JPH06289086A - Fault-state discriminating method - Google Patents

Abstract

PURPOSE:To make it possible to detect the cause of the line-to-ground falut occurring in a power line without visiting the site by analyzing the spectrum of the current waveform of the line-to-ground fault of the power line, and performing the duplicated discrimination analysis based on the power spectrum value. CONSTITUTION:The current waveform of the line-to-ground fault of a power line is detected with a CT. The DC component of the waveform is cut. Dividing computation is performed for the current value of the line-to-ground fault at the maximum amplitude. Normalization is performed so that the maximum amplitude of any waveform becomes equal. The spectrum of the normalized waveform is analyzed. Various peaks appear. The patterns of the peaks are changed by the causes of the line-to-ground faults. When the causes of the faults are equal, the common power- spectrum value is obtained. Thus, the cause is estimated by the analysis using the duplicated discrimination analysis method and a neutral. network method. The analysis is repeated so as to narrow the cause into the upper value. The analysis is performed only for the narrowed causes of the faults, and the final discriminated result is obtained. Information for seasons and data around the site is inputted into an inferrence system, and the cause of the fault is judged in the integrated mode.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention According to the present invention, when a ground fault occurs in a power line such as a transmission line or a distribution line, the cause can be known by easy waveform analysis and discriminant analysis without going to the site. The present invention relates to a possible failure appearance determination method.

[0002]

2. Description of the Related Art The main cause of a ground fault of a power line as described above is that a tree contacts the power line,
Examples include the contact of animals such as birds and snakes with the power line, the contact of cranes with the power line, and the fact that the insulator supporting the power line is contaminated and leakage current flows. it can.

When such a ground fault occurs, it is possible to easily understand that the ground fault has occurred at a monitoring station or the like, but conventionally, in order to identify the cause, complicated arithmetic processing for current and voltage information is required. However, it was greatly affected by the measurement error. Also, there is a limit to the range of failure cause identification, and it is necessary for the worker to visit the site and grasp the cause to identify the cause, but it is not easy to identify the point of occurrence of the ground fault, There was a problem that it took a lot of time and labor because it was necessary to climb onto the tower.

[0004]

SUMMARY OF THE INVENTION The present invention solves the above-mentioned conventional problems and finds out the cause of a ground fault in a power line by easy waveform analysis and discriminant analysis without going to the site. It has been completed in order to provide a failure appearance determination method capable of performing the above.

[0005]

In order to solve the above problems, the inventors of the present invention have studied various methods, and as a result, the waveform of the ground fault current generated in the power line depends on the type of the object in contact with the power line. It was found that the waveforms peculiar to each object can be grasped as numerical values by focusing on the fact that each waveform has a peculiar shape and spectral analysis of this waveform by frequency. The present invention has been completed based on the above findings, and spectrally analyzes the ground fault current waveform of the power line,
A specific frequency is selected based on the power spectrum value, a multiple discriminant analysis of that value is performed, and a discriminant analysis using a neural network is performed. The two results are compared to determine the cause of the final ground fault. It is characterized by making a distinction. The above-mentioned spectrum analysis and determination of the cause of the ground fault can be performed by a system using a computer.

[0006]

The present invention will be described in more detail below with reference to the illustrated embodiments. FIG. 1 is a flow sheet of an embodiment of the present invention. As shown in FIG. 1, first, the ground fault current waveform of the power line is C
Detect by T. This ground-fault current waveform is stored for analysis, and the part where the features are clearly shown is taken out as the analysis part. 2 and 3 show examples of the ground fault current waveforms extracted in this way, and FIG. 2 shows the ground fault current waveforms when a tree, a bird, or a snake contacts the power line. Shows the ground fault current waveforms when insulation failure occurs in the power cable and when the insulator that insulates and supports the power line is soiled.

By the way, such a ground-fault current waveform may contain a DC component, and the maximum amplitude is different. Therefore, it is necessary to normalize the waveform before performing spectrum analysis. For this reason, the DC component is cut off, the ground fault current value is divided by the maximum amplitude, and any ground fault current waveform is normalized so that the maximum amplitude is the same.

Next, the normalized ground fault current waveform is spectrally analyzed. The results are as shown in FIGS.
The peak of the spectrum intensity appears in the commercial frequency part, but various peaks also appear in the part of the frequency (harmonic) that is an integral multiple thereof, and the pattern changes depending on the cause of the ground fault.

For example, in the case of the tree shown in FIG. 4, a peak with a large spectral intensity appears only in the commercial frequency part, and almost no peak appears in other parts. Figure 5
In the case of the bird shown in (1), a slight peak is generated even at the frequency three times the commercial frequency. In addition, in the case of the snake shown in FIG. 6, in addition to the large peak of the commercial frequency part, 1000 Hz
A large number of complex peaks occur in the area up to. On the other hand, in the case of the cable of FIG. 7, the peak of the harmonic is larger than that of the commercial frequency portion, and in the case of the fouling insulator of FIG. 8, it is similar to the case of the snake but much higher. It contains frequency components.

FIGS. 4 to 8 show the results of spectrum analysis of the ground fault current waveforms of FIGS. 2 and 3, and the same waveform is not always generated. However, if the cause of the ground fault accident is the same, a common power spectrum value can be obtained, so the cause is estimated by analyzing this using the multiple discriminant analysis method and the neural network method. The multiple discriminant analysis process is a method of discriminating by showing numerical values the possibility of a plurality of causes, and this analysis process is repeatedly executed to narrow down to a higher rank and to estimate the cause. In this system, the fault current waveform obtained in advance is divided into before and after flashover (flashover) to
Wave and three waves immediately after the flashover are extracted, respectively, Fourier-transformed into a power spectrum, and then the power spectrum value at the frequency of an integral multiple of the commercial frequency and the fault current before the flashover, at the time of the flashover, A group (group) of each failure cause is made with the ground fault current value after flashover as a representative characteristic, and a criterion (discriminant) for dividing them is derived. Here, the newly obtained data is subjected to a discriminant analysis process, and the results are narrowed down to the higher-ranked failure causes (analyzed as likely), and only the narrowed down failure causes are targeted. As a result, the analysis process is performed to obtain the final determination result. Also, in the case of neuro processing, the possibility of each cause is output as a numerical value, and this is a method of performing discrimination.In this system, the power spectrum value and the ground fault current value of the above-mentioned representative characteristics are input data in advance as learning data. A neural network of a back propagation model consisting of three layers of an intermediate layer and an output layer is randomly given to the input layer several thousand times for learning, new data is input to the neural network, and the largest value among the failure causes of the output layer is set. The output result is used as the determination result.

Further, in the case of the embodiment, in addition to reading the results of the multiple discriminant analysis and the results of the neuro processing into an inference system known as an expert system, the information about the season and time around the site is also taken in to determine the cause of the ground fault. Make a comprehensive judgment. In other words, the result of the multiple discriminant analysis method, which has no learning ability but can derive a discriminant with an appropriate amount of data and carry out the analysis, and the enormous amount of data are required to create the network in the initial stage. The results of the neural network method that has the ability to learn each time new data is obtained are weighted by taking advantage of their respective advantages and which analysis result is to be emphasized according to the data amount of the determined failure cause. To judge. At that time, in order to further improve the probability of positive discrimination, the information on the season and time at the time of occurrence of a ground fault is taken in, the impossibility of the cause of the fault is determined before the determination, and the cause of the discrimination is deleted. Limit the results. This is because, for example, if a confusing power spectrum value occurs in the winter even though the snake is hibernating, it may be mistaken for a ground fault due to the snake, or a heavy machine such as a crane (metal This is to prevent erroneous determination that a ground fault is caused by a stick. In addition to this, in case of a subtle final judgment, the actual results of failure causes at the same time in the past and the cumulative number of each failure cause are taken in from the database that stores the failure result data, and used in the comprehensive judgment. More accurate inference can be made by adding weighting to which discrimination result is selected. In this way, the cause of failure can be determined accurately and in detail.

[0012]

As described above, the method of discriminating a failure aspect of the present invention utilizes the fact that the waveform of the ground fault current generated in the power line has a unique shape depending on its cause.
The cause of a ground fault that has occurred in a power line can be known by easy waveform analysis and discriminant analysis without visiting the site. Therefore, by using the present invention, it is possible to know the cause of the ground fault at a remote monitoring station or the like with considerable accuracy, and it is possible to immediately take measures according to the cause. Therefore, the present invention greatly contributes to the industry as a failure aspect determination method that solves the conventional problems.

[Brief description of drawings]

FIG. 1 is a flow sheet showing an example of the present invention.

FIG. 2 is a waveform diagram showing a waveform of a ground fault current.

FIG. 3 is a waveform diagram showing a waveform of a ground fault current.

FIG. 4 is a graph showing a spectrum of a ground fault current waveform generated by a tree.

FIG. 5 is a graph showing a spectrum of a ground fault current waveform generated by a bird.

FIG. 6 is a graph showing a spectrum of a ground fault current waveform generated by a snake.

FIG. 7 is a graph showing a spectrum of a ground-fault current waveform caused by defective insulation of a cable.

FIG. 8 is a graph showing a spectrum of a ground fault current waveform generated by a pollution insulator.

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Masakatsu Arakane 2-7, Noda, Kuwana-shi, Mie 15 (72) Inventor Kazuji Otaneda 1-3, Ishigasaki, Missouri-cho, Okazaki-shi, Aichi (72) Inventor Hiroyuki Katsukawa 62 Inaba, Takagi, Fugo-cho, Niwa-gun, Aichi Prefecture (72) Inventor Koji Tokuyama 259 Matsui-cho, Meito-ku, Nagoya-shi, Aichi (72) Inventor Shikano 4-chome, Obata 4-chome, Moriyama-ku, Aichi Prefecture Issue (72) Inventor Yuzo Ito 1 of 43 Noshicho, Mizuho-ku, Nagoya, Aichi

Claims (1)

[Claims] 1. A ground fault current waveform of a power line is spectrally analyzed, a specific frequency is selected based on the power spectrum value, a multiple discriminant analysis of the value is performed, and a discriminant analysis using a neural network is performed. A method for determining a failure aspect, which comprises comparing two results and determining a cause of a final ground fault.