KR101625087B1 - A electrical equipment fault diagnostic apparatus according to harmonic analysis - Google Patents

Abstract

The present invention relates to a device for diagnosing an electrical equipment breakdown through a harmonic analysis. The device comprises: a measurement device module measuring a current, flowing from a ground of an electrical equipment, through computed tomography (CT); a database module (MySQL) receiving data of the measured current, with a periodic and constant time interval, from the measurement device module to store the data; an intelligent algorithm analysis diagnosing module applying an algorithm of polynomial-based radial basis function neural networks (pRBFNNs) to analyze a waveform pattern and diagnose a breakdown based on the analyzed waveform pattern; and a web-based graphic user interface (GUI) module displaying an analyzed diagnose result of the intelligent algorithm analysis diagnosing module to allow a user to check the analyzed diagnose result. According to the present invention, the device may block and prevent an unexpected breakdown by monitoring a result of a progress and a breakdown status in real time, thereby allowing power to be stably provided.

Description

TECHNICAL FIELD [0001] The present invention relates to an apparatus for diagnosing an electrical equipment failure by harmonic analysis,

The present invention relates to an apparatus for diagnosing an electrical equipment failure by harmonic analysis, and more particularly, to an apparatus for diagnosing an electrical equipment fault by harmonic analysis, which can diagnose a failure of a transformer through an integrated harmonic analysis, .

Generally, with the increase of electric power demand, the transformer is becoming large capacity, super high voltage, and the substation is being unmanned. If an accident occurs in such a large-capacity transformer, the ripple effect is wide, and economic loss and psychological anxiety become enormous, so that there is an increasing need for insulation diagnosis to prevent the accident of a transformer. The transformer is used to supply the high voltage supplied through the pole to the electric power demand such as the factory and the home. The housing is filled with insulating cooling oil having insulation property for cooling the heat generated during the transforming process. And some of them are nonflammable synthetic insulating oils, and there are various properties depending on the application, and they are generally used as having high volume electrical resistance, low viscosity, and stability against oxidation. However, such a transformer is often subjected to thermal deterioration due to high temperature operation, thermal deterioration due to an external short circuit, mechanical deterioration, and other discharge deterioration due to partial discharge, resulting in mechanical deterioration, vibration increase, .

Insulation breakdown of the transformer causes many problems such as explosion, short circuit, etc. of the transformer. In order to minimize the failure, the insulation oil in the transformer is regularly collected and the deterioration of the transformer is checked by measuring the deterioration in the laboratory.

Recently, a constant monitoring device has been developed to constantly measure abnormality, which is the beginning of an accident, in the operating state of the transformer in order to secure the reliability of the transformer and supply the electric power stably. , A preventive diagnosis system is developed to judge the type of abnormality based on the correlation between abnormality detection data and to stop the operation in the event that the abnormality develops due to the abnormality and to take countermeasures have. On-line fault detection technology applied to the constant monitoring of transformer includes gas analysis technology, partial discharge measurement technology and temperature measurement technology of transformer insulation oil. As an associated prior art document, there is disclosed an inflow transformer equipped with a sensor for measuring an oil insulation deterioration of Korean Patent Laid-Open Publication No. 10-2011-0047432. Preventive diagnosis system to prevent sudden failure of electric power equipment and diagnose deterioration condition through such continuous monitoring is inexpensive even though it is expensive equipment.

Since most power transformers are sealed inside with insulating oil, a fault diagnosis system using dissolved gas analysis sensor in insulating oil may be effective. However, the type of gas generated by the type of deterioration occurring in the transformer, It is very difficult to apply the present invention to an apparatus for determining the degree of deterioration of a transformer in a field on a line.

In addition, a diagnosis device using a sensor for detecting a wavelength of ultrasound or UHF band, which occurs when a partial discharge occurs in a breaker or a transformer, is complicated in the process of generating waveforms and filtering out external noise, The reliability is still insufficient.

The present invention has been proposed in order to solve the above-mentioned problems of the previously proposed methods. The present invention measures and acquires the current flowing from the ground of the electric equipment through the CT of the measuring device module, and performs an algorithm of the pRBFNNs And a fault diagnosis device for diagnosing faults of the electric equipment by harmonic analysis to analyze faults of the electric equipment by analyzing the waveform pattern of the separated harmonics included in the measured current data, And the like.

In addition, the present invention can diagnose the failure of the transformer through an integrated harmonic analysis regardless of the type of the load in diagnosing the failure of the electrical equipment through the intelligent algorithm analysis diagnosis module. In addition, It is another object of the present invention to provide an apparatus for diagnosing an electrical equipment failure by harmonic analysis, which can prevent and prevent an unexpected fault through real-time monitoring in advance, thereby enabling stable power supply.

In addition, the present invention constitutes a web-based GUI module so that the user can check the user-controlled and analyzed diagnosis results of the intelligent algorithm analysis diagnostic module, thereby improving user convenience in monitoring failure of the electrical equipment, It is another object of the present invention to provide an apparatus for diagnosing an electric equipment failure by harmonic analysis, which enables real-time remote monitoring of facilities.

According to an aspect of the present invention, there is provided an apparatus for diagnosing an electrical equipment failure by harmonic analysis,

As an apparatus for diagnosing an electrical equipment failure by harmonic analysis,

A measuring device module for measuring the current flowing from the ground of the electric equipment through CT;

A database module (MySQL) for receiving and storing data of the current measured by the measurement device module at regular time intervals;

The data of the measured current stored in the database module is applied to a polynomial-based radial basis function neural network (pRBFNNs) algorithm based on FCM (Fuzzy C-means) An intelligent algorithm analysis diagnostic module that analyzes waveform patterns of harmonics and seventh harmonics and diagnoses failures according to them; And

여기서, Rc는 c(c=1,…, k)번째 규칙, Acn은 클러스터 수, n은 입력변수의 수, fc는 다항식 함수이다.
And a web-based GUI (Graphic User Interface) module for displaying the user-controlled and analyzed diagnosis results of the intelligent algorithm analysis diagnostic module so that the user can confirm the diagnosis result,
Wherein the intelligent algorithm analysis diagnostic module comprises:
A condition part of a polynomial-based Radial Basis Function Neural Networks (pRBFNNs) based on a fuzzy C-means (FCM), a conclusion part and a reasoning part,
The condition part functions to divide the input space into local areas equal to the number of clusters (the number of fuzzy rules) in order to reflect characteristics of the input data, and output the degree of belonging of each local area as a fuzzy set,
The conclusion part forms a rule after " then " in Equation (1) representing the input space divided by the conditional part by a polynomial function of each local area,
The reasoning unit is configured to output the final output of the network by fuzzy reasoning based on an " If-then " fuzzy rule.
[Equation 1]

Here, Rc is the c (c = 1, ..., k) rule, Acn is the number of clusters, n is the number of input variables, and fc is a polynomial function.

Preferably, the measuring device module includes:

And a current probe module for amplifying a current value measured by the CT.

Advantageously, the database module further comprises:

(FFT) with 10,000 pieces of data in units of 1 second through reading and processing of current data under the control of the intelligent algorithm analysis and diagnosis module, receiving and storing current data from the measurement device module, Transform) can be performed to store the separated harmonic data.

delete

delete

delete

delete

delete

delete

delete

delete

Preferably, the web-based GUI module further comprises:

A diagnosis diagnostic setting and diagnosis execution area, a Fourier transform result display area, a diagnostic execution status and result display area, and a time domain trend and a frequency domain graph display area.

More preferably, the failure diagnosis setting and diagnosis performing area includes:

Fault diagnosis is set up in units of 30 minutes as a part to set the interval to diagnose faults periodically by the system through the fault diagnosis setting, and the diagnosis execution area is set to perform the process based on the set time interval after the fault diagnosis setting Function.

More preferably, the Fourier transform result display area includes:

A third harmonic wave (180 Hz), a fifth harmonic wave (300 Hz), and a third harmonic wave (300 Hz) among the values of the frequency domain of the immediately preceding time, and outputs the last Fourier transform result and the new Fourier transform result. , And the seventh harmonic wave (420 Hz), and the new Fourier transform result indicates the value of the fundamental wave (60 Hz), the third harmonic wave (180 Hz ), A fifth harmonic wave (300 Hz), and a seventh harmonic wave (420 Hz).

More preferably, the diagnostic performance status and the result display area include:

The number of times the diagnosis is performed in the performance status is displayed at the beginning of how many diagnoses have been performed since the "execution of diagnosis" button was pressed. The present progress level indicates which part is currently being performed in the diagnosis,

In the result, the classification prediction result shows the pure value outputted from the model, and the final diagnosis result can display the state shown through the condition based on the classification prediction result.

More preferably, the time-domain trend and the frequency-domain graph display region include:

The time domain trend graph display area is a graph window that allows one second data in the time domain to be viewed in a trend format, in which 10,000 current samples are sequentially displayed,

The frequency domain graph display area is a graph window that displays values in the frequency domain, and can function to display the values of the fundamental wave, the third harmonic, the fifth harmonic, and the seventh harmonic part in a graph.

According to the apparatus for diagnosing an electrical equipment failure by the harmonic analysis proposed in the present invention, it is possible to measure the current flowing from the ground of the electrical equipment apparatus through the CT of the measuring apparatus module and obtain an FCM-based algorithm of pRBFNNs By configuring the intelligent algorithm analysis diagnostic module, the waveform pattern of the separated harmonics included in the measured current data can be analyzed to effectively diagnose the failure of the electrical equipment.

According to the present invention, in diagnosing the failure of the electrical equipment through the intelligent algorithm analysis diagnostic module, it is possible to diagnose the failure of the transformer through the integrated harmonic analysis irrespective of the type of the load, It is possible to block and prevent an unexpected failure through real-time monitoring of the power supply, thereby enabling stable power supply.

In addition, the present invention constitutes a web-based GUI module so that the user can check the user-controlled and analyzed diagnosis results of the intelligent algorithm analysis diagnostic module, thereby improving user convenience in monitoring failure of the electrical equipment, Real-time remote monitoring of the facility can be enabled.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a diagram showing an overall system configuration of an apparatus for diagnosing an electrical equipment failure by harmonic analysis according to an embodiment of the present invention; FIG.
BACKGROUND OF THE INVENTION 1. Field of the Invention [0001] The present invention relates to an apparatus for diagnosing an electrical equipment failure by harmonic analysis, and more particularly,
3 is a diagram illustrating a structure of an FCM-based pRBFNNs applied to an intelligent algorithm analysis diagnostic module according to an embodiment of the present invention;
4 is a graph showing current data obtained by an apparatus for diagnosing an electrical equipment failure by harmonic analysis according to an embodiment of the present invention;
5 to 7 are diagrams showing a classification model 1, a classification model 2, and a classification model 3 of the structure of the FCM-based pRBFNNs applied to the intelligent algorithm analysis diagnostic module according to the present invention.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings, in order that those skilled in the art can easily carry out the present invention. In the following detailed description of the preferred embodiments of the present invention, a detailed description of known functions and configurations incorporated herein will be omitted when it may make the subject matter of the present invention rather unclear. In the drawings, like reference numerals are used throughout the drawings.

In addition, in the entire specification, when a part is referred to as being 'connected' to another part, it may be referred to as 'indirectly connected' not only with 'directly connected' . Also, to "include" an element means that it may include other elements, rather than excluding other elements, unless specifically stated otherwise.

FIG. 1 is a diagram showing the overall system configuration of an apparatus for diagnosing an electrical equipment failure by harmonic analysis according to an embodiment of the present invention. FIG. 2 is a schematic diagram of an apparatus for diagnosing an electrical equipment failure by harmonic analysis according to an embodiment of the present invention. FIG. 2 is a diagram showing a GUI zone partitioning of a web-based GUI module applied to the present invention. 1, an apparatus for diagnosing electrical equipment failure 100 by harmonic analysis according to an embodiment of the present invention includes a measurement device module 110, a database module 120, an intelligent algorithm analysis diagnostic module 130 ), And a web-based GUI module 140.

The measurement device module 110 is a device configuration for measuring the current flowing from the ground of the electric equipment device 10 through the CT 111 (Current Transformer). The measurement device module 110 may further include a current probe module 112 for amplifying a current value measured by the CT 111. Here, the measurement device module 110 primarily stores the measured current data (10,000 points per second), and the stored data is stored in the database module 120 to be described later at predetermined time intervals. The electric equipment 10 is a structure for acquiring harmonic data of the electric equipment. In the present invention, for example, a three-phase induction motor which occupies a large part of the electric equipment used in the industry is selected. Since induction motors (generators) are widely used in industrial facilities, they are included in important industrial facilities, and in the case of expensive prime electric motors (generators), maintenance of these equipments is very importantly managed .

The database module 120 is a database (MySQL) that receives and stores data of the current measured by the measurement device module 110 at regular time intervals. The database module 120 receives and stores the current data from the measurement device module 110 and reads and processes the current data under the control of the intelligent algorithm analysis and diagnosis module 130, And performs FFT (Fast-Fourier Transform) with 10,000 data in 1 second through the data to store the separated harmonic data. That is, the database module 120 is a MySQL DB that has been widely used in the Linux OS. The database module 120 fetches and stores the measured current data from the measurement device module 110, performs FFT with 10,000 pieces of data per 1 second, Save to DB. The reading and processing of such data can be executed through a Java program.

The intelligent algorithm analysis and diagnosis module 130 applies a polynomial-based radial basis functional neural network (pRBFNNs) algorithm based on fuzzy C-means (FCM) to the measured current data stored in the database module 120 It analyzes the waveform pattern of DC wave, fundamental wave, third harmonic, fifth harmonic and seventh harmonic, and diagnoses the fault according to it. The intelligent algorithm analysis and diagnosis module 130 includes conditional parts of polynomial-based Radial Basis Function Neural Networks (pRBFNNs) based on fuzzy C-means (FCM), conclusion part and reasoning part, The input space is divided into a number of clusters having a number of c number of clusters (the number of fuzzy rules) and a degree of affiliation of each local region is output as a fuzzy set. The conclusion is that the input space, Then, the reasoning unit outputs the final output of the network by the fuzzy reasoning based on the " If-then " fuzzy rule.

Where R _c is the c (c = 1, ..., k) rule, A _cn is the number of clusters, n is the number of input variables, and f _c is a polynomial function.

The intelligent algorithm analysis diagnostic module 130 is a prediction and classification algorithm developed not only by combining a radial basis function with a conventional polynomial-based neural network theory, but also by fusing it with a fuzzy inference mechanism. The general pRBFNNs structure uses the Gaussian function as the activation function, and the output value of each Gaussian function is applied as the connection weight of the neural network. The connection weight and the polynomial of the neural network are fuzzy relations according to the fuzzy inference method Prediction, and discrimination. That is, the fuzzy rule of the polynomial pRBFNNs based on the FCM rather than the general pRBFNNs model can be expressed as Equation (1) described above.

FIG. 3 shows a structure of an FCM-based pRBFNNs applied to the intelligent algorithm analysis diagnostic module 130 according to the present invention. Here, the conditional part of the FCM-based polynomial RBFNNs does not have a hidden layer having the Gaussian function as an active function, and the entire conditional part is performed through FCM clustering. The conditional function of polynomial RBFNNs using FCM clustering divides the input space into the number of clusters (the number of fuzzy rules) as many as the number of clusters (number of fuzzy rules) and outputs the membership degree of each local domain as a fuzzy set. Here, the FCM clustering method is an algorithm for assigning a degree of belonging to a distance between each data and the center of a specific cluster, and classifying data according to the degree of affiliation.

Also, the conclusion of the FCM-based polynomial RBFNNs is a local session model that expresses the input space divided by the conditional part by a polynomial function of each local area, and forms a rule after "then" in [Equation 1]. f _ji (x) output neurons subscript j on in (= 1, ..., s) to skip a f _i (x) is Equation 2, Equation 3, and Equation 4] under the It has the form of one of three types of functions. That is, the local session model can be expressed as a constant term, a linear equation, or a quadratic equation.

Also, the reasoning part of the FCM-based polynomial RBFNNs obtains the final output of the network by fuzzy reasoning based on "If-then" fuzzy rules. The input signals are multiplied by the neurons labeled " " in the inference portion shown in FIG. 1 and the result is output to the final output of the output layer neuron. Thus, the sequence is the same as the fuzzy inference process, and has the same form as Fuzzy Neural Networks. In conclusion, the final output of the output of the first output in FIG. 1, which shows the structure of the FCM-based pRBFNNs, is expressed by the following Equation 5 by fuzzy inference.

Also, as a Fuzzy C-Means Cluster Algorithm design, the FCM clustering method assigns the degree of belonging to the distance between each data and the center of a specific cluster, and classifies data according to the degree of affiliation. FCM clustering is an objective function By minimizing [Equation 6], a value belonging to each cluster of input data is obtained.

Where c is the number of clusters (number of fuzzy rules), N is the number of input patterns, m is the fuzzification coefficient, and m is greater than 1.0. x _k is the k-th input vector and v _i is the center of the i-th cluster. u _ik is a real number between 0 and 1 indicating the degree of belonging to which the kth data belongs to the i-th cluster, and satisfies the following equations (7) and (8).

In Equation (6), ∥ · ∥ uses a Weighted Euclidean Distance expressed in Equation (9) below.

Where _j is the standard deviation of the j-th input dimension of the input patterns. Weighted cladian distances are widely used because they provide reasonable distance information that is not significantly affected by the data size distribution. An input vector set X = {x ₁ , x ₂ , ..., N) consisting of N patterns in n-dimensional Euclidean space , x _N }, x _k ∈ R ⁿ , 1 ≤ _k ≤ _N and the cluster center v = {v ₁ , v ₂ , ... , v _c }, v _i ∈ R ⁿ , 1 ≤ _i ≤ _c are represented by U = [u _ik ] and u _ik and v _i are expressed by the following equations (10) and (11). ≪ / RTI >

That is, the FCM clustering modifies the belonging matrix U and the center v _i (i = 1, ..., c) of each cluster while iteratively repeating the [Expression 10] and the [Expression 11] The function Q (U, v ₁ , v ₂ , ..., v _c ) is converged to a specific value. The fuzzification coefficient "m" is a very important factor that determines the type of membership function of each cluster. In general, when 2 is used and less than 2 based on 2, overlapping regions between the cluster regions are reduced and regions having membership degrees near 1 and 0 are increased, and the characteristics of a crisp set other than a fuzzy set are shown Area increases. 2, the fuzzy set with a sharp center point of the cluster is generated, and the difference value of belonging to each cluster is decreased.

The web-based GUI module 140 displays the user-controlled and analyzed diagnosis results of the intelligent algorithm analysis diagnostic module 130 so that the user can confirm the diagnosis results. 2, the web-based GUI (Graphic User Interface) module 140 includes a failure diagnosis setting and diagnosis execution area 141, a Fourier transform result display area 142, A situation and result display area 143, and a time-domain trend and a frequency-domain graph display area 144, as shown in FIG. Here, the fault diagnosis setting and diagnosis execution area 141 is a section for setting the interval for diagnosing the fault periodically by the system through the fault diagnosis setting, and the fault diagnosis is set every 30 minutes. And performs the progress based on the set time interval after the progress. Also, the Fourier transform result display area 142 displays the last Fourier transform result and the new Fourier transform result, and the last Fourier transform result indicates the fundamental wave (60 Hz), the third harmonic 180 Hz), the fifth harmonic wave (300 Hz), and the seventh harmonic wave (420 Hz), and the new Fourier transform result indicates the value of the fundamental wave (60 Hz), a third harmonic (180 Hz), a fifth harmonic (300 Hz), and a seventh harmonic (420 Hz). In addition, the number of diagnoses performed in the performance status and result display area 143 is displayed at the beginning of the number of diagnoses performed after the " execute diagnosis " button is initially pressed. The result of the classification prediction is displayed as a pure value outputted from the model, and the final diagnosis result indicates the state shown through the conditional statement based on the classification prediction result. At this time, the diagnosis status can be divided into 1: data read status, 2: graph display status, 3: diagnosis execution and display status, and the final diagnosis result can be divided into normal and inspection needs. The time-domain trend graph and frequency-domain graph display area 144 are graph windows that allow one-second data in the time domain to be viewed in a trend format, in which 10,000 current samples are sequentially displayed , And the frequency domain graph display area is a graph window for displaying the values of the frequency domain, and functions to display the values of the fundamental wave, the third harmonic, the fifth harmonic, and the seventh harmonic part in a graph.

The process of monitoring the diagnosis of failure of the electrical equipment 10 through the web-based GUI module 140 may be performed as follows. First, a user accesses a web server through a web browser, and confirms the facility through the operation of the monitoring system. As an actual process performed in the web server, data is read from the database (DB) module 120 through the JSP when the user accesses the HTML through the Apache server and starts operating the monitoring system, and the JSP reads the read data To HTML (JavaScript), inputs it to the diagnostic algorithm designed in JSP, and sends the diagnostic result back to HTML. Next, HTML displays the data and result data that have been received through JS on the screen. That is, the web-based GUI module 140 may be configured with a PC level graphical user interface (GUI) that helps the user to use the intelligent algorithm analysis and diagnosis module 130. The web-based GUI module 140 is a web-based GUI created for troubleshooting. It is used to create a program based on HTML for quick and easy access, and a JSP (Java Server Pages) and JS (Java Script), and the configuration of the server uses an Apache server using a tool called Tomcat.

The performance verification of the electrical equipment fault diagnosis apparatus by harmonic analysis according to an embodiment of the present invention will be described. First, the data obtained for the development of the fault diagnosis classification model is shown in [Table 1]. One data is the current value measured for 1 sec, which has 10,000 samples. More samples are taken to be meaningless in acquiring the current value and a total of 1,300 samples are used. The Fourier transform of the acquired data is performed based on 1 second, and is converted into an area ranging from 0 to 5,000 Hz from 10,000 data samples.

FIG. 4 is a graph showing current data obtained by an apparatus for diagnosing an electrical equipment failure by harmonic analysis according to an embodiment of the present invention. 4 (a) shows a graph of normal data of the time domain current value measurement, FIG. 4 (b) shows a graph of failure data of the time domain current value measurement, And Fig. 4 (d) shows a graph of the failure data of the frequency domain transformation. Here, the obtained steady state data is difficult to express in the report a large number of acquired data, and is shown as an example as shown in Figs. 4 (a) to 4 (d). These data graphs are composed of 100,000 samples (10 sec) per graph, and the number of current values in the time domain is 130 (60 normal values and 70 high values). The frequency domain transform is a graph obtained by converting the data obtained in the time domain into the frequency domain. As mentioned above, the conversion into the frequency domain is performed based on 1 second, and the number of samples is 10,000. Accordingly, the total number of graphs is 1,300 (600 normal values and 700 high-order graphs).

The developed classification model structure of the intelligent algorithm analysis diagnostic module 130 according to the present invention can be classified into three classification models as shown in FIGS. 5 to 7 according to input factors. The overall classification model structure is designed using the FCM-based polynomial RBFNNs as shown in FIG. 3, and is a classification model having a simple structure change that occurs as the input factor decreases from five to three. First, in the classification model 1 shown in FIG. 5, five types of input factors are selected, namely, DC wave, fundamental wave, third harmonic, fifth harmonic, and seventh harmonic. More precisely, when the Fourier transform is performed, The amplitude of the frequency is the input factor. 6, the input factors of the classification model 3 shown in FIG. 7 have a fundamental wave, a third harmonic, a fifth harmonic, and a seventh harmonic. The input parameters of the classification model 3 are the third harmonic, The fifth harmonic, and the seventh harmonic, respectively. Although each of these classification models has the same overall structure, the performance difference is clearly revealed according to the change of the input factor, and this part will be described in the performance evaluation below. The common parameter values of each of the developed classification models normalize the input factors. That is, since the magnitude of each frequency factor is small and the magnitude of the magnitude of the magnitude of the magnitude of the magnitude of the magnitude of the magnitude of the magnitude of the magnitude of the magnitude is small, do.

Table 2 below shows Max and Min values for each factor.

Classification model For data learning and testing, data partitioning can be expressed as [Table 3] below. The learning and test data are divided by a ratio of 9: 1 using 10-fold Cross Validation, the number of clusters is 8, and the inference method applies linear inference.

Table 4 below shows the center coordinates of each cluster of Classification Model 1. In other words, the center coordinates of each cluster 1 to 8 are shown through FCM clustering. The structure of the classification model 1 has a five-dimensional space when the five input factors are regarded as one point, Are generated as shown in Table 4 below.

[Table 5] and [Table 6] below show the steady state and the failure state of the coefficient parameter of the linear inference formula of the classification model 1. [Table 5] and [Table 6] calculated through computer operation are coefficient parameters corresponding to the linear inference formula [Equation 3] used in the conclusion part of the classification model. This parameter is the key parameter that constitutes the classification model like the cluster center value in [Table 4], and it is the final parameter of the classification model 1 when there are 5 input factors through [Table 4], [Table 5], [Table 6] A classification model is completed.

Table 7 below shows the center coordinates of each cluster in Classification Model 2. The center coordinates of each cluster 1 through 8 are shown through FCM clustering. The structure of the classification model 2 has four input factors, so the center points of each cluster 1 to 8 are generated as shown in [Table 7] below.

[Table 8] and [Table 9] below show the steady state and the failure state of the coefficient parameter of the linear inference formula of the classification model 2. [Table 8] and [Table 9] calculated through computer operation are coefficient parameters corresponding to the linear inference formula [Equation 3] used in the conclusion part of the classification model. This parameter is the key parameter that constitutes the classification model like the cluster center value in [Table 7], and the final result of the classification model 2 when there are four input factors through [Table 7], [Table 8], [Table 9] A classification model is completed.

Table 10 below shows the center coordinates of each cluster in Classification Model 3. The center coordinates of each cluster 1 through 8 are shown through FCM clustering. The structure of the classification model 3 has three input factors. Therefore, the center point of each cluster 1 to 8 having a three-dimensional space can be expressed as [Table 10].

[Table 11] and [Table 12] below show the steady state and the failure state of the coefficient parameter of the linear inference formula of the classification model 3. [Table 11] and [Table 12] calculated through computer operation are coefficient parameters corresponding to the linear inference formula [Equation 3] used in the conclusion part of the classification model. The final classification model of the classification model 3 when the input parameters are three is completed through the key parameters [Table 10], [Table 11], [Table 12].

The following [Table 13] to [Table 15] show the test performance of the classification model 1, the classification model 2, and the classification model 3. As can be seen from the following [Table 13] to [Table 15], the test performance of [Table 13] showing the structure of the classification model 1 is the best, and [Table 14] Next, the test performance is shown in the order of [Table 14] showing the structure of the classification model 3.

The apparatus for diagnosing electrical equipment failure by harmonic analysis according to an embodiment of the present invention monitors and analyzes the current measured through CT differently from the method of diagnosing a fault condition of a transformer using various existing sensors The harmonics analysis is applied to the diagnosis by selecting the harmonics that are expected to be related to the load depending on the type of the load. In order to diagnose the fault, unlike the existing harmonic analysis technology, It is possible to provide differentiated technology to diagnose the fault of transformer through harmonic analysis.

The present invention may be embodied in many other specific forms without departing from the spirit or essential characteristics of the invention.

100: A fault diagnosis apparatus according to an embodiment of the present invention
110: Measuring device module 111: CT (Current Transformer)
112: current probe module 120: database module
130: Intelligent Algorithm Analysis Diagnostic Module
140: Web-based GUI module
141: Fault diagnosis setting and diagnostic execution area
142: Fourier transform result display area
143: Diagnosis performance status and result display area
144: Time domain trend and frequency domain graph display area

Claims (9)

1. An electric equipment fault diagnosis apparatus (100) for harmonic analysis, comprising:
A measuring device module 110 for measuring the current flowing from the ground of the electric equipment 10 through the CT 111;
A database module 120 (MySQL) for receiving and storing data of the current measured by the measurement device module 110 at regular time intervals;
A polynomial-based radial basis functional neural network (pRBFNNs) algorithm based on FCM (fuzzy C-means) is applied to the data of the measured current stored in the database module 120 to generate a DC wave, An intelligent algorithm analysis diagnostic module 130 for analyzing a waveform pattern of the fifth harmonic and the seventh harmonic and diagnosing a failure according to the waveform pattern; And
And a web based GUI (Graphic User Interface) module 140 for displaying user-controlled and analyzed diagnostic results of the intelligent algorithm analysis diagnostic module 130 so that the user can confirm the diagnosis results,
The intelligent algorithm analysis diagnostic module 130,
A condition part of a polynomial-based Radial Basis Function Neural Networks (pRBFNNs) based on a fuzzy C-means (FCM), a conclusion part and a reasoning part,
The condition part functions to divide the input space into local areas equal to the number of clusters (the number of fuzzy rules) in order to reflect characteristics of the input data, and output the degree of belonging of each local area as a fuzzy set,
The conclusion part forms a rule after " then " in Equation (1) representing the input space divided by the conditional part by a polynomial function of each local area,
Wherein the reasoning unit outputs the final output of the network by fuzzy reasoning based on " If-then " fuzzy rule-based fuzzy inference.
[Equation 1]

Here, Rc is the c (c = 1, ..., k) rule, Acn is the number of clusters, n is the number of input variables, and fc is a polynomial function.
The apparatus according to claim 1, wherein the measuring device module (110)
Further comprising a current probe module (112) for amplifying a current value measured by the CT (111).
2. The method of claim 1, wherein the database module (120)
And receives and stores current data from the measurement device module 110. The intelligent algorithm analysis and diagnosis module 130 reads 10,000 pieces of data in units of 1 second through read and processing of current data, Wherein the harmonic wave data is separated by performing an FFT (Fast-Fourier Transform) on the input signal.
delete 4. The web-based GUI module (140) according to any one of claims 1 to 3,
A fault diagnosis setting and diagnosis execution area 141, a Fourier transform result display area 142, a diagnosis execution status and result display area 143, and a time domain trend and frequency domain graph display area 144 Wherein the fault diagnosis apparatus comprises:
6. The system according to claim 5, wherein the failure diagnosis setting and diagnosis execution area (141)
Fault diagnosis is set up in units of 30 minutes as a part to set the interval to diagnose faults periodically by the system through the fault diagnosis setting, and the diagnosis execution area is set to perform the process based on the set time interval after the fault diagnosis setting Wherein the fault diagnosis apparatus comprises:
The display device according to claim 5, wherein the Fourier transform result display area (142)
A third harmonic wave (180 Hz), a fifth harmonic wave (300 Hz), and a third harmonic wave (300 Hz) among the values of the frequency domain of the immediately preceding time, and outputs the last Fourier transform result and the new Fourier transform result. , And the seventh harmonic wave (420 Hz), and the new Fourier transform result indicates the value of the fundamental wave (60 Hz), the third harmonic wave (180 Hz ), The fifth harmonic wave (300 Hz), and the seventh harmonic wave (420 Hz).
6. The system according to claim 5, wherein the diagnostic performance status and result display area (143)
The number of times the diagnosis is performed in the performance status is displayed at the beginning of how many diagnoses have been performed since the "execution of diagnosis" button was pressed. The present progress level indicates which part is currently being performed in the diagnosis,
Wherein the classification prediction result in the execution result indicates a pure value outputted from the model and the final diagnosis result indicates a state represented through a conditional statement based on the classification prediction result.
6. The method of claim 5, wherein the time domain trend and frequency domain graph display area (144)
The time domain trend graph display area is a graph window that allows one second data in the time domain to be viewed in a trend format, in which 10,000 current samples are sequentially displayed,
Wherein the frequency domain graph display region is a graph window for displaying a value in a frequency domain, wherein the value of the fundamental wave, the third harmonic, the fifth harmonic, and the seventh harmonic portion functions as a graph. Electrical equipment fault diagnosis equipment by.

Images