CN105807190A - GIS partial discharge ultrahigh frequency live-line detection method - Google Patents

Abstract

The invention provides a GIS partial discharge ultra-high frequency live detection method, the method comprising: S1 ultra-high frequency sensor receives the electromagnetic pulse signal generated by GIS partial discharge, and converts the electromagnetic pulse signal into a high-frequency voltage signal and then transmits it through a shielded cable to the partial discharge detector; at the same time, the wireless power frequency signal generator transmits the power frequency voltage signal to the partial discharge detector; the S2 partial discharge detector analyzes the high frequency voltage signal and the power frequency voltage signal to obtain the PRPD discharge spectrum and the superhigh Frequency discharge pulse waveform; S3, the partial discharge detector performs PRPD cluster analysis and pulse waveform time-frequency analysis on the PRPD discharge map and UHF discharge pulse waveform respectively, and identifies the discharge type of the GIS partial discharge according to the analysis results. The technical scheme provided by the invention has simple structure and convenient operation, and improves the accuracy and reliability of GIS partial discharge detection.

Description

A GIS Partial Discharge UHF Live Detection Method

technical field

The invention relates to a detection method, in particular to a GIS partial discharge ultra-high frequency charged detection method.

Background technique

GIS is one of the most important power equipment in the power grid system. It has been used more and more widely due to its small footprint and strong insulation level. With the rapid development of power grid technology, the voltage level of GIS is getting higher and higher. If an insulation fault occurs, it will directly damage the main equipment of the substation, cause power supply interruption, and cause a large area of power outage, affecting normal life, production and even social stability.

Insulation performance is an important factor that determines the safe and stable operation of GIS. Bubbles, burrs, scratches, loose screws or even falling off in the process of manufacturing, installation and operation will cause local field strength to increase and generate partial discharge. For example, the 500kV GIS in Jiangmen in China had a flashover shortly after it was put into operation due to the defect on the insulation operating rod. After the insulation test of the 400kV GIS in Daya Bay, it was found that the insulator at the connection between the GIS and the busbar had obvious leakage traces. Therefore, it is necessary to detect the partial discharge of GIS equipment.

When the internal discharge of GIS, due to the rapid transfer of charges at the discharge point, a current pulse with a duration of nanoseconds is formed, and an electromagnetic signal with extremely rich frequency components is generated. Partial discharge detection of the electrical signal generated by sensing partial discharge is not only The sensitivity can be improved, and the early partial discharge can be detected in time. However, because the electrical signal interference on site is mainly bus corona discharge, radio wave, carrier communication and switching action in the system, etc., these interferences are mainly concentrated below 300MHz, and the conventional pulse current method and radio frequency detection method cannot eliminate this well. The UHF detection method adopts the antenna coupling electromagnetic wave method, and its detection frequency band is mainly concentrated in the range of 300MHz-3000MHz, which can effectively avoid the interference received by the conventional pulse current detection method and effectively improve the detection sensitivity.

The existing ultra-high frequency detection method uses ultra-high frequency sensors to monitor GIS partial discharge. The ultra-high frequency sensor sends the partial discharge signal of the detection point to the partial discharge detector through the cable, and the power frequency voltage of the detection point is transmitted by the voltage transformer. The phase signal is sent to the partial discharge detector through the cable; then the partial discharge detection instrument analyzes the power frequency voltage phase signal and the partial discharge signal to obtain the PRPD spectrum, and diagnoses the partial discharge type of the GIS based on the PRPD spectrum.

In actual application, GIS often produces corona discharge in the initial stage of discharge, and the PRPD spectra generated by different defects are relatively similar. However, when there are multiple discharge sources inside a GIS device, the PRPD spectrum of partial discharge will appear partially The method of judging the partial discharge type of GIS equipment based on the PRPD spectrum only by the partial discharge detection instrument may lead to missing or misjudgment due to the overlap of the spectrum.

In addition, the existing ultra-high frequency detection method uses cables to obtain the power frequency voltage phase signal from the voltage transformer at the detection point. In actual use, there may not be available voltage transformers near the on-site detection point. It is also very inconvenient to measure because of the long cable connection.

Therefore, it is necessary to study a new detection method for partial discharge, in order to reduce the detection difficulty of partial discharge in GIS, improve the detection efficiency of partial discharge in GIS, and improve the accuracy of pattern recognition in GIS partial discharge.

Contents of the invention

In order to solve the above-mentioned problems existing in the prior art, the present invention provides a GIS partial discharge ultra-high frequency charged detection method.

The technical solution provided by the present invention is: a GIS partial discharge ultra-high frequency charged detection method, and its improvement is that: the method includes the following steps:

Step S1 The ultra-high frequency sensor receives the electromagnetic pulse signal generated by the partial discharge of the GIS, and converts the electromagnetic pulse signal into a high-frequency voltage signal and transmits it to the partial discharge detector through a shielded cable; at the same time, the wireless power frequency signal generating device transmitting a power frequency voltage signal to the partial discharge detector;

In step S2, the partial discharge detector performs data analysis on the high frequency voltage signal and the power frequency voltage signal to obtain a PRPD discharge spectrum and a UHF discharge pulse waveform;

Step S3, the partial discharge detector performs PRPD cluster analysis and pulse waveform time-frequency analysis on the PRPD discharge spectrum and the UHF discharge pulse waveform respectively, and according to the PRPD cluster analysis results and the pulse waveform Waveform time-frequency analysis results identify the discharge type of the GIS partial discharge.

Preferably, in the step S1, the UHF sensor is a UHF microstrip patch antenna, the high frequency microstrip patch antenna has a length of 5 cm, a width of 3.5 cm, and a thickness of 0.2 cm, which The bandwidth is 0.4GHz-2.7GHz; the high-frequency microstrip patch antenna is pasted on the inner and outer surfaces of the GIS.

Preferably, in the step S1, the wireless power frequency signal generating device is installed next to the UHF microstrip patch, and a low frequency signal receiver is installed in the partial discharge detector, and the partial discharge detector The instrument uses the low frequency signal receiver to receive the power frequency voltage signal sent by the wireless power frequency signal generator.

Preferably, in the step S2, the partial discharge detector performs data analysis on the amplitude and frequency of the high-frequency voltage signal and the phase of the power frequency voltage signal to obtain the PRPD discharge spectrum and superhigh Frequency discharge pulse waveform.

Preferably, the method of the PRPD cluster analysis in the step S3 is: extracting the characteristic parameters of the PRPD discharge spectrum, the characteristic parameters include: type characteristic parameters, moment characteristic parameters, and statistical characteristic parameters;

The type characteristic parameters include: the normalized value f1 of the fractal dimension of the power frequency positive half-wave partial discharge image, the normalized value f2 of the fractal dimension of the power frequency positive half wave partial discharge high value grayscale image, the normalized value f2 of the power frequency positive half wave partial discharge image The normalized value of the second-order generalized dimension of the wave partial discharge image is f3, the normalized value of the fractal dimension of the power frequency negative half-wave partial discharge image is f4, and the normalized value of the fractal dimension of the power frequency negative half-wave partial discharge high-value grayscale image The value f5, the second-order generalized dimension normalized value f6 of the power frequency negative half-wave partial discharge image;

The moment feature parameters include: the gray scale center of gravity coordinates f7 of the power frequency positive half-wave partial discharge image, the gray scale center of gravity coordinates f8 of the power frequency negative half wave partial discharge image, the characteristic parameter f9 of the main axis direction of the power frequency positive half wave partial discharge image, The characteristic parameter f10 of the main axis direction of the frequency negative half-wave partial discharge image;

The statistical characteristic parameters include: the ratio f11 of the power frequency positive half-wave discharge volume to the total discharge volume, the ratio f12 of the power frequency positive half-wave discharge times to the total discharge times, and the normalization of the power frequency positive half-wave initial discharge phase The value f13, the normalized value f14 of the initial discharge phase of the power frequency negative half-wave, and the correlation coefficient f15 of the power frequency positive and negative half-wave discharge images.

Preferably, the pulse waveform time-frequency analysis in the step S3 includes the following steps:

1) The following formula (1) is used to perform Gobor transformation on the UHF discharge pulse waveform z(t) to obtain the time-frequency analysis spectrum G _f (a, b, ω):

${\mathrm{<span\; class="notranslate">\; G</span>}}_{\mathrm{<span\; class="notranslate">\; f</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; a</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; b</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; \&omega;</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>{<span\; class="notranslate">\; \&Integral;</span>}_{<span\; class="notranslate">\; -</span><span\; class="notranslate">\; \&infin;</span>}^{<span\; class="notranslate">\; \&infin;</span>}\mathrm{<span\; class="notranslate">\; z</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; g</span>}}_{\mathrm{<span\; class="notranslate">\; a</span>}}^{<span\; class="notranslate">\; *</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; b</span>}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; e</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; j\&omega;t</span>}}\mathrm{<span\; class="notranslate">\; dt</span>}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>$

in: is the conjugate function of the Gaussian window function g _a (tb); t is time, ω is the angular frequency, f is the frequency, b is the window translation parameter, and a is the window scale adjustment parameter;

2) Extracting the characteristic parameters of the time-frequency analysis spectrum, the characteristic parameters include: the time t _m corresponding to the maximum amplitude point of G _f (a, b, ω), the maximum amplitude of G _f (a, b, ω) The frequency f _m corresponding to the value point, the difference Δt _m between the start time of the discharge pulse and the time at the maximum amplitude.

Preferably, in the step S3, the partial discharge detector uses a BP artificial neural network to train the characteristic parameters of the extracted PRPD discharge spectrum and the characteristic parameters of the time-frequency analysis spectrum, and output the corresponding The discharge type of the GIS PD.

Further, the BP neural network includes an input layer, an intermediate layer and an output layer; the input layer includes 18 corresponding to the 15 characteristic parameters of the PRPD discharge spectrum and the 3 characteristic parameters of the time-frequency analysis spectrum respectively. input layer neurons; the middle layer has 5 typical discharge sample data built in; the output layer includes 5 output layer neurons corresponding to the 5 typical discharge sample data of the middle layer.

Further, the five typical discharge sample data of the intermediate layer include noise sample data, needle point discharge sample data, creeping discharge sample data, air gap discharge sample data and suspension discharge sample data;

The characteristic parameters of the PRPD discharge spectrum and the characteristic parameters of the time-frequency analysis spectrum are input to the input layer of the BP neural network. The discharge sample data is trained to obtain the discharge type corresponding to the characteristic parameters of the PRPD discharge spectrum and the characteristic parameters of the time-frequency analysis spectrum, and output through the output layer of the BP neural network.

Further, the discharge types include noise, needle point discharge, creeping discharge, air gap discharge and levitation discharge.

Compared with the closest technical solution, the present invention has the following remarkable progress:

1) The technical solution provided by the present invention combines PRPD discharge spectrum and UHF discharge pulse waveform, extracts corresponding characteristic parameters through PRPD cluster analysis and pulse waveform time-frequency analysis, and utilizes BP artificial neuron network to carry out characteristic parameters Training, get the discharge type of GIS partial discharge, improve the accuracy and reliability of GIS partial discharge detection;

2) Compared with the traditional UHF antenna, the UHF microstrip patch antenna is used to detect the electromagnetic pulse signal generated by the partial discharge of GIS. It has a smaller volume and can be pasted on the inner and outer surfaces of the GIS cavity, reducing the detection difficulty;

3) The partial discharge detector has a built-in wireless receiving device to receive the power frequency voltage phase signal generated by the on-site wireless power frequency signal generator. The partial discharge detector is only equipped with a cable connected to the UHF sensor, which facilitates on-site operation and reduces detection difficulty.

Description of drawings

Fig. 1 is the structural representation of microstrip patch antenna;

Figure 2 is a schematic diagram of testing the UHF partial discharge charged detection device on a GIS;

Fig. 3 is Gabor time-frequency transformation result figure;

Figure 4 is a structural diagram of the BP artificial neuron network.

detailed description

In order to better understand the present invention, the content of the present invention will be further described below in conjunction with the accompanying drawings and examples.

Fig. 1 is a schematic diagram of the microstrip patch antenna designed by the present invention, the antenna is 5cm long, 3.5cm wide, 0.2cm thick, and has a bandwidth of 0.4GHz to 2.7GHz, which can be installed in the GIS cavity as a built-in sensor , can also be installed as an external sensor on the outer surface of the GIS pot insulator or the outer surface of the GIS cavity.

Fig. 2 is a schematic diagram of on-site detection of a 110kV single-phase GIS according to the present invention. Use ultra-high frequency sensors, wireless power frequency signal generators and partial discharge detectors to detect the discharge type of partial discharge in GIS;

The ultra-high frequency sensor adopts a microstrip patch antenna. The ultra-high frequency sensor is pasted outside the GIS cavity during detection, and the wireless power frequency signal generator is installed next to the ultra-high frequency sensor. The electromagnetic wave signal is converted into a high-frequency voltage signal and sent to the partial discharge detector. The partial discharge detector has a built-in low-frequency signal receiver, and the low-frequency signal receiver receives the power frequency voltage signal sent by the wireless power frequency signal generator. The partial discharge detector The amplitude and frequency of the high-frequency voltage signal sent by the UHF sensor and the phase of the power frequency voltage signal are analyzed to obtain the PRPD spectrum and UHF discharge pulse waveform; and then the PRPD discharge spectrum and UHF discharge are respectively analyzed. Perform PRPD cluster analysis and pulse waveform time-frequency analysis on the pulse waveform; obtain the characteristic parameters of the PRPD map and the time-frequency analysis map, and then use the BP artificial neuron network to train the characteristic parameters to identify the discharge type of GIS partial discharge and the recognition results It is divided into five types: noise, needle point discharge, surface discharge, air gap discharge and suspension discharge.

As shown in Table 1 below: PRPD cluster analysis mainly includes: extracting 15 characteristic parameters of PRPD spectrum's typing features, moment features and statistical features.

Table 1 The characteristic parameters of PRPD spectrum

Pulse waveform time-frequency analysis consists of two steps:

1) Use formula (1) to perform Gobor transformation on the UHF discharge pulse waveform z(t), and obtain the time-frequency analysis spectrum G _f (a,b,ω) of z(t):

${\mathrm{<span\; class="notranslate">\; G</span>}}_{\mathrm{<span\; class="notranslate">\; f</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; a</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; b</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; \&omega;</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>{<span\; class="notranslate">\; \&Integral;</span>}_{<span\; class="notranslate">\; -</span><span\; class="notranslate">\; \&infin;</span>}^{<span\; class="notranslate">\; \&infin;</span>}\mathrm{<span\; class="notranslate">\; z</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; g</span>}}_{\mathrm{<span\; class="notranslate">\; a</span>}}^{<span\; class="notranslate">\; *</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; b</span>}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; e</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; j\&omega;t</span>}}\mathrm{<span\; class="notranslate">\; dt</span>}<span\; class="notranslate">\; ;</span>$

is the conjugate function of Gaussian window function g _a (tb); t is time, f is frequency, w is angular frequency, b is window translation parameter, a is window scale adjustment parameter;

2) After obtaining the time-frequency analysis spectrum, extract the characteristic parameters in the time-frequency analysis spectrum, as shown in Table 2:

Table 2 Time-frequency analysis map characteristic parameters

Characteristic Parameters Concrete meaning t _m The time corresponding to the maximum value point f _m The frequency corresponding to the point of maximum magnitude Δt _m The difference between the start time of the discharge pulse and the time at the maximum amplitude

Fig. 3 is the result of the time-frequency transformation of a discharge pulse signal tested by the present invention, it can be seen that the joint time-frequency analysis has good time-frequency aggregation.

After obtaining the characteristic parameters of the PRPD discharge spectrum and the characteristic parameters of the time-frequency analysis spectrum, the partial discharge detector adopts the BP artificial neural network to train the characteristic parameters of the extracted PRPD discharge spectrum and the characteristic parameters of the time-frequency analysis spectrum , output the discharge type of the corresponding GIS partial discharge according to the training result.

As shown in Figure 4: the BP artificial neural network includes an input layer, an intermediate layer and an output layer; since the PRPD data extracts the clustering feature parameters and the pulse waveform data extracts a total of 18 parameters of the time-frequency joint distribution feature, the input layer of the BP neuron network There are 18 neurons; the middle layer is five typical discharge sample data of built-in noise, needle tip discharge, surface discharge, air gap discharge and suspension discharge; the output layer of BP neuron network has 5 neurons, and the output mode is noise, needle tip There are five types of discharge, creeping discharge, air gap discharge and suspension discharge. After the discharge characteristic data is input in the input layer, it is trained with the sample data until the output error is less than the system setting error, and then the output layer outputs the discharge type corresponding to the discharge characteristic data to realize the pattern recognition of GIS partial discharge.

The above is only an embodiment of the present invention, and is not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention are all pending applications for the rights of the present invention. within the required range.

Claims (10)

1. a GIS partial discharge ultra-high frequency charged detection method, is characterized in that: said method comprises the steps: Step S1 The ultra-high frequency sensor receives the electromagnetic pulse signal generated by the partial discharge of the GIS, and converts the electromagnetic pulse signal into a high-frequency voltage signal and transmits it to the partial discharge detector through a shielded cable; at the same time, the wireless power frequency signal generating device transmitting a power frequency voltage signal to the partial discharge detector; In step S2, the partial discharge detector performs data analysis on the high frequency voltage signal and the power frequency voltage signal to obtain a PRPD discharge spectrum and a UHF discharge pulse waveform; Step S3, the partial discharge detector performs PRPD cluster analysis and pulse waveform time-frequency analysis on the PRPD discharge spectrum and the UHF discharge pulse waveform respectively, and according to the PRPD cluster analysis results and the pulse waveform Waveform time-frequency analysis results identify the discharge type of the GIS partial discharge. 2. a kind of GIS partial discharge ultra-high frequency charged detection method as claimed in claim 1, is characterized in that: In the step S1, the UHF sensor is an UHF microstrip patch antenna, the high frequency microstrip patch antenna has a length of 5 cm, a width of 3.5 cm, a thickness of 0.2 cm, and a bandwidth of 0.4 cm. GHz～2.7GHz; the high-frequency microstrip patch antenna is pasted on the inner and outer surfaces of the GIS. 3. a kind of GIS partial discharge ultra-high frequency charged detection method as claimed in claim 1, is characterized in that: In the step S1, the wireless power frequency signal generating device is installed next to the UHF microstrip patch, and a low frequency signal receiver is installed in the partial discharge detector, and the partial discharge detector uses all The low frequency signal receiver receives the power frequency voltage signal sent by the wireless power frequency signal generator. 4. a kind of GIS partial discharge ultra-high frequency charged detection method as claimed in claim 1, is characterized in that: In the step S2, the partial discharge detector performs data analysis on the amplitude and frequency of the high-frequency voltage signal and the phase of the power frequency voltage signal to obtain the PRPD discharge spectrum and UHF discharge pulse waveform. 5. a kind of GIS partial discharge ultra-high frequency charged detection method as claimed in claim 1, is characterized in that: The method for the PRPD cluster analysis in the step S3 is: extract the characteristic parameters of the PRPD discharge spectrum, the characteristic parameters include: type characteristic parameters, moment characteristic parameters, and statistical characteristic parameters; The type characteristic parameters include: the normalized value f1 of the fractal dimension of the power frequency positive half-wave partial discharge image, the normalized value f2 of the fractal dimension of the power frequency positive half wave partial discharge high value grayscale image, the normalized value f2 of the power frequency positive half wave partial discharge image The normalized value of the second-order generalized dimension of the wave partial discharge image is f3, the normalized value of the fractal dimension of the power frequency negative half-wave partial discharge image is f4, and the normalized value of the fractal dimension of the power frequency negative half-wave partial discharge high-value grayscale image The value f5, the second-order generalized dimension normalized value f6 of the power frequency negative half-wave partial discharge image; The moment feature parameters include: the gray scale center of gravity coordinates f7 of the power frequency positive half-wave partial discharge image, the gray scale center of gravity coordinates f8 of the power frequency negative half wave partial discharge image, the characteristic parameter f9 of the main axis direction of the power frequency positive half wave partial discharge image, The characteristic parameter f10 of the main axis direction of the frequency negative half-wave partial discharge image; The statistical characteristic parameters include: the ratio f11 of the power frequency positive half-wave discharge volume to the total discharge volume, the ratio f12 of the power frequency positive half-wave discharge times to the total discharge times, and the normalization of the power frequency positive half-wave initial discharge phase The value f13, the normalized value f14 of the initial discharge phase of the power frequency negative half-wave, and the correlation coefficient f15 of the power frequency positive and negative half-wave discharge images. 6. a kind of GIS partial discharge ultra-high frequency charged detection method as claimed in claim 1, is characterized in that: The pulse waveform time-frequency analysis in the step S3 comprises the following steps: 1) The following formula (1) is used to perform Gobor transformation on the UHF discharge pulse waveform z(t) to obtain the time-frequency analysis spectrum G _f (a, b, ω): ${\mathrm{<span\; class="notranslate">\; G</span>}}_{\mathrm{<span\; class="notranslate">\; f</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; a</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; b</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; \&omega;</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>{<span\; class="notranslate">\; \&Integral;</span>}_{<span\; class="notranslate">\; -</span><span\; class="notranslate">\; \&infin;</span>}^{<span\; class="notranslate">\; \&infin;</span>}\mathrm{<span\; class="notranslate">\; z</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; g</span>}}_{\mathrm{<span\; class="notranslate">\; a</span>}}^{<span\; class="notranslate">\; *</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; b</span>}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; e</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; j\&omega;t</span>}}\mathrm{<span\; class="notranslate">\; dt</span>}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>$ in: is the conjugate function of the Gaussian window function g _a (tb); t is time, ω is the angular frequency, f is the frequency, b is the window translation parameter, and a is the window scale adjustment parameter; 2) Extracting the characteristic parameters of the time-frequency analysis spectrum, the characteristic parameters include: the time t _m corresponding to the maximum amplitude point of G _f (a, b, ω), the maximum amplitude of G _f (a, b, ω) The frequency f _m corresponding to the value point, the difference Δt _m between the start time of the discharge pulse and the time at the maximum amplitude. 7. a kind of GIS partial discharge ultra-high frequency charged detection method as claimed in claim 1, is characterized in that: In the step S3, the partial discharge detector uses the BP artificial neural network to train the characteristic parameters of the extracted PRPD discharge spectrum and the characteristic parameters of the time-frequency analysis spectrum, and output the corresponding GIS according to the training results The discharge type of partial discharge. 8. a kind of GIS partial discharge ultra-high frequency charged detection method as claimed in claim 7, is characterized in that: The BP neural network includes an input layer, an intermediate layer and an output layer; the input layer includes 18 input corresponding to 15 characteristic parameters of the PRPD discharge spectrum and 3 characteristic parameters of the time-frequency analysis spectrum Layer neurons; the middle layer has built-in five typical discharge sample data; the output layer includes five output layer neurons corresponding to the five typical discharge sample data of the middle layer. 9. a kind of GIS partial discharge ultra-high frequency charged detection method as claimed in claim 8, is characterized in that: The five typical discharge sample data of the intermediate layer include noise sample data, needle point discharge sample data, creeping discharge sample data, air gap discharge sample data and suspension discharge sample data; The characteristic parameters of the PRPD discharge spectrum and the characteristic parameters of the time-frequency analysis spectrum are input to the input layer of the BP neural network. The discharge sample data is trained to obtain the discharge type corresponding to the characteristic parameters of the PRPD discharge spectrum and the characteristic parameters of the time-frequency analysis spectrum, and output through the output layer of the BP neural network. 10. a kind of GIS partial discharge ultra-high frequency charged detection method as claimed in claim 9, is characterized in that: The discharge types include noise, needle point discharge, creeping discharge, air gap discharge and levitation discharge.