JP2013101718A - Plant operation state monitoring method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To simplify computation for obtaining a Mahalanobis distance, and to promptly determine whether or not a plant is normally operated.SOLUTION: A plant operation state monitoring method includes: a step (A) of classifying and dividing state amounts of many monitoring items of the plant into a data group comprising the state amounts of the plurality of monitoring items that are fewer; a Mahalanobis distance computing step (B) of respectively computing the Mahalanobis distance from the state amount for each data group; a Mahalanobis distance individual comparison step (D') of comparing the Mahalanobis distances computed for the respective data groups with a preset threshold; and an abnormality determination step E of determining presence/absence of abnormality by the comparison result. The data group is classified on the basis of measurement positions including a plurality of parts along a circumferential direction of a gas turbine.

Description

  The present invention relates to a plant operation state monitoring method for determining whether or not a plant is operating normally.

  In various plants such as a gas turbine power plant, a nuclear power plant, or a chemical plant, in order to monitor whether or not the plant is operating normally, plant state quantities such as temperature and pressure are acquired and monitored. That is, the state quantities of multiple monitoring items to be monitored are measured at predetermined time intervals, the average and distribution of the state quantities for each monitoring item are calculated and normalized, and the correlation between the state quantities of each monitoring item And the Mahalanobis distance is calculated. When the Mahalanobis distance exceeds a preset threshold value, it is determined that there is a sign of abnormality in the plant.

  Patent Document 1 discloses a facility monitoring method using such Mahalanobis distance. In this facility monitoring method, the result of calculating the system parameters of the mathematical model by giving time series data related to the vibration of the synchronous moving body to the mathematical model predetermined for system identification, and the system in a normal state An index (prediction error fraction) weighted with the autocorrelation coefficient matrix of the original sequence signal with respect to the degree of deviation from the parameter group, and the variance / covariance matrix of the system parameter group at normal time with respect to the degree of deviation A deterioration state of the periodic moving body is comprehensively determined by combining an index (Mahalanobis distance) weighted with an inverse matrix. And even if it has many state quantities of a plant by such an equipment monitoring method, the monitoring of this plant can be performed easily.

JP 59-68644 A

However, in the above-described plant operating state monitoring method using the Mahalanobis distance, the state quantities of a large number of monitoring items are monitored, and therefore it is necessary to repeat many operations in order to obtain the Mahalanobis distance. As a result, there is a problem that the calculation takes a lot of time and the determination of whether or not the plant is operating normally is delayed.
Specifically, when the state quantities are x and y, the Mahalanobis distance is calculated between the monitoring items by a simple calculation such as the square root of x ² + y ^2. For example, if there are 100 monitoring items, each state The number of calculations is required 100 × 100 = 10000 times the quantity, and much time is required for the calculation process.

  The present invention has been made in view of the above-described circumstances, and can speed up the calculation for obtaining the Mahalanobis distance, and can quickly determine whether or not the plant is operating normally. An object of the present invention is to provide a method for monitoring the operating state of a plant.

In order to solve the above problems, the present invention proposes the following means.
In the plant operating state monitoring method of the present invention, the process of classifying the state quantities of many monitoring items of the plant into a data group consisting of a smaller number and a plurality of monitoring item state quantities, and the Mahalanobis distance from the state quantity for each data group A process of comparing the Mahalanobis distance calculated for each data group with a threshold set in advance for each data group, and a process of determining whether there is a plant abnormality based on the comparison result. It is characterized by having.

According to the plant operating state monitoring method of the present invention, state quantities of many monitoring items of a plant are classified into a data group consisting of a smaller number of state quantities of a plurality of monitoring items, and the state quantities are classified for each classified data group. Since the Mahalanobis distance is calculated, there is no need to calculate all the correlations of the state quantities of many monitoring items to determine the Mahalanobis distance as in the past, and as a result, the calculation to calculate the Mahalanobis distance is not required. The speed can be increased, and the abnormality determination work of the plant can be performed promptly.
Also, by classifying the state quantities of many monitoring items in the plant into data groups consisting of a smaller number and a plurality of monitoring item state quantities, the unit space state quantity and signal space state quantity are determined for each data group. The time interval to measure, the time interval to change the unit space can be set individually, and the search criteria of the unit space and the search criterion of the signal space can be set individually for each data group, Various data processing that can improve the reliability of monitoring can be easily performed.

It is a schematic diagram which shows the structural example of the plant operation state monitoring apparatus which concerns on 1st Embodiment of this invention. It is a schematic diagram for demonstrating the process part of FIG. 1 concretely. It is a conceptual diagram which shows the concept of Mahalanobis distance. It is a flowchart which shows the procedure of the plant operation state monitoring method which concerns on this embodiment. It is a schematic diagram for demonstrating specifically the process part which concerns on 2nd Embodiment. It is a schematic diagram for demonstrating specifically the process part which concerns on 3rd Embodiment.

(First embodiment)
Hereinafter, a first embodiment of the present invention will be described with reference to FIGS. The present invention is not limited by the best mode for carrying out the invention (hereinafter referred to as an embodiment). In addition, constituent elements in the following embodiments include those that can be easily assumed by those skilled in the art, those that are substantially the same, and those in a so-called equivalent range. Although this embodiment demonstrates the example which applied this invention to the state monitoring of the power plant of a gas turbine, the object which can apply this invention is not limited to this. For example, the present invention can be applied to all plants having a large number of monitoring items such as nuclear power plants and chemical plants.

FIG. 1 is a schematic diagram illustrating a configuration example of a plant operation state monitoring apparatus according to the present embodiment. The plant operation state monitoring device 10 monitors the operating state of the power plant (gas turbine power plant) 1 using the gas turbine 6 to determine whether or not the gas turbine power plant 1 is operating normally. .
The state quantities for monitoring the gas turbine 6 of the plant are, for example, cavity temperatures measured at a plurality of locations between the rotor and the stator of the gas turbine 6, and measured at a plurality of locations along the circumferential direction at the gas outlet. Blade path temperature, the amount of displacement of the rotor shaft measured at a plurality of locations along the circumferential direction of the rotor, the opening of various valves provided in the gas turbine 6, and the like. These state quantities are shown as monitoring target data, and a plurality of the state quantities, for example, about 10 to 20 points are usually measured at the blade path temperature.

  The gas turbine power plant 1 to be monitored generates power by driving a generator 5 with a gas turbine 6. The gas turbine 6 includes a compressor 2, a combustor 3, and a turbine 4 that rotates the compressor 2. The air drawn from the intake port of the compressor 2 is compressed by the compressor 2 and is led to the combustor 3 as high-temperature and high-pressure air. In the combustor 3, fuel is supplied to high-temperature and high-pressure air and burned. The fuel combusted in the combustor 3 is supplied to the turbine 4 as a high-temperature and high-pressure combustion gas to drive it. As a result, the turbine 4 rotates.

  The output shaft of the gas turbine 6, that is, the rotating shafts of the turbine 4 and the compressor 2 are connected to the generator 5. As a result, the output obtained by operating the gas turbine 6 and rotating the turbine 4 is transmitted to the generator 5. With such a configuration, the gas turbine 6 drives the generator 5 to cause the generator 5 to generate electric power.

  The plant operation state monitoring device 10 monitors the state of the gas turbine power plant 1. In the present embodiment, the plant operation state monitoring device 10 monitors the state of one gas turbine power plant 1, but may monitor the operation state of a plurality of gas turbine power plants 1. The plant operation state monitoring device 10 is, for example, a computer, and includes an input / output unit (I / O) 11, a processing unit 12, and a storage unit 13. The plant operating state monitoring device 10 may be configured using a so-called personal computer, or may be configured by combining a CPU (Central Processing Unit) and a memory.

  The processing unit 12 receives the state quantity of the gas turbine power plant 1 from various state quantity detection means (sensors) attached to the gas turbine power plant 1 via the input / output unit 11. The various state quantity detection means periodically acquire corresponding state quantities at predetermined time intervals from the start of activation, and input them to the processing unit 12 via the input / output unit 11.

  The monitoring target data group indicating the state quantity of the gas turbine power plant 1 is sent to the processing unit 12 of the plant operation state monitoring device 10 in the form of an electrical signal. The processing unit 12 is constituted by, for example, a CPU, and sequentially reads and interprets an instruction sequence called a program (computer program) existing on the storage unit 13, and moves or processes data according to the result.

  Note that the processing unit 12 may be realized by dedicated hardware. Further, a computer program for realizing the function of the processing unit 12 is recorded on a computer-readable recording medium, and the computer program recorded on the recording medium is read into the computer system and executed, thereby executing this embodiment. The processing procedure of the plant operating state monitoring method may be executed. Here, the “computer system” includes an OS and hardware such as peripheral devices.

  The “computer-readable recording medium” refers to a portable medium such as a flexible disk, a magneto-optical disk, a ROM, a CD-ROM, or a recording device such as a hard disk built in a computer system. Further, the “computer-readable recording medium” refers to a computer program that is dynamically used for a short time, such as a communication line when a computer program is transmitted via a communication line such as the Internet or a telephone line. What is held, and what holds a computer program for a certain period of time, such as a volatile memory inside a computer system serving as a server or a client in that case, are included. Further, the computer program may be for realizing a part of the above-described functions, and even if the functions described above can be realized in combination with a computer program already recorded in the computer system. Good.

  The plant operating state monitoring method according to the present embodiment can be realized by executing a computer program prepared in advance on a computer such as a personal computer or a workstation. This computer program can be distributed via a communication line such as the Internet. The computer program is recorded on a computer-readable recording medium such as a hard disk, a flexible disk (FD), a CD-ROM, an MO, and a DVD, and is executed by being read from the recording medium by the computer. May be.

  In the processing unit 12, as shown in FIG. 2, the data division process A for classifying the monitoring target data indicating the state quantity of the plant into a plurality of data groups (NO.1, 2, 3,... N), and the data The Mahalanobis distance calculation process B for calculating the Mahalanobis distance for each data group divided in the division process A, the maximum value selection process C for selecting the maximum value of the Mahalanobis distance calculated in each data group, and the selected Each arithmetic unit performs processing such as a comparison process D in which the maximum value of the Mahalanobis distance is compared with a threshold value, and an abnormality determination process E in which presence / absence of an abnormality is determined based on a comparison result between the maximum value of the Mahalanobis distance and the threshold value. .

  The concept of Mahalanobis distance is shown in FIG. As an example, FIG. 3 shows the correlation between two parameters with the generator output on the horizontal axis and the blade path temperature on the vertical axis. That is, if the generator output increases, the blade path temperature also increases. Each measurement data varies depending on the atmospheric conditions, operating conditions, and the like, but there is a correlation between the generator output and the blade path temperature, and it falls within a specific range. A unit space serving as a reference is created using these as reference data. Also in each of the other state quantities, the correlation can be obtained like the generator output and the blade path temperature. Then, whether the data to be determined is normal or abnormal is determined based on the Mahalanobis distance for the unit space.

The above-described Mahalanobis unit space can be obtained by the following items that are predetermined in the present embodiment.
(1) A unit space of Mahalanobis based on the monitoring target data indicating the state quantity of the gas turbine power plant 1 in the past period from the time when the state of the gas turbine power plant 1 is evaluated to the past before a predetermined period. Is calculated.
(2) Predict the future state of the gas turbine power plant 1 based on the monitoring target data indicating the state quantity at the time of evaluating the state of the gas turbine power plant 1, and determine the Mahalanobis unit space based on the predicted value. Calculate.
(3) The future state of the gas turbine power plant 1 is predicted based on the monitoring target data indicating the state quantity at the time of evaluating the state of the gas turbine power plant 1 and the control target set value set at the time of starting the plant. The unit space of Mahalanobis is calculated based on the predicted value.

  In addition, when determining whether the gas turbine power plant 1 is normal using Mahalanobis distance, multidimensional data is re-baked into one-dimensional data using Mahalanobis distance. Then, the difference between the unit space and the signal space (data to be compared with the unit space, for example, a state quantity when the state of the gas turbine power plant 1 is evaluated) is seen by the Mahalanobis distance. In the present embodiment, the Mahalanobis distance of the signal space is obtained using a matrix created from the unit space. As a result, data anomalies can be expressed.

  A control panel 14 serving as output means is connected to the input / output unit 11 of the plant operating state monitoring apparatus 10. The control panel 14 is provided with a display 14D as display means and an input means 14C for inputting a command to the plant operating state monitoring device 10. The storage unit 13 of the plant operation state monitoring device 10 includes, for example, a volatile memory such as a RAM (Random Access Memory), a non-volatile memory such as a ROM (Read Only Memory), a hard disk device, and a magneto-optical disk. A device, a storage medium that can only be read, such as a CD-ROM, or a combination thereof. The storage unit 13 stores a computer program, data, and the like for realizing the plant operation state monitoring method according to the present embodiment. The processing unit 12 implements the plant operation state monitoring method according to the present embodiment or controls the operation of the gas turbine power plant 1 using these computer programs and data. The storage unit 13 may be provided outside the plant operation state monitoring device 10 so that the plant operation state monitoring device 10 can access the storage unit 13 via a communication line.

Here, a calculation formula for calculating a general Mahalanobis distance D will be described.
First, the total number of a plurality of state quantities representing the state of the gas turbine power plant 1 is u, each state quantity is assigned to a variable X, and u state quantities are defined by variables X1 to Xu (where u is 2 or more). Integer). Next, in the operation state of the gas turbine power plant 1 as a reference, a total of v (two or more) state quantities (indicated by monitoring target data) of the variables X1 to Xu are collected.

  The average value Mi and the standard deviation σi (the degree of variation of the reference data) of each of the variables X1 to Xu are obtained by the formulas (1) and (2). Note that i is the number of items (the number of state quantities, an integer), which is a value corresponding to the variables X1 to Xu set to 1 to u. j takes any value (integer) from 1 to v, meaning that the number of each state quantity is v. For example, when 60 state quantities are acquired, v = 60. Here, the standard deviation is the square root of the expected value of the difference between the state quantity and its average value squared.

  Next, the standardization of converting the original variables X1 to Xu into x1 to xu by the following formula (3) using the above-described average value Mi and standard deviation σi which are state quantities that are calculated and indicate characteristics. Do. That is, the state quantity of the gas turbine power plant 1 is converted into a random variable having an average of 0 and a standard deviation of 1. In the following mathematical formula (3), j takes any value (integer) from 1 to v, meaning that the number of each state quantity is v.

  Next, in order to analyze the data with the variables standardized to mean 0 and variance 1, the correlation between the variables X1 to Xu, that is, the covariance matrix (correlation matrix) R indicating the relationship between the variables, and the covariance An inverse matrix R-1 of the matrix (correlation matrix) is defined by the following mathematical formula (4). In the following mathematical formula (4), k is the number of items (the number of state quantities), which is u here. Further, i and p indicate values in the respective state quantities, and take values of 1 to u here.

  After such a calculation process, the Mahalanobis distance D, which is a state quantity indicating a feature, is obtained based on the following formula (5). In Equation (5), j takes any value (integer) from 1 to v, meaning that the number of each state quantity is v. Further, k is the number of items (the number of state quantities), which is u here. Further, a11 to akk are coefficients of the inverse matrix R-1 of the covariance matrix R shown in the above-described equation (4).

  The Mahalanobis distance D is the reference data, that is, the average value of the Mahalanobis distance D in the unit space is 1, and is approximately 3 or less when the state quantity of the gas turbine power plant 1 is normal. However, when the state quantity of the gas turbine power plant 1 becomes abnormal, the value of the Mahalanobis distance D increases. As described above, the Mahalanobis distance D has a property that the value increases according to the degree of abnormality of the state quantity of the gas turbine power plant 1 (the degree of separation from the unit space).

The above formulas (1) to (5) are calculation formulas for calculating a general Mahalanobis distance D, but in this embodiment, state quantities (for example, for monitoring the gas turbine 6 of the plant described above) The cavity temperature measured at a plurality of locations between the rotor and the stator of the gas turbine 6, the blade path temperature measured at a plurality of locations along the circumferential direction at the gas outlet, and the circumferential direction of the rotor to detect the axial vibration of the rotor A group of data composed of a smaller number of monitored items such as displacement amounts of the rotor shaft measured at a plurality of locations along the opening, opening amounts of various valves provided in the gas turbine 6, etc. The Mahalanobis distance is calculated for each divided data group.
Each of the plurality of data groups is classified by a state quantity measurement position (for example, a plurality of cabty temperatures, a plurality of blade path temperatures, a plurality of rotor shaft displacement amounts, or opening amounts of various valves). It is also possible to classify according to the sampling period of these state quantities (for example, every 1 minute, every 10 minutes, or every 1 hour sampling period), and the state quantity type (temperature Or monitoring items such as pressure, displacement, deviation between command value and measurement value). It is particularly preferable to classify data groups by grouping highly correlated monitoring items.
Note that the division pattern for dividing such monitoring target data into each data group is stored in the storage unit 13 in advance, and a monitoring target based on the division pattern is issued to the processing unit 12 by issuing a division command. Let the data split processing be executed.

  Next, the procedure of the plant operation state monitoring method according to this embodiment will be described. The plant operation state monitoring method according to the present embodiment is realized by the processing unit 12 of the plant operation state monitoring apparatus 10 shown in FIG.

FIG. 4 is a flowchart showing the procedure of the plant operating state monitoring method according to the present embodiment.
In step S1, first, monitoring target data indicating a state quantity is obtained from the gas turbine power plant 1 in the current state quantity obtaining period. For example, the state quantity is periodically acquired from various sensors attached to the gas turbine power plant 1 at predetermined time intervals and stored in the storage unit 13 of the plant operation state monitoring device 10.

  In step S2, the monitoring target data is divided into a plurality of data groups and grouped based on the monitoring target data division command (described above) set in advance. As described above, these data groups may be divided and grouped for each monitoring target such as temperature, pressure, vibration, rotation speed, etc., or monitoring items with particularly high correlation are selected and grouped. It may be formed by forming.

  In steps S3-1 to S3-N, the Mahalanobis distance is calculated from each divided data group according to the above formula, and then in the next step S4, the maximum of the Mahalanobis distance in these steps S3-1 to S3-N is calculated. Select a value.

In step S5, it is determined whether or not the maximum value of the Mahalanobis distance exceeds the threshold value by comparing the preset threshold value with the maximum value of the Mahalanobis distance obtained in the previous step S4. . Thereafter, in step S6, based on the determination result in step S5, when the maximum value of the Mahalanobis distance exceeds YES, it is determined as “abnormal” (step S6), and the maximum value of the Mahalanobis distance is determined. However, it is determined as “normal” in the case of NO that does not exceed the threshold (step S7).
As described above, the Mahalanobis distance increases as the distance from the unit space increases, depending on the degree of abnormality. The Mahalanobis distance D is approximately 3 or less when the reference data, that is, the average value of the unit space is 1, and the state quantity of the gas turbine power plant 1 is normal. Therefore, for example, the threshold value can be appropriately set to a value larger than the maximum value of the unit space. In addition, the threshold value is preferably a set value that takes into consideration the unique characteristics of the gas turbine power plant 1, manufacturing variations of the gas turbine power plant 1, and the like.

As described above in detail, according to the plant operation state monitoring method shown in the present embodiment, the state quantities of many monitoring items of the plant are classified into a data group composed of a smaller number and a plurality of monitoring item state quantities, Since the Mahalanobis distance is calculated for each state quantity of each classified data group, it is not necessary to calculate all the correlations of the state quantities of many monitoring items to obtain the Mahalanobis distance as in the past, and as a result The calculation for obtaining the Mahalanobis distance can be speeded up, and the plant abnormality determination work can be quickly performed. Specifically, when the state quantities are x and y, the Mahalanobis distance is calculated between the monitoring items by a simple calculation such as the square root of x ² + y ^2. For example, if there are 100 monitoring items, each state The number of calculations is required 100 × 100 = 10000 times the quantity, and much time is required for the calculation process.
However, in this embodiment, the state quantity of the monitoring item 100 is divided into data groups each consisting of 10 monitoring items, and 10 × 10 = 100 operations for the state quantity for each data group are performed 10 times. Then, a total of 1000 calculations are required, and this can speed up the calculation for obtaining the Mahalanobis distance, and the abnormality determination work of the plant can be performed quickly.

  Also, by classifying the state quantities of many monitoring items in the plant into data groups consisting of a smaller number and a plurality of monitoring item state quantities, the unit space state quantity and signal space state quantity are determined for each data group. The time interval for measuring and the time interval for changing the unit space can be set individually, and the search criteria for the unit space and the search criteria for the signal space can be set individually for each data group. It is possible to easily perform various data processing that can improve reliability.

(Second Embodiment)
A second embodiment of the present invention will be described with reference to FIG. The plant operating state monitoring method shown in the second embodiment is different from the first embodiment in the data processing method after calculating the Mahalanobis distance based on each divided data group.
That is, as shown in FIG. 5, instead of the maximum value selection process C for selecting the maximum value of the Mahalanobis distance shown in the first embodiment, in the second Mahalanobis distance calculation process indicated by the reference symbol B ′, The Mahalanobis distance of the Mahalanobis distance of each group calculated in the Mahalanobis distance calculation process B is obtained.

Specifically, as shown in FIG. 5, the processing unit 20 classifies and divides the state quantities of many monitoring items of the plant into data groups composed of a smaller number and a plurality of monitoring item state quantities. And a Mahalanobis distance calculation process B for calculating the Mahalanobis distance for each data group divided in the data division process A, and a Mahalanobis distance of the Mahalanobis distance calculated in each group of the Mahalanobis distance calculation process B. 2 Mahalanobis distance calculation process B ′, a comparison process D comparing the Mahalanobis distance of the Mahalanobis distance calculated in the second Mahalanobis distance calculation process B ′ with a threshold value, and a comparison result between the maximum value of the Mahalanobis distance and the threshold value. Each calculation unit performs processing such as an abnormality determination process E for determining whether or not there is an abnormality.
In this processing unit 20, in order to know what distribution the plurality of Mahalanobis distances calculated in each group of the Mahalanobis distance calculation process B have, the second Mahalanobis distance calculation process B ′ Calculate the Mahalanobis distance of the Mahalanobis distance. Then, after comparing the calculation result in the second Mahalanobis distance calculation process B ′ with a preset threshold value in the comparison process D, the Mahalanobis distance of the previous Mahalanobis distance is set to the threshold value in the abnormality determination process E. Whether or not there is an abnormality in the plant is determined based on whether or not the value is exceeded.

And also in the plant operation state monitoring method shown in the second embodiment, as in the first embodiment, the state quantities of many monitoring items are classified into a data group composed of a smaller number of state quantities of a plurality of monitoring items. Because the Mahalanobis distance is calculated for each state quantity of each classified data group, it is not necessary to calculate all the correlations of the state quantities of many monitoring items to obtain the Mahalanobis distance as in the past. As a result, the computation for obtaining the Mahalanobis distance can be speeded up, and the plant abnormality determination work can be quickly performed.
In the second Mahalanobis distance calculation process B ′, the Mahalanobis distance of the plurality of Mahalanobis distances calculated in each group of the Mahalanobis distance calculation process B is calculated and compared with a predetermined threshold value, thereby calculating the Mahalanobis distance calculation. The abnormal value of the computed Mahalanobis distance in the process B can be accurately detected, and the occurrence of the plant abnormality can be accurately determined.
Also in the present embodiment, as in the first embodiment, by classifying the state quantities of many monitoring items of the plant into data groups consisting of a smaller number of state quantities of a plurality of monitoring items, for each data group. In addition, the unit space state amount, the time interval for measuring the signal space state amount, and the time interval for changing the unit space can be set individually, and for each data group, the unit space search criteria, signal Spatial search criteria can be individually set, and various data processing that can improve the reliability of monitoring can be easily performed.

(Third embodiment)
A third embodiment of the present invention will be described with reference to FIG. The plant operation state monitoring method shown in the three embodiments differs from the previous embodiment in the data processing method after calculating the Mahalanobis distance based on each divided data group.
That is, as shown in FIG. 6, instead of the maximum value selection process C and the comparison process D shown in the first embodiment, a Mahalanobis distance individual comparison process indicated by reference numeral D ′ is provided.

Specifically, as illustrated in FIG. 6, the processing unit 21 classifies and divides the state quantities of many monitoring items of the plant into data groups including a smaller number and a plurality of monitoring item state quantities. The Mahalanobis distance calculation process B for calculating the Mahalanobis distance for each data group divided in the data division process A and the Mahalanobis distance calculated in each group of the Mahalanobis distance calculation process B are set in advance. Each arithmetic unit performs processing such as a Mahalanobis distance individual comparison process D ′ for comparison with a predetermined threshold value and an abnormality determination process E for determining the presence / absence of an abnormality based on a comparison result in the Mahalanobis distance individual comparison process D ′. Do.
In this processing unit 21, at least one Mahalanobis distance is set as a predetermined threshold in each comparison performed with each Mahalanobis distance in each data group and a predetermined threshold set in the Mahalanobis distance individual comparison process D ′. Is exceeded, it is determined that the plant is “abnormal” in the next abnormality determination process E.
In addition, the determination of the plant abnormality in the abnormality determination process E can be appropriately changed by the user. For example, in comparison with the Mahalanobis distance to be performed in large numbers and a predetermined threshold value set in advance, When at least half of the Mahalanobis distance exceeds a predetermined threshold value, it may be determined that the plant is “abnormal” in the next abnormality determination process E.

In the plant operation state monitoring method shown in the third embodiment, as in the previous embodiment, the state quantities of many monitoring items are classified into a data group consisting of a smaller number of state quantities of a plurality of monitoring items. Because the Mahalanobis distance is calculated for each state quantity of each classified data group, it is not necessary to calculate all the correlations of the state quantities of many monitoring items to obtain the Mahalanobis distance as in the past. As a result, the computation for obtaining the Mahalanobis distance can be speeded up, and the plant abnormality determination work can be quickly performed.
In addition, in the Mahalanobis distance individual comparison process D ′, when at least one Mahalanobis distance exceeds a predetermined threshold value in each comparison with each Mahalanobis distance in each data group and a predetermined threshold value, the following is performed. By determining that the plant is “abnormal” in the abnormality determination process E, it is possible to quickly determine the abnormality in each monitoring target, and as a result, to perform the abnormality determination work of the plant quickly. Can also be obtained.
Also in this embodiment, as in the previous embodiment, by classifying the state quantities of many monitoring items of the plant into data groups consisting of a smaller number of state quantities of a plurality of monitoring items, each data group In addition, the unit space state amount, the time interval for measuring the signal space state amount, and the time interval for changing the unit space can be set individually, and for each data group, the unit space search criteria, signal Spatial search criteria can be individually set, and various data processing that can improve the reliability of monitoring can be easily performed.

  As mentioned above, although embodiment of this invention was explained in full detail with reference to drawings, the concrete structure is not restricted to this embodiment, The design change etc. of the range which does not deviate from the summary of this invention are included.

1 Gas Turbine Power Plant 6 Gas Turbine 12 Processing Unit 20 Processing Unit 21 Processing Unit

Claims (4)

The process of classifying the state quantities of many monitoring items in the plant into a data group consisting of state quantities of a smaller number and a plurality of monitoring items;
The process of calculating the Mahalanobis distance from the state quantity for each data group,
A process of comparing the Mahalanobis distance calculated for each data group with a threshold set in advance for each data group,
A process for determining the presence or absence of a plant abnormality based on the comparison result, and a plant operation state monitoring method comprising:
In the classification process of the data group, the data group is classified based on the measurement position of the state quantity,
The plant operation state monitoring method, wherein the measurement positions include a plurality of locations along the circumferential direction of the rotating shaft of the gas turbine.   The classification process of the data group according to claim 1, wherein in the classification process of the data group, the data is classified into a data group including a combination of a monitoring item having a high correlation with the measurement position of the state quantity and the measurement position of the state quantity. Plant operating condition monitoring method.   The plant operation state monitoring method according to claim 1, wherein in the classification process of the data group, the data group is classified based on a sampling period of the state quantity.   The plant operation state monitoring method according to claim 1, wherein in the classification process of the data group, the data group is classified based on a type of the state quantity.

Images