JPH04310102A - Control method and device - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a control method and apparatus, and more particularly to a control method and apparatus suitable for controlling the same control object using a plurality of different control methods.

0002

[Prior Art] As a conventional technology that employs a plurality of different control methods to control one control object, Japanese Patent Application Laid-Open No. 1999-1-1
There is one described in Publication No. 77604. In this conventional technology, when the deviation between the target value and the controlled amount is large, control is performed using a control device with excellent quick response, and when the deviation is small, the control device is switched, and a control device with excellent stability is used. I'm trying to control it. Although this prior art uses two different control devices, both control devices belong to linear control.

Fuzzy control and neural network control, which are suitable for nonlinear control, have been attracting attention as control techniques in recent years. This nonlinear control device is good at imitating the control operations of an expert by learning, but it has not yet been established because there is insufficient data to learn the conditions for situations that even experts rarely encounter. It cannot be said that it is a sophisticated technology. Therefore, it is used in combination with conventional linear control, which is an established technology. For example, JP-A-1-2
In the conventional technology described in Publication No. 50103 and Japanese Unexamined Patent Publication No. 2-8903, a fuzzy control device and a linear control device are used together, and when one controlled object is stable, PID control, that is, linear control is performed. Fuzz when it gets bad
y control is being performed.

0004

[Problem to be Solved by the Invention] For example, when controlling the rotation speed of a loaded motor to a target value, the conventional technique described in Japanese Patent Application Laid-Open No. 1-177604 mentioned above is applied, and the actual rotation speed and target value are controlled. By using a configuration that uses a control device with excellent quick response when the deviation from be able to. However, there is no guarantee that overshoot will not occur when switching control devices. Further, in the above example, since there is only one state quantity to be controlled, "rotation speed deviation", it is easy to configure a control device corresponding to it. However, in large-scale plants such as chemical plants, steel plants, and power generation plants, there are many state quantities to be controlled, and controlling all of these state quantities with a linear controller requires complicated control device settings. . Furthermore, in plants, etc., there are many non-linear controls in which the plant state quantity rapidly approaches a target value, such as when the plant is started up, and when it approaches the target value, it is stably controlled to the target value. Therefore, it is desired to introduce fuzzy control and neural network control, which are effective in nonlinear control.

[0005] The prior art described in Japanese Patent Application Laid-open No. 1-250103 and Japanese Patent Application Laid-open No. 2-8903, which introduce fuzzy control, use a fuzzy control device as a backup in case of an emergency. However, in this case, the characteristics of fuzzy control, which is effective in nonlinear control such as during startup, are not utilized. Furthermore, it is difficult to collect the amount of data that determines the membership function when an abnormality occurs in the plant and it is unclear what will happen, and fuzzy control for such a situation is still not an established technology. More importantly, it does not guarantee that the amount of transient response, ie, overshoot or undershoot, will be within a predetermined tolerance when switched to backup.

An object of the present invention is to provide a control method, device, and state quantity display method that takes advantage of the characteristics of nonlinear control devices such as Fuzzy and neural networks, and also places emphasis on safety, and can guarantee a transient response that always falls within an allowable range. Our goal is to provide the following.

[0007]

[Means for solving the problem] The above purpose is to
A nonlinear control device such as the above and a conventional linear control device that performs PID control are used together to preset a closed region in the space (state space) formed by the state quantities in the closed loop system between the linear control device and the controlled object. However, this is achieved by having the nonlinear control device perform control when the state quantity is within this closed region, and by switching to the linear control device when the state quantity is about to go outside the region.

[0008]

[Operation] Even with a non-linear control device, control within the closed region where a sufficient amount of data can be collected to determine the membership function can be controlled as reliably as a linear control device. can. Furthermore, this can be said to be a control region suitable for the characteristics of the nonlinear control device. When the state quantity deviates from this closed region, the stability is controlled by a reliable and conventionally established linear control device, so safety is high. Furthermore, since the linear control device is not switched to the linear control device after the state quantity leaves the closed region, but is switched to the linear control device when it is about to go out, problems such as overshoot do not occur.

[0009]

DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a configuration diagram of a control device according to an embodiment of the present invention. The control device according to this embodiment is a Fuzzy
A nonlinear controller (shown as controller 1) 11 such as a computer or a neural network, and a linear controller (shown as controller 1) 11 that performs conventional PID control.
It is illustrated as a controller 2. ) 12 and contribution rate determining means 13, the outputs of both controllers 11 and 12 are determined by the contribution rate determining means 13.
The signal is supplied to the plant 14, which is the object of control, via the. Both controllers 11 and 12 take in plant state quantities including the output (control amount) of the plant 14, and each generates a control output for independently controlling the plant 14. On the other hand, the contribution rate determining means 13 monitors both the internal state quantities of each controller 11 and 12 and the state quantity of the plant 14, and contributes the control outputs from both controllers 11 and 12 in an appropriate ratio. , generates a control output value. This control output value becomes the manipulated variable actually input to the plant 14.

FIG. 2 is a block diagram of one embodiment of the above-mentioned contribution rate determination means 13. In this embodiment, each controller 11,
The contribution rate of 12 is set to 0% or 100% by the switch 21.
It is decided to be one of the following. In other words, control during normal operation of the plant 14 is performed by the output of only the nonlinear controller 11, and as will be described in detail later, when the plant state quantity is about to exceed a certain range, control is performed only by the output of the linear controller 12. The plant 14 is controlled by the output. Note that instead of setting the contribution rate between 0% and 100%, a configuration may be adopted in which the contribution rate is switched in stages between the two. That is, if large changes in control input are disliked, a configuration may be adopted in which the control input is switched from 0% to 100% step by step as soon as possible. In any case, the timing of this switching is determined by the switching determining means 22.
This means 22 monitors and judges each state quantity as described above. The basis for the switching decision made here,
The principle will be described in detail below.

The basis for determining the switching is the transient response formed by the closed loop between the linear controller 12 and the plant 14, which becomes effective immediately after switching. Therefore, the previous Fuz
Switching is performed regardless of what kind of control the nonlinear controller 11 such as zy is executing. This allows the linear controller 12 to function as an excellent backup means. The problem here is that
The test is whether the transient response of the loop formed by the linear controller 12 and the plant 14 falls within a predefined range of overshoot and undershoot. Generally, the transient response of a linear system is influenced by the initial value of the state quantity at the start of control. Regarding this point, I would like to describe the findings related to the present invention next.

In the following, linear controller 12 and plant 14
The state quantities including x1, x2, . . . are expressed as x1, x2, . . . , and when simply written as Further, it is assumed that the output of the linear controller 12, that is, the control input to the plant 14 after switching, is represented by u1, u2, . . . or collectively as a column vector U.

Now, the differential equation regarding the state quantity X of the plant, that is, the state equation is:

[0014]

[Math 1]

##EQU1## where A and B are system matrices characterizing the dynamic characteristics of the plant 14. On the other hand, the linear controller 12 generates an output U according to Equation 2 below.

[0016]

[Math 2]

[0017] Here, F is a feedback gain matrix. The following equation 3 is obtained from these two equations 1 and 2.

[0018]

[Math 3]

[0019] Here, Ac is a system matrix corresponding to a closed-loop system. When this equation 3 is solved under the initial value X(0), a transient response X(t) expressed by the following equation 4 is obtained as the solution.

[0020]

[Math 4]

As shown in Equation 4, the transient response of the state quantity X depends on its initial value X(0), and even if the closed-loop stability of the linear control system is guaranteed by the value of the matrix Ac. However, depending on the initial value, large overshoots or undershoots may occur. The initial value referred to here is the value of the state quantity at the moment when control is switched from the nonlinear controller 11 to the linear controller 12. Therefore, in order to ensure that the transient response falls within a pre-allowed range, at least attention must be paid to the conditions that the state quantity must satisfy at the time of switching. If we were to exhaustively examine all possible state quantities, there would normally be several to several dozen state quantities, so the number of combinations would be enormous. In other words, it is necessary to check the transient response using Equation 4 using each point on a state space of several tens of dimensions as an initial value, which is extremely difficult to implement.

However, by utilizing the convexity of the initial value dependence of the transient response, it is possible to find the conditions that the initial value should satisfy using an extremely simple procedure. This convexity is demonstrated as follows. First, point Xα on the state space
Transient response Xα(t) with (0) as the initial value (the following formula 5
), and the transient response X with Xβ(0) as the initial value
Assume that β(t) (shown in Equation 6 below) falls within the allowed range.

[0023]

[Math 5]

[0024]

[Math 6]

Next, the midpoint Xγ on the straight line connecting both initial values
Let us consider the properties of the transient response Xγ(t) with (0) as the initial value. The midpoint Xγ(0) is defined by Equation 7 below.

[0026]

[Math 7]

According to the above equation 7, as shown in the following equation 8,
The transient response Xγ(t) is also expressed by both transient responses Xα(t) and Xβ(
It can be seen that it has properties intermediate between t).

[0028]

[Math. 8]

That is, neither overshoot nor undershoot exceeds the former two (Xα(t), Xβ(t)), and always falls within the allowable range. From this, it immediately follows that all points on a straight line connecting two points that are acceptable as initial values are acceptable as initial values, and furthermore, the set of such acceptable points is a convex region in the state space. It is also obvious that this will happen. Assuming that the state quantities are X1 to X3, the image will be as shown in FIG. 3.

In practice, it is convenient to implement as shown in FIG. That is, in FIG. 4, the convex region is represented by a polyhedron,
Points P1, P2, found along the coordinate axes on the state space
... are connected to each other to form a polyhedron. In order to find these points, the transient response is confirmed by simulation or the like, but it is sufficiently practical because it is sufficient to find at least twice the number of points as there are coordinate axes.

As a more specific example, a two-dimensional (two state quantities) plant shown by the following equation 9 will be explained. There are two types of state quantities, x1 and x2, and x2(t
) is unrestricted, but the response of x1(t) is assumed to have a tolerance range of −1 to +1 with zero as the reference.

[0032]

[Math. 9]

The linear controller 12 performs feedback as shown in equation 10 below.

[0034]

[Math. 10]

[0035] This gain [100, 20] is selected in advance so that the eigenvalue of the closed loop is a double root of -10.0. Therefore, although it is sufficiently stable as a closed-loop system, the transient response varies depending on the initial value. In Figure 5, x1
, x2 as the coordinate axes, and the results of calculating the transient response of x1(t) using several points on the state space as initial values are also shown.

As shown in FIG. 5, two points (1,0) and (-1,0) are allowed on the x1 coordinate axis (-
(falls within the range of 1 to +1). On the other hand, on the x2 coordinate axis,
(0,28) and (0,-28) are acceptable limits. When these four points are connected to form a quadrilateral, everything on its periphery and inside is allowed. As an example, the results of calculating the transient response by newly selecting four points on the circumference are also shown in the same figure.

The method described above may be further modified in several ways. First, using a circle or an ellipsoid included in a polyhedron as a convex region in the state space is an effective technique for speeding up calculation processing. In other words, in the case of a convex polyhedral region, in order to determine whether a point is inside it, it is necessary to examine as many linear inequalities as there are faces, but in the case of a circle or an ellipsoid, one quadratic inequality must be examined. Since you only need to look up the inequalities, you can save time and effort in calculations.

[0038] When performing sample value control using a digital calculator, the state quantity at the time of a sample one step in the future is predicted using a state equation, and the value of the predicted state quantity is used to determine switching. It is effective to do so. In addition to the embodiments that have been described so far that automatically determine switching, we also display the position of the state quantity on the state space to the operator, and when the operator is about to exceed the allowable range (when approaching the boundary), ) An example of generating an alarm is also conceivable. Such displays and alarms are also effective in plant control panels that use only conventional linear controllers, and the configuration of such an embodiment is shown in FIG.

In FIG. 6, the state quantity monitoring means 62 monitors the state quantities of the plant 14 and the controller 61, and plots the values of the state quantities in the state space and displays them on the display device 62. When approaching the boundary of the above-mentioned tolerance range, the alarm generator 63 is used to notify the operator. In this embodiment, the controller 61 is a linear controller, and the allowable range is set according to this controller 61. When some kind of large disturbance occurs, the state quantity momentarily shifts significantly,
This is an effective method for generating an alarm when it is predicted that if control is continued, overshoot or undershoot will exceed the allowable range.

FIG. 7 shows yet another embodiment in which the present invention is applied to water supply control for a nuclear reactor. In Figure 7, plant 7
1 represents the dynamic characteristics of the water supply system including the pump. It includes three state quantities (x1: water level deviation, x2: flow rate deviation, x3: flow rate deviation change). On the other hand, the control device includes a fuzzy controller 72 and a linear controller 73, which are switched and used by a state quantity monitoring/switching means 74. The switching means 74 uses a display means 7 to clearly explain to the operator the basis for switching and the condition of the state quantity.
It is equipped with 2.

The linear controller 73 is a so-called PI controller, and has a proportional gain Kp and an integral gain Ki. For this purpose, a state quantity x4: water level deviation integral value is included inside. The state equation of the plant is as shown in Equation 11 below, and the linear feedback law is as shown in Equation 12.

[0042]

[Math. 11]

[0043]

[Math. 12]

These equations 11 and 12 correspond to the above-mentioned equations 1 and 2, respectively. Similar to the method described above, x1 to x
A 16-hedral region can be created in the state space with 4 as the coordinate axis. If the plant state is inside this region,
Neural network control and Fu that are trained to operate by experts
Perform nonlinear control using zzy control.

As in this embodiment, when a linear controller is used as a backup means in case of an emergency, the normal Fuz
It is an essential requirement to provide so-called transparency so that the operator is clearly informed on what basis the zy controller is switching over. An example of the display screen of the display means 72 for this purpose is shown in FIG. A polyhedron in a four-dimensional space is divided into two two-dimensional coordinate spaces and displayed. The combination of coordinate axes for this division is the same no matter how you decide, but as an example,
x1, x2 and x3, x4 are each combined. The bright spots displayed in each two-dimensional space represent the values of the current state quantities, and the rectangular frames indicate the planes that cut the aforementioned polyhedron with the planes that include the current state quantities. There is. In other words, quadrilaterals P1 to P4 are polyhedrons cut by planes parallel to the x1 and x2 axes, where x3 = (x3 value of current state quantity), x4 = (x4 value of current state quantity) . Therefore, the size of this quadrilateral changes depending on the current value of the state quantity. Similarly, for quadrilaterals P5 to P7, the results of cutting the polyhedron along planes parallel to the x3 and x4 axes are shown. If a bright spot is inside both quadrilaterals, fuzzy control can be used with confidence, and the operator can also confirm this.

As yet another example of display, there is a three-dimensional graph shown in FIG. This becomes effective when the control law is determined on the premise that the state quantity x4 is always started from zero immediately after switching to linear control. in this case,
Since it is sufficient to consider a three-dimensional polyhedron obtained by cutting the four-dimensional polyhedron described above along the plane of x4=0, it is easy to visually display this. If the internal state quantities of the plant are even larger than those in this embodiment, it is also effective to divide the graph into a plurality of three-dimensional graphs and display them.

It should be noted that the present invention is applicable not only to the above-mentioned power plant water supply control system, but also to general control including other nonlinear controls. For example, it can also be applied to output control of a nuclear power plant. In this case, the state quantities used for control are preferably main steam flow rate deviation, core pressure deviation, neutron flux deviation, core flow rate deviation, feed water flow rate deviation, etc. When applied to control of the rotation angle of a motor, the rotation angle deviation, rotation angular velocity, differential value of the rotation angle deviation, and integral value of the rotation angle deviation are used as state quantities. In addition, when applied to control the rotational angular velocity of a motor,
Let the rotational angular velocity deviation and the integral value of the rotational angular velocity deviation be the state quantities. Furthermore, when applied to active suspension control of automobiles, vibration displacement, vibration velocity, and vibration acceleration are used as state quantities. By switching between nonlinear control and linear control using these state quantities, or
By changing the contribution rate, automatic control that is highly safe and imitates control by an expert becomes possible.

[0048]

[Effects of the Invention] According to the present invention, it is possible to guarantee a transient response that always falls within the allowable range for any controller including fuzzy, neural net, etc., and it is possible to improve safety. .

[Brief explanation of drawings]

FIG. 1 is a configuration diagram of a control device according to an embodiment of the present invention.

FIG. 2 is a configuration diagram of contribution rate determining means shown in FIG. 1;

FIG. 3 is an explanatory diagram of the operating principle of the present invention.

FIG. 4 is an explanatory diagram of a specific implementation method of the present invention.

FIG. 5 is an explanatory diagram of a specific operation of the present invention.

FIG. 6 is a configuration diagram of a control device according to another embodiment of the present invention.

FIG. 7 is a configuration diagram of a control device according to yet another embodiment of the present invention.

FIG. 8 is a configuration diagram of a display screen in the control device shown in FIG. 7;

FIG. 9 is a configuration diagram of a display screen different from that shown in FIG. 8;

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 11... Nonlinear control controller, 12... Linear controller, 13... Contribution rate determination means, 14... Plant, 62... State quantity monitoring means, 63... Display device, 64... Alarm generator.

Claims (14)

[Claims] Claim 1: When controlling a controlled object by two or more control means, a convex polyhedral region is set in advance in a space (state space) whose coordinate axis is the state quantity of the closed-loop system, and the state quantity of the closed-loop system is , a control method characterized in that the contribution of each control means output is adjusted so as not to go out from inside this region. 2. When controlling a controlled object by two or more control means, a hyperellipsoidal region on the state space is set in advance in the state space of the closed-loop system, and the state quantity of the closed-loop system is set in advance in the state space of the closed-loop system. A control method characterized in that the contribution of each control means output is adjusted so that the output does not come out from inside. 3. The control method according to claim 1, wherein the method for setting the convex polyhedron region includes first setting permissible limit points on each coordinate axis, and forming a convex polyhedron with these points as vertices. 4. A system comprising a controlled object and its control device, wherein means for presetting a convex polyhedral region or a hyperellipsoid region in a state space of a closed-loop system, and a means for presetting a convex polyhedral region or a hyperellipsoid region in a state space of a closed-loop system; 1. A control device characterized by comprising: means for generating an alarm when the device exits or is predicted to exit from the inside. 5. A system comprising a controlled object and its control device, comprising: a setting means for presetting a convex polyhedral region or a hyperellipsoid region in a state space of a closed-loop system; A control device comprising means for displaying the value of the state quantity together with the region set by the setting means so that it can be seen whether the state quantity is within the range. 6. In a control device comprising fuzzy control means and linear control means for controlling a controlled object, and means for switching between the two control means, a convex closed region in the state space of the closed-loop system is set in advance, and the closed-loop system A control method characterized in that when the state quantity of is within this region, the control method is switched to the fuzzy control means, and when the state quantity is outside the region, the control method is switched to the linear control means. 7. In a control device comprising neural network control means and linear control means for controlling a controlled object, and means for switching between both control means, a convex closed region in a state space of a closed loop system is set in advance, A control method characterized by switching to a neural network control means when the state quantity of a closed-loop system is inside this region, and switching to a linear control means when it is outside this region. 8. A water supply control system control device for a nuclear power plant, comprising a fuzzy control means, a linear control means, and a means for switching between the two control means, the water level deviation, the feed water flow rate deviation, the change in the feed water flow rate deviation, and the water level. A convex closed region is set in advance in a space whose coordinate axis is the integral value of the deviation, and when these values are inside this region, Fu
A control method characterized in that a zzy control means is switched to a linear control means externally. 9. A control device comprising at least two types of control means, a nonlinear control means and a linear control means, and controlling a controlled object by switching between the two control means in accordance with a plurality of state quantities of the controlled object. means for setting a required convex region in a multidimensional space defined by state quantities; and when the state quantity of the controlled object is within the convex region, the controlled object is controlled by the nonlinear control means; A control device comprising: a control means switching means that causes the linear control means to control a controlled object when the state quantity of is outside the convex region or is predicted to be outside the convex region. 10. A control device that controls a controlled object according to a plurality of state quantities of the controlled object, wherein a multidimensional space defined by the plurality of state quantities is displayed in coordinates, and the state quantity of the controlled object is displayed on the coordinates. A state quantity display method characterized by displaying the position of. 11. In a control device to which the state quantity display method of claim 10 is applied, when the state quantity of the controlled object leaves the closed area set on the coordinates or is predicted to leave the closed area. A control device characterized by comprising means for issuing an alarm. 12. In a control device that controls a plant and includes a fuzzy controller and a PI controller, the fuzzy controller controls the plant under normal operating conditions of the plant so that the state of the plant is maintained at the normal operating state. A control device comprising controller switching means for causing the PI controller to control the plant when the PI controller is out of the operating state or when it is predicted that the PI controller is out of the operating state. 13. The control device according to claim 12, further comprising display means for informing a plant operator whether the controller controlling the plant is a fuzzy controller or a PI controller. 14. Claim 12, wherein the controller is a PI
and means for outputting a switching command to the controller switching means at a plant state quantity such that the control initial value of the PI controller when switched to the controller is a value that does not cause overshoot or undershoot due to PI control. A control device characterized by: