CN1484034A

CN1484034A - On-line intelligent monitoring system for transformer and intelligent analysis diagnosis method thereof

Info

Publication number: CN1484034A
Application number: CNA021396248A
Authority: CN
Inventors: 张冠军; 尼; 严璋; 欧阳南尼; 吴文强; 王英华; 郭长泰; 赵文彬; 魏铁军
Original assignee: TEBIAN ELECTROTECHNICAL CO Ltd XINJIANG
Current assignee: TEBIAN ELECTROTECHNICAL CO Ltd XINJIANG
Priority date: 2002-09-18
Filing date: 2002-09-18
Publication date: 2004-03-24

Abstract

The invention refers to a transformer on-line intelligent monitoring system and its intelligent diagnosis method. It includes all kinds of sensors for the transformer's dynamic capability monitoring device, some electric signals collected by the sensors connected with A/D converting circuit through the amplifier and regulating circuit, some ones connected with industrial computer through A/D collecting card or interface circuit to be processed by the foreground computer, then transmitted into the background CPU transformer fault diagnosis system for diagnosis and analysis, to obtain the monitoring results. The method: inputting various signal data into the computer, making the computer combined with the stored information database to carry out on-line real-time diagnosis and alarm on the abnormal condition.

Description

Transformer on-line intelligence monitoring system and intellectual analysis diagnostic method thereof

Technical field

The invention belongs to the transformer technology field, relate to a kind of transformer on-line intelligence monitoring system and intellectual analysis diagnostic method thereof.

Background technology

Maintenance in China to transformer from entirely passive type decades ago, waits and just keeps in repair after having an accident, and has developed in a planned way and has regularly keeped in repair, and this is very big progress.History is best witness, detects owing to adopt regularly, has found many defectives, has avoided many accidents, has obtained huge economy and social benefit.But along with the raising of place capacity, rated voltage, many contradictions are also more outstanding.For example, when regularly having a power failure test, still add 10kV and exchange survey tg δ, add 1 ~ 5kV direct current survey insulation resistance etc. 110kV and above power equipment.Because trial voltage is far below working voltage, sensitivity is very low, often only can find that some ubiquities are made moist or the existing significant deficiency that approaches the penetrability passage; And, wasted great amount of manpower and material resources, but be difficult to guarantee operational reliability no matter equipment state how, nearly all is regularly detection, periodic maintenance of clean cut ground.

Summary of the invention

Technical matters to be solved by this invention is that a kind of transformer on-line intelligence monitoring system and intellectual analysis diagnostic method thereof will be provided, the various dynamic properties that align the transformer that is in operation are monitored and analyzing and diagnosing in real time, in time reflect diagnostic result, can in time find out transformer fault, thereby prevent the generation of transformer fault and can in time get rid of transformer fault.

The technical scheme that solution the technology of the present invention problem is adopted is the sensor that this transformer on-line intelligence monitoring system comprises the various dynamic property monitoring devices of transformer, what the electric signal of being gathered by various sensor had inserts the modulus capture card after amplifier and modulate circuit processing, have directly through the modulus capture card, all inserting in the industrial computer behind interface circuit of having carried out data processing by the foreground computing machine, draw the performance data that records, thereafter this digital signal is transported to backstage analyzing and diagnosing computing machine, system carries out diagnostic analysis by the transformer fault real-time diagnosis, draws analysis result.This transformer on-line intelligence analysis and diagnosis method is that the various signal datas of transformer online monitoring are sent in the analyzing and diagnosing computing machine, store the equipment nameplate parameter information that monitored transformer is arranged in this analyzing and diagnosing computing machine, the true empirical rule information of transformer in the past, the technical parameter rules and the standard of on-line monitoring, the technical parameter threshold value and the operating experience of on-line monitoring, have in the analyzing and diagnosing computing machine can according to automatic input come in the various signal datas of transformer online monitoring carry out Dynamic Data Processing and trend analysis, the reasoning of state comprehensive diagnos, the transformer state comprehensive assessment, support the final analysis and diagnosis system software of making a decision of user, by the analyzing and diagnosing computing machine according to automatic input come in the various signal datas of transformer online monitoring carry out the transformer line real time diagnosis in conjunction with the information stored database, the result is shown, the abnormal conditions of transformer are reported to the police.

The present invention can unify to consider when manufacturing and designing, and organically combines with the transformer body, and intelligent degree height, compact conformation is reliable, the ratio of performance to price is also more reasonable.The status flag amount such as shelf depreciation, oil gas, water content in oil, core current, sleeve pipe insulating property, winding hottest spot temperature that can be used for the on-line monitoring transformer, can carry out the analysis-by-synthesis diagnosis to transformer fault, advantageously remedy the deficiency of offline inspection and periodic maintenance, made the user can grasp the truth of equipment more timely, more delicately.Both improved the transformer running status had been monitored and the computer-aided diagnosis level in real time, and helped the user to reach more reliable, the more economical target of equipment operation again.

Description of drawings

The present invention is described in further detail below in conjunction with the embodiment accompanying drawing.

Fig. 1 general principles block diagram of the present invention

The theory diagram of Gas in Oil of Transformer monitoring sensor among Fig. 2 the present invention

The theory diagram of moisture content monitoring sensor in the transformer oil among Fig. 3 the present invention

Fig. 4 system architecture synoptic diagram of the present invention

Partial discharge of transformer monitoring device structural representation among Fig. 5 the present invention

Analog signal processing block scheme among Fig. 6 the present invention

Analog signal conditioner circuit (10) figure among Fig. 7 the present invention

Fig. 8 middle sleeve three-phase imbalance of the present invention electric current summing circuit (31) figure

The functional module block scheme of Fig. 9 transformer fault real-time diagnosis system

Figure 10-1, Figure 10-2, Figure 10-3 transformer fault real-time diagnosis system flowchart

Three layers of B P of Figure 11 Gas in Oil of Transformer gas chromatography network structure

Figure 12 transformer resultant fault decision tree diagram

：1- 2- 3- 4- 5- 6- 7- 9- 10- 11- 12- 13-RS232 14- 15- 16- 17- 18- 19- 20- 21- 22- 23- 24-DB37 25- 26- 27- 28- 29- 30- 31- 32- 33- 34- 35- 36- 37- 38- 39-40- 41- 42- 43- 44- 45-46- 47- 48- R- RW- C- U- LC- ( A4N25 ) A- J- GND-

Embodiment

In the present embodiment, be status flag amounts such as gas in the shelf depreciation, oil of transformer, water content in oil, core current, sleeve pipe insulating property, winding hottest spot temperature to the transformer online monitoring project, below the monitoring of these projects introduced respectively.

1, partial discharge of transformer monitoring device

To the statistical results show of transformer physical fault, insulation fault is the main cause that influences the normal operation of transformer.At present the large-scale power transformer of 110kV and above grade still mainly adopts the paper oil insulation structure, and it is aging that insulation fault usually originates from the paper oil insulation that shelf depreciation causes.Paper oil insulation is the weak link of power transformer, and shelf depreciation is to cause the key factor of insulation fault.Shelf depreciation is meant because Electric Field Distribution is irregular, and internal field is too high, causes electrical discharge or punch-through in the subrange in the insulating medium.It may be created in the solid insulation hole, in the fluid insulation bubble or between the insulation course of different medium characteristic.If electric field intensity is higher than the particular value that medium has, also may occur in the liquid or solid insulation.Shelf depreciation develops gradually, by insulating medium around it is constantly corroded, finally causes the inefficacy of whole insulation system, so shelf depreciation is to cause the main cause of insulation degradation.It also is the important sign and the form of expression of insulation degradation simultaneously, closely related with the deterioration and the breakdown process of insulating material, so can reflect the fault of transformer built-in electrical insulation effectively, especially much effective than methods such as stratographic analyses in dielectric loss measurement, the oil to the early detection of catastrophic discontinuityfailure.Therefore it is significant for the safe and stable operation of power equipment shelf depreciation to be carried out effective on-line monitoring.

To the partial discharge of transformer on-line monitoring, make people can understand the shelf depreciation situation of the transformer in the duty at any time among the present invention,, prevent the generation of transformer insulated fault for the normal operation of control transformer provides foundation.This partial discharge of transformer on-Line Monitor Device comprises sensor probe (41), the rodmeter (35) that is installed in oil tank of transformer fuel outlet valve place (37), bottom (also can be the leak valve place at top), rodmeter is connected with prime amplifier (32) by ultrahigh frequency cable (33), be connected with SHF receiver (34) again, SHF receiver (34) with signal change, processing is after modulus capture card (11) inserts foreground industrial computer (14).So just the shelf depreciation situation in the transformer running status in time can be monitored the ultrahigh frequency electric signal that receives come out through amplification with after handling by sensor (1), when unusual circumstance, make people can in time take regulation measure, thereby avoid the generation of transformer fault.

The sensor element product in the detection system instrument is put in the UHF-TMS office that the Dutch KEMA of high-frequency local discharging sensor (1) employing company produces, this sensor can effectively be got rid of dried excellent in hyper band, the phase place that reflects inner shelf depreciation in the transformer on-line operation truly, discharge time, the discharge capacity size, by relatively drawing four kinds of dissimilar partial discharges fault (along the insulating board surface-discharge with fingerprint base, the discharge of transformer coil turn-to-turn insulation, the suspension metal causes discharge, metal tip discharge under the noble potential), measuring the partial discharge quantity guarantee value is 20PC, minimum value reaches 10PC, even can reach 5PC.

According to the IEC of International Electrotechnical Commission 60270, there are a variety of methods to carry out the detection of shelf depreciation.Based on the detection technique on the electricity-magnetic emission basis of shelf depreciation, on a SHF band, utilize broad-band antenna and broadband reception device to be achieved.Sensor probe (41) can be installed on the transformer bottom fuel outlet valve (37), and the signal that it receives is transmitted to foreground industrial computer (14) by prime amplifier (32), SHF receiver (34) and modulus capture card (11) by sensor (1).Wherein, the tank wall of transformer can also be as the screen layer of electromagnetic interference (EMI).Prime amplifier (32) and SHF receiver (34) adopt UHF-TMS office to put in the detection system instrument and the suitable parts product of sensor (1) equally, the receiver that this SHF receiver (34) has two different SHF band, it is that the receiver of 40MHz can be regulated from 300～600MHz for the 10MHz. broadband range that the whole step pitch of frequency adjustable is detected at the center, and broadband range is that the receiver of 80MHz can be regulated from 600～1200MHz.Overvoltage protection is carried out in input, and noise level is lower than-65dBm, and sensitivity is 100mV/dB, and 60dBm is configured to 0V (so 1V=-50dBm, 2V=-40dBm etc.), and signal should not surpass 60dBm.The output signal that input pulse is put in the office of a weak point is determined by pulse shape, input signal is converted to the standard output signals of about 500ns of cycle.

SHF receiver (34) is furnished with the N type plug of high frequency cable special use, and sensor (1) must directly be linked prime amplifier (32).Recommend to use high frequency cable (33), because the decay of signal is very serious in the crosslinked cable of standard as RG 213U one class.

Before the probe (41) of sensor (1) is installed, the outer flange of transformer valve should be removed (valve is closed always).To be inserted into the probe rod of sensor (1) and how long analyze, design thickness according to oil tank wall, fuel tap, valve orchid, calculate probe rod and insert length, so that sensor probe (41) head can not exceed oil tank of transformer (36) tank wall 3cm after inserting transformer inside, rodmeter (35) is installed in the flange (38) of transformer and goes up (perhaps being installed on the leak valve at top).

Ultrahigh frequency cable (33) is connected to the end of rodmeter (35), and the other end of cable (33) is connected to the input end of prime amplifier (32).At stube cable (33) before, cable conductor must pass through the short circuit of a ground connection object, in case prime amplifier (32) damages owing to the static electric charge of cable (33).The input end of SHF receiver (34) is received in the output of prime amplifier (32), and the acquisition channel of the modulus capture card (11) of foreground industrial computer (14) is received in the output of receiver (34).

When using the monitoring of this partial discharge of transformer monitoring device,, need to measure period and be set at hundreds of times (being generally 200 times) in order to react the true discharge scenario of transformer itself more accurately.So just can produce lot of data after measurement, measured data are stored with independent data file.Such way can effectively be avoided obscuring because of repeatedly measuring the data that caused, and also is convenient to each measurement is inquired about simultaneously.Yet, bring very big trouble reading of data will certainly for the backstage diagnostic routine at this storage mode.Be directed to this, we have developed partial discharge of transformer data read software, have overcome when data read the obstacle by file layout caused.The modulus capture card adopts GaGe, data handling machine adopts CPU:PIII 850 as foreground industrial computer (14), after having backstage diagnostic process computing machine (15) (adopting CPU:PIII 866) to be connected foreground industrial computer (14), in foreground industrial computer (14), utilize the foreground capture program, at first digital signal is carried out software filtering and smoothing processing, then according to the requirement of backstage diagnostic process computing machine (15) to the data form, digital signal is carried out format conversion, leave it in appointed positions at last.Backstage diagnostic process computing machine (15) utilizes independently developed comprehensive diagnos program after obtaining needed data to appointed positions, and partial discharge of transformer signal and other desired signal are carried out overall treatment, provides diagnostic result at last.As when being judged as transformer operation shelf depreciation overshoot scope, then send alerting signal, reminds people should be taked the generation of corresponding measure prevention transformer fault.

2, transformer iron core grounding current on-Line Monitor Device

In high voltage, powerful transformer, be the discharge that prevents that the iron core floating potential from causing, transformer core ground wire in service is often drawn through sleeve pipe, under the normal running (operation) conditions, iron core grounding current is a milliampere level, when causing overheating fault unshakable in one's determination when iron core generation multipoint earthing, iron core grounding current can obviously increase.Rule of thumb, under normal circumstances, iron core grounding current has only tens～tens milliamperes, if just should take appropriate measures when iron core grounding current is increased to 100mA; When if iron core grounding current is increased to 1A, just must take necessary means.Therefore to the on-line monitoring of iron core grounding current, can find iron core generation multipoint earthing fault effectively.And also lack the effective means that transformer core current is monitored at present, make this work fail to launch always.

Transformer iron core grounding current on-Line Monitor Device among the present invention makes people can understand the iron core grounding current state of the transformer in the duty at any time, for the normal operation of control transformer provides foundation.

This transformer iron core grounding current on-Line Monitor Device comprises the iron core grounding current extension line, this ground current extension line is connected with earth terminal after passing sensor, the induced signal of sensor inserts signal amplification circuit, connect the analog signal conditioner circuit again, analog signal conditioner circuit and modulus capture card (modulus capture card major function be digital signal with analog signal conversion) are connected the back and insert data handling machine.This device goes out electric signal to transformer iron core grounding current by sensor sensing, after amplifying, can be come out by monitoring in time, when abnormal conditions occurring, make people in time take regulation measure or handle automatically, avoid the generation of transformer fault by automatic control system.

Select condenser type punching sensor (model YA-XX-E4) to make current sensor (5), adopt the current sensor of two kinds of ranges: A:0-200mA, B:0-2A.When the varying in size of ground current, the sensors that adopt different amounts to claim are measured.Thereby guarantee the accuracy of data.When measuring with the sensor of two different ranges, measured result has certain deviation certainly.Want from both to select one as final reading, will do and judge: when reading wide range sensor reading＞200mA, output wide range sensor reading, otherwise output journey sensor reading in a small amount.Because when using this current sensor to measure, measured value is an analog quantity, so in the process of computing machine Data transmission, be also noted that two problems: the one, the conditioning of signal (finishing) by the modulate circuit plate; The 2nd, analog quantity is converted to can be for the digital quantity (being finished by the modulus capture card) of computing machine use.

(1), the conditioning of signal

Acquired signal is nursed one's health, be meant after sensor (5) induces simulating signal, the pre-amplification circuit that carries through sensor (5) tentatively amplifies, simultaneously current signal is converted to voltage signal, through 4 cores (just, negative, signal, ground) shielded cable (22) transmits to conditioning rack (9); Then by the linearity modulate circuit (27) in the modulate circuit plate (10) with the input signal linearityization, photoelectric isolating circuit (28) in modulate circuit plate (10) carries out isolation processing (its role is to prevent the impact of external high voltage to this detection system) then, through amplification filtering circuit (29) simulating signal is carried out filtering and noise reduction and processing and amplifying.Analog signal conditioner circuit (10) carries out linearity conditioning, filtering, denoising and processing and amplifying to this simulating signal.

In order to guarantee the versatility and the interference free performance of analog signal conditioner circuit (10), require the output signal of sensor (5) to unite, be defined as the voltage signal of 0-12V.Analog signal conditioner circuit (10) comprises linearity modulate circuit (27), electrical isolation circuit (28), amplification filtering circuit (29), and electrical isolation circuit (28) adopts photoelectrical coupler.The function that the analog signal conditioner circuit need be finished has: signal linearization, electrical isolation, amplification (decay), filtering and functions such as power supply are provided.

1) linearity conditioning

Because the photoelectricity isolation has been passed through in the transmission of signal, undistorted for the signal that guarantees the photoelectrical coupler two ends, so need carry out the conditioning of the voltage signal linearity, the circuit that is adopted among the present invention has guaranteed the undistorted of photoelectrical coupler two end signals preferably.Shown in circuit among Fig. 7 27.

2) electrical isolation

Because measuring-signal is from the higher transformer of voltage, so must isolate.Through what consider to use is that photoelectrical coupler is proper.

3) amplification filtering

The amplification filtering circuit is designed to the link that can regulate enlargement factor according to the gain situation of front stage circuits, adopts traditional activated amplifier, wave filter.

(2) analog to digital conversion and data processing

Modulus capture card (12) adopts ADLINKPCI-9113, data handling machine adopts CPU:PIII 850 as foreground working machine (14), after having backstage diagnostic process computing machine (15) (adopting CPU:PIII 866) to be connected Foreground Data process computer (9), in foreground industrial computer (14), utilize the foreground capture program, at first digital signal is carried out software filtering and smoothing processing, then according to the requirement of backstage diagnostic process computing machine (15) to the data form, digital signal is carried out format conversion, leave it in appointed positions at last.Backstage diagnostic process computing machine (15) utilizes independently developed comprehensive diagnos program after obtaining needed data to appointed positions, and iron core grounding current signal and other desired signal are carried out overall treatment, provides diagnostic result at last.As when being judged as transformer operation overshoot scope, then send alerting signal, reminds people should be taked corresponding measure prevention transformer

The generation of fault.

3, transformer oil dissolved gas on-Line Monitor Device

The insulation of 110kV and above electric pressure power transformer is oil-paper composite insulation structure,, heat, under the synergy of machinery equal stress, insulating oil, insulating material such as paper and cardboard will be degraded aging gradually, produce various low molecular weight gas thereby decompose, these gases will be dissolved in the transformer oil to a certain extent, utilize the described method of IEC567 publication, can from oil sample, isolate these dissolved gases, and can further utilize chromatographic technique that it is carried out quantitative test, Here it is dissolved gas analysis (Dissolved Gas Analysis, be called for short DGA), or be called the oil chromatography analysis.Oil dissolved gas has reflected the fault state of transformer to a certain extent, and therefore the result according to stratographic analysis can diagnose the type and the order of severity of fault.To one of conventional important detection means of using of oil immersion power equipment, effect is remarkable on inner latency fault of diagnosis transformer and development degree thereof as electric system at present for it.Conventional stratographic analysis test is drawn oil sample on the transformer at the scene, delivers to testing laboratory again and carries out stratographic analysis, and the test period is long, and measurement result is easy to generate deviation.Particularly for those speed of developments insulation fault faster, often be difficult to timely discovery, if can carry out on-the-spot on-line monitoring to the transformer oil dissolved gas, in advance before transformer breaks down, capture some fault tendency information timely, each power supply unit to operation power all is highly significant so.

Transformer oil dissolved gas on-Line Monitor Device among the present invention can be monitored out oil tank of transformer oil dissolved gas amount continuously, in time understands the inner latency fault of transformer, takes preventive measures and avoids the generation of fault.This transformer oil dissolved gas on-Line Monitor Device comprises and is installed in gas intelligence sensor (2) in the oil that transformer bottom fuel outlet valve (37) locates, the electric signal of intelligence sensor (2) inserts the foreground industrial computer (14) that can show gas content through RS232 interface (13) and serial communication line (23), also industrial computer (14) can be connected with backstage diagnostic process computing machine (15) afterwards.This device can carry out on-line continuous to Gas in Oil of Transformer content constantly monitors, and whether contained gas flow in the transformer oil is exceeded standard, and can understand at any time, can in time take corresponding measure when situation occurring exceeding standard, avoids the generation of transformer fault.

Intelligence sensor can adopt U.S. SERVERON company production and sales " gas chromatogram sensor in the oil " product, and this sensor can detect eight kinds of gas markers of latent transformer fault, and measurement range and precision are as follows: hydrogen (H ₂) 10-2,000PPM ± 10% acetylene (C ₂H ₂) 5-500PPM ± 1PPM or ± 10% ethene (C ₂H ₄) 7-2,000PPM ± 1PPM or ± 10% methane (CH ₄) 70-2,000PPM ± 10% carbon monoxide (CO) 5-2,000PPM ± 1PPM or ± 10% ethane (C ₂H ₆) 10-2,000PPM ± 10% oxygen (O ₂) 10-1,400PPM ± 10% carbon dioxide (CO ₂) 5-7,000PPM ± 1PPM or ± 10%

Diagnostic detection data comprise other seven kinds of gas contents and the total hydrocarbon content (total hydrocarbon content=methane+ethane+ethene+acetylene) that deoxygenation gas is outer among the present invention.

Also adoptable another kind of intelligence sensor be Canadian SYPROTEC company produce and " Hydran201i gas monitoring system " equipment of sale in " intellectual transmitters of H201Ti " product (the Hydran201i gas monitoring system comprises several parts such as communication controller that are called the intellectual transmitters of H201Ti and are called H201Ci series, employed harvester is exactly a hydran201Ti intellectual transmitters part among the present invention), the situation that is used for total appropriate hydrocarbon gas in the gas on-line monitoring transformer obtains hydrogen (H as needs ₂), acetylene (C ₂H ₂), the rare (C of second ₂H ₄), carbon monoxide (CO), methane (CH ₄), ethane (C ₂H ₆), carbon dioxide (CO ₂) several individual event gas contents, adopt traditional oil sample chromatogram analysis method to obtain each individual event gas content, and, participate in the comprehensive diagnos analysis by in the friendly interface input system.

Intellectual transmitters H201Ti (2) among the Hydran201i, it has a little column type shell, this shell can directly install to the fuel outlet valve place of transformer by a brass connector, and a sensor, a sleeving temperature regulating system and a smart electronics forwarder that has microprocessor are housed in this shell.Also have terminal row in the shell for connecting AC power, the connector that a RS232 also is housed in the shell is connected with a serial communication line (23) with foreground industrial computer (14).The principle of work of H201Ti intellectual transmitters (2) is summarized as follows: the inflammable gas in the transformer oil enters interior electrochemical gas detector---the heating chamber (43) of H201Ti intellectual transmitters (2) by the film (42) that one deck has the selectivity ventilating performance, produces an electric signal that is directly proportional with reaction rate with the oxygen reaction in electrochemical gas detector.This electric signal is converted to digital signal by analog signal conditioner circuit (44) and the single-chip microcomputer (45) in the H201Ti intellectual transmitters (2).

Foreground industrial computer (14) is a data handling machine, adopt CPU:PIII 850, there is backstage diagnostic process computing machine (15) to adopt CPU:PIII 866, after being connected foreground industrial computer (14), (14) utilize the foreground capture program in the industrial computer of foreground, at first digital signal are carried out software filtering and smoothing processing, then according to the requirement of backstage diagnostic process computing machine (15) to the data form, digital signal is carried out format conversion, leave it in appointed positions at last.After backstage diagnostic process computing machine (15) is obtained needed data by coaxial cablet (25) to appointed positions, utilize independently developed comprehensive diagnos program, total hydrocarbon gas content electric signal and other desired signal are carried out overall treatment, provide diagnostic result at last.As when being judged as transformer operation overshoot scope, then send alerting signal, reminds people should be taked the generation of corresponding measure prevention transformer fault.Its course of work: the content (digital quantity of from transformer oil, measuring total appropriate hydrocarbon gas by Intelligence sensor (2), use the ppm value representation), by serial communication line (23) the THC gas flow (being digital signal this moment) that is measured is transferred to this digital quantity in the interior Access database of foreground industrial computer (14) by RS232 serial ports (13) again, software by establishment is shown by the industrial computer (14) that is sent to the foreground after in the Access database of industrial computer (14) digital quantity of THC gas being read and this digital quantity is sent in the SQL database of foreground industrial computer, transfer to by coaxial network interface card PCI-2029 transmission line (25) in the SQL database of backstage diagnostic process computing machine (15), as the foundation of data analysis diagnosis.

4, bushing shell for transformer (19) insulating property on-Line Monitor Device

By 240 transformer fault reasons are added up, finding just has 17 owing to what sleeve pipe (19) capacitance plate insulation breakdown was discharged, since the end of sleeve pipe (19) screen capacitance plate punctures or over the ground discharge just have 4, its failure rate has reached 8.75%, so from above data and data presentation, the state of bushing shell for transformer (19) occupies an important position in the status monitoring of transformer, by timely monitoring to bushing shell for transformer (19) insulating property state, can in time understand the transformer fault that will occur, prevent the generation of transformer burst accident.Bushing shell for transformer among the present invention (19) insulating property on-Line Monitor Device makes people can understand sleeve pipe (19) the insulating property state of the transformer in the duty at any time, for the normal operation of monitoring transformer provides foundation.This bushing shell for transformer (19) insulating property on-Line Monitor Device comprises condenser type punching sensor (6), the end screen ground wire of three-phase sleeve pipe passes condenser type punching sensor (6) respectively, induce three voltage signals of the end screen of three-phase sleeve pipe respectively through condenser type punching sensing (6) device, induced signal is imported the modulate circuit plate through 4 core shielded cables (22), carry out linearity conditioning by modulate circuit (10), amplify (decay), Filtering Processing, export as three-phase bottom shielding of bushing current signal, and the three-phase output signal is input to the three-phase voltage summing circuit, obtain its three-phase imbalance voltage U 0, as the 4th output signal, with first three signal, by the DB37 shielded cable, be sent to modulus capture card (ADLINK PCI-9113) (12), is digital signal through modulus capture card (12) with analog signal conversion, be sent to foreground industrial computer (14), call the foreground capture program, digital signal is carried out software filtering and smoothing processing and digital signal carried out format conversion deposit in the computing machine of assigned address and demonstration.

Under sleeve pipe (19) insulation normal operation, the plenary capacitance electric current of sleeve pipe (19) is a milliampere level, and when insulation fault took place sleeve pipe (19), the plenary capacitance electric current can obviously increase.The electric signal that induces through sensor (6) can be come out by monitoring in time through amplifying, and makes people in time take regulation measure or is handled automatically by automatic control system, avoids the generation of fault.

The range of the condenser type punching sensor (6) that we select is 0-25mA, and model is YA-XX-E4.

(1), the conditioning of signal

Acquired signal is nursed one's health, be meant after sensor sensing goes out simulating signal, at first will transmit to conditioning rack (9) through 4 cores (just, bearing signal, ground) shielded cable (22); By modulate circuit plate (10) this simulating signal is carried out linearity conditioning, photoelectricity isolation, filtering, denoising and processing and amplifying then.

In order to guarantee the versatility and the interference free performance of analog signal conditioner circuit (10), require the output signal of all the sensors to unite, be defined as the voltage signal of 0-12V.Below conditioning circuit board (10) is carried out system description:

Analog signal conditioner circuit (10) comprises linearity modulate circuit (27), electrical isolation circuit (28), amplification filtering circuit (29), and electrical isolation circuit (28) adopts photoelectrical coupler.The function that the analog signal conditioner circuit need be finished has: signal linearization, electrical isolation, amplification filtering and functions such as power supply are provided.Photoelectric isolating circuit (28) is because measuring-signal is from the higher transformer of voltage, so must carry out electrical isolation.Proper through considering to use photoelectrical coupler.The design has satisfied the function that photoelectricity is isolated.

1) linearity conditioning

2) electrical isolation

3) amplify (decay) filtering

Amplification (decay), filtering circuit are designed to the link that can regulate enlargement factor according to the gain situation of front stage circuits, adopt traditional activated amplifier, wave filter.

(2) neutral point voltage summing circuit:

The three-phase voltage signal of amplification filtering circuit (29) output is input to three-phase imbalance voltage summing circuit (31) among Fig. 8, as shown in Figure 8, at first form differential filtering circuit input signal is carried out Filtering Processing by operational amplifier, carry out phase adjusted by the two-stage slide rheostat then, make three-phase voltage signal output phase each differ from 120 the degree, after the summation filtering circuit three-phase signal is sued for peace and Filtering Processing, and as neutral point unbalance voltage signal (U ₀) output.

(3) analog to digital conversion and data processing

Front end three-phase voltage simulating signal is sent to the modulus capture card (10) that is arranged in foreground industrial computer (14) by DB37 shielded cable (24), is digital signal by modulus capture card (10) with analog signal conversion; Then, calling corresponding software by foreground industrial computer (14) carries out software filtering and smoothing processing and digital signal is carried out format conversion being stored in assigned address and showing at foreground industrial computer (14) to digital signal; Connect backstage diagnostic process computing machine (15) then.Modulus capture card (10) adopts ADLINKPCI-9113, foreground industrial computer (14) adopts CPU:PIII 850, backstage diagnostic process computing machine (15) adopts CPU:PIII 866, in Foreground Data process computer (14), utilize independently developed foreground capture program, at first digital signal is carried out software filtering and smoothing processing, then according to the requirement of backstage diagnostic process computing machine (15) to the data form, digital signals format is converted to the desired storage format of SQL server database, leaves it in appointed positions at last.After backstage diagnostic process computing machine (15) is obtained needed data to appointed positions, utilize independently developed comprehensive diagnos program, the signal of three-phase bottom shielding of bushing current signal and three-phase imbalance voltage signal and other introducings is carried out overall treatment, provide diagnostic result at last.

5, coiling hot point of transformer temperature on-line monitoring device

Large-scale power transformer cryogenic overheating problem is that operation and manufacturing sector all very pay close attention to also one of issue to be resolved.From transformer insulated operation life, it is generally acknowledged that should follow six degree rules is that year-round average temperature is 98 ℃, every rising or reduce by 6 ℃, then transformer life reduces half or prolongs one times.The transformer focus allows that limit value is 118 ℃.When hot-spot temperature of transformer surpassed 140 ℃, oil on every side can decomposite gas, causes CO, CO ₂Content increases, and paper insulation also can cracking produce furfural simultaneously, will reduce insulating property, influences the normal operation of transformation.Therefore hot spot temperature of winding is carried out the hot(test)-spot temperature situation that on-line monitoring helps in time understanding transformer, go wrong and in time to solve, help prolonging the serviceable life of transformer.And the hot spot temperature of winding that pair transformer is not arranged is at present as yet carried out the effective means of on-line monitoring.

Coiling hot point of transformer temperature on-line monitoring device among the present invention can carry out the real-time continuous monitoring to the coiling hot point of transformer temperature, improves the monitoring level that aligns the transformer that is in operation.This coiling hot point of transformer temperature on-line monitoring device comprises the ring current transformer that transformer condenser type insulating sleeve (19) afterbody is installed, with the leading-out terminal of this ring current transformer with a cable short circuit, this short-circuiting cable passes load current sensor (7), the induced signal of load current sensor (7) inserts signal conditioning circuit (10), modulate circuit (10) is converted to standard signal with this current signal and inserts foreground industrial computer (14), other has a probe that the temperature controller (4) that is inserted into transformer upper temp meter seat is installed, its induced signal inserts signal conditioning circuit (10) by cable (22), modulate circuit (10) is converted to standard signal with this current signal and inserts foreground industrial computer (14), and industrial computer (14) is connected with backstage diagnostic process computing machine (15).

The IEC354 of International Electrotechnical Commission standard and National Standard oil-immersed power transformer load guide rule " in, having provided the temperature rise of hot spot computing formula, this is traditional classical way of present coiling hot point of transformer temperature estimation.The present invention makes according to this principle.Computing formula according to stable state hot(test)-spot temperature under the arbitrary load: in the formula: θ _a-top-oil temperature; H _Gr-focus is to the temperature rise of winding top oil; K-load factor is by I/I _HObtain, I is the actual loading electric current, gets maximal value in the threephase load electric current; I _HBe nominal load current, can directly obtain from transformer nameplate parameter; Y-winding temperature index; H _{Gr, y}Even load performance data value such as table 1 show.Therefore, only need measure top-oil temperature and load current I (K=I/I _H), just can estimate hot spot temperature of winding.Record each phase actual loading electric current I from each phase casing current transformer (7), and record top-oil temperature θ by top-oil temperature sensor (4) _aThe coiling hot point of transformer temperature is gone out current signal by sensor senses can be come out by monitoring in time, thereby can in time understand the coiling hot point of transformer temperature variations, when abnormal conditions appear in temperature, make people in time take regulation measure or handle automatically, avoid the generation of transformer fault by automatic control system.

Select condenser type punching sensor (model YA-XX-E4) to make to measure the sensor (7) of load current, it is XMT-288FC temperature controller (4) that top-oil temperature is measured used sensor.

(1), load current is measured:

Transformer condenser type insulating sleeve (19) afterbody is equipped with ring current transformer (7), with the leading-out terminal of this ring current transformer (7) with a cable short circuit, this short-circuiting cable passes condenser type punching sensor of the present invention (7), condenser type punching sensor (7) induces voltage signal, by 4 core shielded cables (22) signal is sent to conditioning rack (9), carry out the conditioning of signal by analog signal conditioner circuit (10), and signal is carried out amplification filtering handle, send modulus capture card (12) in the foreground industrial computer (14) to by DB37 cable (24) then, is digital signal by modulus capture card (12) with analog signal conversion, call foreground industrial computer (14) call signal capture program then, be converted to load current value I by certain linear scaling, by coaxial network interface card (26) (PCI-2029) and coaxial cablet (25) send backstage diagnostic process computing machine (15) to, like this, backstage diagnostic process computing machine (15) has just obtained the actual loading electric current I.

1), analog signal conditioner circuit (10) is described:

Acquired signal is nursed one's health, be meant after sensor sense (7) should go out simulating signal, at first will transmit to conditioning rack (9) through 4 cores (just, bearing signal, ground) shielded cable (22); By modulate circuit plate (10) this simulating signal is carried out linearity conditioning, filtering, denoising and processing and amplifying then.The modulate circuit of load current signal as shown in Figure 7.

In order to guarantee the versatility and the interference free performance of analog signal conditioner circuit (10), require the output signal of sensor (7) to unite, be defined as the voltage signal of 0-12V.Analog signal conditioner circuit (10) comprises linearity modulate circuit (27), electrical isolation circuit (28), amplification filtering circuit (29), and electrical isolation circuit (28) adopts photoelectrical coupler.The function that the analog signal conditioner circuit need be finished has: signal linearization, electrical isolation, amplification (decay), filtering and functions such as power supply are provided.

1) linearity conditioning

2) electrical isolation

3) amplify (decay) filtering

2). analog to digital conversion and data processing

Front end load current simulating signal I is sent to modulus capture card (12) in the foreground industrial computer (14) by DB37 shielded cable (24), is digital signal by modulus capture card (12) with analog signal conversion; Then, calling corresponding software by foreground industrial computer (14) carries out software filtering and smoothing processing and digital signal is carried out format conversion being stored in assigned address and showing at foreground industrial computer (14) to digital signal; Connect backstage diagnostic process computing machine (15) then.Modulus capture card (12) adopts ADLINKPCI-9113, foreground industrial computer (14) adopts CPU:PIII 850, backstage diagnostic process computing machine (15) adopts CPU:PIII 866, in foreground industrial computer (14), utilize independently developed foreground capture program, at first digital signal is carried out software filtering and smoothing processing, then according to the requirement of backstage diagnostic process computing machine (15) to the data form, digital signals format is converted to the desired storage format of SQL server database, leaves it in appointed positions at last.

(2), top-oil temperature is measured:

It is XMT-288FC temperature controller (4) that top-oil temperature is measured used sensor, the probe of XMT-288FC temperature controller (4) is inserted into transformer upper temp meter seat, the signal of obtaining oil temperature carries prime amplifier by XMT-288FC temperature controller (4) after by signal, current signal is converted to voltage signal, by 4 core shielded cables (22) signal is sent to conditioning rack (9) again, carry out the conditioning of signal by analog signal conditioner circuit (10), and signal is carried out the linearity nurse one's health, amplification filtering is handled, and its analog signal conditioner circuit (10) is consistent with the load current signal conditioning circuit.Send modulus capture card (12) in the foreground industrial computer (14) to by DB37 cable (24) then, is digital signal by modulus capture card (12) with analog signal conversion, by foreground industrial computer (14) call signal capture program digital signal is carried out software filtering and smoothing processing, and be converted to the top-oil temperature value by certain linear scaling, then according to the requirement of backstage diagnostic process computing machine (15) to the data form, digital signals format is converted to the desired storage format of SQL server database, leave it in appointed positions at last, and show at foreground industrial computer (14); Description in the collection of its detailed process and load current is consistent.Then with the load current signal, by coaxial network interface card (26) (PCI-2029) and coaxial cablet (25) send backstage diagnostic process computing machine (15) to, like this, backstage diagnostic process computing machine (15) has just obtained top-oil temperature value θ _a

At last, backstage diagnostic process computing machine (15) is with the actual loading electric current I and the top-oil temperature value θ that record _aCalculate coiling hot point of transformer temperature θ _h, and show that when this temperature surpasses the setting value that gives, then send alerting signal, reminds people should be taked the generation of corresponding measure prevention transformer fault.

6, oil tank of transformer water content in oil on-Line Monitor Device

Water content in oil fixes on really in the maintenance schedule of transformer and occupies vital part in the oil tank of transformer.Thereby oil is invaded the formula transformer and is relied on the insulation of the circulation protection transformer of oil not to be subjected to breakage; after water content in oil surpasses 35ppM; just can cause aging and intensity insulation of insulating medium also can reduce; under serious situation; the reduction of this insulation ag(e)ing and dielectric strength can cause the distortion and the turn-to-turn short circuit of coil; produce the transformer operation troubles, shorten the serviceable life of transformer greatly.Therefore to the on-line monitoring of oil tank of transformer water content in oil, can predict the fault that transformer will take place early.And do not have at present water content in oil in pair oil tank of transformer to carry out the effective means of on-line monitoring as yet.

Oil tank of transformer water content in oil on-Line Monitor Device among the present invention is in time monitored the changing condition of oil tank of transformer water content in oil, accurately prevents the appearance of transformer fault, guarantees that the safety of transformer is normally moved.This oil tank of transformer water content in oil on-Line Monitor Device comprises bottom flowline (20), the intelligence sensor (3) that one group of refrigeratory of transformer links to each other with transformer, with after the RS232 interface is connected with foreground industrial computer (14) with serial communication line (23), foreground industrial computer (14) can be connected with backstage diagnostic process computing machine (15) by Ethernet the electric signal that intelligence sensor (3) will be installed in the sensor probe (46) on bottom flowline (20) through amplification or decay, rectification, filtering and analog to digital conversion.This device can in time be monitored the variation of oil tank of transformer (36) water content in oil, when abnormal conditions occurring, makes people in time take regulation measure or is handled automatically by automatic control system, avoids the generation of transformer fault.

The intelligence sensor (3) that monitoring is adopted in this device is to adopt the HMP228 hygronom product of Finland Visala company production and selling as intelligence sensor, this intelligence sensor (3) is made up of sensor probe (46), analog signal conditioner circuit (47) and (48) three of the single-chip microcomputers part that has a RS232 interface, sensor probe (46) is installed on one group of refrigeratory and transformer body (18) the bottom flowline (20), in this flow circuits, the measured water cut of sensor (3) only needs a few minutes can reach balance.This sensor probe (46) can be measured active two simulating signals of oily gentle water of sensor (3) position, through analog signal conditioner circuit (47) amplify or decay, rectification, filtering after, give single-chip microcomputer (48) and carry out analog to digital conversion, and calculate water content in oil content according to active two measured values of the gentle water of oil, be sent to foreground industrial computer (14) by RS232 interface and serial communication line (23), can also sampling rate, unit, the output format of sensor (3) be provided with by this RS232 interface (13).Foreground industrial computer (14) is a data handling machine, adopt CPU:PIII 850 as the foreground working machine, there is backstage diagnostic process computing machine (15) to adopt CPU:PIII 866, after being connected foreground industrial computer (14), in foreground industry control (14) machine, utilize the foreground capture program, at first digital signal is carried out software filtering and smoothing processing, then according to the requirement of backstage diagnostic process computing machine (15) to the data form, digital signal is carried out format conversion (digital signal is converted to the desired data memory format of SQL Server), leave it in appointed positions at last.Backstage diagnostic process computing machine (15) by coaxial network interface card (26) (PCI-2029) and coaxial cablet (25), read little water number according to backstage diagnostic process computing machine (15) in the mode of database communication from foreground industrial computer (14) assigned address by Ethernet interface, backstage diagnostic process computing machine (15) first step is handled the raw data that exactly front end is transmitted and is carried out pre-service, promptly reject wild point, smoothing processing, measures such as rate of change calculating are to weaken environmental factor and sensor itself, the influence that Network Transmission is brought, data are carried out various analyses, for reliable data are extracted in follow-up judgement, and setting warning value, if signal exceeds warning value, promptly provide alerting signal, remind the staff to handle.

A total database is arranged in foreground industrial computer (14), more than after six kinds of signals enter industrial computer (14), according to predefined category classification storage, reading of data during each is classified to the Foreground Data storehouse successively by backstage diagnostic process computer program, be sent to classification and storage in the database of backstage diagnostic process computing machine, and arrange according to the time order and function of data acquisition.

Transformer on-line intelligence analysis and diagnosis method of the present invention is to carry out the analysis-by-synthesis diagnosis according to the functional block diagram among Fig. 9, and wherein each functions of modules is described below:

(1) System Privileges is safeguarded: the user of diagnostic system is registered and authority is authorized, cancelled.Be the basis of user's all operations, thereby make system possess certain security.

(2) Transformer Plant Test data management: the equipment test data are handled, analyze, inquired about and retrieve.

(3) the online detection data management of equipment: the centralized and unified management that online detection data is carried out the database server level.

(4) Dynamic Data Processing and trend analysis: the detection of dynamic data of each on-line parameter of transformer are removed pre-service such as wild point, smoothing processing, rate of change calculating and carried out statistical study, for the accumulation of follow-up diagnosis and on-site experience provides the technical support platform.

(5) the fault real-time diagnosis system knowledge base is safeguarded: finish typing, the renewal of the required knowledge and experience of transformer running status diagnosis, the source of these knowledge, experience is relevant national standard, field operation experiences, expertise and field feedback.In addition, this module be responsible for to realize that also knowledge base effectively, share safely.

(6) state comprehensive diagnos reasoning: system core module.Extract the data of the various on-line measurements of transformer,, call knowledge, experience and rule in the knowledge base, finish the comprehensive diagnos process of transformer state by the inference machine of state real-time diagnosis system through necessary calculating, conversion and pre-service etc.

(7) equipment running status comprehensive assessment: according to the reasoning of state comprehensive diagnos, produce diagnosis or suggestion with certain reference value, the diagnosis report of output certain format uses for decision support.The final goal of this module is to determine transformer state, reports to the police under the abnormal conditions, and produces certain fault type and maintenance suggestion.

(8) support the user finally to make a decision: for the user provides happening part, notable feature and the harmfulness information of the resulting fault type of diagnostic reasoning, assisted user is rule of thumb assessed transformer state and is made final resolution.This module can prevent to fail to judge and judge by accident.Support information can be adjusted by study, improves the accuracy and the validity of information.

The network structure and the database structure of transformer fault real-time diagnosis system among the present invention:

(1) network structure of fault real-time diagnosis system

Adopt client/server system (C/S) structural regime diagnostic expert system.This structure has proposed higher requirement to the performance of server host and the connectivity of network, especially ought be connected to the client of server simultaneously many time.But this system mature and reliable with regard to present both at home and abroad large and medium-sized enterprise, has developed and has operated in business management system on its Intranet or MIS etc., most characteristics that still meet the C/S system.

(2) system running environment and developing instrument

The operation platform of system also is divided into the client-server part, is 32 bit platforms.At server side, operation Chinese edition Windows 2000 Server/NT Server 4.0; At client-side, operation Windows 95/98/NT (Server or Station).The background data base server of system is based on MS SQL Server 7.0 (for Windows NT), and it has higher performance.And the document of SQL Server is comprehensive, and technical support also is reasonable.Comprehensive above advantage, and consider the data type conversion that data-interface may face when adopting disparate databases, its database server as fault diagnosis system is selected in decision for use.The system foreground adopts PowerBuilder 7.0 (Enterprise Edition/32) to develop.Main consideration for user interface, server data access performance, programmed control dirigibility.

The data structure of transformer fault real-time diagnosis system among the present invention:

In the state-detection diagnostic expert system, knowledge and data message are the design cores of all functions, operation---the realization of each step operation of system, each function all is based on the data on the server.Whether data structure is reasonable, practical on the server, will directly have influence on progress and the complexity of realization and the workload that follow-up system is safeguarded of system development.For ease of exploitation and management, total data is divided into three classes in this diagnostic system: 1. equipment operation parameter delivery test data; 2. online detection data of equipment; 3. transformer state diagnostic system knowledge base.The index of every delivery test and online detection data is avoided redundant information again simultaneously as far as possible for convenience, all is provided with primary key in each database table of Various types of data.

Transformer presence real time diagnosis technology and specific implementation thereof among the present invention

1, online detection data processing

The fault diagnosis first step is exactly that the raw data that transmits of the various signal datas to transformer online monitoring is carried out pre-service by the sliding window method, to reject wild point, obtain believable measured value, then credible measured value is asked a day moving average, month moving average and day mobile rate of change, month mobile rate of change.

The sliding window method is to be that the window pointwise of N point moves forward with a width, then with the data in the window by the processing of sorting of ascending order, the intermediate value of getting window is as output valve.N gets value according to determining that ground unrest is big more according to the size of ground unrest, and the value of N is also big more, and in embodiments of the present invention, N equals 3.The sliding window method can effectively reduce the signal high frequency background noise, and is very strong to the high-frequency impulse inhibiting effect, can slacken the influence that environmental factor and sensor itself, Network Transmission bring, and it is that follow-up sign is extracted reliable data that data are carried out various analyses.

The algorithm of day moving average is will handle the credible measured value that obtains through the sliding window method to sort by the Measuring Time sequencing, after each new credible measured value input, with this new credible measured value is starting point, in sequence, get continuous 24 credible measured values forward and be one group and average, promptly obtain a day moving average.Month moving average is with identical method, is starting point with new credible measured value, gets continuous 720 credible measured values forward and be one group and average in sequence, promptly obtains a moon moving average.Short-term, the secular variation of day moving average and the respectively corresponding transformer state of monitoring of month moving average, promptly correspondence the paroxysmal fault of transformer and the potential fault of being monitored.

The algorithm of day mobile rate of change be with when the day before yesterday moving average ask poor with day moving average of synchronization (or immediate moment) before 24 hours, with difference divided by ought the day before yesterday moving average, promptly obtain a day mobile rate of change.According to identical principle, with current month moving average with 30 days before moon moving average of synchronization (or immediate moment) ask poor, difference divided by current moon moving average, is promptly obtained a mobile rate of change of the moon.

2, threshold value determines

Threshold value is to adopt traditional comparison overall approach that transformer fault is judged the ground foundation.The whether correct and system of diagnostic rule can judge effectively transformer fault depend on to a great extent threshold value determine whether reasonable.Threshold value among the present invention can be divided into two classes according to the acquisition mode, one class threshold value obtains by statistics, core current, little water in the oil, casing current, the threshold value preparation method of on-line measurement characteristic quantity such as gas gross is in the oil: at monitoring system of the present invention and the transformer of being monitored after longer a period of time of normal operation, add up the credible measured value of above-mentioned five kinds of on-line measurement amounts in this time period respectively, day moving average, month moving average, a day mobile rate of change and a month mobile rate of change maximal value and a standard variance separately, maximal value is added and subtracted K times of standard variance promptly obtain separately threshold value, what parameter K reflected is the sensitivity that the present invention reports to the police, be that practical operation situation and expertise according to institute's monitoring transformer acquires, enlarge the value of K to a certain extent, threshold value can be divided into multistage threshold value.In the present embodiment, divide that firsts and seconds threshold value, K value are arranged is 1.5 to be the one-level threshold value, the K value be 3 o'clock be the secondary threshold value, corresponding respectively early warning and fault two states.It is directly given by relevant criterion and expertise that another kind of threshold value must obtain, the threshold value of hot(test)-spot temperature directly obtains according to six degree rules, and in the present embodiment, 104 degrees centigrade is the one-level threshold value of hot spot temperature of winding, 110 degrees centigrade is the secondary threshold value, and 116 degrees centigrade is three grades of threshold values.The demand value that the threshold value of individual event gas content is recommended according to national standard " melting gas analysis and judgment criterion in the transformer oil " (GB/T 7252-2001) in the oil is given.

3, transformer fault diagnosis

Transformer fault diagnosis system adopts the fault decision tree.The fault decision tree is a kind of differentiation network with tree structure, wherein each node is represented concrete failure modes rule, to the classification of fault is root node from the fault decision tree, finding out corresponding attribute of this object and rule compares, select satisfactory branch, move to until object till the leaf node of fault decision tree.Adopt the method for fault decision tree to help defective is progressively segmented.Figure 12 is a transformer resultant fault decision tree.As shown in Figure 12, the resultant fault decision tree is made up of two parts, part among Figure 12 in the bottom square frame is Dan Zhishu, can be referred to as rudimentary diagnosis, and be stratographic analysis (DGA) decision-making number in the top square frame, be by analyzing many structures that data and other online characteristic parameter data are carried out fault diagnosis according to oil chromatography, can being referred to as high level diagnostics.It is pointed out that such design is the applicability that strengthens this diagnostic system, makes the present invention can both be suitable for current two kinds of gas sensors preferably.Adopt " gas chromatogram sensor in the oil " product of U.S. SERVERON company production and sales as monitoring system of the present invention, because this sensor can directly obtain each individual event gas content, therefore can be real-time carry out high level diagnostics and rudimentary diagnosis, and in the present embodiment, what adopt is " intellectual transmitters of H201Ti " in Canadian SYPROTEC company " Hydran201i gas monitoring system " equipment of producing and selling, this sensor can only obtain the total content of gas in the oil, can not directly obtain the content of each individual event gas, at this moment, the function of high level diagnostics part can not be used when monitoring in real time, only move rudimentary diagnosis part, as find fault and warning, can obtain each individual event gas content data by the method for oil analysis, in predefined interface input diagnostic system, then can enable senior fault diagnosis functions, analyze more detailed fault type.

In the decision tree reasoning process, three kinds of diagnostic methods have been used: traditional comparison overall approach; Three ratio compiling methods of improvement; BP combination neural net method.Sorter shown in Figure 12 is represented the fault judgement rule that adopts these three kinds of analytical approachs respectively, sorter 1, sorter 5, sorter 6, sorter 7, sorter 8, sorter 9 are relatively overall approach judgment rule of tradition, sorter 2 is the judgment rule of three ratio compiling methods of improvement, and sorter 3 is BP neural network method judgment rules.Sorter 4: the partial discharge monitoring data join in the fault tree diagnosis by relatively drawing four kinds of dissimilar partial discharges fault (causing that along the discharge of insulating board surface-discharge, transformer coil turn-to-turn insulation, suspension metal the metal tip under discharge, the noble potential discharges) with fingerprint base.The function of sorter 4 is that the sensor element that UHF-TMS office that Dutch KEMA of the present invention company produces puts in the detection system instrument provides.

Above-mentioned diagnostic method there is following elaboration:

Traditional comparison overall approach: directly will compare with corresponding threshold value, to judge the transformer running status through the every monitoring feature value that obtains after the online detection data processing.Concise and to the point step is as follows:

The first step: extract a day moving average, day mobile rate of change, month moving average, the corresponding threshold value of month mobile rate of change.

Second step: with day moving average, day mobile rate of change do comprehensive comparison with corresponding threshold value, provide day state outcome and according to different conditions as a result assignment be 0,1,2 (0 expression is normal, 1 expression early warning or may be unusual, 2 expression fault alarms).

The 3rd step: just month moving average, month mobile rate of change are done comprehensive comparison with corresponding threshold value, provide moon state outcome and according to different conditions as a result assignment be 0,1,2 (0 expression is normal, 1 expression early warning or may be unusual, 2 expression fault alarms).

The 4th step: get the assignment of the big person of day state outcome value and month state outcome value as the diagnostic result state.

It should be noted that, in order to reduce the influence of wild point to diagnostic result, when carrying out traditional comparison overall approach diagnosis, present embodiment is only when three continuous diagnostic result ground state values are non-0, just provide corresponding diagnostic result, carry out early warning or warning.

The IEC three ratio compiling methods of improvement: the simulation test that is based on thermal cracking and discharging cracking, and the internal check result of physical fault transformer, what propose serves as the method that oil immersion equipment failure type is judged on the basis with relative content between characteristic gas, and this diagnostic method has detailed regulation in " transformer dissolved gas analysis and judgement guide rule " (GB/T 7252-2001).The coding rule and the determination methods of IEC three-ratio method are as shown in table 2.

BP neural network method: the BP neural network is a kind of of ANN (artificial neural network).ANN is a kind of network that is made of some interconnective nodes, and different connection weights are arranged between the different contacts.The knowledge that is stored among the ANN just is distributed among these different connection weights---these weights be the result who utilizes a large amount of typical samples that network is trained.Simultaneously, network possesses certain border " softening " or " obfuscation " ability, and this has improved the performance of system aspect the processing uncertain inference virtually.In fault diagnosis and the pattern recognition problem, error backpropagation algorithm (Back Propagation is called for short BP) is that the BP neural network is ripe relatively, has obtained using preferably at present.By collecting nearly thousand test figures that failure transformer arranged and suspended core testing result, a large amount of oil gas analysis of hplc (DGA) data are trained, so accuracy of diagnosis can improve greatly as sample.In the present invention, the BP neural network is mainly used in the diagnosis of realization to transformer oil gas chromatographic analysis (DGA).Be to reduce the influence that brings owing to the dispersiveness of chromatographic data, adopted the multivariate statistical analysis technology that the input vector of artificial neural network has been carried out normalized; Be about to various measurement numbers and be distributed in the unified computational space, all inputs are converted in the data interval of [0,1], just declare rate with what improve network.Figure 11 is three layers of BP network structure.

4, user interface

User interface is the window that system and user carry out message exchange.This diagnostic system adopts the multi-menu form on INTERFACE DESIGN, accomplish clear layer, beautiful interface, function practicality.Main interface is divided into three districts: comprehensive diagnos results display area, the current confidence values that real time on-line monitoring display field and each individual event Monitoring Data are arranged and corresponding states viewing area.The comprehensive diagnos results display area shows the transformer operation conditions with literal, and is provided with the image of reporting to the police, and when transformer takes place in time to report to the police when unusual, becomes the bright early warning of " Huang " color table with " green " look, and it is alert to become " red " color table contract quotation.Main layout setting has factory data inquiry, Monitoring Data inquiry, diagnostic result to inquire about, simply diagnose and menus such as high level diagnostics switching and user management.Select arbitrary menu can enter the corresponding function module.When carrying out the inquiry of Monitoring Data, data such as the confidence values of the data that detect, day moving average, day mobile rate of change, month moving average, month mobile rate of change show in the mode of changing trend diagram and form simultaneously, convenient relatively, consult.

Expert diagnosis level menu then activates the expert diagnosis result and to the final support information of making a decision of user etc.

Table 1 load characteristic data

Characteristic	Substation transformer	Large and medium-sized transformer
	Substation transformer	Large and medium-sized transformer			????ONAN	????ONAF	?OFWFOFAF	?ODWFODAF
	Focus is to the temperature rise Hgr (K) of winding top oil	????1.6	????1.6	????1.6	????ONAN	????ONAF	?OFWFOFAF	?ODWFODAF	????2.0
Winding temperature index y	Focus is to the temperature rise Hgr (K) of winding top oil	????1.6	????1.6	????1.6	????23	????26	????22	????29	????2.0

Table 2 IEC three ratio tables

What quote in the ※ table only is representative value, may have the ratio combination that is not included in the table in actual applications.

Claims

1, a kind of transformer on-line intelligence monitoring system, the sensor that comprises the various dynamic property monitoring devices of transformer, it is characterized in that by the electric signal that various sensor is gathered have through amplifier and modulate circuit handle back access analog to digital conversion circuit, have carry out data processing through modulus capture card, behind interface circuit, all inserting in the industrial computer of having by the foreground computing machine, draw the performance data that records, thereafter this digital signal is transported to backstage central processing unit transformer fault real-time diagnosis system and carries out diagnostic analysis, draw monitoring result.

2, transformer on-line intelligence monitoring system according to claim 1 is characterized in that the various dynamic property monitoring devices of transformer are partial discharge of transformer monitoring device, transformer iron core grounding current on-Line Monitor Device, transformer oil dissolved gas on-Line Monitor Device, bushing shell for transformer insulating property on-Line Monitor Device, coiling hot point of transformer temperature on-line monitoring device, oil tank of transformer water content in oil on-Line Monitor Device.

3, transformer on-line intelligence monitoring system according to claim 2, it is characterized in that the partial discharge of transformer monitoring device comprises sensor probe, the rodmeter at the leak valve place that is installed in oil tank of transformer fuel outlet valve place, bottom or top, rodmeter is connected with prime amplifier by the ultrahigh frequency cable, be connected with SHF receiver again, SHF receiver with signal change, processing is after the modulus capture card inserts data handling machine.

4, transformer on-line intelligence monitoring system according to claim 2, it is characterized in that the transformer iron core grounding current on-Line Monitor Device comprises the iron core grounding current extension line, this ground current extension line is connected with earth terminal after passing sensor, the induced signal of sensor inserts signal amplification circuit, connect the analog signal conditioner circuit again, the analog signal conditioner circuit is connected the back and inserts data handling machine with the modulus capture card.

5, transformer on-line intelligence monitoring system according to claim 2, it is characterized in that transformer oil dissolved gas on-Line Monitor Device comprises gas intelligence sensor in the oil that is installed in fuel outlet valve place, transformer bottom, the electric signal of intelligence sensor inserts the industrial computer that can show gas content through RS232 interface and cable transmission line, and industrial computer is connected with backstage diagnostic process computing machine.

6, transformer on-line intelligence monitoring system according to claim 2, it is characterized in that bushing shell for transformer insulating property on-Line Monitor Device comprises condenser type punching sensor (CT), the end screen ground wire of three-phase sleeve pipe passes condenser type punching sensor (CT) respectively, go out the end screen current signal of three-phase sleeve pipe through condenser type punching sensor sensing, induced signal is imported the modulate circuit plate through 4 core shielded cables, carry out linearity conditioning by modulate circuit, amplify (decay), Filtering Processing, export as three-phase casing current signal, simultaneously the three-phase output signal is input to the three-phase voltage summing circuit, obtain its three-phase imbalance voltage U 0, as the 4th output signal, with first three signal, by the DB37 shielded cable, be sent to modulus capture card (ADLINK PCI-9113), is digital signal through the modulus capture card with analog signal conversion, be sent to the foreground industrial computer, call the foreground capture program, digital signal is carried out software filtering and smoothing processing and digital signal carried out format conversion deposit in the computing machine of assigned address and demonstration.

7, transformer on-line intelligence monitoring system according to claim 2, it is characterized in that the coiling hot point of transformer temperature on-line monitoring device comprises the ring current transformer that transformer condenser type insulating sleeve afterbody is installed, with the leading-out terminal of this ring current transformer with a cable short circuit, this short-circuiting cable passes the load current sensor, the sensor current signal of load current sensor inserts signal conditioning circuit, modulate circuit is converted to standard signal with this current signal and inserts the foreground industrial computer, other has a probe that the temperature controller that is inserted into transformer upper temp meter seat is installed, its induced signal inserts signal conditioning circuit by cable, modulate circuit is converted to standard signal with this current signal and inserts the foreground industrial computer, and industrial computer is connected with data handling machine.

8, transformer on-line intelligence monitoring system according to claim 2, it is characterized in that comprising bottom flowline, intelligence sensor that one group of refrigeratory of transformer links to each other with transformer, with after the RS232 interface is connected with the foreground industrial computer with 3 core shielded cables, the foreground industrial computer can be connected with backstage diagnostic process computing machine by Ethernet the electric signal that intelligence sensor will be installed in the sensor probe of bottom on the flowline through amplification or decay, rectification, filtering and analog to digital conversion.

9, a kind of transformer on-line intelligence analysis and diagnosis method, be that various signal datas with transformer online monitoring enter in the analyzing and diagnosing computing machine, store the equipment nameplate parameter information that has at line transformer in this analyzing and diagnosing computing machine, the true empirical rule information of transformer in the past, the technical parameter rules and the standard of on-line monitoring, the technical parameter threshold value and the operating experience of on-line monitoring, have in the analyzing and diagnosing computing machine can according to automatic input come in the various signal datas of transformer online monitoring carry out Dynamic Data Processing and trend analysis, the reasoning of state comprehensive diagnos, the transformer state comprehensive assessment, support the final analysis and diagnosis system software of making a decision of user, by the analyzing and diagnosing computing machine according to automatic input come in the various signal datas of transformer online monitoring carry out the transformer line real time diagnosis in conjunction with the information stored database, the abnormal conditions of transformer are reported to the police.

10, transformer on-line intelligence analysis and diagnosis method according to claim 9, it is characterized in that the analyzing and diagnosing computing machine according to automatic input come in the various signal datas of transformer online monitoring carry out the transformer line real time diagnosis in conjunction with the information stored database, its process raw data that to be the first step transmit the various signal datas of transformer online monitoring is carried out pre-service, promptly reject wild point, smoothing processing, measures such as rate of change calculating are to slacken environmental factor and sensor itself, the influence that Network Transmission is brought, it is that follow-up sign is extracted reliable data that data are carried out various analyses, to the raw data that the is obtained processing of at first will sorting; Second step was that threshold value is asked for; The 3rd step was set up underlying database; The 4th step adopted the fault decision tree that transformer fault is diagnosed, and in the decision tree reasoning process, used three kinds of diagnostic methods: three ratio compiling methods of improvement; Traditional comparison overall approach; BP combination neural net method.