CN1924595A - Virtual instrument technique based gas insulation combined electric appliances online detecting method - Google Patents

Abstract

This invention relates to gas insulation combination apparatus online monitor method based on virtual device in the device insulation monitor technique field. This invention uses the computer and gas insulation combination apparatus local online test and positioning device and through virtual device technique and comprises parameter setting, online monitor, data analysis, history research four units module computer program and all-day GIS local discharging for online monitoring with multiple data analysis, warning and fault finding and requiring function.

Description

Gas insulation combined electric appliances online detecting method based on virtual instrument technique

Technical field

The invention belongs to the insulation of electrical installation monitoring technical field, relate to local discharge of gas-insulator switchgear ultrahigh frequency on-line monitoring method, particularly based on the local discharge of gas-insulator switchgear ultrahigh frequency on-line monitoring method of virtual instrument technique.

Background technology

Gas insulated combined electrical equipment in the electrical equipment (GIS) is in operation, because the existence of the various defectives of GIS inside causes GIS to cause various forms of shelf depreciations, as corona, impurity discharge etc.The inner shelf depreciation of GIS continues for some time and can develop into Fault of Insulating Breakdown, causes power outage, influences the safe operation of electric system, brings about great losses to national economy.The shelf depreciation of the early stage insulation fault of GIS has the characteristic of ultrahigh frequency, and its equivalent frequency is very high, can reach 1GHz.Because GIS is good waveguiding structure, the ultra-high frequency signal of shelf depreciation generation can be propagated therein thus, and leak in the ring flange junction by its insulator, so at the inner local discharge signal that detects local discharge signal or leak of GIS in its external detection, all can in time find the inherent vice of GIS, take appropriate measures, can avoid the generation of electrical equipment malfunction, guarantee the safe operation of electric system.

As Chinese patent application number is 200310118942.1 disclosed " gas insulated transformer substation high-frequency wideband partial discharge monitoring methods ", its local discharge detection device is the enclosure that a bottom surface has the slit, being packaged with frequency band range in the box is the broadband detection antenna of 10～3000MHz, and antenna ends is useful on the SMA high-frequency coaxial connector of feed; The one end is connected with 50 ohm matched impedance.But because the frequency of corona discharge is usually at 100-200MHz in the air, this sensor low-frequency range is low excessively, can not overcome to detect on-the-spot strong corona interference, and influence easily is interfered.Its diagnosis unit application wavelet packet and fractal dimension come the local feature of quantitative test local discharge signal, realization is to the quantitative analysis of local discharge signal waveform character, but owing to always have various noises at the on-line monitoring scene, the existence of noise has great influence to fractal dimension, causes calculating inaccurate.

And for example the application number of the applicant's application is the localization method of 200610054229.9 disclosed " online detection and location device of local discharge of gas-insulator switchgear and localization methods ", though can accurately locate to Partial Discharge Sources, but it can not make accurate judgement to the fault type of shelf depreciation, can not make assessment to the development trend of shelf depreciation, be difficult to more references to the maintainer.

Summary of the invention

The objective of the invention is to weak point at existing GIS Partial Discharge Detection, a kind of combination of gases electrical equipment on-line monitoring method based on virtual instrument technique is provided, have characteristics such as efficient, that use is flexible, powerful, realized the virtual instrument of " software is exactly instrument ", in conjunction with digital processing technologies such as modern signal analyzing and processing and databases, help the GIS staff of transformer station to carry out the judgement of GIS discharge fault type, repair and maintenance decision-making, be convenient to those skilled in the art and easily use.

The application number that the present invention utilizes the applicant to apply for is " online detection and location device of local discharge of gas-insulator switchgear and localization method " disclosed device of 200610054229.9, based on virtual instrument technique, by LabVIEW software platform and high-speed figure collector, adopt the mode bus that is fit to, utilize virtual control to control bottom hardware, the collection, analyzing and processing and the result that are finished signal by powerful software show, have realized functions such as on-line monitoring, data analysis, data in real time preservation, historical query, fault alarm, demonstration.And use special programming software to save a lot of man power and materials' input, shorten product development cycle, develop the product of superior performance.

The object of the present invention is achieved like this: a kind of combination of gases electrical equipment on-line monitoring method based on virtual instrument technique, the application number that utilizes computing machine and the applicant application is the sensor array that the micro-strip paster antenna of locating device in 200610054229.9 " online detection and location device of local discharge of gas-insulator switchgear and localization method " is formed, the multi-way intelligence selector switch, amplifilter, high-speed figure collector and control module, carry out partial discharge monitoring by the GIS on-line monitoring computer program based on virtual instrument technique, this routine package containing parameter is provided with, on-line monitoring, data analysis, four unit modules of historical query.Its concrete method step is as follows:

(1) parameter setting

At first enter the parameter setting: mainly carry out sample frequency, sampling length, regularly sampling parameter setting such as length, the activation threshold value setting, the alarm threshold value setting, monitoring mode (comprise manually and automatically) is provided with, and filtering parameter such as is provided with at the software initial setting up.

(2) systemic-function is selected

After (1) step, the parameter setting was finished, enter the function selecting unit, and automatic selection function 1, enter the on-line monitoring module; When data acquisition finished, selection function 2 entered data analysis module automatically; When operating personnel patrolled and examined, selection function 3 entered the historical query module; In the time of need stopping to monitor, operating personnel's selection function 4, system withdraws from, EOP (end of program).

1) when selecting 1, enters the on-line monitoring module, carry out on-line monitoring.The on-line monitoring unit at first according to the sampling parameter setting in the parameter set unit, carries out initialization to the high-speed figure collector.Carry out monitoring mode then and judge, when monitoring mode was automated manner, the local discharge signal to the sensor acquisition at the tested disc insulator in full station place carried out circulatory monitoring automatically.When monitoring mode is manual mode, select the sensor signal at some tested disc insulators place to concentrate monitoring, manual mode is used to needing when local discharge signal surpasses early warning value to occur the GIS interval of primary part observation.When monitoring mode is automated manner: according to sampling timing the interval is set and starts timer, when interval time then, computing machine is at first by total line traffic control multi-way intelligence switch, micro-strip paster antenna in the gating sensor array in turn, start the high-speed figure collector simultaneously, conversion of signals with each micro-strip paster antenna is a digital signal successively, be sent to by bus again and carry out data processing in the computing machine, and data are shown and stores processor, patrol and examine and return wait when finishing and regularly arrive next time.When monitoring mode was manual mode: computing machine was at first by certain micro-strip paster antenna in total line traffic control multi-way intelligence switching gate sensor array, start the high-speed figure collector simultaneously, conversion of signals with this micro-strip paster antenna is a digital signal successively, be sent to by bus again and carry out data processing in the computing machine, and data are shown and stores processor.Memory module has realized preserving automatically for the real-time raw data of the shelf depreciation of measuring, and database adopts general relevant database, and memory contents comprises local discharge signal waveform and device number, acquisition time and maximum pd quantity etc.After on-line monitoring is finished, withdraw from judgement, when withdrawing from, returned for (2) step and carry out function selecting; When not withdrawing from, selection function 2 enters data analysis module automatically.

2) when selecting 2, enter data analysis module, carry out data analysis.Data analysis unit is used for multiple small echo denoising and the spectrum analysis to local discharge data, and fault type is carried out pattern-recognition and Partial Discharge Sources location.

1. small echo denoising again: the application number that utilizes the inventor to invent is the multiple Wavelet noise-eliminating method in 200510057243.x " online detector for partial discharge of gas-insulated substation and denoising method ", adopt multiple wavelet technique that the sensor signal of gathering is carried out denoising, undesired signals such as the on-the-spot white noise that may exist, arrowband are detected in the place to go.According to signal amplitude after the multiple small echo denoising, when greater than alarm threshold value, then carry out sound and light alarm.

2. spectrum analysis: the change procedure that joint time frequency analysis can careful portrayal discharge signal be taken place on time-frequency plane, local discharge signal is carried out spectrum analysis, can reflect the time-frequency characteristic of local discharge signal, be provided with the back and further analyze.

3. pattern-recognition: pattern-recognition is carried out fault diagnosis to the GIS shelf depreciation, so that accurately grasp the defect property and the guide maintenance of GIS inside.Pattern-recognition at first to carry out feature extraction, then the feature of extracting is carried out pattern-recognition by neural network.The concrete steps of its feature extraction are as follows:

1〉PD Signal Pretreatment: the local discharge signal sample is carried out multiple small echo denoising, and with the signal normalization after the denoising;

2〉the multiple wavelet decomposition of processing back signal: pretreated signal is carried out multiple wavelet transformation, and multiple wavelet decomposition number of plies n is the 3-5 layer, obtains the multiple wavelet coefficient of each yardstick;

3〉each scale coefficient fuzzy clustering: to the Fuzzy C-Means Clustering of carrying out of each scale coefficient (real part R, imaginary part I and the composite information R|I| of multiple wavelet coefficient), cluster numbers c is 2-4.

C={c _{I, 1}, c _{I, 2}..., c _{I, n}For waiting the i yardstick wavelet coefficient of classifying; V={v ₁, v ₂..., v _cBe c cluster centre, v _p∈ R ^PBe subordinate to matrix U=[u _{P, k}] _{N * c}, and satisfy u _{P, k}∈ [0,1], $\underset{p=1}{\overset{c}{\mathrm{\Σ}}}{u}_{p,k}=1,$ Given preliminary classification matrix U _{C * n} ^(l), wherein l is an iterations.

The cluster centre vector of preliminary classification then:

$V_p^{\left(l\right)}=\frac{\underset{k=1}{\overset{n}{\mathrm{\Σ}}}\left[{\left(u_{p,k}^{\left(l\right)}\right)}^m{C}_k\right]}{\underset{k=1}{\overset{n}{\mathrm{\Σ}}}{\left(u_{p,k}^{\left(l\right)}\right)}^m}---\left(1\right)$

Renewal is subordinate to matrix:

$u_{p,k}^{(l+1)}=\frac{1}{\underset{j=1}{\overset{c}{\mathrm{\Σ}}}{\left(\frac{\left|\right|C_k-V_p^{\left(l\right)}\left|\right|}{\left|\right|C_k-V_j^{\left(l\right)}\left|\right|}\right)}^{\frac{2}{m-1}}}---\left(2\right)$

Objective function J _bBe defined as:

$J_b(U,V)=\underset{k=1}{\overset{n}{\mathrm{\Σ}}}\underset{p=1}{\overset{c}{\mathrm{\Σ}}}{u}_{\mathrm{pk}}{\left|\right|c_{l,k}-v_p\left|\right|}^2---\left(3\right)$

With formula (1-3) repeated calculation cluster centre, classification matrix (being the degree of membership matrix), up to objective function J _bReach minimum.

4〉form each scale feature amount: as characteristic quantity, each shelf depreciation sample generates n * c characteristic quantity with the energy of the wavelet coefficient of each cluster under each yardstick.

$f_{i,k}=\sqrt{d_{i,l}^2+\·\·\·+d_{i,m}^2},k=\mathrm{1,2},\·\·\·,n---\left(4\right)$

F wherein _{I, k}Be the characteristic quantity of k cluster under the i yardstick, d _{I, l}..., d _{I, m}Be all wavelet coefficients in k cluster under the i yardstick.

When carrying out pattern-recognition, at first dissimilar shelf depreciation waveform training samples is formed the characteristic quantity storehouse by above-mentioned characteristic extraction step, and neural network training; The waveform sample that the shelf depreciation of on-line monitoring collection in worksite need be classified forms characteristic quantity by above-mentioned characteristic extraction step then, and input neural network carries out the identification of Partial Discharge Sources defect type, recognition result confession technician reference.

4. shelf depreciation source location: the application number that utilizes the inventor to invent carries out the shelf depreciation source location for the Partial Discharge Sources localization method in 200510057243.x " online detection and location device of local discharge of gas-insulator switchgear and localization method ".

In (2)-2) after data analysis finishes, withdraw from judgement, when withdrawing from, returned for (2) step and carry out function selecting; When not withdrawing from, selection function 1 enters the on-line monitoring module automatically.

3) when selecting 3, enter the historical query module, carry out historical query.For making things convenient for the operation conditions of power equipment in staff of transformer station and the upper management personnel long-term observation station, monitoring system provides query function.At first inquire about, generate historical data report and historical discharge signal amplitude changing trend diagram at historical shelf depreciation data, and formulation equipment year, the moon and day trend curve inquiry.Because the signal amplitude that uhf sensor is measured is closely related with travel path, so demarcate very big difficulty with discharge capacity PC.The inventor adopts the signal amplitude of measuring UHF local discharge signal waveform by on-line continuous, the historical variations trend of observation signal amplitude, judge the order of severity of shelf depreciation, when signal amplitude is changed significantly, expression built-in electrical insulation degradation trend is remarkable, report to the police to the monitoring personnel with acousto-optic, the technician should in time handle.After historical query is finished, withdraw from judgement, when withdrawing from, returned for (2) step and carry out function selecting; When not withdrawing from, proceed historical query.

4) when selecting 4, system withdraws from, EOP (end of program).

The present invention adopts technique scheme, based on graphical programming language LabView and Database Systems, realized a kind of based on virtual instrument technique GIS on-line monitoring computer program, the variation of can be round-the-clock automatic monitoring GIS transformer station multi-point partial discharge signal, carry out the several data analysis, and have early warning, fault diagnosis and a positioning function, multiple query function.Show that through trial run this software has stable, friendly interface is easy to operate, the characteristics that function is complete.The present invention can be widely used in the GIS partial discharge monitoring of generating plant, transformer station.

Description of drawings

Fig. 1 is the theory diagram of the hardware unit of the present invention's utilization;

Fig. 2 is the program flow chart of the inventive method;

Fig. 3 is the program flow diagram of the on-line monitoring of present embodiment;

Fig. 4 is the program flow diagram that the shelf depreciation waveform character of present embodiment extracts;

Fig. 5 is the program flow diagram of the historical query of present embodiment.

Embodiment

Below in conjunction with embodiment, further specify the present invention.

Shown in Fig. 1～5, a kind of gas insulation combined electric appliances online detecting method based on virtual instrument technique, the application number that utilizes computing machine and the applicant application is the device of 200610054229.9 " online detection and location device of local discharge of gas-insulator switchgear and localization method ", carry out the on-line monitoring of GIS shelf depreciation by program, concrete method step is as follows:

(1) parameter setting

At first enter the parameter setting: mainly carry out sample frequency, sampling length, regularly sampling parameter setting such as length, the activation threshold value setting, the alarm threshold value setting, monitoring mode (comprise manually and automatically) is provided with, and filtering parameter such as is provided with at the software initial setting up.

(2) systemic-function is selected

After (1) step, the parameter setting was finished, enter the function selecting unit, and automatic selection function 1, enter the on-line monitoring module; When data acquisition finished, selection function 2 entered data analysis module automatically; When operating personnel patrolled and examined, selection function 3 entered and goes through official's enquiry module; In the time of need stopping to monitor, operating personnel's selection function 4, system withdraws from, EOP (end of program).

1) when selecting 1, enters the on-line monitoring module, carry out on-line monitoring according to program circuit shown in Figure 3.The on-line monitoring unit at first according to the sampling parameter setting in the parameter set unit, carries out initialization to the high-speed figure collector.Carrying out monitoring mode then judges, when monitoring mode is automated manner: according to sampling timing the interval is set and starts timer, when interval time then, computing machine is at first by total line traffic control multi-way intelligence switch, first micro-strip paster antenna in the gating sensor array, start the high-speed figure collector simultaneously, conversion of signals with each micro-strip paster antenna is a digital signal successively, be sent to by bus again and carry out data processing in the computing machine, and data are shown and stores processor, this back that disposes judges once to patrol and examine whether finish, when patrolling and examining when not finishing, next micro-strip paster antenna in the gating sensor array carries out signals collecting, waveform shows and data storage processing; When patrolling and examining end, return and wait for that fixed time interval arrives next time.When monitoring mode was manual mode: computing machine was at first by certain micro-strip paster antenna in total line traffic control multi-way intelligence switching gate sensor array, start the high-speed figure collector simultaneously, conversion of signals with this micro-strip paster antenna is a digital signal successively, be sent to by bus again and carry out data processing in the computing machine, and data are shown and stores processor.Memory module has realized preserving automatically for the real-time raw data of the shelf depreciation of measuring, and database adopts general relevant database, and memory contents comprises local discharge signal waveform and device number, acquisition time and maximum pd quantity etc.After on-line monitoring is finished, withdraw from judgement, when withdrawing from, returned for (2) step and carry out function selecting; When not withdrawing from, selection function 2 enters data analysis module automatically.

2) when selecting 2, enter data analysis module, carry out data analysis.Data analysis unit is used for multiple small echo denoising and the spectrum analysis to local discharge data, and fault type is carried out pattern-recognition and Partial Discharge Sources location.

1. small echo denoising again: the application number that utilizes the inventor to invent is the multiple Wavelet noise-eliminating method in 200510057243.x " online detector for partial discharge of gas-insulated substation and denoising method ", adopt multiple wavelet technique that the sensor signal of gathering is carried out denoising, undesired signals such as the on-the-spot white noise that may exist, arrowband are detected in the place to go.According to signal amplitude after the multiple small echo denoising, when greater than alarm threshold value, then carry out sound and light alarm.

2. spectrum analysis: the change procedure that joint time frequency analysis can careful portrayal discharge signal be taken place on time-frequency plane, local discharge signal is carried out spectrum analysis, can reflect the time-frequency characteristic of local discharge signal, be provided with the back and further analyze.

3. pattern-recognition: pattern-recognition at first to carry out feature extraction, then the feature of extracting is carried out pattern-recognition by neural network.Wherein the concrete steps of feature extraction are as shown in Figure 4:

1〉PD Signal Pretreatment: the local discharge signal sample is carried out multiple small echo denoising, and with the signal normalization after the denoising;

2〉the multiple wavelet decomposition of processing back signal: pretreated signal is carried out multiple wavelet transformation, and multiple wavelet decomposition number of plies n is 5 layers, obtains the multiple wavelet coefficient of each yardstick;

3〉each scale coefficient fuzzy clustering: to the Fuzzy C-Means Clustering of carrying out of each scale coefficient (real part R, imaginary part I and the composite information R|I| of multiple wavelet coefficient), cluster numbers c is 3.

C={c _{I, 1}, c _{I, 2}..., c _{I, n}For waiting the i yardstick wavelet coefficient of classifying; V={v ₁, v ₂..., v _cBe c cluster centre, v _p∈ R ^PBe subordinate to matrix U=[u _{P, k}] _{N * c}, and satisfy u _{P, k}∈ [0,1], $\underset{p=1}{\overset{c}{\mathrm{\Σ}}}{u}_{p,k}=1,$ Given preliminary classification matrix U _{C * n} ^(l), wherein l is an iterations.

The cluster centre vector of preliminary classification then:

$V_p^{\left(l\right)}=\frac{\underset{k=1}{\overset{n}{\mathrm{\Σ}}}\left[{\left(u_{p,k}^{\left(l\right)}\right)}^m{C}_k\right]}{\underset{k=1}{\overset{n}{\mathrm{\Σ}}}{\left(u_{p,k}^{\left(l\right)}\right)}^m}---\left(1\right)$

Renewal is subordinate to matrix:

$u_{p,k}^{(l+1)}=\frac{1}{\underset{j=1}{\overset{c}{\mathrm{\Σ}}}{\left(\frac{\left|\right|C_k-V_p^{\left(l\right)}\left|\right|}{\left|\right|C_k-V_j^{\left(l\right)}\left|\right|}\right)}^{\frac{2}{m-1}}}---\left(2\right)$

Objective function J _bBe defined as:

$J_b(U,V)=\underset{k=1}{\overset{n}{\mathrm{\Σ}}}\underset{p=1}{\overset{c}{\mathrm{\Σ}}}{u}_{\mathrm{pk}}{\left|\right|c_{l,k}-v_p\left|\right|}^2---\left(3\right)$

With formula (1-3) repeated calculation cluster centre, classification matrix (being the degree of membership matrix), up to objective function J _bReach minimum.

4〉form each scale feature amount: as characteristic quantity, each shelf depreciation sample generates 15 characteristic quantities with the energy of the wavelet coefficient of each cluster under each yardstick.

$f_{i,k}=\sqrt{d_{i,l}^2+\·\·\·+d_{i,m}^2},k=\mathrm{1,2},3---\left(4\right)$

F wherein _{I, k}Be the characteristic quantity of k cluster under the i yardstick, d _{I, l}, d _{I, m}Be all wavelet coefficients in k cluster under the i yardstick.

When carrying out pattern-recognition, at first dissimilar shelf depreciation waveform training samples is formed the characteristic quantity storehouse by above-mentioned characteristic extraction step, and neural network training; The waveform sample that the shelf depreciation of on-line monitoring collection in worksite need be classified forms characteristic quantity by above-mentioned characteristic extraction step then, and input neural network carries out the identification of Partial Discharge Sources defect type, recognition result confession technician reference.

4. shelf depreciation source location: the application number that utilizes the inventor to invent carries out the shelf depreciation source location for the Partial Discharge Sources localization method in 200510057243.x " online detection and location device of local discharge of gas-insulator switchgear and localization method ".

In (2)-2) after data analysis finishes, withdraw from judgement, when withdrawing from, returned for (2) step and carry out function selecting; When not withdrawing from, selection function 1 enters the on-line monitoring module automatically.

3) when selecting 3, enter the historical query module, carry out historical query according to program circuit shown in Figure 5, its concrete steps are as follows:

At first select year, the moon and day date of historical query, and need to select the sensor corresponding equipment number in the sensor array of inquiry, the judgement of whether inquiring about then, when inquiry, the historical shelf depreciation data query that carries out this equipment is handled and is shown; When not inquiring about, return and re-enter the Query Dates selection.Displaying contents comprises historical shelf depreciation data, historical data report and historical discharge signal amplitude changing trend diagram.Operating personnel judge the order of severity of shelf depreciation by observing the historical variations trend of discharge amplitude, and when the discharge amplitude was changed significantly, expression built-in electrical insulation degradation trend was remarkable, reports to the police to the monitoring personnel with acousto-optic, and the technician should in time handle.After historical query is finished, withdraw from judgement, when withdrawing from, returned for (2) step and carry out function selecting; When not withdrawing from, proceed historical query.

4) when selecting 4, system withdraws from, EOP (end of program).

Claims (2)

1, a kind of combination of gases electrical equipment on-line monitoring method based on virtual instrument technique, the application number that utilizes computing machine and the applicant application is " online detection and location device of local discharge of gas-insulator switchgear and localization method " middle locating device of 200610054229.9, carry out partial discharge monitoring by the GIS on-line monitoring computer program based on virtual instrument technique, its concrete method step is as follows:

(1) parameter setting

At first enter the parameter setting: mainly carry out sample frequency, sampling length, regularly sampling parameter setting such as length, the activation threshold value setting, the alarm threshold value setting, monitoring mode (comprise manually and automatically) is provided with, and filtering parameter such as is provided with at the software initial setting up;

(2) systemic-function is selected

After (1) step, the parameter setting was finished, enter the function selecting unit, and automatic selection function 1, enter the on-line monitoring module; When data acquisition finished, selection function 2 entered data analysis module automatically; When operating personnel patrolled and examined, selection function 3 entered the historical query module; In the time of need stopping to monitor, operating personnel's selection function 4, system withdraws from, EOP (end of program);

1) when selecting 1, enter the on-line monitoring module, carry out on-line monitoring, the on-line monitoring unit is at first according to the sampling parameter setting in the parameter set unit, the high-speed figure collector is carried out initialization, carrying out monitoring mode then judges, when monitoring mode is automated manner, automatically the local discharge signal to the sensor acquisition at the tested disc insulator in full station place carries out circulatory monitoring, when monitoring mode is manual mode, select the sensor signal at some tested disc insulators place to concentrate monitoring, manual mode is used to needing when local discharge signal surpasses early warning value to occur the GIS interval of primary part observation, when monitoring mode is automated manner: according to sampling timing the interval is set and starts timer, when interval time then, computing machine is at first by total line traffic control multi-way intelligence switch, micro-strip paster antenna in the gating sensor array in turn, start the high-speed figure collector simultaneously, conversion of signals with each micro-strip paster antenna is a digital signal successively, be sent to by bus again and carry out data processing in the computing machine, and data are shown and stores processor, patrol and examine and return wait regularly arrival next time when finishing, when monitoring mode was manual mode: computing machine was at first by certain micro-strip paster antenna in total line traffic control multi-way intelligence switching gate sensor array, start the high-speed figure collector simultaneously, conversion of signals with this micro-strip paster antenna is a digital signal successively, be sent to by bus again and carry out data processing in the computing machine, and data are shown and stores processor, memory module has realized preserving automatically for the real-time raw data of the shelf depreciation of measuring, database adopts general relevant database, memory contents comprises local discharge signal waveform and device number, acquisition time and maximum pd quantity etc., after on-line monitoring is finished, withdraw from judgement, when withdrawing from, returned for (2) step and carry out function selecting; When not withdrawing from, selection function 2 enters data analysis module automatically;

2) when selecting 2, enter data analysis module, carry out data analysis, data analysis unit is used for multiple small echo denoising and the spectrum analysis to local discharge data, and fault type is carried out pattern-recognition and Partial Discharge Sources location;

1. small echo denoising again: the application number that utilizes the inventor to invent is the multiple Wavelet noise-eliminating method in 200510057243.x " online detector for partial discharge of gas-insulated substation and denoising method ", adopt multiple wavelet technique that the sensor signal of gathering is carried out denoising, undesired signals such as the on-the-spot white noise that may exist, arrowband are detected in the place to go, according to signal amplitude after the multiple small echo denoising, when greater than alarm threshold value, then carry out sound and light alarm;

2. spectrum analysis: the change procedure that joint time frequency analysis can careful portrayal discharge signal be taken place on time-frequency plane, local discharge signal is carried out spectrum analysis, can reflect the time-frequency characteristic of local discharge signal, be provided with the back and further analyze;

3. pattern-recognition: pattern-recognition is carried out fault diagnosis to the GIS shelf depreciation, so that accurately grasp the defect property and the guide maintenance of GIS inside, pattern-recognition at first to carry out feature extraction, then the feature of extracting is carried out pattern-recognition by neural network, the concrete steps of its feature extraction are as follows:

1〉PD Signal Pretreatment: the local discharge signal sample is carried out multiple small echo denoising, and with the signal normalization after the denoising;

2〉the multiple wavelet decomposition of processing back signal: pretreated signal is carried out multiple wavelet transformation, and multiple wavelet decomposition number of plies n is the 3-5 layer, obtains the multiple wavelet coefficient of each yardstick;

3〉each scale coefficient fuzzy clustering: to the Fuzzy C-Means Clustering of carrying out of each scale coefficient (real part R, imaginary part I and the composite information R|I| of multiple wavelet coefficient), cluster numbers c is 2-4;

C={c _{I, 1}, c _{I, 2}..., c _{I, n}For waiting the i yardstick wavelet coefficient of classifying; V={v ₁, v ₂..., v _cBe c cluster centre, v _p∈ R ^PBe subordinate to matrix U=[u _{P, k}] _{N * c}, and satisfy u _{P, k}∈ [0,1], $\underset{p=1}{\overset{c}{\mathrm{\Σ}}}{u}_{p,k}=1,$ Given preliminary classification matrix U _{C * n} ^(l), wherein l is an iterations;

The cluster centre vector of preliminary classification then:

$V_p^{\left(l\right)}=\frac{\underset{k=1}{\overset{n}{\mathrm{\Σ}}}\left[{\left(u_{p,k}^{\left(l\right)}\right)}^m{C}_k\right]}{\underset{k=1}{\overset{n}{\mathrm{\Σ}}}{\left(u_{p,k}^{\left(l\right)}\right)}^m}---\left(1\right)$

Renewal is subordinate to matrix:

$u_{p,k}^{(l+1)}=\frac{1}{\underset{j=1}{\overset{c}{\mathrm{\Σ}}}{\left(\frac{\left|\right|C_k-V_p^{\left(l\right)}\left|\right|}{\left|\right|C_k-V_j^{\left(l\right)}\left|\right|}\right)}^{\frac{2}{m-1}}}---\left(2\right)$

Objective function J _bBe defined as:

$J_b(U,V)=\underset{k=1}{\overset{n}{\mathrm{\Σ}}}\underset{p=1}{\overset{c}{\mathrm{\Σ}}}{u}_{\mathrm{pk}}{\left|\right|c_{i,k}-v_p\left|\right|}^2---\left(3\right)$

With formula (1-3) repeated calculation cluster centre, classification matrix (being the degree of membership matrix), up to objective function J _bReach minimum;

4〉form each scale feature amount: as characteristic quantity, each shelf depreciation sample generates n * c characteristic quantity with the energy of the wavelet coefficient of each cluster under each yardstick;

$f_{i,k}=\sqrt{d_{i,l}^2+\·\·\·+d_{i,m}^2},k=\mathrm{1,2},\·\·\·,n---\left(4\right)$

F wherein _{I, k}Be the characteristic quantity of k cluster under the i yardstick, d _{I, l}..., d _{I, m}Be all wavelet coefficients in k cluster under the i yardstick;

When carrying out pattern-recognition, at first dissimilar shelf depreciation waveform training samples is formed the characteristic quantity storehouse by above-mentioned characteristic extraction step, and neural network training; The waveform sample that the shelf depreciation of on-line monitoring collection in worksite need be classified forms characteristic quantity by above-mentioned characteristic extraction step then, and input neural network carries out the identification of Partial Discharge Sources defect type, recognition result confession technician reference;

4. shelf depreciation source location: the application number that utilizes the inventor to invent carries out the shelf depreciation source location for the Partial Discharge Sources localization method in 200510057243.x " online detection and location device of local discharge of gas-insulator switchgear and localization method ";

In (2)-2) after data analysis finishes, withdraw from judgement, when withdrawing from, returned for (2) step and carry out function selecting; When not withdrawing from, selection function 1 enters the on-line monitoring module automatically;

3) when selecting 3, enter the historical query module, carry out historical query, for making things convenient for the operation conditions of power equipment in staff of transformer station and the upper management personnel long-term observation station, monitoring system provides query function, at first inquire about at historical shelf depreciation data, generate historical data report and historical discharge signal amplitude changing trend diagram, and formulation equipment year, the moon and day trend curve inquiry, because the signal amplitude that uhf sensor is measured is closely related with travel path, so demarcate very big difficulty with discharge capacity PC, the inventor adopts the signal amplitude of measuring UHF local discharge signal waveform by on-line continuous, the historical variations trend of observation signal amplitude, judge the order of severity of shelf depreciation, when signal amplitude was changed significantly, expression built-in electrical insulation degradation trend was remarkable, report to the police to the monitoring personnel with acousto-optic, the technician should in time handle, and after historical query is finished, withdraws from judgement, when withdrawing from, returned for (2) step and carry out function selecting; When not withdrawing from, proceed historical query;

4) when selecting 4, system withdraws from, EOP (end of program).

2, according to the described gas insulation combined electric appliances online detecting method of claim 1 based on virtual instrument technique, it is characterized in that, a kind of gas insulation combined electric appliances online detecting method based on virtual instrument technique, the application number that utilizes computing machine and the applicant application is the device of 200610054229.9 " online detection and location device of local discharge of gas-insulator switchgear and localization method ", carry out the on-line monitoring of GIS shelf depreciation by program, concrete method step is as follows:

(1) parameter setting

At first enter the parameter setting: mainly carry out sample frequency, sampling length, regularly sampling parameter setting such as length, the activation threshold value setting, the alarm threshold value setting, monitoring mode (comprise manually and automatically) is provided with, and filtering parameter such as is provided with at the software initial setting up;

(2) systemic-function is selected

After (1) step, the parameter setting was finished, enter the function selecting unit, and automatic selection function 1, enter the on-line monitoring module; When data acquisition finished, selection function 2 entered data analysis module automatically; When operating personnel patrolled and examined, selection function 3 entered the historical query module; In the time of need stopping to monitor, operating personnel's selection function 4, system withdraws from, EOP (end of program);

1) when selecting 1, enter the on-line monitoring module, carry out on-line monitoring according to program circuit shown in Figure 3, the on-line monitoring unit is at first according to the sampling parameter setting in the parameter set unit, the high-speed figure collector is carried out initialization, carrying out monitoring mode then judges, when monitoring mode is automated manner: according to sampling timing the interval is set and starts timer, when interval time then, computing machine is at first by total line traffic control multi-way intelligence switch, first micro-strip paster antenna in the gating sensor array, start the high-speed figure collector simultaneously, conversion of signals with each micro-strip paster antenna is a digital signal successively, is sent to by bus and carries out data processing in the computing machine, and data are shown and stores processor, this back that disposes judges once to patrol and examine whether finish, when patrolling and examining when not finishing, next micro-strip paster antenna in the gating sensor array carries out signals collecting, waveform shows and data storage processing; When patrolling and examining end, return and wait for that fixed time interval arrives next time, when monitoring mode was manual mode: computing machine was at first by certain micro-strip paster antenna in total line traffic control multi-way intelligence switching gate sensor array, start the high-speed figure collector simultaneously, conversion of signals with this micro-strip paster antenna is a digital signal successively, be sent to by bus again and carry out data processing in the computing machine, and data are shown and stores processor, memory module has realized preserving automatically for the real-time raw data of the shelf depreciation of measuring, database adopts general relevant database, memory contents comprises local discharge signal waveform and device number, acquisition time and maximum pd quantity etc. after on-line monitoring is finished, withdraw from judgement, when withdrawing from, returned for (2) step and carry out function selecting; When not withdrawing from, selection function 2 enters data analysis module automatically;

2) when selecting 2, enter data analysis module, carry out data analysis, data analysis unit is used for multiple small echo denoising and the spectrum analysis to local discharge data, and fault type is carried out pattern-recognition and Partial Discharge Sources location;

1. small echo denoising again: the application number that utilizes the inventor to invent is the multiple Wavelet noise-eliminating method in 200510057243.x " online detector for partial discharge of gas-insulated substation and denoising method ", adopt multiple wavelet technique that the sensor signal of gathering is carried out denoising, undesired signals such as the on-the-spot white noise that may exist, arrowband are detected in the place to go, according to signal amplitude after the multiple small echo denoising, when greater than alarm threshold value, then carry out sound and light alarm;

2. spectrum analysis: the change procedure that joint time frequency analysis can careful portrayal discharge signal be taken place on time-frequency plane, local discharge signal is carried out spectrum analysis, can reflect the time-frequency characteristic of local discharge signal, be provided with the back and further analyze;

3. pattern-recognition: pattern-recognition at first to carry out feature extraction, then the feature of extracting is carried out pattern-recognition by neural network, wherein the concrete steps of feature extraction as shown in Figure 4:

1〉PD Signal Pretreatment: the local discharge signal sample is carried out multiple small echo denoising, and with the signal normalization after the denoising;

2〉the multiple wavelet decomposition of processing back signal: pretreated signal is carried out multiple wavelet transformation, and the multiple wavelet decomposition number of plies is that n is 5 layers, obtains the multiple wavelet coefficient of each yardstick;

3〉each scale coefficient fuzzy clustering: to the Fuzzy C-Means Clustering of carrying out of each scale coefficient (real part R, imaginary part I and the composite information R|I| of multiple wavelet coefficient), cluster numbers c is 3;

C={c _{I, 1}, c _{I, 2}..., c _{I, n}For waiting the i yardstick wavelet coefficient of classifying; V={v ₁, v ₂..., v _cBe c cluster centre, v _p∈ R ^PBe subordinate to matrix U=[u _{P, k}] _{N * c}, and satisfy u _{P, k}∈ [0,1], $\underset{p=1}{\overset{c}{\mathrm{\Σ}}}{u}_{p,k}=1,$ Given preliminary classification matrix U _{C * n} ^(l), wherein l is an iterations;

The cluster centre vector of preliminary classification then:

$V_p^{\left(l\right)}=\frac{\underset{k=1}{\overset{n}{\mathrm{\Σ}}}\left[{\left(u_{p,k}^{\left(l\right)}\right)}^m{C}_k\right]}{\underset{k=1}{\overset{n}{\mathrm{\Σ}}}{\left(u_{p,k}^{\left(l\right)}\right)}^m}---\left(1\right)$

Renewal is subordinate to matrix:

$u_{p,k}^{(l+1)}=\frac{1}{\underset{j=1}{\overset{c}{\mathrm{\Σ}}}{\left(\frac{\left|\right|C_k-V_p^{\left(l\right)}\left|\right|}{\left|\right|C_k-V_j^{\left(l\right)}\left|\right|}\right)}^{\frac{2}{m-1}}}---\left(2\right)$

Objective function J _bBe defined as:

$J_b(U,V)=\underset{k=1}{\overset{n}{\mathrm{\Σ}}}\underset{p=1}{\overset{c}{\mathrm{\Σ}}}{u}_{\mathrm{pk}}{\left|\right|c_{i,k}-v_p\left|\right|}^2---\left(3\right)$

With formula (1-3) repeated calculation cluster centre, classification matrix (being the degree of membership matrix), up to objective function J _bReach minimum,

4〉form each scale feature amount: as characteristic quantity, each shelf depreciation sample generates 15 characteristic quantities with the energy of the wavelet coefficient of each cluster under each yardstick;

$f_{i,k}=\sqrt{d_{i,l}^2+\·\·\·+d_{i,m}^2},k=\mathrm{1,2},\·\·\·,n---\left(4\right)$

F wherein _{I, k}Be the characteristic quantity of k cluster under the i yardstick, d _{I, l}..., d _{I, m}Be all wavelet coefficients in k cluster under the i yardstick;

When carrying out pattern-recognition, at first dissimilar shelf depreciation waveform training samples is formed the characteristic quantity storehouse by above-mentioned characteristic extraction step, and neural network training; The waveform sample that the shelf depreciation of on-line monitoring collection in worksite need be classified forms characteristic quantity by above-mentioned characteristic extraction step then, and input neural network carries out the identification of Partial Discharge Sources defect type, recognition result confession technician reference;

4. shelf depreciation source location: the application number that utilizes the inventor to invent carries out the shelf depreciation source location for the Partial Discharge Sources localization method in 200510057243.x " online detection and location device of local discharge of gas-insulator switchgear and localization method ";

In (2)-2) after data analysis finishes, withdraw from judgement, when withdrawing from, returned for (2) step and carry out function selecting; When not withdrawing from, selection function 1 enters the on-line monitoring module automatically;

3) when selecting 3, enter the historical query module, carry out historical query according to program circuit shown in Figure 5, its concrete steps are as follows:

At first select year, the moon and day date of historical query, and need to select the sensor corresponding equipment number in the sensor array of inquiry, the judgement of whether inquiring about then, when inquiry, the historical shelf depreciation data query that carries out this equipment is handled and is shown; When not inquiring about, return and re-enter the Query Dates selection, displaying contents comprises historical shelf depreciation data, historical data report and historical discharge signal amplitude changing trend diagram, operating personnel are by observing the historical variations trend of discharge amplitude, judge the order of severity of shelf depreciation, when the discharge amplitude is changed significantly, expression built-in electrical insulation degradation trend is remarkable, reports to the police to the monitoring personnel with acousto-optic, and the technician should in time handle, after historical query is finished, withdraw from judgement, when withdrawing from, returned for (2) step and carry out function selecting; When not withdrawing from, proceed historical query;

4) when selecting 4, system withdraws from, EOP (end of program).

Images