KR102072836B1 - Data classification method based on correlation, and a computer-readable storege medium having program to perform the same - Google Patents

Abstract

A method of classifying data in consideration of correlation and a computer readable storage medium storing a program for performing the method are provided. According to an embodiment of the present invention, a data classification method considering a correlation may include: receiving a data set including first and second data, which are operation data received from equipment of a plant, determining a correlation factor of the data set And setting the data set as a first group based on the correlation factors.
In a computer-readable storage medium according to an embodiment of the present invention, a program for performing a data classification method in consideration of the correlation is stored.

Description

DATA CLASSIFICATION METHOD BASED ON CORRELATION, AND A COMPUTER-READABLE STOREGE MEDIUM HAVING PROGRAM TO PERFORM THE SAME}

The present invention relates to a data classification method and a computer readable storage medium in consideration of correlation. More particularly, the present invention relates to a grouping method according to similarity between data and a computer readable storage medium storing a program for performing the method.

In general, there are many facilities in various industrial plant facilities, and they monitor whether they are working properly so that actions can be taken before serious problems occur.

For example, for power plants, ancillary equipment such as turbines and auxiliary systems, generators and auxiliary systems, boilers and auxiliary systems, main water supply systems, condensate systems, fuel supply systems, cooling water systems, circulating water systems and auxiliary steam systems In the case of turbine and auxiliary system, the high pressure turbine, medium pressure turbine, low pressure turbine, main steam control valve system, main steam shutoff valve system, turbine speed control system, turbine bleed system, turbine bearing lubricant system, etc. Each of these systems, in turn, consists of unit devices or sub-systems, and these facilities work together to produce electricity. If the operation of these facilities is out of normal or degrades, an alarm will occur or the entire plant or plant will be forcibly shut down when it is dangerous to operate the equipment.

Therefore, in order to produce the desired product at the desired quality level and cost, it is necessary to continuously monitor the operation status of the above-mentioned facilities constituting the plant in real time and to maintain optimal operation and performance.

The plant facility includes a plurality of modules and a plurality of sensors for monitoring the plurality of modules, and should receive as many as tens of thousands of detection signals per unit time from the plurality of sensors to determine whether the plant is operating normally. In a power plant plant installation, minor failures in one module can cause serious malfunctions throughout the plant. If a plant stops operating, it can lead to large power outages such as blackouts in the high-demand season, as well as the maintenance costs of the plant itself, which can also be serious.

When an abnormal signal occurs in an existing industrial plant, experts check the plant facility schematic (P & ID) to identify the cause of the abnormal signal, but due to the characteristics of the plant facility, it observes tens of thousands of operating signals and analyzes the linkage between them. It is very difficult and time consuming to find the faulty equipment, so the monitoring system can improve the stability of plant operation by quickly detecting the faulty equipment and the system ripple effect by taking appropriate corrective action when abnormal signal occurs. Systemization skills are required.

SUMMARY OF THE INVENTION The present invention has been made in view of the above-mentioned problems, and a technical classification method for grouping data on the basis of correlation factors by determining correlation factors between received data and a computer-readable program storing a program for performing the method. It is to provide a storage medium.

Another technical problem to be solved by the present invention is to provide a data classification method for filtering abnormal data in a dataset based on correlation factors and a computer-readable storage medium storing a program for performing the method.

Technical problems of the present invention are not limited to the aforementioned technical problems, and other technical problems not mentioned will be clearly understood by those skilled in the art from the following description.

In order to solve the above-mentioned technical problems, the data classification method in consideration of the correlation according to an embodiment of the present invention, receiving a data set including the first and second data that is the operating data received from the equipment of the plant Determining a correlation factor of the dataset and setting the dataset as a first group based on the correlation factor.

The determining may include determining coordinates for each unit time of the data set and showing the coordinates.

The determining may further include extracting a continuous section of the illustrated coordinates, and the correlation factor may be determined based on the slope of the continuous section.

The determining may further include determining a distribution of the illustrated coordinates.

In the setting, the first group may be set based on an absolute value of the correlation factor.

The setting may further include setting the data set into a plurality of groups including a second group.

In the setting, the abnormal data of the data set may be filtered based on the correlation factor.

In a computer-readable storage medium according to an embodiment of the present invention, the method comprising the steps of: receiving a data set including first and second data which is operation data received from a plant equipment, determining a correlation factor of the data set And a program for performing a data classification method in consideration of correlations, including setting the data set as a first group based on the correlation factors.

According to the present invention as described above, it is possible to improve the learning efficiency of the data received from the plurality of equipment constituting the plant by grouping the received data set into a plurality of groups based on the correlation factor, and filter the abnormal data to plant You can more accurately monitor the dataset that represents the state of.

1 is a view showing a schematic flow of a data classification method in consideration of a correlation according to an embodiment of the present invention.
FIG. 2 is a diagram illustrating a process of determining a correlation factor in a data classification method considering a correlation according to an embodiment of the present invention.
3 is a diagram illustrating a grouping process of data sets in a data classification method considering a correlation according to an embodiment of the present invention.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. Advantages and features of the present invention, and methods for achieving them will be apparent with reference to the embodiments described below in detail in conjunction with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below, but can be implemented in various forms, and only the embodiments are intended to complete the disclosure of the present invention, and the general knowledge in the art to which the present invention pertains. It is provided to fully convey the scope of the invention to those skilled in the art, and the present invention is defined only by the scope of the claims. Like reference numerals refer to like elements throughout.

Unless otherwise defined, all terms used in the present specification (including technical and scientific terms) may be used in a sense that can be commonly understood by those skilled in the art. In addition, terms that are defined in a commonly used dictionary are not ideally or excessively interpreted unless they are specifically defined clearly.

The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. In this specification, the singular also includes the plural unless specifically stated otherwise in the phrase. As used herein, “comprises” and / or “comprising” does not exclude the presence or addition of one or more other components in addition to the mentioned components.

Hereinafter, a method of classifying data in consideration of correlation and a computer readable storage medium storing a program for performing the method will be described with reference to the accompanying drawings.

Referring to FIG. 1, a schematic flow of a data classification method considering a correlation according to an embodiment of the present invention is disclosed. 1 is a view showing a schematic flow of a data classification method considering a correlation according to an embodiment of the present invention, Figure 2 is a method of determining a correlation factor in a data classification method considering a correlation according to an embodiment of the present invention 3 is a diagram illustrating a grouping process of data sets in a data classification method considering a correlation according to an embodiment of the present invention.

In detail, the data classification method considering the correlation according to the present embodiment includes: receiving a data set including first and second data, which is operation data received from equipment of a plant (S10), and a correlation factor of the data set. Determining (S20, S30) and setting the dataset as a first group based on the correlation factors (S40).

Plant plants often have a plurality of modules that are organically combined to closely influence each other. Therefore, a plant facility includes a plurality of modules and a plurality of sensors for monitoring them in real time. As a method for determining whether the plant is operating normally, in this embodiment, it is possible to manage the history of the data set received from the plurality of equipment constituting the plant. The plant facility may learn the received datasets and compare them with the current state of the plant to determine or predict whether the plant is operating normally or abnormally.

However, it is necessary to determine the normal operation of the plant equipment by receiving as many as tens of thousands of detection signals per unit time from the plurality of sensors, and abnormal data may cause a decrease in learning efficiency. Therefore, there is a need for a method for effectively learning the received dataset and a method for improving the accuracy of the dataset to be learned.

In the present embodiment, the received data set may be grouped according to similarity in advance to selectively learn the data group required for the corresponding equipment, and the abnormal data group may be filtered in advance to improve the learning efficiency.

To this end, first, a step S10 of receiving a data set including first and second data, which is operation data received from a plurality of equipments constituting the plant, may be performed. As described above, the plurality of operation data included in the received data set may be different in units or scale. For example, the equipment constituting the power plant includes a cooling water pump, a high-pressure turbine, a boiler, a current collector, a generator, and the like. Operation data received from such equipment is various units or scales such as ℃, Kw, tons, rpm, and the like. May have

Accordingly, in the present embodiment, data distortion due to units or scales can be prevented through a scale correction process between data. The scale correction process may include a method of extracting a transform factor using an average of maximum and minimum values and correcting a data scale based on the transform factor, but is not limited thereto. Various correction or interpolation techniques may be used.

Next, an operation (S20) of extracting the first data and the second data from the received data set may be performed. As a criterion for extracting the first data and the second data, various criteria such as extracting in order of importance of determining whether the plant or equipment is operating properly, randomly selecting from a data set, or extracting time-series data, etc. Applicable

In operation S30, a correlation factor between the extracted first data and the second data may be determined. In the present embodiment, the correlation factor may be determined using the slope, distribution, and the like, but is not limited thereto, and the correlation factor may be determined in various ways. A detailed process of determining the correlation factor will be described later with reference to FIG. 2.

Finally, the step S40 of grouping the datasets based on the determined correlation factors may be performed. By comparing the correlation factors generated from each operation data with the reference correlation factors, operation data having similar correlations can be grouped. In this case, in order to group the correlation factors having the same size but having the negative value and the correlation values having the positive value into the same group, each correlation factor may perform the grouping step S40 based on the absolute value.

In this embodiment, although the K-NN and the Gaussian mixture model (GMM) are used as classifiers for classifying the driving data, the present invention is not limited thereto. The data may be grouped using various classification algorithms such as grouping data having similar correlations sequentially into a plurality of groups.

The K-Neighbor classifier finds K data in the order of closest distance from the given data and assigns the group to the largest group among them. K-Neighbors are a nonparametric learning algorithm and a kind of distance-based classifier.

Gaussian mixture models (GMMs) are also useful for classifying data and are commonly used for parametric learning algorithms. The Gaussian mixture model is a density estimation method that improves the method of modeling the distribution density of a given dataset as a probability density function. The Gaussian mixture model can provide each group of data as a linear combination of Gaussian distributions. Parameters can be inferred from the learned data using a maximum-minimum algorithm.

In addition, in the grouping step S40, the abnormal operation data may be filtered in the data set based on the correlation factor. By comparing correlation factors, we can filter out abnormal data having a certain degree of correlation or higher for more precise plant monitoring and data learning. In this case, the abnormal data may mean data that does not reach a predetermined correlation, and abnormal data included in the abnormal category may be sequentially grouped into the plurality of groups according to the correlation. In the case of learning using the grouped abnormal data, it can be determined whether the abnormal operation is compared with the current state of the plant equipment.

Referring to FIG. 2, a process of determining a correlation factor in a data classification method considering a correlation according to an embodiment of the present invention is disclosed.

First, an operation (S100) of extracting first data and second data from a received data set may be performed. The first data may be used as reference data, and the remaining data sets including the second data may be used as comparison data. As described above, the method of selecting the reference data may be data that is an important indicator for plant learning or monitoring, but is not limited thereto. The reference data may be determined based on various criteria.

Next, a step (S200) of determining coordinates for each unit time of the first data and the second data may be performed. Since the data generated by the equipment of the plant is generally time series data, each data can be arranged corresponding to the time domain. In an embodiment with reference to FIG. 1, the dataset may not be affected by a unit or scale through a scale correction process.

In operation S300, coordinates of the first data and the second data may be performed. For example, the first data can be arranged on the x-axis and the second data on the y-axis, and the coordinates for each unit time can be shown in this two-dimensional area. If the value of the first data corresponding to the first unit time is 3 and the value of the second data is 7, it can be shown as D1 (3,7). In this way, the coordinates of the first and second data corresponding to the unit time can be shown.

The distributed first data and the second data may be performed to extract a continuous section of coordinates according to a reference (S410) or to determine a distribution of the coordinates (S450). The former step S410 may perform a process of serializing the coordinates of the discontinuous data. When the continuous section is extracted (S410), the step of extracting the slope of the continuous section may be performed (S420).

By performing the steps of S410 to S450, it is possible to perform the step (S500) of determining the correlation factor using the slope or distribution.

In the present embodiment, the correlation factor may be determined using a Pearson correlation coefficient, but is not limited thereto, and the correlation factor may be determined based on various criteria such as characteristics of each device and importance weight of the data.

Correlation factors are statistics representing the covariate relationship between two variables. Generally speaking, correlation factor or correlation coefficient means Pearson's correlation coefficient. Pearson's product moment correlation coefficient, Pearson's r (Pearson's r), r, and R are all other terms that describe Pearson's correlation coefficient. Pearson's correlation coefficient is in the range of -1 ~ 1. If the correlation factor between two variables is positive, it is called positive correlation. In other words, when the first variable increases, the second variable also increases.

For two variables x and y, n observations are given by (x1, y1), (x2, y2), ..., (xn, yn), and the mean of x and y is m (x), If m (y), the method of determining the correlation factor r for two variables x and y is as follows.

First, the deviation between each variable and the mean can be obtained. The deviation of x1 in the first observation (x1, y1) is x1-m (x), the deviation of y1 is y1-m (y), and in the same way the deviation of x and y for the i th observation is xi-m (x), yi-m (x).

Next, the covariance can be obtained. If variance can determine how the population is distributed based on the mean of one variable, covariance is an indicator of the relationship between two variables rather than one. Covariance can be obtained by multiplying the deviations of two variables and taking their average. The equation for calculating the covariance is as follows.

In Equation 1, n-1 is used instead of n in the denominator to calculate n averages, which is due to degrees of freedom. As mentioned above, xi-m (x), yi-m (y) represent deviations, and the sum of these deviations should always be zero. Therefore, Equations 2 and 3 must be satisfied.

If the sum of the deviations is zero, either value may be determined and fixed by all other values such that the sum of the deviations is zero even if all observations change. Here, the number of freely varying observations may be degrees of freedom. In this embodiment, the degree of freedom may be determined to be n-1.

Finally, the standard deviations of the two variables can be obtained to find the correlation factor by dividing the sum of the covariances by the product of the standard deviations. The sum of standard deviations for xi may be expressed as in Equation 4.

Similarly, the degrees of freedom allow the denominator to be n-1 rather than n. The correlation factor r can be determined as follows.

Here, the n-1 terms included in the numerator and denominator are canceled and can be ignored. Therefore, the correlation factor r can also be calculated by the following equation.

As described above, the value of the correlation factor r may have a range equal to or greater than -1 and equal to or less than 1. Correlators have no units and can be defined independently of the unit being measured. Also, the correlator has no directivity. That is, the correlation coefficient of x and y is equal to the correlation coefficient of y and x. The correlation factor between the first and second variables represented by the slope or distribution chart may be determined using the above-described process of obtaining the correlation factor.

Finally, in operation S600, the correlation factors of the first data and the data set may be compared and grouped based on the correlation factors of the first data and the second data.

Referring to FIG. 3, a grouping process of datasets is disclosed in a data classification method considering a correlation according to an embodiment of the present invention.

The reference data is determined from the plurality of operation data included in the data set 10, and the reference correlation factor 20 is determined from the first data which is the reference data and the second data which is the comparison data. A correlation factor between the first data and operation data included in the remaining data sets is determined, and the data set is grouped into a plurality of groups 12 including the first group according to the degree of correlation by comparing with the reference correlation factor 20. can do. As a method for classifying datasets, K-Near or Gaussian Mixed Model (GMM) can be used.

On the other hand, the data that does not meet the predetermined correlation can be determined and filtered as the abnormal data (11). The remaining data group 12 except the filtered abnormal data 11 may be selectively learned according to the characteristics of the plant or the equipment.

Alternatively, by selectively learning the data that does not meet the predetermined degree of correlation can be determined whether the abnormal operation compared to the current state of the plant equipment.

Meanwhile, the present invention may be implemented by storing computer readable codes in a computer readable storage medium. The computer readable storage medium includes all kinds of storage devices for storing data that can be read by a computer system.

The computer readable code, when read and executed by a processor from the computer readable storage medium, is configured to perform the steps of implementing a data classification method considering the abnormal correlation according to the present invention. The computer readable code may be implemented in various programming languages. And the functional program, code and code segments for implementing the embodiments of the present invention can be easily programmed by those skilled in the art.

Examples of computer-readable storage media include ROM, RAM, CD-ROM, magnetic tape, floppy disk, optical data storage, and the like, and also include implementing in the form of a carrier wave (eg, transmission over the Internet). The computer readable storage medium can also be distributed over network coupled computer systems so that the computer readable code is stored and executed in a distributed fashion.

Although embodiments of the present invention have been described above with reference to the accompanying drawings, those skilled in the art to which the present invention pertains may implement the present invention in other specific forms without changing the technical spirit or essential features thereof. I can understand that. Therefore, it should be understood that the embodiments described above are exemplary in all respects and not restrictive.

10: dataset
11: Abnormal data
12: Data group
20: correlation factor

Claims (8)

In the data classification method in consideration of the correlation performed by the device interworking with the plant equipment,
Receiving a dataset from the plant equipment, the dataset comprising a plurality of operational data;
Determining a correlation factor of the data set; And
Grouping the dataset into a plurality of groups based on the correlation factors;
Receiving the data set,
When the plurality of operation data included in the data set have different scales, scale correction between the plurality of operation data is performed based on a scale conversion factor,
Grouping into the plurality of groups,
Grouping the plurality of operation data included in the data set into a plurality of groups according to the degree of correlation between the plurality of operation data and the correlation factor; And
And selectively learning according to the characteristics of the plant equipment based on the plurality of operation data for each of the plurality of groups. The method of claim 1,
Determining the correlation factor,
Extracting reference data and comparison data from the data set; And
Determining a correlation factor between the reference data and the comparison data, and determining a reference correlation factor from the reference data. The method of claim 2,
Grouping into the plurality of groups,
And comparing the correlation factors with the reference correlation factors to sequentially group data having similar correlation among the plurality of operation data into the plurality of groups. The method of claim 2,
Grouping into the plurality of groups,
And comparing the correlation factors with the reference correlation factors to sequentially group data having different correlation degrees among the plurality of operation data into the plurality of groups. The method of claim 1,
And comparing the current state of the plant equipment with the current state of the plant equipment based on the data selectively learned according to the characteristics of the plant equipment. The method of claim 1,
And comparing the current state of the plant equipment with the current state of the plant equipment based on the data selectively learned according to the characteristics of the plant equipment. The method of claim 1,
And monitoring the dataset indicative of the condition of the plant equipment. A computer readable storage medium having stored thereon a program for performing the method of claim 1.

Images