CN104034289A - Condition monitoring method and device for windings of power transformer - Google Patents

Abstract

The invention provides a condition monitoring method for windings of a power transformer. The method is applied to the power transformer which operates currently. Multiple vibrating sensors are arranged on the power transformer and used for collecting multiple sets of vibration signals of the windings in the power transformer within a preset time period, vibration amplitudes of preset base frequencies in all target vibration signals in each set of vibration signals are extracted, winding vibration variations between every two adjacent vibration amplitudes in each set of vibration amplitudes are obtained, a vibration correlation model is generated according to the winding vibration variations, and then principal component analysis is carried out on the vibration correlation model so that the conditions of the windings of the power transformer can be determined. Compared with the prior art, the conditions of the windings can be obtained by collecting and analyzing vibration data of the power transformer in the normal operation process without the need of stopping operation of the power transformer, on-line monitoring is achieved, and influences on operation of a whole electric system are small.

Description

A kind of Winding in Power Transformer state monitoring method and device

Technical field

The application relates to Power Transformer Faults detection technique field, relates in particular to a kind of Winding in Power Transformer state monitoring method and device.

Background technology

Power transformer, is the visual plant in electric system, and it connects generating plant and electric substation, can boosted voltage electric energy be sent to electricity consumption district, can by lower voltage, be also the voltages that use at different levels, to meet different needs for electricity.If power transformer breaks down, can interrupt power transmission network and cause serious economic loss and serious social influence.Therefore, need to monitor power transformer, to find in time and to fix a breakdown, guarantee the normal operation of electric system.

Transformer comprises winding and iron core, and wherein, frequency that winding breaks down is higher and harm is also larger.The form of expression of winding failure is winding deformation, particularly, is divided into radial deformation and axioversion.After winding deformation, Short Circuit withstand impulsive force sharply declines, and easily causes the serious consequence of winding interturn short-circuit.Therefore, the monitoring to Winding in Power Transformer fault, is mainly to judge whether winding exists distortion.

At present, conventional winding deformation monitoring method comprises frequency response method, the method is in winding one end, to inject swept-frequency signal, at the other end, measure the signal that winding passes over, according to the transport function of the calculated signals winding measuring, finally according to the variation of winding transport function before and after being subject to impacting, judge whether winding is out of shape.Yet the method need to stop the course of work of transformer, can not realize on-line monitoring.

Summary of the invention

In view of this, the application provides a kind of Winding in Power Transformer state monitoring method and device, can not realize the problem of on-line monitoring in order to solve existing monitoring method.

For achieving the above object, the application provides following technical scheme:

A Winding in Power Transformer state monitoring method, the power transformer that is applied to moving, is all furnished with at least two vibration transducers in the tank surface that in described power transformer, each winding is corresponding, and the method comprises:

Obtain one group of vibration signal that vibration transducer gathers separately in Preset Time section described in each;

Extract the vibration amplitude of presetting fundamental frequency described in every group in vibration signal in each target vibration signal, obtain many group vibration amplitudes; Wherein, described target vibration signal is according to the default definite vibration signal of sampling interval;

For every group of vibration amplitude, obtain the basket vibration variable quantity between every two adjacent vibration amplitudes in this group;

The basket vibration variable quantity described in each of take is foundation, generates vibration correlation models;

Described vibration correlation models is carried out to principal component analysis (PCA), to determine the winding state of described power transformer; Wherein, described winding state comprises normal condition or deformation state.

Said method, preferably, comprises basket vibration and core vibration in described vibration signal;

Correspondingly, extract the vibration amplitude of presetting fundamental frequency described in every group in vibration signal in each target vibration signal, comprising:

Utilize following formula (1), extract in vibration signal described in every group the default fundamental vibration amplitude of basket vibration in each target vibration signal;

α _w∝I ² (1)

Wherein, α _wfor the default fundamental vibration amplitude of basket vibration, I is default electric current corresponding to fundamental frequency;

Utilize following formula (2), extract in vibration signal described in every group the default fundamental vibration amplitude of core vibration in each target vibration signal;

α _c＝f(U) (2)

Wherein, α _cfor the default fundamental vibration amplitude of core vibration, f (U) is the function of default fundamental frequency corresponding voltage.

Said method, preferably, described for every group of vibration amplitude, obtain the basket vibration variable quantity between every two adjacent vibration amplitudes in this group, comprising:

Utilize following formula (3), according to the default fundamental vibration amplitude of described basket vibration, and the default fundamental vibration amplitude of described core vibration, obtain overall default fundamental vibration amplitude;

Wherein, α _tfor overall default fundamental vibration amplitude, for the vector angle between basket vibration and core vibration;

Utilize following formula (4), for every group of overall default fundamental vibration amplitude, obtain every two adjacent described basket vibration variable quantities between fundamental vibration amplitudes of totally presetting in this group;

$\mathrm{\Δw}=\sqrt{{\mathrm{\α}}_{t1}^2+{\mathrm{\α}}_{t2}^2-{2\mathrm{\α}}_{t1}{\mathrm{\α}}_{t2}\cos \mathrm{\β}}---\left(4\right)$

Wherein, Δ w is basket vibration variable quantity, α _t1, α _t2be two adjacent overall default fundamental vibration amplitudes, β is the phase place angle between described two overall default fundamental vibration amplitudes in current phase alignment situation.

Said method, preferably, the vibration correlation models of described generation is correlation matrix X, and

Wherein, described Δ w _nmfor basket vibration variable quantity corresponding to adjacent two vibration amplitudes, n is the sequence number of basket vibration variable quantity, the sequence number that m is vibration transducer.

Said method, preferably, describedly carries out principal component analysis (PCA) to described vibration correlation models, to determine the winding state of described power transformer, comprising:

Described correlation matrix X is normalized;

Utilize following formula (5), obtain covariance matrix C corresponding to correlation matrix X after described normalized _xxeigenwert;

$C_{\mathrm{xx}}=\frac{1}{m-1}{X}^{T}X=\mathrm{U\Λ}{U}^T---\left(5\right)$

Wherein, the proper vector of classifying described covariance matrix as comprising in U matrix, Λ is for getting diagonal element computing, obtains described in each proper vector characteristic of correspondence value λ separately ₁λ ₂λ _m, and described each eigenwert meets λ ₁>=λ ₂>=...>=λ _m;

Utilize following formula (6), according to described eigenwert, obtain the basket vibration relevance parameter MPC of described power transformer;

$\mathrm{MPC}={\mathrm{\λ}}_1/\underset{i=1}{\overset{m}{\mathrm{\Σ}}}{\mathrm{\λ}}_i---\left(6\right)$

When described basket vibration relevance parameter MPC surpasses parameter preset threshold value, determine that described winding is normal condition; Wherein, described parameter preset threshold value is to be greater than 0 numerical value that is less than 1;

When described basket vibration relevance parameter MPC does not surpass described parameter preset threshold value, determine that described winding is deformation state.

The application also provides a kind of Winding in Power Transformer state monitoring apparatus, and the power transformer that is applied to moving is all furnished with at least two vibration transducers in the tank surface that in described power transformer, each winding is corresponding, and this device comprises:

Vibration signal acquiring unit, for obtaining one group of vibration signal that vibration transducer gathers separately in Preset Time section described in each;

Vibration amplitude extraction unit, for extracting described in every group the vibration amplitude of the default fundamental frequency of each target vibration signal in vibration signal, obtains many group vibration amplitudes; Wherein, described target vibration signal is according to the default definite vibration signal of sampling interval;

Vibration variable quantity acquiring unit, for for every group of vibration amplitude, obtains the basket vibration variable quantity between every two adjacent vibration amplitudes in this group;

Correlation models generation unit, is foundation for take basket vibration variable quantity described in each, generates vibration correlation models;

Winding state determining unit, for carrying out principal component analysis (PCA) to described vibration correlation models, to determine the winding state of described power transformer; Wherein, described winding state comprises normal condition or deformation state.

Said apparatus, preferably, comprises basket vibration and core vibration in described vibration signal, described vibration amplitude extraction unit comprises:

Basket vibration magnitude extraction subelement, for utilizing following formula (1), extracts in vibration signal described in every group the default fundamental vibration amplitude of basket vibration in each target vibration signal;

α _w∝I ² (1)

Wherein, α _wfor the default fundamental vibration amplitude of basket vibration, I is default electric current corresponding to fundamental frequency;

Core vibration magnitude extraction subelement, for utilizing following formula (2), extracts in vibration signal described in every group the default fundamental vibration amplitude of core vibration in each target vibration signal;

α _c＝f(U) (2)

Wherein, α _cfor the default fundamental vibration amplitude of core vibration, f (U) is the function of default fundamental frequency corresponding voltage.

Said apparatus, preferably, described vibration variable quantity acquiring unit comprises:

Overall vibration amplitude is obtained subelement, for utilizing following formula (3), and according to the default fundamental vibration amplitude of described basket vibration, and the default fundamental vibration amplitude of described core vibration, obtain overall default fundamental vibration amplitude;

Wherein, α _tfor overall default fundamental vibration amplitude, for the vector angle between basket vibration and core vibration;

Vibration variable quantity obtains subelement, for utilizing following formula (4), for every group of overall default fundamental vibration amplitude, obtains every two adjacent described basket vibration variable quantities between fundamental vibration amplitudes of totally presetting in this group;

$\mathrm{\Δw}=\sqrt{{\mathrm{\α}}_{t1}^2+{\mathrm{\α}}_{t2}^2-{2\mathrm{\α}}_{t1}{\mathrm{\α}}_{t2}\cos \mathrm{\β}}---\left(4\right)$

Wherein, Δ w is basket vibration variable quantity, α _t1, α _t2be two adjacent overall default fundamental vibration amplitudes, β is the phase place angle between described two overall default fundamental vibration amplitudes in current phase alignment situation.

Said apparatus, preferably, the vibration correlation models that described correlation models generation unit generates is correlation matrix X, and

Wherein, described Δ w _nmfor basket vibration variable quantity corresponding to adjacent two vibration amplitudes, n is the sequence number of basket vibration variable quantity, the sequence number that m is vibration transducer.

Said apparatus, preferably, described winding state determining unit comprises:

Normalized subelement, for being normalized described correlation matrix X;

Proper value of matrix is obtained subelement, for utilizing following formula (5), obtains covariance matrix C corresponding to correlation matrix X after described normalized _xxeigenwert;

$C_{\mathrm{xx}}=\frac{1}{m-1}{X}^{T}X=\mathrm{U\Λ}{U}^T---\left(5\right)$

Wherein, the proper vector of classifying described covariance matrix as comprising in U matrix, Λ is for getting diagonal element computing, obtains described in each proper vector characteristic of correspondence value λ separately ₁λ ₂λ _m, and described each eigenwert meets λ ₁>=λ ₂>=...>=λ _m;

Relevance parameter is obtained subelement, for utilizing following formula (6), according to described eigenwert, obtains the basket vibration relevance parameter MPC of described power transformer;

$\mathrm{MPC}={\mathrm{\λ}}_1/\underset{i=1}{\overset{m}{\mathrm{\Σ}}}{\mathrm{\λ}}_i---\left(6\right)$

Winding is normally determined subelement, for when described basket vibration relevance parameter MPC surpasses parameter preset threshold value, determines that described winding is normal condition; Wherein, described parameter preset threshold value is to be greater than 0 numerical value that is less than 1;

Winding deformation is determined subelement, for when described basket vibration relevance parameter MPC does not surpass described parameter preset threshold value, determines that described winding is deformation state.

Compared with prior art, the present invention has following beneficial effect:

From above scheme, Winding in Power Transformer state monitoring method provided by the invention, the power transformer that is applied to moving, on this power transformer, be furnished with vibration transducer, obtain one group of vibration signal that vibration transducer gathers separately in Preset Time section described in each, and extract the vibration amplitude of presetting fundamental frequency described in every group in vibration signal in each target vibration signal, obtain many group vibration amplitudes; Wherein, described target vibration signal is according to the default definite vibration signal of sampling interval, for every group of vibration amplitude, obtain the basket vibration variable quantity between every two adjacent vibration amplitudes in this group, and to take described each basket vibration variable quantity be foundation, generate vibration correlation models, and then described vibration correlation models is carried out to principal component analysis (PCA), to determine the winding state of described power transformer.Compared with prior art, the method that the embodiment of the present invention provides does not need to stop the operation of power transformer, by the vibration data that gathers and analyze in its normal course of operation, can know winding state, realize on-line monitoring, less to the influence on system operation of whole electric system.

Accompanying drawing explanation

In order to be illustrated more clearly in the embodiment of the present invention or technical scheme of the prior art, to the accompanying drawing of required use in embodiment or description of the Prior Art be briefly described below, apparently, accompanying drawing in the following describes is only embodiments of the invention, for those of ordinary skills, do not paying under the prerequisite of creative work, other accompanying drawing can also be provided according to the accompanying drawing providing.

The process flow diagram of the Winding in Power Transformer state monitoring method that Fig. 1 provides for the embodiment of the present invention;

The layout schematic diagram of the vibration transducer that Fig. 2 provides for the embodiment of the present invention;

The another process flow diagram of the Winding in Power Transformer state monitoring method that Fig. 3 provides for the embodiment of the present invention;

Relation model figure between the winding fundamental vibration that Fig. 4 provides for the embodiment of the present invention, iron core fundamental vibration and overall fundamental vibration three;

The overall fundamental vibration amplitude of two samples that Fig. 5 provides for the embodiment of the present invention and the relation model figure between basket vibration variable quantity three;

Basket vibration relative coefficient schematic diagram corresponding to different power transformers that Fig. 6 provides for the embodiment of the present invention;

Between the basket vibration relative coefficient that Fig. 7 provides for the embodiment of the present invention and default sampling interval, be related to schematic diagram;

The layout schematic diagram of vibration transducer in the test experiments that Fig. 8 provides for the embodiment of the present invention;

The vibration signal schematic diagram of the normal winding that Fig. 9 provides for the embodiment of the present invention;

The vibration signal schematic diagram of the abnormal winding that Figure 10 provides for the embodiment of the present invention;

Basket vibration variable quantity schematic diagram corresponding to normal winding that Figure 11 provides for the embodiment of the present invention;

Basket vibration variable quantity schematic diagram corresponding to abnormal winding that Figure 12 provides for the embodiment of the present invention;

The structured flowchart of the Winding in Power Transformer state monitoring apparatus that Figure 13 provides for the embodiment of the present invention.

Embodiment

Below in conjunction with the accompanying drawing in the embodiment of the present invention, the technical scheme in the embodiment of the present invention is clearly and completely described, obviously, described embodiment is only the present invention's part embodiment, rather than whole embodiment.Embodiment based in the present invention, those of ordinary skills, not making the every other embodiment obtaining under creative work prerequisite, belong to the scope of protection of the invention.

Referring to Fig. 1, it shows the flow process of the Winding in Power Transformer state monitoring method that the embodiment of the present invention provides, the power transformer that the method is applied to moving, in the tank surface that in described power transformer, each winding is corresponding, be all furnished with at least two vibration transducers, as shown in Figure 2, power transformer includes three windings, and respectively corresponding two vibration transducers of each winding, particularly, winding A corresponding No. #1 and #2 vibration transducer, winding B corresponding No. #3 and #4 vibration transducer, winding C corresponding No. #5 and #6 vibration transducer.Wherein, the method comprises the following steps:

Step S101: obtain one group of vibration signal that vibration transducer collects in Preset Time section described in each.

Wherein, power transformer, under different load currents, has the vibration of different amplitudes.Described vibration signal is the vibration of the described power transformer that gathers of vibration transducer under different loads electric current, in Preset Time Duan Ruyi days, each vibration transducer according to the sample frequency of self can continuous acquisition to one group of vibration signal.

Step S102: extract the vibration amplitude of presetting fundamental frequency described in every group in vibration signal in each target vibration signal, obtain many group vibration amplitudes; Wherein, described target vibration signal is according to the default definite vibration signal of sampling interval.

Wherein, the one group of vibration signal collecting for each vibration transducer, every default sampling interval, chooses a vibration signal as sampled point, and this sampled point is defined as to target vibration signal.Alternatively, described default sampling interval is 5 minutes.In Chinese electric system, the sine wave that winding current is 50Hz, the default fundamental frequency of vibration signal is 100Hz.In each target vibration signal, by Fourier transform, extract the vibration amplitude of default fundamental frequency.

Step S103: for every group of vibration amplitude, obtain the basket vibration variable quantity between every two adjacent vibration amplitudes in this group.

Wherein, each vibration transducer is corresponding one group of vibration amplitude all, and the vibration amplitude that comprises a plurality of default fundamental frequencies in each group vibration amplitude,, in every group of amplitude, obtains the winding variable quantity between every two adjacent vibration amplitudes.It should be noted that, basket vibration variable quantity refers to the variable quantity of vibration amplitude between adjacent two vibration amplitudes.For example, comprise vibration amplitude 1, vibration amplitude 2, vibration amplitude 3, vibration amplitude 4 in one group of vibration amplitude, every two adjacent vibration amplitudes refer to: vibration amplitude 1 and 2, vibration amplitude 2 and 3, vibration amplitude 3 and 4.

Step S104: described each basket vibration variable quantity of take is foundation, generates vibration correlation models.

Wherein, corresponding two vibration transducers of each winding, that is to say, two vibration transducers can gather the vibration signal of same winding, have correlativity between vibration signal.In addition, the variable quantity that vibration variable quantity is connected between two sampled points, described variable quantity characterizes the vibration correlativity of former and later two adjacent time points.

Step S105: described vibration correlation models is carried out to principal component analysis (PCA), to determine the winding state of described power transformer; Wherein, described winding state comprises normal condition or deformation state.

Wherein, described principal component analysis (PCA) refers in judgement vibration correlation models whether have a major component component, if, the vibrational energy that vibration correlation models is described is concentrated, each winding of power transformer is in normal vibration state, otherwise, illustrate that vibrational energy disperses, and exists the winding of distortion in transformer.

From above technical scheme, the Winding in Power Transformer state monitoring method that the embodiment of the present invention provides, the power transformer that is applied to moving, on this power transformer, be furnished with a plurality of vibration transducers, to gather many groups vibration signal of winding in this power transformer in Preset Time section, extract the vibration amplitude of presetting fundamental frequency described in every group in vibration signal in each target vibration signal, obtain many group vibration amplitudes, for every group of vibration amplitude, obtain the basket vibration variable quantity between every two adjacent vibration amplitudes in this group, and to take described each basket vibration variable quantity be foundation, generate vibration correlation models, and then described vibration correlation models is carried out to principal component analysis (PCA), to determine the winding state of described power transformer.Compared with prior art, the method that the application provides does not need to stop the operation of power transformer, by the vibration data that gathers and analyze in its normal course of operation, can know winding state, realize on-line monitoring, less to the influence on system operation of whole electric system.

Meanwhile, the embodiment of the present invention does not need to carry out any electrical connection with transformer, utilizes basket vibration mechanical principle to realize the monitoring to winding state, even if slight distortion also can accurately detect, degree of accuracy is higher.

It should be noted that, the main part of power transformer is provided with winding and iron core, in iron core column, be set with winding, the vibration of power transformer can be thought to be formed by winding and the two-part vibration of iron core, therefore, in vibration signal, include basket vibration signal, also include core vibration signal.

Referring to Fig. 3, it shows the another flow process of the Winding in Power Transformer state monitoring method that the embodiment of the present invention provides, alternatively, the step S102 in said method embodiment extracts in vibration signal described in every group the vibration amplitude of default fundamental frequency in each target vibration signal and can realize in the following manner:

Step S202: utilize following formula (1), extract in vibration signal described in every group the default fundamental vibration amplitude of basket vibration in each target vibration signal;

α _w∝I ² (1)

Wherein, α _wfor the default fundamental vibration amplitude of basket vibration, I is default electric current corresponding to fundamental frequency.

Particularly, winding is the circuit part of transformer, and iron core is the magnetic circuit part of transformer, and basket vibration is produced by electromagnetic force, and the size of electromagnetic force and square being directly proportional by electric current in winding.In Chinese electric system, winding current is the sine wave of 50Hz, thereby electromagnetic force is the sinusoidal excitation power of 100Hz, and correspondingly, default fundamental frequency is 100Hz.If winding is equivalent to mass-spring system, the steady-state vibration frequency of winding is 100Hz, and meanwhile, the vibration amplitude of default fundamental frequency is directly proportional to current effective value square, concrete manifestation form (1) formula that sees above.

Step S203: utilize following formula (2), extract in vibration signal described in every group the default fundamental vibration amplitude of core vibration in each target vibration signal;

α _c＝f(U) (2)

Wherein, α _cfor the default fundamental vibration amplitude of core vibration, f (U) is the function of default fundamental frequency corresponding voltage.

Particularly, core vibration is that magneto-striction phenomenon by ferromagnetic material causes.Magneto-striction phenomenon can simply be expressed as: the variation ferromagnetic material with magnetic field changes shape and size.Power transformer magnetic field is directly related with the voltage at winding two ends, and because voltage keeps substantially invariable 50Hz frequency, the vibration amplitude of iron core under default fundamental frequency (100Hz) is the function relevant to voltage, concrete manifestation form (2) formula that sees above.

Correspondingly, the step S103 in said method embodiment is for vibration transducer described in each, and according to default sampling interval, obtaining every basket vibration variable quantity corresponding to adjacent two vibration amplitudes can realize by following manner:

Step S204: utilize following formula (3), according to the default fundamental vibration amplitude of described basket vibration, and the default fundamental vibration amplitude of described core vibration, obtain overall default fundamental vibration amplitude;

Wherein, α _tfor overall default fundamental vibration amplitude, for the vector angle between basket vibration and core vibration.

Concrete, the vibration on power transformer surface can be thought and mainly comes from winding and iron core, therefore, sets up the relation between winding fundamental vibration, iron core fundamental vibration and overall fundamental vibration three, described relation is (3) formula as above, and three's relational model can be referring to Fig. 4.

Step S205: utilize following formula (4), for every group of overall default fundamental vibration amplitude, obtain every two adjacent described basket vibration variable quantities between fundamental vibration amplitudes of totally presetting in this group;

$\mathrm{\Δw}=\sqrt{{\mathrm{\α}}_{t1}^2+{\mathrm{\α}}_{t2}^2-{{2\mathrm{\α}}_{t1}\mathrm{\α}}_{t2}\cos \mathrm{\β}}---\left(4\right)$

Wherein, Δ w is basket vibration variable quantity, α _t1, α _t2be two adjacent overall default fundamental vibration amplitudes, β is the phase place angle between described two overall default fundamental vibration amplitudes in current phase alignment situation.

Particularly, power transformer is when normal operation, and the voltage on winding remains unchanged substantially, and the phasing degree between electric current and voltage (power factor angle) also remains unchanged substantially.To preset sampling interval, as 5 minutes, determine the default fundamental vibration amplitude of sampling.Because the time interval of sampling is less, can think that the core vibration relevant to voltage is constant, and the vector angle between basket vibration and core vibration also remain unchanged, thereby obtain according to above-mentioned (3) formula the overall default fundamental vibration amplitude that each sampled point is corresponding.Further, according to above-mentioned (4) formula, obtaining the variable quantity between adjacent two overall default fundamental vibration amplitudes, is the basket vibration variable quantity between adjacent two sampled points.The overall default fundamental vibration amplitude of two samples is with the relation between variable quantity three vibrated as formula (4) above, and three's relational model can be referring to Fig. 5.

Alternatively, the vibration correlation models generating in said method embodiment step S104 is correlation matrix X, and

Wherein, described Δ w _nmfor basket vibration variable quantity corresponding to adjacent two vibration amplitudes, n is the sequence number of basket vibration variable quantity, the sequence number that m is vibration transducer.

Particularly, for each vibration transducer, according to default sampling interval, can obtain adjacent two overall default basket vibration variable quantities corresponding to fundamental vibration amplitude corresponding with this vibration transducer, described basket vibration variable quantity is generated to above-mentioned correlation matrix X, wherein, n represents to preset the number of the basket vibration variable quantity that sampling interval collects, and m represents total number of vibration transducer.

According to basket vibration, produce principle, each basket vibration variable quantity and current effective value square variable quantity be directly proportional, the concrete form of expression is:

From above-mentioned correlation matrix, can find out to there is correlativity between each row of this matrix.Therefore, correspondingly, said method embodiment step S105 carries out principal component analysis (PCA) to described vibration correlation models, to determine that the winding state of described power transformer can realize in the following manner:

Step S207: described correlation matrix X is normalized;

Step S208: utilize following formula (5), obtain covariance matrix C corresponding to correlation matrix X after described normalized _xxeigenwert;

$C_{\mathrm{xx}}=\frac{1}{m-1}{X}^{T}X=\mathrm{U\Λ}{U}^T---\left(5\right)$

Wherein, the proper vector of classifying described covariance matrix as comprising in U matrix, Λ is for getting diagonal element computing, obtains described in each proper vector characteristic of correspondence value λ separately ₁λ ₂λ _m, and described each eigenwert meets λ ₁>=λ ₂>=...>=λ _m.

In this enforcement, Λ=diag{ λ ₁λ ₂λ _m, for U matrix is got to diagonal element computing, thereby obtain each eigenwert, described each eigenwert is carried out to descending sort, that is to say λ ₁for maximum eigenwert.

Step S209: utilize following formula (6), according to described eigenwert, obtain the basket vibration relevance parameter MPC of described power transformer;

$\mathrm{MPC}={\mathrm{\λ}}_1/\underset{i=1}{\overset{m}{\mathrm{\Σ}}}{\mathrm{\λ}}_i---\left(6\right)$

Wherein, first each λ is added and tries to achieve and be worth, then by λ ₁divided by described and value, try to achieve described basket vibration relevance parameter, this formula is in order to represent maximum major component energy λ ₁account for the proportion of total energy.

Step S210: when described basket vibration relevance parameter MPC surpasses parameter preset threshold value, determine that described winding is normal condition; Wherein, described parameter preset threshold value is to be greater than 0 numerical value that is less than 1.

Step S211: when described basket vibration relevance parameter MPC does not surpass described parameter preset threshold value, determine that described winding is deformation state.

Wherein, parameter preset threshold value is to be greater than 0 numerical value that is less than 1, and preferably, this parameter preset threshold value is 0.8.The value of MPC can characterize winding state, when this value surpasses parameter preset threshold value, more close to 1 o'clock, illustrates that winding state is better, when this value does not surpass parameter preset threshold value, illustrates that winding exists distortion.

To basket vibration relative coefficient MPC, can represent that the principle of winding state is introduced below:

For the winding in normal condition in power transformer, according to quality spring damping model, reach a conclusion: α _w∝ I ², the default fundamental vibration amplitude of basket vibration and current effective value square is directly proportional.While there are the abnormal conditions such as distortion when winding, vibration signal there will be uncertainty, and therefore, the default fundamental vibration amplitude that basket vibration is corresponding also just no longer meets above-mentioned rule.

According to basket vibration, producing principle reaches a conclusion: being basket vibration variable quantity is directly proportional to the variable quantity of current effective value square, and from the angle of above-mentioned correlation matrix X, each row in this correlation matrix X are all proportional.Therefore, under winding normal condition, in this matrix, only have a major component component, and then the value of the basket vibration relative coefficient MPC of acquisition is also just larger, and more close to 1.Otherwise if winding is in being out of shape abnormality, the major component component of this matrix is not unique, and then the value of the basket vibration relative coefficient MPC obtaining is smaller.As can be seen here, whether this basket vibration relative coefficient can characterize the major component component of correlation matrix unique, and then whether reflection winding is in normal condition.

From above technical scheme, the embodiment of the present invention obtains overall default fundamental vibration amplitude by the default fundamental vibration amplitude of basket vibration and the default fundamental vibration amplitude of core vibration, generate the basket vibration variable quantity between every two adjacent overall default fundamental vibration amplitudes, and then generation correlation matrix, this correlation matrix is carried out obtaining basket vibration relevance parameter after principal component analysis (PCA), described basket vibration relative coefficient and parameter preset threshold value are compared to the rear final state of determining Winding in Power Transformer.

It should be noted that, about the parameter preset threshold value in above-mentioned each embodiment of the method, inventor adds up many experiments result, obtain the statistical form shown in Fig. 6, as can be seen from the figure, for dissimilar power transformer, vibration relative coefficient in normal condition winding nearly all surpasses threshold value 0.8, therefore, preferably, parameter preset threshold value is 0.8.

In addition, the accuracy of above-mentioned basket vibration relative coefficient MPC is subject to the impact of default sampling interval, and inventor finds relation between the two as shown in Figure 7 through overtesting.Shown in Figure 7, monitoring total duration is 300s, when default sampling interval too small, as while being less than 1s, the basket vibration relative coefficient of normal winding does not trend towards 1, and the basket vibration relative coefficient of abnormal winding unstable, the larger error that can cause like this, measurement.Therefore, need to select applicable default sampling interval, alternatively, can select to be more than or equal to the numerical value of 1s.Certainly, in actual monitoring process, can, according to the default sampling interval of the corresponding selection of the length of monitoring time, if monitoring time is one day, can be set as 5 minutes.

Below by one group of complete experimentation, Winding in Power Transformer state monitoring method provided by the invention is introduced.

For the validity of verification method, the vibration performance by same winding under normal condition and failure condition is compared.Wherein, fault winding carries out short-circuit impact realization by normal winding.Particularly, transformer short-circuit test is a kind of specially for the method for testing of winding, and it passes through low pressure end short circuit, and at high-pressure side making alive, makes Transformer Winding electric current reach ratings.

Experimental subjects is the three-phase oil-immersed power transformer of a 110KV, select the higher vibration transducer of sensitivity, and, in order to guarantee the vibratory response of vibration transducer within sampling filter frequency band, by the mode that magnetic support adsorbs or glue is bonding, vibration transducer is fixed on power transformer fuel tank sidewall.Particularly, vibration transducer comprises the modules such as preposition amplification, anti-aliasing filter, AD sampling, wherein, at least 12 of AD sampling resolutions, frequency overlapped-resistable filter cutoff frequency is 2000Hz.In experiment, sample frequency is 10000Hz, and AD sampling resolution is 16, and adopts continuous sampling pattern to record the overall process of experiment.

(1) arrange that vibration transducer is as vibration measuring point.

As shown in Figure 8, on power transformer fuel tank surface, arrange 7 measuring points.Because each winding of transformer all needs to be covered by least two vibration transducers, arrangement shown in Fig. 8 can not guarantee that winding C is capped, therefore, and after once having tested, adjust the position of vibration transducer to winding C place, to measure the vibration signal of winding C.Particularly, see Fig. 8, after A phase winding is destroyed by short circuit, as the winding of abnormality, on oil tank wall corresponding to this phase winding, be furnished with five vibration transducers, B phase winding is furnished with two, usings as the A contrast winding of abnormal winding mutually.

(2) vibration signal of acquisition abnormity winding and normal winding.

First, power transformer is carried out to short circuit experiment: progressively improve on high-tension side voltage to increase the electric current on winding, and the ratio that every primary current increases is 10%, until ratings.In addition, after the each increase of electric current, keep 30s to stablize constant.Vibration transducer utilizes continuous sampling pattern, records the vibration signal of each winding.As shown in Figure 9, vibration transducer #4 collects the vibration signal of winding under normal condition.

Then, A phase winding is carried out to short-circuit impact so that it produces distortion, and again power transformer is carried out the short circuit experiment of said process, record equally the vibration signal of each winding.As shown in figure 10, vibration transducer #4 collects the vibration signal of winding under abnormal (distortion) state.

(3) obtain the basket vibration variable quantity between sample in each vibration transducer.

The vibration signal collecting for each sensor, utilizes the relational model shown in Fig. 4, obtains the vibration amplitude of overall default fundamental frequency (100Hz).And then, according to default sampling interval, as 1s, in described each totally default fundamental vibration amplitude, determine sampled point, and utilize the relational model shown in Fig. 5, obtain the basket vibration variable quantity between adjacent two samplings.It should be noted that, in short circuit experiment, owing to there is no core vibration, the β angle shown in Fig. 5 is 0.

Above-mentioned steps gathers the vibration signal in 300 seconds in (2) altogether, and default sampling interval is 1s, can obtain 299 basket vibration variable quantities, and 50 that choose are wherein carried out analytic statistics.See Figure 11, it shows 50 basket vibration variable quantities that normal winding is corresponding, sees Figure 12, and it shows 50 basket vibration variable quantities that abnormal winding is corresponding.By comparison Figure 11 and Figure 12, can find out, under normal condition, the basket vibration variable quantity consistance that each measuring point obtains is better, and under abnormality, basket vibration variable quantity consistance is poor.

(4) utilize each basket vibration variable quantity, calculate basket vibration relative coefficient.

First, above-mentioned each basket vibration variable quantity is generated to correlation matrix, and this matrix is carried out to zero-mean and normalized.Then, the matrix after normalized is carried out to principal component analysis (PCA), obtain the eigenwert of covariance matrix, particularly, the eigenwert between each composition is as shown in following table table 1.Utilize described eigenwert to calculate vibration relative coefficient.

Table 1

	λ1	λ2	λ3	λ4	λ5	λ6	λ7
								Normal winding	109	3.1	0.58	0.23	0.09	0.03	0
Abnormal winding	76.6	20.9	10.6	2.4	1.2	0.72	0.56

Although from above-mentioned Figure 11 and Figure 12, the basket vibration situation of the normal condition that can visually see and abnormality, step (4) is calculated the basket vibration relative coefficient obtaining and is presented experimental result with the form quantizing.

The Winding in Power Transformer the state monitoring apparatus below embodiment of the present invention being provided is introduced, and it should be noted that, the explanation of relevant Winding in Power Transformer state monitoring apparatus refers to above-mentioned Winding in Power Transformer status method, at this, does not repeat.

Referring to Figure 13, it shows the structure of the Winding in Power Transformer state monitoring apparatus that the embodiment of the present invention provides, this application of installation is in the power transformer moving, in the tank surface that in described power transformer, each winding is corresponding, be all furnished with at least two vibration transducers, as shown in Figure 2, power transformer includes three windings, and respectively corresponding two vibration transducers of each winding, particularly, winding A corresponding No. #1 and #2 vibration transducer, winding B corresponding No. #3 and #4 vibration transducer, winding C corresponding No. #5 and #6 vibration transducer.This device comprises:

Vibration signal acquiring unit 100, for obtaining one group of vibration signal that vibration transducer gathers separately in Preset Time section described in each;

Vibration amplitude extraction unit 200, for extracting described in every group the vibration amplitude of the default fundamental frequency of each target vibration signal in vibration signal, obtains many group vibration amplitudes; Wherein, described target vibration signal is according to the default definite vibration signal of sampling interval;

Vibration variable quantity acquiring unit 300, for for every group of vibration amplitude, obtains the basket vibration variable quantity between every two adjacent vibration amplitudes in this group;

Correlation models generation unit 400, is foundation for take basket vibration variable quantity described in each, generates vibration correlation models;

Winding state determining unit 500, for carrying out principal component analysis (PCA) to described vibration correlation models, to determine the winding state of described power transformer; Wherein, described winding state comprises normal condition or deformation state.

The Winding in Power Transformer state monitoring apparatus that the embodiment of the present invention provides, the power transformer that is applied to moving, on this power transformer, be furnished with a plurality of vibration transducers, to gather many groups vibration signal of winding in this power transformer in Preset Time section, extract the vibration amplitude of presetting fundamental frequency described in every group in vibration signal in each target vibration signal, obtain many group vibration amplitudes, for every group of vibration amplitude, obtain the basket vibration variable quantity between every two adjacent vibration amplitudes in this group, and to take described each basket vibration variable quantity be foundation, generate vibration correlation models, and then described vibration correlation models is carried out to principal component analysis (PCA), to determine the winding state of described power transformer.Compared with prior art, the device that the application provides does not need to stop the operation of power transformer, by the vibration data that gathers and analyze in its normal course of operation, can know winding state, realize on-line monitoring, less to the influence on system operation of whole electric system.

Alternatively, comprise basket vibration and core vibration in described vibration signal, described vibration amplitude extraction unit comprises:

Basket vibration magnitude extraction subelement, for utilizing following formula (1), extracts in vibration signal described in every group the default fundamental vibration amplitude of basket vibration in each target vibration signal;

α _w∝I ² (1)

Wherein, α _wfor the default fundamental vibration amplitude of basket vibration, I is default electric current corresponding to fundamental frequency;

Core vibration magnitude extraction subelement, for utilizing following formula (2), extracts in vibration signal described in every group the default fundamental vibration amplitude of core vibration in each target vibration signal;

α _c＝f(U) (2)

Wherein, α _cfor the default fundamental vibration amplitude of core vibration, f (U) is the function of default fundamental frequency corresponding voltage.

Alternatively, described vibration variable quantity acquiring unit comprises:

Overall vibration amplitude is obtained subelement, for utilizing following formula (3), and according to the default fundamental vibration amplitude of described basket vibration, and the default fundamental vibration amplitude of described core vibration, obtain overall default fundamental vibration amplitude;

Wherein, α _tfor overall default fundamental vibration amplitude, for the vector angle between basket vibration and core vibration;

Vibration variable quantity obtains subelement, for utilizing following formula (4), for every group of overall default fundamental vibration amplitude, obtains every two adjacent described basket vibration variable quantities between fundamental vibration amplitudes of totally presetting in this group;

$\mathrm{\Δw}=\sqrt{{\mathrm{\α}}_{t1}^2+{\mathrm{\α}}_{t2}^2-{{2\mathrm{\α}}_{t1}\mathrm{\α}}_{t2}\cos \mathrm{\β}}---\left(4\right)$

Wherein, Δ w is basket vibration variable quantity, α _t1, α _t2be two adjacent overall default fundamental vibration amplitudes, β is the phase place angle between described two overall default fundamental vibration amplitudes in current phase alignment situation.

Alternatively, the vibration correlation models that described correlation models generation unit generates is correlation matrix X, and

Wherein, described Δ w _nmfor basket vibration variable quantity corresponding to adjacent two vibration amplitudes, n is the sequence number of basket vibration variable quantity, the sequence number that m is vibration transducer.

Alternatively, described winding state determining unit comprises:

Normalized subelement, for being normalized described correlation matrix X;

Proper value of matrix is obtained subelement, for utilizing following formula (5), obtains covariance matrix C corresponding to correlation matrix X after described normalized _xxeigenwert;

$C_{\mathrm{xx}}=\frac{1}{m-1}{X}^{T}X=\mathrm{U\Λ}{U}^T---\left(5\right)$

Wherein, the proper vector of classifying described covariance matrix as comprising in U matrix, Λ is for getting diagonal element computing, obtains described in each proper vector characteristic of correspondence value λ separately ₁λ ₂λ _m, and described each eigenwert meets λ ₁>=λ ₂>=...>=λ _m;

Relevance parameter is obtained subelement, for utilizing following formula (6), according to described eigenwert, obtains the basket vibration relevance parameter MPC of described power transformer;

$\mathrm{MPC}={\mathrm{\λ}}_1/\underset{i=1}{\overset{m}{\mathrm{\Σ}}}{\mathrm{\λ}}_i---\left(6\right)$

Winding is normally determined subelement, for when described basket vibration relevance parameter MPC surpasses parameter preset threshold value, determines that described winding is normal condition; Wherein, described parameter preset threshold value is to be greater than 0 numerical value that is less than 1;

Winding deformation is determined subelement, for when described basket vibration relevance parameter MPC does not surpass described parameter preset threshold value, determines that described winding is deformation state.

Claims (10)

1. a Winding in Power Transformer state monitoring method, is characterized in that, the power transformer that is applied to moving is all furnished with at least two vibration transducers in the tank surface that in described power transformer, each winding is corresponding, and the method comprises:

Obtain one group of vibration signal that vibration transducer gathers separately in Preset Time section described in each;

Extract the vibration amplitude of presetting fundamental frequency described in every group in vibration signal in each target vibration signal, obtain many group vibration amplitudes; Wherein, described target vibration signal is according to the default definite vibration signal of sampling interval;

For every group of vibration amplitude, obtain the basket vibration variable quantity between every two adjacent vibration amplitudes in this group;

The basket vibration variable quantity described in each of take is foundation, generates vibration correlation models;

Described vibration correlation models is carried out to principal component analysis (PCA), to determine the winding state of described power transformer; Wherein, described winding state comprises normal condition or deformation state.

2. method according to claim 1, is characterized in that, comprises basket vibration and core vibration in described vibration signal;

Correspondingly, extract the vibration amplitude of presetting fundamental frequency described in every group in vibration signal in each target vibration signal, comprising:

Utilize following formula (1), extract in vibration signal described in every group the default fundamental vibration amplitude of basket vibration in each target vibration signal;

α _w∝I ² (1)

Wherein, α _wfor the default fundamental vibration amplitude of basket vibration, I is default electric current corresponding to fundamental frequency;

Utilize following formula (2), extract in vibration signal described in every group the default fundamental vibration amplitude of core vibration in each target vibration signal;

α _c＝f(U) (2)

Wherein, α _cfor the default fundamental vibration amplitude of core vibration, f (U) is the function of default fundamental frequency corresponding voltage.

3. method according to claim 2, is characterized in that, described for every group of vibration amplitude, obtains the basket vibration variable quantity between every two adjacent vibration amplitudes in this group, comprising:

Utilize following formula (3), according to the default fundamental vibration amplitude of described basket vibration, and the default fundamental vibration amplitude of described core vibration, obtain overall default fundamental vibration amplitude;

Wherein, α _tfor overall default fundamental vibration amplitude, for the vector angle between basket vibration and core vibration;

Utilize following formula (4), for every group of overall default fundamental vibration amplitude, obtain every two adjacent described basket vibration variable quantities between fundamental vibration amplitudes of totally presetting in this group;

$\mathrm{\Δw}=\sqrt{{\mathrm{\α}}_{t1}^2+{\mathrm{\α}}_{t2}^2-{{2\mathrm{\α}}_{t1}\mathrm{\α}}_{t2}\cos \mathrm{\β}}---\left(4\right)$

Wherein, Δ w is basket vibration variable quantity, α _t1, α _t2be two adjacent overall default fundamental vibration amplitudes, β is the phase place angle between described two overall default fundamental vibration amplitudes in current phase alignment situation.

4. method according to claim 1, is characterized in that, the vibration correlation models of described generation is correlation matrix X, and

Wherein, described Δ w _nmfor basket vibration variable quantity corresponding to adjacent two vibration amplitudes, n is the sequence number of basket vibration variable quantity, the sequence number that m is vibration transducer.

5. method according to claim 4, is characterized in that, described described vibration correlation models is carried out to principal component analysis (PCA), to determine the winding state of described power transformer, comprising:

Described correlation matrix X is normalized;

Utilize following formula (5), obtain covariance matrix C corresponding to correlation matrix X after described normalized _xxeigenwert;

$C_{\mathrm{xx}}=\frac{1}{m-1}{X}^{T}X=\mathrm{U\Λ}{U}^T---\left(5\right)$

Wherein, the proper vector of classifying described covariance matrix as comprising in U matrix, Λ is for getting diagonal element computing, obtains described in each proper vector characteristic of correspondence value λ separately ₁λ ₂λ _m, and described each eigenwert meets λ ₁>=λ ₂>=...>=λ _m;

Utilize following formula (6), according to described eigenwert, obtain the basket vibration relevance parameter MPC of described power transformer;

$\mathrm{MPC}={\mathrm{\λ}}_1/\underset{i=1}{\overset{m}{\mathrm{\Σ}}}{\mathrm{\λ}}_i---\left(6\right)$

When described basket vibration relevance parameter MPC surpasses parameter preset threshold value, determine that described winding is normal condition; Wherein, described parameter preset threshold value is to be greater than 0 numerical value that is less than 1;

When described basket vibration relevance parameter MPC does not surpass described parameter preset threshold value, determine that described winding is deformation state.

6. a Winding in Power Transformer state monitoring apparatus, is characterized in that, the power transformer that is applied to moving is all furnished with at least two vibration transducers in the tank surface that in described power transformer, each winding is corresponding, and this device comprises:

Vibration signal acquiring unit, for obtaining one group of vibration signal that vibration transducer gathers separately in Preset Time section described in each;

Vibration amplitude extraction unit, for extracting described in every group the vibration amplitude of the default fundamental frequency of each target vibration signal in vibration signal, obtains many group vibration amplitudes; Wherein, described target vibration signal is according to the default definite vibration signal of sampling interval;

Vibration variable quantity acquiring unit, for for every group of vibration amplitude, obtains the basket vibration variable quantity between every two adjacent vibration amplitudes in this group;

Correlation models generation unit, is foundation for take basket vibration variable quantity described in each, generates vibration correlation models;

Winding state determining unit, for carrying out principal component analysis (PCA) to described vibration correlation models, to determine the winding state of described power transformer; Wherein, described winding state comprises normal condition or deformation state.

7. device according to claim 6, is characterized in that, comprises basket vibration and core vibration in described vibration signal, and described vibration amplitude extraction unit comprises:

Basket vibration magnitude extraction subelement, for utilizing following formula (1), extracts in vibration signal described in every group the default fundamental vibration amplitude of basket vibration in each target vibration signal;

α _w∝I ² (1)

Wherein, α _wfor the default fundamental vibration amplitude of basket vibration, I is default electric current corresponding to fundamental frequency;

Core vibration magnitude extraction subelement, for utilizing following formula (2), extracts in vibration signal described in every group the default fundamental vibration amplitude of core vibration in each target vibration signal;

α _c＝f(U) (2)

Wherein, α _cfor the default fundamental vibration amplitude of core vibration, f (U) is the function of default fundamental frequency corresponding voltage.

8. device according to claim 7, is characterized in that, described vibration variable quantity acquiring unit comprises:

Overall vibration amplitude is obtained subelement, for utilizing following formula (3), and according to the default fundamental vibration amplitude of described basket vibration, and the default fundamental vibration amplitude of described core vibration, obtain overall default fundamental vibration amplitude;

Wherein, α _tfor overall default fundamental vibration amplitude, for the vector angle between basket vibration and core vibration;

Vibration variable quantity obtains subelement, for utilizing following formula (4), for every group of overall default fundamental vibration amplitude, obtains every two adjacent described basket vibration variable quantities between fundamental vibration amplitudes of totally presetting in this group;

$\mathrm{\Δw}=\sqrt{{\mathrm{\α}}_{t1}^2+{\mathrm{\α}}_{t2}^2-{-2{\mathrm{\α}}_{t1}\mathrm{\α}}_{t2}\cos \mathrm{\β}}---\left(4\right)$

Wherein, Δ w is basket vibration variable quantity, α _t1, α _t2be two adjacent overall default fundamental vibration amplitudes, β is the phase place angle between described two overall default fundamental vibration amplitudes in current phase alignment situation.

9. device according to claim 6, is characterized in that, the vibration correlation models that described correlation models generation unit generates is correlation matrix X, and

Wherein, described Δ w _nmfor basket vibration variable quantity corresponding to adjacent two vibration amplitudes, n is the sequence number of basket vibration variable quantity, the sequence number that m is vibration transducer.

10. device according to claim 9, is characterized in that, described winding state determining unit comprises:

Normalized subelement, for being normalized described correlation matrix X;

Proper value of matrix is obtained subelement, for utilizing following formula (5), obtains covariance matrix C corresponding to correlation matrix X after described normalized _xxeigenwert;

$C_{\mathrm{xx}}=\frac{1}{m-1}{X}^{T}X=\mathrm{U\Λ}{U}^T---\left(5\right)$

Wherein, the proper vector of classifying described covariance matrix as comprising in U matrix, Λ is for getting diagonal element computing, obtains described in each proper vector characteristic of correspondence value λ separately ₁λ ₂λ _m, and described each eigenwert meets λ ₁>=λ ₂>=...>=λ _m;

Relevance parameter is obtained subelement, for utilizing following formula (6), according to described eigenwert, obtains the basket vibration relevance parameter MPC of described power transformer;

$\mathrm{MPC}={\mathrm{\λ}}_1/\underset{i=1}{\overset{m}{\mathrm{\Σ}}}{\mathrm{\λ}}_i---\left(6\right)$

Winding is normally determined subelement, for when described basket vibration relevance parameter MPC surpasses parameter preset threshold value, determines that described winding is normal condition; Wherein, described parameter preset threshold value is to be greater than 0 numerical value that is less than 1;

Winding deformation is determined subelement, for when described basket vibration relevance parameter MPC does not surpass described parameter preset threshold value, determines that described winding is deformation state.