KR19990070710A - Partial discharge measuring device to remove noise in real time using neural network - Google Patents

Abstract

By using the partial discharge measuring device to detect the partial discharge signal and noise generated from the measuring object such as cable, transformer, etc. to remove the noise and to remove the noise in real time by using neural network that can output only the pure partial discharge signal. A partial discharge measuring device is disclosed. When the voltage is applied after the measurement object is mounted on the measurement object mounting part, the partial discharge signal and noise generated from the measurement object are applied to the detection impedance circuit part. The partial discharge signal and noise applied to the detection impedance circuit portion are applied to the neural network device through an amplifier. The neural network apparatus compares the set signal with the signal applied from the measurement object and applies only the signal corresponding to the set signal to the computer so that noise is not applied to the computer. Therefore, the computer displays only the partial discharge signal generated from the measurement object, so that the measurer can check the partial discharge signal quickly and accurately.

Description

Partial discharge measuring device to remove noise in real time using neural network

The present invention relates to a partial discharge measuring device, and more particularly, by using a partial discharge measuring device to detect a partial discharge signal and noise generated from a measurement object such as a cable, transformer, etc. to remove noise and output only a pure partial discharge signal. The present invention relates to a partial discharge measuring device for removing noise in real time using a neural network.

A conventional partial discharge measuring apparatus is shown in FIG. That is, Figure 1 is a view showing a conventional partial discharge measuring device, a power supply unit 10 for applying a voltage; A measurement object mounting part 12 connected to the power supply part 10 in parallel with the capacitor C; A detection impedance circuit portion 16 connected to the measurement object mounting portion 12; It consists of an oscilloscope 20 connected to the detection impedance circuit section 16 via an amplifier 18.

In the conventional partial discharge measuring apparatus configured as described above, first, the measurement object mounting part 12 is mounted on the measurement object 14. The measurement object 14 is an insulator and corresponds to a cable, a transformer, a joint box, and the like. After the measurement object 14 is mounted on the measurement object mounting part 12, a voltage is applied to the measurement object 14 using the power supply unit 10. When a voltage is applied from the power supply unit 10 to the measurement object 14, the noise generated by the measurement object 14 and the partial discharge signal are applied to the detection impedance circuit unit 16. The noise and the partial discharge signal are generated in the measurement object 14. Since the measurement object 14 is an insulator, partial discharge should not occur, but in the case where foreign matter is attached to the measurement object 14, current is temporarily applied to the foreign material. Charging and discharging occurs. Since the measurement object 14 is an insulator, when the partial discharge signal is generated, the insulation function is destroyed and the normal function is not performed, so the partial discharge signal must be accurately measured and removed.

However, the partial discharge signal generated in the measurement object 14 and applied to the detection impedance circuit unit 16 is applied together with the noise. The detection impedance circuit unit 16 amplifies the partial discharge signal and noise applied from the measurement target 14 through the amplifier 18 and then applies it to the oscilloscope 20. The oscilloscope 20 simultaneously displays the partial discharge signal and noise amplified by the amplifier 18. Therefore, the measurer must check the partial discharge signal displayed on the oscilloscope 20 to determine the presence or absence of the partial discharge signal generated from the measurement object 14, but there is a problem in that the partial discharge signal cannot be quickly and accurately determined due to noise.

The present invention has been made to solve the above problems, the object of the present invention is to remove the noise by detecting the partial discharge signal and noise generated from the measurement object, such as cable, transformer using a partial discharge measuring device The present invention provides a partial discharge measuring apparatus for removing noise in real time using a neural network capable of outputting only a partial discharge signal.

1 is a view showing a conventional partial discharge measuring device.

2 is a view showing a partial discharge measuring apparatus according to the present invention.

3A is a waveform diagram schematically illustrating an example of a partial discharge signal applied to a partial discharge measuring apparatus.

3B is a waveform diagram schematically showing an example of a noise signal applied to the partial discharge measuring apparatus.

4 is a waveform diagram illustrating a state in which a partial discharge signal is applied to a neural network controller.

Explanation of symbols for main parts of the drawings

50: power supply unit 60: measuring object mounting unit

62: measuring object 70: detection impedance circuit

80: amplifier 90: neural network device

92: A / D converter 94: Neural network controller

96: computer

The present invention provides a power supply for applying a voltage to achieve the above object; A measurement object mounting part connected to the power supply part in parallel with the capacitor; A detection impedance circuit portion connected to the measurement object mounting portion; A neural network device connected to said detection impedance circuit part via an amplifier; And it provides a partial discharge measuring device that is connected to the neural network device, and removes the noise in real time using a neural network consisting of a computer for displaying a signal applied from the neural network device to the outside.

According to the present invention, when the voltage is applied after the measurement object is mounted on the measurement object mounting portion, the partial discharge signal and noise generated from the measurement object are applied to the detection impedance circuit portion. The partial discharge signal and noise applied to the detection impedance circuit portion are applied to the neural network device through an amplifier. The neural network apparatus compares the set signal with the signal applied from the measurement object and applies only the signal corresponding to the set signal to the computer so that noise is not applied to the computer. Therefore, the computer displays only the partial discharge signal generated from the measurement object, so that the measurer can check the partial discharge signal generated from the measurement object quickly and accurately.

Hereinafter, the present invention will be described with reference to the accompanying drawings.

2 is a view showing a partial discharge measuring apparatus according to the present invention.

3A is a waveform diagram schematically illustrating an example of a partial discharge signal applied to a partial discharge measuring apparatus.

3B is a waveform diagram schematically showing an example of a noise signal applied to the partial discharge measuring apparatus.

4 is a waveform diagram illustrating a state in which a partial discharge signal is applied to a neural network controller.

Referring to FIGS. 2 to 4, a partial discharge measuring apparatus for removing noise in real time using a neural network will be described. First, the measurement object mounting unit is parallel to the power supply unit 50 to which a voltage is applied. 60 is connected. The measurement impedance mounting section 60 is connected to a detection impedance circuit section 70 for detecting and processing a partial discharge signal generated from the measurement target 62 mounted on the measurement mounting section 60. The neural network device 90 is connected to the detection impedance circuit unit 70 through an amplifier 80. A computer 96 is connected to the neural network device 90, and the computer 96 displays a signal applied from the neural network device 90. The neural network device 90 is composed of an A / D converter 92 and a neural network controller 94, and the A / D converter 92 is a digital signal applied to the analog signal applied through the amplifier 80. Is converted to and applied to the neural network controller 94. The neural network controller 94 learns in advance the partial discharge signal and noise signal desired by the measurer. The neural network controller 94 is composed of 128 neurons in the input layer, 8 neurons in the hidden layer, and two neurons in the output layer, and uses a back propagation algorithm.

Hereinafter, an embodiment of the present invention will be described.

First, the waveform corresponding to the partial discharge signal (FIG. 3 (a)) and the waveform corresponding to the noise signal (FIG. 3 (b)) are inputted to the neural network control unit 94 of the neural network device 90 to learn in advance. . The signal input to the neural network controller 94 repeats equation (1) until the desired output value is obtained to obtain the connection strength W1 between the input layer and the hidden layer, and the connection strength W2 between the hidden layer and the output layer.

Z = f2 (W2 * f1 ((W1 * X + b1) + b2) ------------- (1)

(f1, f2 = evaluation function, W1, W2 = connection strength, b1, b2 = bias, X = input value, Z = output value)

The connection strength thus obtained determines the relationship between neurons and neurons, and when data is input, the value changes according to the connection strength and appears as an output. For example, if the output value is 1 and 0 as the partial discharge signal, and if the output value is 0 or 1 as the noise signal, the output value is 0.76 or 0.1, this can be determined as the partial discharge signal. That is, if the output value is 0.5 or more, the determination is accepted. If the output value is 0.5 or less, the signal is excluded.

Referring to the process of inputting the input signal to the neural network device 90, the waveform corresponding to the partial discharge signal of FIG. 4 is triggered at zero potential to sample the waveform with the A / D converter. That is, 128 pieces of sampled data are substituted into the input layer of the neural network device for learning or determination. Each input pulse signal is thus determined, and only the signal determined as the partial discharge signal is transmitted to the computer 96 together with the phase value at that time. The signal transmitted to the computer 96 is represented by the magnitude of the partial discharge signal with respect to the phase using computer graphics.

In more detail, the measurer inputs the partial discharge signal and noise signal shown in FIGS. 3 (a) and 3 (b) to the neural network control unit 94 constituting the neural network device 90 to learn in advance. After the measurement object 62 is mounted on the measurement object mounting part 60, a voltage is applied through the power supply unit 50. When a voltage is applied from the power supply unit 50, when a foreign matter is adhered to the measurement object 62, a partial discharge signal and a noise signal are generated from the measurement object 62. The noise signal and the partial discharge signal generated from the measurement target 62 are detected by the detection impedance circuit unit 70, amplified into a signal having a predetermined magnitude through the amplifier 80, and then applied to the neural network device 90. The partial discharge signal and the noise signal, which are analog signals applied to the neural network device 90, are converted into digital signals by the A / D converter 92 and then applied to the neural network controller 94. The partial discharge signal and the noise signal applied to the neural network controller 94 are compared with the patterns corresponding to the previously learned partial discharge signal and the noise signal. The neural network controller 94, when the partial discharge signal and noise signal applied from the A / D converter 92 satisfies the previously learned pattern and determines that the partial discharge signal is the partial discharge signal, only the partial discharge signal corresponding to the computer 96 is detected. The noise signal is not applied to the computer 96. Therefore, since only the partial discharge signal is displayed on the computer 96, the partial discharge signal can be confirmed quickly and accurately.

As can be seen from the above description, in the present invention, when the voltage is applied after the measurement object is mounted on the measurement object mounting portion, the partial discharge signal and noise generated from the measurement object are applied to the detection impedance circuit unit. The partial discharge signal and noise applied to the detection impedance circuit portion are applied to the neural network device through an amplifier. The neural network apparatus compares the set signal with the signal applied from the measurement object and applies only the signal corresponding to the set signal to the computer so that noise is not applied to the computer. Therefore, the computer displays only the partial discharge signal generated from the measurement object, so that the measurer can check the partial discharge signal quickly and accurately.

Claims (3)

A power supply unit 50 for applying a voltage; A detection impedance circuit unit 70 connected to the power supply unit 50 in parallel with the capacitor C and connected to the measurement object mounting unit 60 on which the measurement object 62 is mounted; A neural network device (90) connected to the detection impedance circuit unit (70) through an amplifier (80); And Partial discharge measurement device that is connected to the neural network device (90) and removes noise in real time using a neural network comprising a computer (96) for displaying a signal applied from the neural network device (90) to the outside. 2. The neural network device (90) according to claim 1, wherein the neural network device (90) is applied from an A / D converter (92) for converting an analog signal applied from the amplifier (80) into a digital signal and from the A / D converter (92). Partial discharge to remove noise in real time using a neural network, characterized in that the neural network control unit 94 for comparing the partial discharge signal and the noise signal to the pattern corresponding to the pre-learned partial discharge signal and noise signal Measuring device. The neural network controller 94 receives a determination if the neural network output of the partial discharge signal and the noise signal applied from the A / D converter 92 is 0.5 or more, and outputs the result to the computer 96. , Partial discharge measurement device for removing noise in real time using a neural network, characterized in that the output is not output to the computer (96) except 0.5 or less.