JPH0656640B2 - Abnormality diagnosis method of plant - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial field of application] The present invention has a large number of parameters to be monitored, and a plant abnormality detection of a nuclear power plant, a thermal power plant, etc., in which these parameters are mutually complicatedly connected. The present invention relates to an abnormality diagnosis system for a plant that identifies the cause of an abnormality.

[Conventional technology]

In general, many announcers are used in plants such as nuclear power plants and thermal power plants to output an alarm when the plant state quantity exceeds a prescribed limit value and notify the operator of the occurrence of an abnormality. . For example, in the case of a nuclear power plant, etc., the number of this announcer reaches several thousand, and even in the case of a normal minor failure, a large number of lamps flash and it is necessary to provide only necessary information to the operator. It's getting harder. For this reason, systems for diagnosing abnormalities using computers are being developed all over the world, and typical systems include Japan's nuclear power generation support system and West Germany's.
Gesellschaft fr Reaktorischerheit (GRS), a STA developed by the Halden project in Norway
R, ERPI (US Electric Power Research Institute), DAS, etc. developed by consignment.

In these systems, an abnormality diagnosis method called the cause-effect logic method (CCT) is used. For example, in CCT, the relationship between the cause and the result of an abnormality is expressed in a tree shape, and when an abnormality occurs, the state of the result is displayed. If you follow the branches of this tree in the opposite direction, you can find the cause.

[Problems to be Solved by the Invention]

By the way, since the CCT search starts when the announcer set to a level of ± 10 to 20% of the wave height of the plant state quantity during the normal operation starts to operate, the abnormality occurs for several tens of seconds. There is a possibility that the operator may not be able to carry out an appropriate recovery operation when propagating at a high speed.

The present invention is to solve the above-mentioned problems, and detects the abnormality at the level of ± 3 to 7% of the wave height of the plant state quantity during normal operation before the announcer operates, thereby determining the cause of the abnormality. The purpose is to identify and enable operators to perform appropriate recovery operations with sufficient time margin, and to improve plant availability and stability.

[Means for Solving the Problems]

Therefore, the plant abnormality diagnosis method of the present invention, a change in the plant state quantity at the time of abnormality is obtained by simulation, means for creating a coupling mode between the plant state quantity by autoregressive analysis of the change in the plant state quantity, Means for obtaining a fuzzy relation matrix from the coupling pattern between the plant state quantities, means for storing the obtained fuzzy relation matrix data as a table, and coupling mode between the plant state quantity and the plant state quantity when an actual plant abnormality is detected And a means for collating the above table with a pattern indicating the coupling pattern between plant state quantities created when an actual plant abnormality is detected, and the cause of the abnormality is identified by the collation result. This is a system abnormality diagnosis method.

[Action]

According to the plant abnormality diagnosis method of the present invention, a fuzzy relation matrix R is obtained from a pattern showing a coupling mode between plant state quantities obtained by performing autoregressive analysis of plant state quantities at the time of abnormality by simulation. The cause of the small anomaly is obtained by identifying the cause of the anomaly by entering the fuzzy relational expression into the pattern showing the coupling mode between the plant state quantities created after the anomaly is detected and the plant state quantity and the value of the relation matrix R. Can be identified early, and the operator can perform recovery operation with sufficient time margin to improve the plant operating rate and improve the plant safety.

〔Example〕

 Embodiments will be described below with reference to the drawings.

FIG. 1 is a block diagram for explaining a procedure for diagnosing abnormalities in a plant, FIG. 2 is a diagram showing an overall configuration of an advanced conversion reactor (ATR), FIG. 3 is a diagram showing a fuzzy membership function, and FIG. Fig. 5 is a block diagram showing the hardware configuration for diagnosing abnormalities in the plant. Fig. 5 (a) shows a decrease in reactivity, Fig. 5 (b) shows a decrease in the feed water flow rate, and Fig. 5 (c) shows a steam drum gas. Diagram showing frequency spectrum correlation between neutron flux and feedwater flow rate when phase pressure is increased, Fig. 6 (a) shows decreased reactivity, Fig. 6 (b) shows decreased feedwater flow rate, Fig. 6 (C)
FIG. 7 is a diagram showing a cause identification result in the case where the pressure increase in the vapor drum vapor phase portion is directly caused. In the figure, 1 is a calandria tank, 2 is a control rod, 3 is a pressure pipe assembly, 4 is a heavy water cooling system,
5 is a steam drum, 6 is a coolant recirculation system, 7 is a water drum, 8 is a recirculation pump, 10 is a water supply system, 11 is a main steam system, 12 is an outlet pipe, 13 is an inlet pipe, 21 and 24 are electronic. A computer, 22 is a pattern storage device, 23 and 31 are plants, 32 is an electronic computer, 33 is an external storage device, and 34 is an alarm device.

First, the principle of the plant abnormality diagnosis method according to the present invention will be described.

In diagnosing a plant abnormality, first, the plant response characteristic at the time of abnormality is obtained. In order to obtain the plant response characteristics at the time of this abnormality, it is not possible to actually generate an abnormality in the plant, so a simulation is performed using the plant dynamic characteristic analysis code, and the plant at the time of abnormality when disturbance is applied. Predict response.

For example, if "Lower steam drum water level" is selected as the type of anomaly, this is regarded as a peak event, and the cause event that is considered to be the result of this event, and the cause event that is considered to be the result of this cause event Faults obtained by connecting event-causing events in a tree-like fashion
From the tree, for example, the following events that can input the plant dynamic characteristic analysis code are selected.

(B) Decrease in reactivity (b) Decrease in feed water flow rate (c) Increase in pressure of steam drum vapor phase part (d) Decrease in feed water temperature (e) Increase in main steam flow rate Occurrence is, for example, the second
In the advanced conversion reactor (ATR) as shown in the figure, the water that removes the heat generated in the core disappears,
Since this corresponds to an anomaly that may lead to core melting, the operator constantly monitors the "steam drum water level" with great care. The AT shown in FIG.
At R, heavy water as a moderator is circulated in the calandria tank, heat generated in the moderator is removed by the heavy water cooling system 4, and carbon dioxide gas is filled between the calandria tube and the pressure tube to provide a heat insulating material. Then, the cooling water recirculation system 6 circulates the cooling water through the pressure pipe, while the water is supplied by the water supply system 10, the steam generated in the steam drum 5 is taken out and used in the main steam system 11.

Next, with respect to the plant state quantity obtained from the plant dynamic characteristic analysis, the deviation from the steady operation is obtained as shown in the equation (1).

Where x _{i, t} : plant state quantity type i, plant state quantity deviation at time t y _{i, o} : plant state quantity type i value of plant state quantity during rated operation y _{i, t} : plant It is the value of the plant state quantity at time t of the state quantity type i. In addition, the equation (1) is used so that plant state quantities having different dimensions such as pressure, temperature, and flow rate can be compared with each other.
Normalized and dimensionless.

Next, for the plant state quantity normalized by the computer using the autoregressive analysis code, a pattern showing the correlation between the frequency spectra of the state quantity is created. In the autoregressive analysis code, first, the time series data X (t) can be expressed by autoregressive equation (2).

However, transpose matrix x _k (t) of X (t): [x ₁ (t), x ₂ (t), ... x _k (t)]: deviation value U of plant state quantity of kind k at time t (T): White noise Next, the spectrum density function P (f) can be obtained from the residual covariance matrix Σ and the autoregressive coefficient A (m) as shown in the following equation.

A (f) is a matrix whose (j, s) element is A _is (f), P
(F) is a matrix whose (i, j) element is P _ij (f),
() ^-1 is the inverse matrix, () ^T is the transposed matrix, Indicates the complex conjugate number. Note that A _js (f) is the Fourier transform of the autoregressive coefficient A (m) as shown in the following equation.

The correlation function between the frequency spectra of the plant state quantities i and j can be expressed by the following coherence function.

COH _ij = | P (f) _ij | / (P (f) _ii · P (f) _jj ) ... (5) The type of plant state quantity for examining the correlation between these frequency spectra is, for example, as shown in FIG. In ATR, neutron flux in recirculation system, recirculation flow rate, water level in steam drum, feed water flow rate in feed water system, feed water temperature, main steam flow rate in main steam system, mixing section pressure, main steam control valve opening degree , Like turbine output in a turbine system. Among these state quantities, for example, the correlation (coherence function) between the neutron flux and the frequency spectrum of the feed water flow rate is as shown in FIG.

FIG. 5 (a) is a diagram in the case of a decrease in reactivity, FIG. 5 (b) is a diagram in the case of a decrease in the feed water flow rate, and FIG. 2.70 × 10 ^-2 $ / for reduced reactivity
The abnormal speed was changed at a speed of 1.62 × 10 ^-2 $ / sec, 0.41 × 10 ^-2 $ / sec. In case of water supply flow rate, 3%, 5%, 7
The flow rate was reduced in a step function up to%. In the case of increasing the pressure of the vapor phase of the vapor drum, the pressure was also increased stepwise to 2%, 4% and 6%. In the figure, the horizontal axis represents frequency and the vertical axis represents the magnitude of correlation.

From the perspective of these figures, it can be said that the pattern does not depend on the speed or magnitude of the abnormality, but differs depending on the cause of the abnormality. However, locally, it can be said that these patterns fluctuate when the speed or magnitude of the abnormality is changed. That is, the correlation value has a certain width, that is, ambiguity, with respect to one certain frequency.

Such an abnormal time pattern is stored as a reference abnormal pattern, and in the case of identifying whether or not there is an abnormality by collating with a pattern obtained for an actual plant state quantity that changes from moment to moment, in the present invention, Uses the solution of the inverse problem of fuzzy relations.

First, we give an overview of fuzzy, fuzzy relational expressions, and the solution method of inverse problems of fuzzy relational expressions, and finally show an example of the solution.

The fuzzy English is replaced by the ambiguous Japanese. For example, "a large group of people", "a beautiful group of people", "slow speed", etc. are not included in the conventional set theory. Number 5 is "large number"
Whether it is included in or not depends on the subjectivity of the individual, and an ambiguous intermediate state that cannot be determined by either appears. In fuzzy set theory, such a concept of unclear boundaries is expressed as a set concept using a membership function. An example of the membership function is shown in FIG.

Next, the fuzzy relational expression will be described.

Think of abnormal signs and symptoms as a result of a cause, and diagnose as identifying the cause. Now, let b = {y ₁ , y ₂ , ..., Y _n } be a set of symptom items and a = {x ₁ , x ₂ , ..., x _m } be a set of cause items.

When diagnosing a plant abnormality, {y ₁ , y ₂ , ..., y _n } is a pattern (coherence function) showing the correlation of the frequency spectrum of the plant state quantity, {x ₁ , x ₂ , ..., x _m } Causes abnormalities such as (a) decrease in reactivity (b) decrease in feed water flow rate, and (c) increase in pressure of vapor drum vapor phase part. Further, r _ij is a matrix of m rows and n columns that represents the presence or absence or the depth of association between x _i and y _j, and is a relation matrix R. At this time, the fuzzy relational expression is expressed by Expression (6), which means the operation of Expression (7).

Here, each element x _i , r _ij , and y _j of a, R, and b is [0
1], the degree of the causal relationship between the cause i and the symptom j and the degree of occurrence of the symptom j are respectively represented when the cause i occurs. T represents transposition.
The circles are max-min composite operations, ∨ is max operation and ∧ is min operation.

Now, assuming that S _ij = a _i ∧r _ij , the ∧ operation is S _ij = min (X _i , Y _ij ) i = 1 to m, j = 1 to n) (8). ∨ The operation is b _j = max (S _ij , ... S _ij , ... S _mj ) (j = 1 to n) ...
… (9).

The inverse problem of fuzzy relations can be solved by using ε composition and composition. Here, ε composition and composition are as follows (10) (1
It is defined by the equation 1).

The inverse problem of the fuzzy relational expression in this case is as follows.
It can be solved by the algorithm of (iii).

(I) An m × n matrix U = {U _ij }, V = {V _ij } is obtained.

Here, b _j is a symptom and r _ij is a matrix element of the fuzzy relational expression.

(Ii) m × n matrix Ask for.

However, Is given for every j columns of W ^k as

This is because i for which U _ij is not φ is arbitrarily selected for each j column, _Be U _ij and i of other rows Indicates that V _ij .

The index k represents any combination of W _ij for which U _ij ≠ φ can be arbitrarily selected.

(Iii) The solution to the inverse problem of the fuzzy relational expression is as follows.

Set K is The intersection set for J of is a set of indexes K that is not empty for all i.

The existence conditions of the solution are shown below.

In this way, the cause a ^K _i when the symptom b is given can be calculated.

Next, a method for performing pattern matching will be shown. First, the pattern (y ₁ , y ₂₎ showing the correlation of the frequency spectrum of the plant state quantity created by using the simulator.
... y _n ) is used to determine the value of the relation matrix r _{ij in} the fuzzy relational expression (6). And these r _ij
Is stored as a reference abnormal time pattern, and a pattern (y ₁ ′, y) obtained from the actual plant state quantity that changes momentarily when a decrease in the steam drum water level is detected.
₂ ′ ... Y _n ′), the stored relation matrix r _ij is put into the fuzzy relational expression (6), and the inverse problem is solved to solve the cause of the abnormality (x ₁ ′, x ₂ ′ ... x _m ′) Is identified. As shown in Fig.5, to identify the cause of abnormality from an ambiguous pattern in which the correlation value has a certain width with respect to one frequency, it is necessary to use the "solution of the inverse problem of the fuzzy relational expression". convenient.

Since it is possible that noise is superimposed on the state quantity of the actual plant, while superimposing white noise with a standard deviation of 5% of the normal wave height on the plant state quantity analyzed using a simulator, The result of identifying the cause of the abnormality is shown in FIG.

FIG. 6 (a) shows the case where the reactivity decreases, FIG. 6 (b) shows the case where the feed water flow rate decreases, and FIG. 6 (c) shows the case where the pressure in the vapor phase portion of the steam drum increases. It can be said that the first cause of the abnormality is identified by the degree of association of 0.9 to 1.0. The first cause of the abnormality means an abnormality of the decrease in reactivity when the decrease in reactivity is input as the disturbance of the simulation in the case of FIG. 6A as an example. The degree of association with other causes is at most 0.4. The correct cause of the abnormality can be identified from the difference between these numerical values.

FIG. 1 is a block diagram for explaining a procedure of a plant abnormality diagnosis based on such a principle. In the figure, 21 is a reference abnormal time pattern creation function block, and 22 is a reference abnormal time pattern storage function block. , 23 is a plant, 2
Reference numeral 4 is a plant abnormality diagnosis function block.

In the figure, in a functional block 21, an abnormal plant state quantity pattern is created by an electronic computer. In step, the simulator (plant dynamic characteristic analysis code) is used to analyze the plant response at the time of abnormality, in step, the pattern showing the correlation of the frequency spectrum of the plant state quantity is created, and in this step, the fuzzy relational expression is used. A relationship matrix R is created. The function block 22 is constituted by an external storage device, such as a magnetic desk, and in the step, the pattern created in the step, that is, the value of r _ij which is an element of the relation matrix R in the fuzzy relational expression is set. The table is stored as a reference abnormal pattern. The function block 24 is performed by a computer for plant abnormality diagnosis,
The data of the plant state quantity which changes from moment to moment in steps is taken in from the plant 23 at a speed of, for example, about once / second. With respect to the data input to the plant abnormality diagnosis electronic computer, a coupling mode pattern between the plant state quantities is created in the same step as in the case of creating a pattern indicating the coupling mode between the plant state quantities of the step.

The step detects an abnormality, for example, when the indication value of the water level meter of the steam drum becomes smaller than the normal value. In this case, the threshold value for presence / absence of abnormality is slightly larger than the value (X ₂ ± δX ₂ ) fluctuating due to noise mixed in the measurement line (X ₂ ) of the water level gauge.

Next, in the step, the pattern matching with the pattern at the abnormal time for reference is performed by utilizing the method of solving the inverse problem of the fuzzy relational expression and the pattern of the coupling manner between the plant state quantities of the step.

Next, in a step, the result of identifying the abnormality is output to a printer or the like of the computer. As a result of pattern matching with the reference abnormal pattern, if there is a matching pattern, the alarm in the plant is turned on. If there is no match, the process returns to the start point of the plant abnormality diagnosis. When actually constructing a plant abnormality diagnosis system, if no matching pattern is found, the degree of matching is shown as a percentage as "probability", or "confirmation operation" is presented to the operator. May be taken.

As described above, for example, the neutron flux of the reactor plant,
Using the solution method, the inverse of the fuzzy relational equation is used to find a variation of the coupling state of plant state quantities that can be easily monitored in the central control room such as temperature, flow rate, pressure, valve opening, etc.
The cause of abnormality can be identified.

FIG. 4 is a diagram showing a hardware configuration for performing such abnormality diagnosis. 11 is a plant, 12 is an electronic computer,
Reference numeral 13 is an external storage device, and 14 is a printer, which perform the operations described above.

〔The invention's effect〕

As described above, according to the present invention, since the overall state of the plant is comprehensively determined, a small abnormality is detected, and the cause of this abnormality is identified and notified to the operator, thereby recovering early. The operation can be performed, and not only the plant operation rate can be improved but also the plant safety can be improved.

[Brief description of drawings]

FIG. 1 is a block diagram for explaining a procedure for diagnosing abnormalities in a plant, FIG. 2 is a diagram showing an overall configuration of an advanced conversion reactor (ATR), FIG. 3 is a diagram showing a fuzzy membership function, and FIG. Fig. 5 is a block diagram showing the hardware configuration for diagnosing abnormalities in the plant. Fig. 5 (a) shows a decrease in reactivity, Fig. 5 (b) shows a decrease in the feed water flow rate, and Fig. 5 (c) shows a steam drum gas. Diagram showing frequency spectrum correlation between neutron flux and feedwater flow rate when phase pressure is increased, Fig. 6 (a) shows decreased reactivity, Fig. 6 (b) shows decreased feedwater flow rate, Fig. 6 (C)
FIG. 7 is a diagram showing a cause identification result in the case where the pressure increase in the vapor drum vapor phase portion is directly caused. 21, 24 ... Electronic computer, 22 ... Pattern storage device, 2
3, 31 ... Plant, 32 ... Electronic computer, 33 ... External storage device, 34 ... Alarm device.

Claims (3)

[Claims] 1. A means for obtaining a change in the plant state quantity at the time of an abnormality by simulation, performing autoregressive analysis of the change in the plant state quantity to create a coupling mode between the plant state quantities, and a coupling between the created plant state quantities. Means for obtaining a fuzzy relation matrix from the form, means for storing the obtained fuzzy relation matrix data as a table, and means for creating a coupling form between plant state quantities from plant state quantities when an actual plant abnormality is detected, An abnormality diagnosis of a plant, characterized by comprising means for collating the above-mentioned table with a pattern indicating a coupling pattern between plant state quantities created when an actual plant abnormality is detected, and identifying the cause of the abnormality based on the collation result. method. 2. The abnormality diagnosis system for a plant according to claim 1, wherein the coupling mode between the plant state quantities is a coherence spectrum showing a correlation between frequency spectra of the plant state quantities. 3. The fault diagnosis system for a plant according to claim 1, wherein the pattern and the table are collated by using a solution of an inverse problem of a fuzzy relational expression.