JPS59186002A - Plant controlling device - Google Patents

Abstract

PURPOSE:To make stable plant control coping with time change of plant characteristic by providing a control inference section that receives information from a plant characteristic change detecting section and determines controlling method by inference basing on knowledge on plant control. CONSTITUTION:A difference signal calculating section 1 sends a difference signal (b) between set value (a) of reactor water level and measured value (d) of reactor water level to a controlling section 2. The controlling section 2 operates the signal (b) and outputs controller output (c) that determines number of rotation of a water supply pump in the reactor water supply system 3. A plant characteristic change detecting section 4 takes in output (c) and measured value (d) successively while the system 3 is in operation and appraises with plant operation status quantity (e) such as reactor water supply flow rate etc. from the system 3, and compares model with actual dynamic characteristic. When there is large discrepancy, the model is determined again. A control inference section 5 receives dynamic characteristic information (f), continues simulation until stable time transition of reactor water level is obtained, and rewrites rule for control. Numerical values of rewritten control rule are sent to the controlling section 2 as controlling section revised information (g).

Description

[Detailed description of the invention] [Field of application of the invention] The present invention relates to a plant control unit #.

[Prior art]

When a plant becomes a multi-input, multi-output system, is it best to use the conventional PID and other arithmetic methods for its nailing system?
Although it is difficult to realize, a control method in which operations are determined by inference according to the plant state is effective.If the control characteristic r of the control unit @ is fixed and used for plant control f11, If there are fluctuations in plant characteristics, control characteristics will deteriorate. This situation? In order to avoid this, adaptive control has been proposed in arithmetic control methods such as PID, which adjusts parameters of the side wheel I device in accordance with temporally changing plant characteristics. The objects to which this method can be easily applied are limited to linear systems with single-input and multiple-input multiple-output systems; in the case of multiple-input and multiple-output systems, the configuration of the control device is complex and the real-time It is difficult to realize adaptive behavior. On the other hand, for the multi-input/output system side landing gear system, Fukagawa developed a control method that is determined by inference of manipulated variables from the plant state, based on the knowledge gained from experience regarding plant interlocking and control. However, even when implementing adaptive control using the I (I oblique method) based on this inference, conventional parameter adjustment methods cannot be applied, so adaptive control of multi-input, multi-output systems cannot be applied.
There was no way to make it happen.

[Purpose of the invention]

The purpose of the present invention is to provide a plant side leg device suitable for the change in plant characteristics over time in the direction of plant ti+lJ [corresponding to the current state of stable control characteristics]. It is about providing.

[Summary of the invention]

The purpose of the above? In order to achieve this, the device of the present invention performs repeated learning to reconfigure and apply control elements while driving the identified plant dynamic characteristic model every time there is a change in plant characteristics. Optimal configuration of control elements with high speed inference? It has seven functions of determining.

[Embodiments of the invention]

Hereinafter, an embodiment of the present invention will be explained with reference to FIG.

FIG. 1 is a schematic diagram of an example of the plant control system of the present invention applied to a KBWR reactor feed water control system.

The reactor water supply system 3 receives information from the control unit 2 regarding the opening degree of the water supply valve? In response to the side leg gain output Ct to be determined, the difference signal calculation unit 1 outputs a difference signal b between the reactor water level setting 111t a and the reactor water level measurement d2 received from the reactor water supply system 3 to the control unit 2. send. The control unit 2 also controls the difference signal b? f: Outputs the controller output ci which determines the rotation speed of the feed water pump in the reactor water supply system 3 through arithmetic processing. A feature of the present invention is that the following two components are added to the i configuration of the closed loop control system above.

(1) Plant characteristic change detection section 4 (2) Control inference section 5 An example of the processing procedure of the plant characteristic change detection section 4 and control inference section 5 in this embodiment will be described below.

The processing procedure in the plant characteristic change detection unit 4 is shown in a flowchart as shown in FIG. The plant characteristic change detection unit 4 receives the controller output C5 and the actual reactor water level value d2 from time to time while the reactor water supply system 3 is in operation. Evaluate the recirculation flow rate, reactor main steam flow, etc. together with the plant operating state sensitivity, and compare the dynamic characteristics of the reactor water supply system model with the actual reactor water system dynamic characteristics to determine the degree of discrepancy. ?
For example, when the difference between the measured reactor water level and the model's reactor water level setting becomes 5% or more in the relative cage, the reactor water supply system model? Re-identify using actual measured values and plant operating state variables.

Identification methods include (1) a method of describing plant dynamic characteristics using a Kurayoshi differential equation with a finite number of parameters according to an all-physical model, and identifying the parameters; (2) a method of identifying model parameters using a linear autoregressive model; Of course, the method of identifying, etc. can be applied, but if the method (2) above is used in this embodiment, the dynamic characteristics of the reactor water supply system 3 are described as follows.

+ WFS WM ”・−(2) Here, t is the reactor water level, Wy is the water supply flow rate, WrsVi is the water supply flow rate, WR is the reactor decay circulation flow before, and WR8 is the reactor recirculation flow i. , rated value, WM is reactor main steam K, m'd
iigs Ha output, “I * a2+ a3+゛τM,
τF is a model parameter. The model parameters are
9 is selected for the case where the time course of the reactor water level determined by the above equation (1) fits the time course of the reactor water level central fAtI value. Identification of the reactor water supply system after characteristics change ff1
The model parameters of the reactor water supply system model are sent to the control inference unit 5VC as reactor water supply system dynamic characteristic information f every time the plant characteristic change detection unit 4 finishes processing the same model. Start the process.

Next, an exemplary embodiment of the processing of the inference unit 5 will be described below. The control inference section 5 is a central section for achieving the object of the present invention, and has the function of controlling the control section 2r.

Upon receiving the reactor water supply system dynamic characteristic information f from the plant custom-made change detection unit 4, the control reasoning unit 5 performs the following steps (, 1).

(1) Reactor water supply system - parameters included in characteristic information f? f: Assigned to the reactor water level time transition calculation unit that calculates the reactor water level in the side landing gear inference unit.

(2) Prepare the control unit 2 that was used before the change in plant characteristics by expressing the knowledge.

(3) The Kawabe Shida Seiken rules in (2) are used to determine the human power (controller output m) in the simulation using the reactor water (i time transition calculation unit) created in (1), and the following process is performed. This is done by searching for rules (inference processing).

(a) Nomulation until the reactor water level shows unstable behavior 4? continue.

(b) Infer the type of unstable behavior I7J and the control law that caused it.

(C) 1- Make a simplistic inference on how to correct the part of the control law that should be corrected as a result of (b).

(d) When the reactor water level is unstable, the simulation can be carried out using the corrected 1ltlJ8 rule and return to the previous time and summer L (values shown above).

(e) If unstable behavior appears, repeat the above correction work until a stable reactor water level change over time is achieved.

Concrete l The above inference processing (Il-driving rules and 7
, old)' is an example of the expression, when the reactor water level r is compared stepwise, the valve opening adjustment used for the water supply j'L constant A Hj in the reactor water supply system. )
If there is a characteristic of - next delay and a sign reduction between delay - 1, 1IIIJ (goods) rule σ) 4th correction is performed? The targets are shown below.

Simulation k hs 4n f rule (If
(AND (Vxsxst) (■u$u$t) (=$tt (-1-6t time)) (simu
la, tor $X 8u gxx)) (y x sx
x 5tt)) When the water level is $X for two hours, if the control wholesale output $ur per unit time is input, the water level will be g xx at time $tt.

(I f −(AND (Ve 5eea sad
5tt) (VX $XX @tt ) (VX $X St ) (=$ t (-1-$ t t t t in+e )
)(2$ e(-1, o 5x)) (=$d (/(-1e 5ee) tin]e))(
check @e gd St $C)) (Ve $e
a scJ 5t)), advance the simulation to time $t, and before searching for the next constraint, calculate the difference from the difference $e between the water level $ If the behavior is stable, use it as a condition for the $e gd 2 control law search. If it is unstable, control law modification processing is performed depending on the type of unstable behavior, and a complete rule search is started.

(If' +AND (Ve$e dad I
t) fVL1$u, (e gel Be2) (d s
ai sd2) (> se 5ei) <<se 5ez
)(>sd 5a1)(<sd sd2)(Vu$LI
$t)) ; Lateral leg rule search? From the conditions of Be and sd at time $t,
The side leg law busy behavior stored as knowledge, the output $u of O units of time at time $t? decide. Note that the determination of the control talk output in the side leg section 2 is also based on Be and sd.
Infer directly from this.

An example of a control law that has been replaced as knowledge'('V u PBl (eo, s tatami) (do, 1
(tatami)); When o, s<se and 0, l<sd, it is assumed to be the side leg protrusion break PBI per unit time. PBxVc
is a number (directly substitute 2.0).

'(Vu NM(eO,20,8)(d4$-0,l)
): 0.2<Be<σ, 8 and sd<-Q, l (At 7J, the controller output kNM(-1) for each unit time.

Example of reactor water level unstable behavior detection rule (If (AND (Ve 5eed $dd$t
t) (=-8t (-14tt time) )(>
Be O,0) (>sd O,0) (<$dd 0.0>) (checJ< Be sd 6t cl)n;
The type of unstable/unstable behavior in which the reactor water level rises, approaches the set value, and then continues to fall is named C1.

Control law “modification part” Rules for inference (I f (
AND H) u 8u, st @c) (V j
ump su ($V $W $X)) (V mo
diby su ($v $w$x)It $c);
The type of unstable behavior $C is determined by comparing the control law that resulted in a favorable water level rise time transition t and the control law that led to unstable behavior.
K, therefore, $W v) #i, 2 amended.

Good water level rise time transition? Example of control law knowledge to be provided (VjumpNM(eO12d)): After su reaches NMk, the reactor water level moves in the direction in which e decreases with Be at the boundary value of 0.2 Te.

Example of rule for implementing control law modification (If (AND (Vmodjby su
(e$wd7$t C1) (V u su Ie $
W Be2) (d $dl sd2)) (<su
o) (= $tt (-It time)) (1) e Be d $ (1 It C1) (V e $
ee d $dd It t) (= , $W f/(+
$e 5ee) 2. L))))(AND(St (V
u su (esww$e2) (cl 5ctisc
tz)n(del (vusu (e se 15ez)
) (d $dl 1Bd2) ) (rewind I
t)): When the type of unstable behavior is cl and su is negative, Be
1. In order to speed up the red line, the control law is set so that the threshold value I comes between Be and $ee at the previous time? Rewrite and erase previous knowledge.

Go back to time t again and continue the simulation.

As a result of the simulation, when a good response is obtained and there is no longer a need to change the control law, the reactor water level unstable behavior detection rule (If (AND (
Vu$LI It) (= $01.0) (>$ee,) (<ge e,) (>Bdao) (<$d-do) ) (chec, k Be sd It 5table))
? Then, the inference process ends. Rewritten 1DI
The J number law c/J number side is sent to the control unit 2 as side leg eruption positive information g and substituted into the parameter of f17Il Misakija.

A number of examples according to this embodiment are shown in FIGS. 3 and 4.

Figure 3 shows that the reactor water level system responded well to a 20% water level rise disturbance? 1 is shown.

In Figure 4, when the control law that gave the good response characteristics in Figure 3 is applied to the system after the characteristics have changed, it shows unstable behavior as indicated by failure. , the unstable behavior was detected around 24 seconds after the simulation, and it was concluded that the unstable behavior was the type c1, and the control law knowledge'(Vu NM (e O, 20,
8) 0.2 on the 0.2 part of (d'--0,1) ).
By modifying K to 3, the stable response r shown by the solid line (1)
This shows the results of the inference processing that led to the given result.

As can be seen from the above calculation, the ability of the control reasoning section 5 has the effect of being able to respond more flexibly to changes in plant timing.

Historically, the target flint that can be used is not limited to the reactor water supply system, but also suitable for the reactor recirculation slope flow site Kirimoe J system.
1 can be done. Due to the non-linear customization, which depends on the rotating branch of the fluid flow control hand, when changing the output at different reactor outputs, on the other hand, the 161J gave a nice response.
Output change of other output level V? If X is used, the response may deteriorate, but with the device of the present invention, '+l1l
J groan] Good 'lx fIIIJ by Noh Tawara of IL theory department
The effect is that automatic correction to the control opening side that gives the vertical characteristics can be performed.

[The young fruits of invention]

The present invention is a plant control system for "-General L1", and the control department's side leg rule inference VC! whenever the characteristics of the plant change. Since the correction is made by recursive learning, stable control characteristics can be obtained over all degrees and types of changes in plant characteristics even after changes have been made. Furthermore, the present invention has the advantage that adaptive operation of plant control having multiple inputs and multiple outputs can be realized.

[Brief explanation of drawings]

Fig. 1 shows an overview of an example of a plant control system using the plant control device r of the present invention. ′
i Control by an example of a plant control device according to the present invention II
FIG. 4 is a diagram showing an example of control by an example of the plant-side leg device according to the present invention. l...Difference old name calculation section, 2...Side leg section, 3...Reactor water supply system, 4...Plant characteristic change detection section, 5...
・System @] 1 picture of rice porridge

Claims (1)

[Claims] 1. A difference signal calculation unit between the output of the plant and the output setting;
In a control device consisting of a control unit that receives the calculated difference signal and determines the manipulated variable r to the plant, it includes a plant characteristic change detection unit that detects changes in plant characteristics from the input/output relationship of the plant, and a plant characteristic change detection unit. A control inference section that receives plant characteristic changes as plant control characteristic information from the control section and inferentially determines a control method for the control section based on knowledge about runt control from this information. A plant control device with the following features: 2. In the plant characteristic change detection section of the plant control device, a plant dynamic characteristic identification section that identifies plant dynamic characteristics using the plant state tv, and a plant dynamic characteristic simulation line that performs simulation calculations based on the identified plant dynamic characteristics. ? The difference signal between the plant output and the plant simulation output from the f7U characteristic simulation unit (plan) is used as a plant characteristic change detection signal, and the magnitude of this difference is completely determined and activated on the plant dynamic characteristic identification unit to detect the newly identified durant movement. Characteristic? Should it be output to the control inference unit as plant control characteristic information? A plant control device according to claim 1, characterized in that: 3. In the fijU control section of the plant control device, the manipulated variable is determined from the input difference signal and its time change value. 2. The plant control device according to claim 1, further comprising a storage unit for storing the determined control law as data, and a function of referencing the control law in inference processing. 4. The control inference section of the plant control device receives the control characteristic information r from the plant characteristic change detection section, inputs the manipulated variable t from the control section to the plant, and performs independent control in the plant dynamic characteristic simulation section. Amount of operation from the department? Switching function to guide?
Have you prepared? A plant control device according to claim 1, characterized in that: 5. Operation amount from an independent control section in the control inference section of the plant control device? Based on the plant simulation output from the plant dynamic characteristics simulation unit used, it is determined whether the control law correction processing of the control unit is complete and the independence of the control unit is determined. The plant groan control device as set forth in claim 1, characterized in that it is provided with a function of releasing.