so that the control system can be operated without adjustment.

[Brief explanation of drawings]

Fig. 1 is a block diagram of a control method according to an embodiment of the present invention, Fig. 2 is an example of a conventional control method, Fig. 3 is an example of a mathematical expression of a controlled object, and Fig. 4 is a control device of Fig. 2. 5 is an example of the control storage mechanism shown in FIG. 2, FIG. 6 is a schematic diagram of a rolling mill, and FIG. 7 is an example of the transfer function shown in FIG. 6. 1... Controlled object, 7... Timing control mechanism, 3.
...Identification mechanism, 4...Design calculation mechanism, 13...Parameter control/storage mechanism, 5...Control device.

Claims (1)

[Claims] 1. In a multi-input, multi-output plant control method that detects a plurality of state quantities of a plant to be controlled and performs control by outputting a number of control outputs to the controlled object, The method is characterized in that parameters of the controlled object are identified from the relationship of outputs, parameters of a control device for controlling the controlled object are determined by calculation using the identified parameters, and the parameters are changed and determined online. A plant control method that 2. In claim 1, a plurality of parameters are stored as calculated parameters of a control device for controlling the controlled object, and the storage is performed based on the input/output relationship of the controlled object. 1. A plant control method, characterized in that the selected parameters are selected from among the parameters that have been set and used as parameters for the control device. 3. A plant control method according to claim 2, characterized in that a writing cycle for storing the parameters and a cycle for reading or changing the parameters of the control device are different. 4. A plant control method according to claim 2, characterized in that the frequency of selected use of the parameter is measured and the parameter is managed. 5. In claim 2, if the input-output relationship at the time of using the control parameters has not been determined before then, the plant uses data similar to the known input-output relationship as the control parameters. Control method. 6. A plant control method according to claim 5, which selects a control parameter that minimizes a deviation between a known input-output relationship and a corresponding input-output relationship at the time when the control parameters are used. 7. A plant control system according to claim 1, wherein the input/output relationship is collected and a storage means is provided, thereby making it possible to change the time for determining the input/output relationship and the timing of using the parameters.