CN113899968B - Voltage transformer monitoring method - Google Patents

Abstract

The invention discloses a voltage transformer monitoring method, which comprises the following steps: s1, data acquisition and analysis are carried out on a tested transformer, and the data acquisition and analysis are divided into the following two cases: A. when the tested transformer is provided with the grounding quantity collection, calculating the metering error of the transformer and analyzing the insulation state according to the secondary quantity and the grounding quantity; B. when the tested transformer does not have the ground quantity acquisition or the ground leakage quantity acquisition fails and the data fluctuation quantity is large, and the data cannot be calculated by the individual transformer error, adopting a mode of joint analysis of a plurality of transformers; s2, comprehensively evaluating the relevance of metering error fluctuation and insulation state: and synchronously evaluating the metering and the insulation state by adopting a neural network algorithm, and evaluating the reason of the metering out-of-tolerance according to the insulation index. The invention discloses a voltage transformer monitoring method without collecting high-voltage side voltage and current signals, which is compatible with electromagnetic and capacitive voltage transformers and solves a plurality of problems in the existing voltage transformer metering error monitoring.

Description

Voltage transformer monitoring method

Technical Field

The invention relates to a transformer, in particular to a voltage transformer monitoring method.

Background

The transformer is used as important power equipment, and has the characteristics of a large number of maintenance tasks, equipment faults affecting metering and protection, and the like. Early transformer monitoring is mainly means of insulation monitoring, partial secondary electrical parameter monitoring and secondary combined monitoring.

The secondary combination mainly comprises the steps of synchronously collecting the current or voltage of the high-voltage side and synchronously comparing the current or voltage with the current or voltage of the secondary loop; however, since the high voltage needs to be contacted once, the insulation is more studied, and the technology is generally only used for a low-voltage transformer.

With the development of artificial intelligence technology and statistical computing technology, related researches based on relative metering errors of secondary voltage analysis are developed in recent years, but the researches mainly stay at the aspect of more remarkable secondary voltage abnormality at present, and a mathematical model is not well established for the relationship between the secondary voltage and the errors; part of research results synchronously collect secondary voltages of a plurality of transformers and analyze confidence intervals, so that data deviating from a target confidence interval can be found, but the method is limited by the number of samples, and does not well process abnormal voltage fluctuation (such as various interference waveforms introduced by switch closing, load adjustment, lightning waves, operation waves, secondary oscillation and overhaul) existing in normal system operation, wherein the abnormal waves are not characteristics of faults or defects of the transformers and are easy to form misguidance on the existing statistical calculation method; in addition, the choice of the opposite communication interval in the statistical algorithm does not form a calculation method related to the measured data; the mechanism of the error overrun of the transformer is not provided with timely excavation and analysis of the overrun cause due to the lack of synchronous insulating parameter analysis means, and the fault can be accumulated and extended, so that the fault phenomenon can only be found only by the method of carrying out power failure and disassembly test on the transformer after the overrun, and the real source of the fault can not be reasonably explained. There is thus a need for an improvement to the above practical problems.

Disclosure of Invention

The invention aims to overcome the defects of the prior art, provides a voltage transformer monitoring method which does not need to collect high-voltage side voltage and current signals and solves a plurality of problems faced by the existing voltage transformer metering error monitoring.

The aim of the invention is realized by the following technical scheme: a voltage transformer monitoring method comprising the steps of:

s1, data acquisition and analysis are carried out on a tested transformer, and the data acquisition and analysis are divided into the following two cases:

A. when the tested transformer is provided with the grounding quantity collection, calculating the metering error of the transformer and analyzing the insulation state according to the secondary quantity and the grounding quantity;

B. When the tested transformer does not have the ground quantity collection or the ground quantity collection fails and the data fluctuation quantity is large, and the data cannot be calculated by the individual transformer errors, a mode of joint analysis of a plurality of transformers is adopted;

s2, comprehensively evaluating the relevance of metering error fluctuation and insulation state: and synchronously evaluating the metering and the insulation state by adopting a neural network algorithm, and evaluating the reason of the metering out-of-tolerance according to the insulation index.

Further, the tested transformer in the step S1 includes an electromagnetic voltage transformer and a capacitive voltage transformer.

When the tested transformer is provided with the grounding quantity collection, the collection and analysis modes are as follows:

the method comprises the following steps of firstly, collecting secondary quantity and grounding quantity of a measured transformer and auxiliary quantity; the auxiliary quantity comprises temperature, humidity and vibration data;

If the tested transformer is an electromagnetic voltage transformer, collecting at least one path of secondary voltage and one path of overall grounding current of the voltage transformer when collecting secondary quantity and grounding quantity;

If the measurement transformer is a capacitive voltage transformer, collecting at least one path of secondary voltage, one path of integral grounding leakage current and one path of low-voltage capacitance grounding current when collecting secondary quantity and grounding quantity; or at least collecting one path of secondary voltage, one path of low-voltage capacitor grounding current and one path of step-down transformer primary grounding current;

And secondly, respectively calculating the error of the voltage transformer and evaluating the insulation state according to the collected secondary quantity and the ground quantity.

Further, when the measured voltage transformer is an electromagnetic voltage transformer, the acquisition and analysis method comprises the following steps:

(a) The waveform Ig of the secondary voltage V2 and the whole grounding current of the transformer is collected at fixed time, and the amplitude, the phase and the frequency spectrum are calculated; temperature, humidity, vibration data;

(b) Calculating the metering instantaneous error: windowing is carried out on the V2 and Ig waveform data, and the direct current components V2dc and Igdc of the V2 and Ig and the ratio kdc =v2dc/Igdc of the direct current components V2dc and Igdc are calculated by adopting wavelet or short-time Fourier transform STFT; when Igdc is carried out, V2dc is increased by more than 10% compared with the fault-free historical average data or kdc is smaller than 90% of the fault-free historical average data, kdc is marked as an abnormal value, and the line is considered to have operation overvoltage or thunder overvoltage or line resonance, so that the voltage V2 is not used as a basis for calculating errors, and collected V2 data is abandoned; when the ratio V2/Ig of the effective value or root mean square value V2 and Ig is increased by more than 10% compared with the fault-free historical average value, the secondary voltage of the voltage transformer is regarded as open circuit, the voltage V2 is not used as the basis of metering error, the collected data of V2 are abandoned, when Igdc, the absolute value of the difference between V2dc and the historical value which is lower than the historical average value 110% or kdc is larger than the historical average value 95%, kdc is recorded as a normal value, and the phase difference between V2 and Ig is calculated The ratio of V2 to ig×m is different, vr 0= (V2 x Nr-Ig x M)/Ig x M = V2 x Nr/(Ig x M) -1 = kg x Nr/M-1,/>Dielectric loss/>Nr is a rated transformation ratio or a transformation ratio value of a history test, M is a reference insulation resistance value or a history test value, kg=v2/Ig; when the M value cannot be determined, vr0 cannot be calculated, and (c) is entered, and the comparison difference and insulation assessment are completed by (c) + (d).

(C) Calculating the measurement mean error offset:

Obtaining kdc as normal data, secondary voltages V2N, ground currents Ign, n=1, 2 … N at a plurality of time points, calculating kgn =V2n/Ign, recording the phase angle of the V2 voltage as θ2n, recording the phase angle of the ground current Ig as θgn, Calculating a normal distribution P (kg 0) of kgn based on kg0, δkg with a reference standard value V20, ig0 amplitude ratio kg0 = V20/Ig0 as a standard deviation of kgn;

based on the magnitude ratio of the mean value of V2n, kgn 0=v2avg/Igavg, as the expected value μ, δkg is the standard deviation of kgn, the normal distribution P (kgn) is calculated for kgn.

Respectively with reference valueAverage values are respectively taken as expected values mu,/>Standard deviation/>Calculation/>Obtain/>

Setting a confidence interval, calculating a specific difference offset DeltaP=P (kgn) -P (kg 0) and recording the specific difference offset as the specific difference offset; offset of angular difference

When the confidence interval cannot be set, carrying out Fourier transformation on the amplitude or phase angle difference of N (N > 8) secondary voltages V2N and the grounding current Ign to obtain frequency domain data, calculating the proportion of the maximum absolute value of the frequency domain data to the arithmetic sum of all the data, wherein the proportion corresponds to the duty ratio coefficients Sv2 and Sign corresponding to the secondary voltages and the grounding currents, and taking the minimum value in the [ Sv2 and Sign ] as the confidence interval.

(D) Carrying in the absolute quantity of the ratio difference and the angular difference, and calculating the absolute quantity of the ratio difference and the angular difference of the current running state;

And D1, recording the power failure state verification ratio difference and angle difference data, namely when the power failure data is lost, adopting the power failure data mean value of the ratio difference and angle difference of other voltage transformers with the same specification in the same operation substation environment to replace the power failure data mean value, recording the obtained power failure verification ratio difference and angle difference data or the calculated ratio difference and angle difference data as Vr0 and delta 0, and then calculating the ratio difference V2=Vr0+DeltaP and the angle difference delta=δ0+Deltadelta of the current voltage transformer to be tested. Wherein Δp, Δδ is the data obtained in step (c);

D2: when the ratio difference and the angle difference obtained by the D1 are in the standard value range, continuously increasing vibration data or when the vibration amplitude is larger than the history normal value, considering that the external bolt of the voltage transformer is loosened; when the ratio difference and the angle difference obtained by the D1 exceed the standard value range, and when the vibration data is continuously increased or the vibration amplitude is larger than the history normal value, the voltage transformer is considered to have overvoltage, discharge or insulation abnormality;

(e) Taking environmental influence into consideration, historical data of ratio difference, angle difference, temperature, humidity and vibration of the voltage transformer are analyzed at fixed time, and normal distribution is calculated

And drawing a normal distribution curve based on the ratio difference, the angle difference, the temperature, the humidity and the vibration.

When the ratio difference and the angle difference are over-limit, the temperature and humidity data are in the historical average value or 95% confidence interval, and the voltage transformer is considered to be over-limit in metering error due to internal reasons; when the ratio difference and the angle difference are over-limited, at least one of the temperature, the humidity and the vibration data exceeds the historical average value or the 95% confidence interval, the voltage transformer is considered to have measurement errors which are over-limited due to the influence of the environment or the over-limitation of the errors caused by external insulation or internal insulation.

Further, when the measured voltage transformer is a capacitive voltage transformer, the collecting and analyzing method comprises the following steps:

(a) Waveforms of a secondary voltage V2, a secondary capacitor grounding current IC2, an exciting transformer primary grounding current IP1 and a transformer integral grounding current Ig are collected safely at fixed time, and amplitudes, phases and frequency spectrums of the V2, the IC2, the IP1 and the Ig are calculated; temperature, humidity, vibration data;

(b) Judging whether a fault exists or not;

b1, calculating the voltage, the ratio difference and the angle difference of a high-voltage capacitor end of the capacitive transformer;

Firstly, calculating the current vector value of the high-voltage capacitor, wherein when the acquisition of the Ig fails, fails or has no data, the current vector value of the high-voltage capacitor is calculated through IC2 and IP1, and the current vector value of the high-voltage capacitor is calculated through IC1=IC2+IP1, namely the amplitude IC1 and the phase of the high-voltage current are obtained through vector addition

Secondly, calculating the amplitude and the phase of the capacitor terminal voltages VC1 and VC2, wherein the calculation formula is as follows:

Vc=ic×zc, zc=1/(2×3.141592×f×c), f is the calculated frequency value, C is the high voltage capacitor C1, the nominal value of the nameplate of the high voltage capacitor C2 or the historical verification data; IC is a low-voltage capacitor IC2, high-voltage power Rong Dianliu ic1=ig or ic1=ic2+ip1.

Zc1＝1/(2*3.141592*f×C1)，Zc2＝1/(2*3.141592*f×C2)。

VC1＝IC1×Zc1,VC2＝IC2×Zc2,

The line voltage is then calculated: vline=vc1+vc2;

the line voltage Vline phase angle is recorded as

And finally, calculating a ratio difference: vr= (Nr V2-VLINE)/VLINE

Angular difference: Is the phase angle of the secondary voltage; nr is the rated transformation ratio or the transformation ratio value of the historical test.

B2, judging whether partial discharge phenomenon exists according to harmonic content:

When harmonic parameters appear in the frequency spectrum of the capacitor grounding current IC2 in a plurality of acquisition time periods and the harmonic content is more than 2% of the total frequency spectrum, or harmonic parameters appear in the Ig frequency spectrum and the harmonic content is more than 3% of the total frequency spectrum, judging that the partial discharge phenomenon occurs in the capacitor voltage transformer to be tested; or when harmonic parameters appear in the Ig frequency spectrum, the harmonic content is more than 3% of the total frequency spectrum, and the IP1 harmonic content is more than the IC2 harmonic content, judging that the partial discharge phenomenon occurs in the step-down transformer; otherwise, entering b3;

b3: based on whether there is a fault in the phase angle difference:

calculating phase angle differences delta vc of V2 and IC2, and phase angle differences delta vp of V2 and IP 1; the phase angle difference δcp of IC2, IP 1; if IC2 is greater than 10% of rated value and |90- δvc| is less than the threshold value, this indicates that the high voltage capacitor C1 has a capacity drop; if IP1 is more than 20% of rated value and |90-delta cp| is less than a threshold value, indicating that the primary winding of the step-down transformer or the smoothing compensation reactor or the lightning arrester or the high-voltage capacitor C1 fails; when IP1 is larger than 20% of rated value and |90-delta cp| is larger than the threshold value, the step-down transformer is proved to be faulty;

b4: estimating capacitance

Carrying out time domain to frequency domain conversion on waveforms of secondary voltage V2 of the transformer, or grounding current IC2 of the capacitor of the voltage transformer, or grounding current IP1 of primary side of the step-down transformer or grounding current Ig of the whole transformer, searching whether a stable non-power frequency signal fr exists in the range of 20Hz-150Hz on frequency domain data, if fr exists, calculating deviation delta f of fr and power frequency signal f0, wherein the power frequency signal f0 takes a frequency value corresponding to the maximum gain of 50Hz or 60Hz or the frequency domain data:

When IC2 deviates by more than 5% from the standard value, ig or V2 deviates by less than 2% from the standard value, the capacitor C2 is considered to deviate; when IC2 is deviating, C2 is reduced, and delta C is smaller than 0; when IC2 deviates negatively, C2 increases, and delta C is more than 0; Δc=0, fr=f0;

first, the standard value f0, the nominal capacitance of C2, Calculating an L value:

then the L value, the C2 value, the Fr value instead of f0 is taken into the following equation,

In the equation, C2 directly replaces C, and then only Δc needs to be calculated to obtain the latest c2' value=c2+Δc;

When IC2 is deviating more than 5% from the standard value, ig or V2 deviates more than 2% from the standard value, it is considered that the reduction of the high voltage capacitor C1 results in the reduction of the transformation ratio N of the voltage transformer, resulting in the rise of the output voltage and the rise of IC 2; it is assumed that the high-voltage capacitor and the low-voltage capacitor do not fail at the same time, and therefore the situation belongs to the failure of the high-voltage capacitor;

the same method is used next for calculating C1.

Firstly, the standard value f0 is brought in, C is the nominal capacitance of the high-voltage power supply C1,Calculate L' =l value.

Then the L' value, the C1 value, the Fr expects the f0 value to be brought into the following equation,

Δc' is calculated. The latest high-voltage capacitor C1 'has a value of C1+ΔC';

The above standard values refer to design manufacturing values or historical test data or average values of fault-free historical operating data.

It follows that the key point of the above algorithm is the image frequency fr, and the amount of change in capacitance can be calculated from the equation only when the image frequency is present, i.e. fr+.f0. When no image frequency exists, the capacitance is considered to be unchanged, and the estimated capacitance value is the capacitance of the prior known test.

Of course, the precondition of the above algorithm is that C1 and C2 do not shift in magnitude at the same time, wherein the calculated L inductance value is also an equivalent value to C1 and C2, and the calculated equivalent inductance value is different when C1 and C2 are calculated. Obviously, if the capacitance is determined or assumed to be unchanged and an accurate capacitance value is obtained, the same equation can be carried into f0 and C to calculate the fault-free inductance L, and then carried into fr and C to calculate the offset of the equivalent inductance. Since the equivalent inductance is mainly from the electromagnetic unit and the compensating reactor, the offset of the equivalent inductance can be used to find defects or faults of the electromagnetic unit and the compensating reactor.

When the tested transformer does not have the ground quantity collection or the ground leakage quantity collection fails and the data fluctuation quantity is large, and the data cannot be calculated by individual transformers, a mode of joint analysis of a plurality of transformers with the same voltage level is adopted, namely, the probability distribution calculation error of secondary voltages of the plurality of voltage transformers is adopted, and the method specifically comprises the following steps:

(1) Setting the synchronously acquired transformer signals as Ak, wherein k=1, 2 … N, k is the data of a plurality of transformers acquired at the same time point, and N is the maximum number of acquisition time points; ak is the phase and amplitude of the secondary voltage of the electromagnetic voltage transformer or the secondary voltage of the capacitive voltage transformer and the C2 current waveform signal; or the phase and amplitude of waveform signals of secondary voltage and ground current Ig of the capacitive voltage transformer;

(2) Performing quantization processing on the Ak to obtain Bk, wherein Bk=Ak/Akmax, akmax is the maximum value in the Ak, so that the quantized range of the Bk is 0-1, and then performing windowing processing on the Bk by applying a window function to perform time domain-frequency domain transformation to obtain frequency domain characteristic data Ck; taking the modulus value of Ck to obtain the absolute value Ck, and calculating the sum of the modulus values Searching the maximum value | Ckmax | in the |Ck|, calculating the ratio p= | Ckmax |/SAU, when p > = lambda, considering that the whole data is in a good state, when p < lambda considers that the data fluctuation is increased, and taking p as a confidence interval when abnormal data exists; the lambda value interval is 0.8000-0.9999, and the default value is 0.9600.

The window function is applied to Bk to perform windowing, and any one of a Hamming window, a Hanning window, a Gaussian window, a Blackman window, a Black-Harries window, a Kernel window, a flat top window, a rectangular window and a triangular window is adopted.

(3) The Ak normal distribution is then calculated.

(4) Calculating Ak data and Ak average value of the corresponding confidence interval p as error offset; the obtained data includes:

Electromagnetic voltage transformer: the magnitude offset and the angle offset of the secondary voltage V2 are denoted as PT [. DELTA.ε1,. DELTA.ε2], and DELTA.ε1 is regarded as the specific difference offset, and DELTA.ε2 is the angular difference offset.

Capacitive voltage transformer: the magnitude offset of the secondary voltage V2, IC2 or Ig is denoted as [ DELTAepsilon 1, DELTAepsilon 2]; v2 angular offset, IC2 or Ig angular offset [. DELTA.epsilon.3,. DELTA.epsilon.4 ], then calculating the ratio and angular difference offsets:

differential offset: delta epsilon 1-Delta epsilon 2 (direct subtraction arithmetic difference)

Angular difference offset: delta epsilon 3-Delta epsilon 4 (direct subtraction arithmetic difference)

The idea here of calculating the offset of the ratio difference and the angular difference is that there is a deviation in the synchronicity of the secondary voltage and the ground in response to the primary voltage change, which is reflected at least in the ratio difference or the angular difference.

And when the calculated ratio difference and the calculated angular difference offset exceed the specified value of the precision grade of the voltage transformer, the error is regarded as overrun.

Further, the step S2 includes:

A. Neural network rack for establishing following parameters

A1, when the tested equipment is a universal voltage transformer:

the input quantity is as follows: v2peak, V2rms, V2dc, igpeak, igrms, frg,Mfg,N

The output is: verr, δerr, Z, tan delta, PD, YN

A2. when the tested equipment is a capacitive voltage transformer:

The input quantity is as follows: v2peak, V2rms, V2dc, ic2, ip1, Igpeak,Igrms,frg,Mfg,C10,C20,N

The output is: verr, δerr, Z, C2, C1, tan delta, PD, YN

Wherein V2peak = secondary voltage peak; v2rms=secondary voltage effective value; v2dc=secondary voltage dc voltage; igpeak =voltage transformer ground current peak value, igrms =voltage transformer ground current effective value, frg =mirror frequency of current Ig, is non-system power frequency signal of frequency of second largest gain in 20-150Hz range; mfg is the ratio of the sum of other frequency gains greater than the power frequency signal to the sum of all frequency gains, so 0= < Mfg < = 1;

ic2 is the effective current value or root mean square value or peak value of the low-voltage capacitor of the capacitive voltage transformer;

ip1 is the primary grounding current of the electromagnetic transformer of the capacitive voltage transformer;

Is the phase angle of the secondary voltage V2,/> For the phase angle of the low-voltage current IC2,/>The Ig phase angle of the grounding current;

C10 is nameplate or high voltage power supply value of historical test;

C20 is the nameplate or the historically tested low voltage power value.

A low voltage power output by the C2 neural network;

the high-voltage power output by the C1 neural network;

Whether YN has insulation abnormality or overall state abnormality or not, if so, taking 1, and if not, taking 0;

PD partial discharge capacity, unit pc or mV or dB;

B. Training the samples by adopting a neural network algorithm, and building training sample frames with different input amounts and output amounts under the condition of known output amounts, wherein the number of the training samples is not less than 3; if the input quantity is partially missing, the input quantity is marked as 0 value or takes other uniformly fixed numbers, and if the output quantity is known to be partially missing, the input quantity is marked as 0 value or takes other uniformly fixed numbers;

C. according to the input quantity monitored on line, invoking the trained neural network algorithm in the step B to calculate the output quantity, observing the measurement error data and the insulation indexes of the discharge quantity and capacitance according to the output quantity, comparing the synchronism of the error overrun and the insulation abnormal parameter, and explaining the reason of the error overrun;

The basic classification is as follows:

(c1) When the metering error and the dielectric loss are out of standard at the same time and YN=1, judging that the metering is out of tolerance caused by aging of an insulating medium, wetting of the medium or sealing defect;

(c2) When the metering error and the partial discharge amount PD exceed the standard at the same time, and YN=0, judging that the error exceeds the standard and is caused by insulation discharge or discharge temperature rise;

(c3) When the metering error and the partial discharge amount PD exceed the standard at the same time and YN=1, judging that the error exceeds the standard and is caused by heat loss and aging caused by insulation discharge or accumulation of temperature rise of discharge;

(c4) When the insulation resistance Z is normal and the metering error exceeds the limit, but PD exceeds the limit, YN=0, and the metering error exceeds the limit caused by discharge or vibration gaps caused by dirt on the surface of the transformer or poor contact of wiring terminals of high and low voltage layers of the transformer is judged.

The beneficial effects of the invention are as follows: (1) The problem of algorithm defect caused by single synchronous data distribution monitoring of secondary voltages of a plurality of transformers is solved; because the standard value is lacking in the samples, the confidence interval can be changed when any voltage transformer fails, and even if the number of the samples is increased, the individual samples are obviously changed, failure equipment cannot be sensitively captured in the total samples, and only the difference of the whole confidence and probability distribution can be reflected; (the analysis method can be attributed to a lateral statistical algorithm)

(2) The problem of longitudinal data analysis of a plurality of time single devices is solved;

(3) The method solves the problem that the confidence interval calculation error is large due to insufficient sample of the monitored transformer.

(4) Solving the problem of the value of the optimal confidence interval;

(5) The problem of the timing risk of analysis many mutual-inductors secondary excessively rely on the GPS system is solved, in case GPS system trouble appears, like interference, shielding, satellite signal malfunction, monitoring terminal's GPS module inefficacy, or under the condition of not disposing GPS synchronous monitoring, still can obtain error data through the algorithm of this patent.

(6) The correlation analysis problem of errors and insulation is solved; and timely and reasonable explanation is provided for the reasons of error out-of-tolerance. And can be used for continuous on-line monitoring, short-time on-line monitoring or portable device on-site live line inspection, laboratory simulation error testing, simulation, training systems and the like of the transformer.

Drawings

FIG. 1 is a flow chart of the method of the present invention.

Detailed Description

The technical solution of the present invention will be described in further detail with reference to the accompanying drawings, but the scope of the present invention is not limited to the following description.

As shown in fig. 1, a voltage transformer monitoring method includes the following steps:

s1, data acquisition and analysis are carried out on a tested transformer, and the data acquisition and analysis are divided into the following two cases:

A. when the tested transformer is provided with the grounding quantity collection, calculating the metering error of the transformer and analyzing the insulation state according to the secondary quantity and the grounding quantity;

B. When the tested transformer does not have the ground quantity collection or the ground quantity collection fails and the data fluctuation quantity is large, and the data cannot be calculated by the individual transformer errors, a mode of joint analysis of a plurality of transformers is adopted;

s2, comprehensively evaluating the relevance of metering error fluctuation and insulation state: and synchronously evaluating the metering and the insulation state by adopting a neural network algorithm, and evaluating the reason of the metering out-of-tolerance according to the insulation index.

The tested transformer in the step S1 comprises an electromagnetic voltage transformer and a capacitive voltage transformer.

When the tested transformer is provided with the grounding quantity collection, the collection and analysis modes are as follows:

the method comprises the following steps of firstly, collecting secondary quantity and grounding quantity of a measured transformer and auxiliary quantity; the auxiliary quantity comprises temperature, humidity and vibration data;

If the tested transformer is an electromagnetic voltage transformer, collecting at least one path of secondary voltage and one path of overall grounding current of the voltage transformer when collecting secondary quantity and grounding quantity;

If the measurement transformer is a capacitive voltage transformer, collecting at least one path of secondary voltage, one path of integral grounding leakage current and one path of low-voltage capacitance grounding current when collecting secondary quantity and grounding quantity; or at least collecting one path of secondary voltage, one path of low-voltage capacitor grounding current and one path of step-down transformer primary grounding current;

And secondly, respectively calculating the error of the voltage transformer and evaluating the insulation state according to the collected secondary quantity and the ground quantity.

Further, when the measured voltage transformer is an electromagnetic voltage transformer, the acquisition and analysis method comprises the following steps:

(a) The waveform Ig of the secondary voltage V2 and the whole grounding current of the transformer is collected at fixed time, and the amplitude, the phase and the frequency spectrum are calculated; temperature, humidity, vibration data;

(b) Calculating the metering instantaneous error: windowing is carried out on the V2 and Ig waveform data, and the direct current components V2dc and Igdc of the V2 and Ig and the ratio kdc =v2dc/Igdc of the direct current components V2dc and Igdc are calculated by adopting wavelet or short-time Fourier transform STFT; when Igdc is carried out, V2dc is increased by more than 10% compared with the fault-free historical average data or kdc is smaller than 90% of the fault-free historical average data, kdc is marked as an abnormal value, and the line is considered to have operation overvoltage or thunder overvoltage or line resonance, so that the voltage V2 is not used as a basis for calculating errors, and collected V2 data is abandoned; when the ratio V2/Ig of the effective value or root mean square value V2 and Ig is increased by more than 10% compared with the fault-free historical average value, the secondary voltage of the voltage transformer is regarded as open circuit, the voltage V2 is not used as the basis of metering error, the collected data of V2 are abandoned, when Igdc, the absolute value of the difference between V2dc and the historical value which is lower than the historical average value 110% or kdc is larger than the historical average value 95%, kdc is recorded as a normal value, and the phase difference between V2 and Ig is calculated The ratio difference, angle difference, dielectric loss of V2 and IgxM, the corresponding algorithm is as follows:

the ratio difference Vr 0= (V2 Nr-Ig M)/(Ig M) =v2 Nr/(Ig M) -1=kg Nr/M-1

Angular difference

Dielectric loss

Nr is a rated transformation ratio or a transformation ratio value of a history test, M is a reference insulation resistance value or a history test value, kg=v2/Ig;

In the embodiment of the application, the measured CVT is of precision 0.2 level, voltage class 220kV, setting m=11gohm, v2=58.11v, nr= 3666.7, Ig＝19.3mA。

The ratio difference= (58.11×3666.7-0.0193×11000000)/(0.0193×11000000) = 0.00364 =0.364%.

Dielectric loss

From the error data analysis, the voltage transformer was greater than 0.2%, which has been out of tolerance.

(C) Calculating the measurement mean error offset:

Obtaining kdc as normal data, secondary voltages V2N, ground currents Ign, n=1, 2 … N at a plurality of time points, calculating kgn =V2n/Ign, recording the phase angle of the V2 voltage as θ2n, recording the phase angle of the ground current Ig as θgn, Calculating a normal distribution P (kg 0) of kgn based on kg0, δkg with a reference standard value V20, ig0 amplitude ratio kg0 = V20/Ig0 as a standard deviation of kgn;

Based on the magnitude ratio of the mean value of V2n, kgn 0=v2avg/Igavg, as the expected value μ, δkg is the standard deviation of kgn, the normal distribution P (kgn) is calculated for kgn.

Respectively with reference valueAverage value is the expected value mu,/>Standard deviation/>Calculating corresponding normal distribution to obtain/>

Setting a confidence interval, calculating a specific difference offset DeltaP=P (kgn) -P (kg 0) and recording the specific difference offset as the specific difference offset; offset of angular difference

When the confidence interval cannot be set, carrying out Fourier transformation on the amplitude or phase angle difference of N (N > 8) secondary voltages V2N and the grounding current Ign to obtain frequency domain data, calculating the proportion of the maximum absolute value of the frequency domain data to the arithmetic sum of all the data, obtaining the corresponding duty ratio coefficients Sv2, sign of the corresponding secondary voltages and grounding currents, and taking the minimum value in the [ Sv2, sign ] as the confidence interval.

The normal distribution algorithm is as follows

The above equation: y represents V2, ig, P (y) is a probability density equation, P (a < y < b) is a probability distribution function, mu is an expected value of V20 calculated corresponding to P0, ig0 or V2a calculated corresponding to Pa, iga data, a mean value or a set standard value is taken, delta is a standard deviation of V2n and Ign, and a and b are the amplitude and the phase of V2n, ign or the ratio V2n/Ign of the V2n and Ign of the corresponding confidence interval to be calculated respectively;

let the confidence interval be 95%,. DELTA.P=P (kgn) -P (kg 0) =0.00035 (no units); Unit division

If it is uncertain whether 95% is the best confidence interval, a fourier transform method can be adopted, the number n=64 of collected V2N and Ign is set, the gain in the frequency domain range is obtained after fourier transform is performed, the maximum value V2N and the SUM of all arithmetic gains are recorded as SUM (V2N), then V2N/SUM (V2N) =0.966.

The reason for selecting the minimum value is to fully consider the boundary effect, and consider the maximum transformation threshold value as a confidence interval, so that the sensitivity of the diagnostic error offset can be improved.

(D) Carrying in the absolute quantity of the ratio difference and the angular difference, and calculating the absolute quantity of the ratio difference and the angular difference of the current running state;

And D1, recording the power failure state verification ratio difference and angle difference data, namely when the power failure data is lost, adopting the power failure data mean value of the ratio difference and angle difference of other voltage transformers with the same specification in the same operation substation environment to replace the power failure data mean value, recording the obtained power failure verification ratio difference and angle difference data or the calculated ratio difference and angle difference data as Vr0 and delta 0, and then calculating the ratio difference V2=Vr0+DeltaP and the angle difference delta=δ0+Deltadelta of the current voltage transformer to be tested. Wherein Δp, Δδ is the data obtained in step (c);

D2: when the ratio difference and the angle difference obtained by the D1 are in the standard value range, continuously increasing vibration data or when the vibration amplitude is larger than the history normal value, considering that the external bolt of the voltage transformer is loosened; when the ratio difference and the angle difference obtained by the D1 exceed the standard value range, and when the vibration data is continuously increased or the vibration amplitude is larger than the history normal value, the voltage transformer is considered to have overvoltage, discharge or insulation abnormality;

(e) Repeating the above process for multiple voltage transformers, analyzing the ratio difference, angle difference, temperature, humidity and vibration data of multiple voltage transformers with the same specification at regular time, and calculating normal distribution

And drawing a probability curve, a ratio difference, an angle difference, a temperature and humidity curve based on time. When the ratio difference and the angle difference do not exceed the limit, the probability distribution functions Pv and the deviation DeltaS of Pdelta are calculated respectively corresponding to the ratio difference and the angle difference in the same time period;

When the delta S has positive and negative polarity changes in a certain time period, taking an arithmetic average value as delta Savg;

When the ratio difference and the angle difference are over-limited, the temperature and humidity curve of the corresponding time point is in the historical average value or 95% confidence interval, and the metering error of the voltage transformer is considered to be over-limited; when the ratio difference and the angle difference are over-limited, at least one of the temperature, the humidity and the vibration data of the corresponding time points exceeds the historical average value or the 95% confidence interval, the voltage transformer is considered to have metering error over-limit and insulation abnormality. When the ratio difference and the angle difference have overrun, and the absolute value of the deviation DeltaSy-DeltaSavg of the probability distribution function value of the ratio difference and the angle difference is larger than 0.1 and kdc is normal, the potential insulation hazard or insulation fault is considered to occur; in embodiments of the application

Setting the time period 1 to collect 100 times of data, calculating a ratio difference curve and an angle difference curve, and calculating a ratio deviation distribution function value Pv= 0.9829 and an angle difference distribution function value Pdelta=0.9779; Δs=0.005.

Setting the temperature and the humidity to be in a 95% confidence interval, wherein the ratio difference and the angle difference are not overrun;

if at least one of the ratio difference or the angle difference exceeds the limit and the temperature and the humidity are in the 95% interval, the metering error of the voltage transformer is considered to exceed the limit; when at least one of the ratio difference and the angle difference is out of the 95% confidence interval, the ratio difference and the angle difference have little meaning, and the end point is to analyze the insulation condition.

Further, when the measured voltage transformer is a capacitive voltage transformer, the collecting and analyzing method comprises the following steps:

(a) Waveforms of a secondary voltage V2, a secondary capacitor grounding current IC2, an exciting transformer primary grounding current IP1 and a transformer integral grounding current Ig are collected safely at fixed time, and amplitudes, phases and frequency spectrums of the V2, the IC2, the IP1 and the Ig are calculated; temperature, humidity, vibration data;

(b) Judging whether a fault exists or not;

b1, calculating the voltage, the ratio difference and the angle difference of a high-voltage capacitor end of the capacitive transformer;

Firstly, calculating the current vector value of the high-voltage capacitor, wherein when the acquisition of the Ig fails, fails or has no data, the current vector value of the high-voltage capacitor is calculated through IC2 and IP1, and the current vector value of the high-voltage capacitor is calculated through IC1=IC2+IP1, namely the amplitude IC1 and the phase of the high-voltage current are obtained through vector addition

Secondly, calculating the amplitude and the phase of the capacitor terminal voltages VC1 and VC2, wherein the calculation formula is as follows:

Vc=ic×zc, zc=1/(2×3.141592×f×c), f is the calculated frequency value, C is the high voltage capacitor C1, the nominal value of the nameplate of the high voltage capacitor C2 or the historical verification data; IC is a low-voltage capacitor IC2, high-voltage power Rong Dianliu ic1=ig or ic1=ic2+ip1.

Zc1＝1/(2*3.141592*f×C1)，Zc2＝1/(2*3.141592*f×C2)。

VC1＝IC1×Zc1,VC2＝IC2×Zc2,

The line voltage is then calculated: vline=vc1+vc2;

the line voltage Vline phase angle is recorded as

And finally, calculating a ratio difference: vr= (Nr V2-VLINE)/VLINE

Angular difference: Is the phase angle of the secondary voltage; nr is the rated transformation ratio or the transformation ratio value of the historical test.

B2, judging whether partial discharge phenomenon exists according to harmonic content:

When harmonic parameters appear in the frequency spectrum of the capacitor grounding current IC2 in a plurality of acquisition time periods and the harmonic content is more than 2% of the total frequency spectrum, or harmonic parameters appear in the Ig frequency spectrum and the harmonic content is more than 3% of the total frequency spectrum, judging that the partial discharge phenomenon occurs in the capacitor voltage transformer to be tested; or when harmonic parameters appear in the Ig frequency spectrum, the harmonic content is more than 3% of the total frequency spectrum, and the IP1 harmonic content is more than the IC2 harmonic content, judging that the partial discharge phenomenon occurs in the step-down transformer; otherwise, entering b3;

b3: based on whether there is a fault in the phase angle difference:

Calculating phase angle differences delta vc of V2 and IC2, and phase angle differences delta vp of V2 and IP 1; the phase angle difference δcp of IC2, IP 1; if IC2 is greater than 10% of rated value and |90- δvc| is less than the threshold value, this indicates that the high voltage capacitor C1 has a capacity drop; if IP1 is more than 20% of rated value and |90-delta cp| is less than a threshold value, indicating that the primary winding of the step-down transformer or the smoothing compensation reactor or the lightning arrester or the high-voltage capacitor C1 fails; when IP1 is larger than 20% of rated value and |90- δcp|, δvp is larger than a threshold value, indicating that the step-down transformer has faults;

b4: estimating capacitance

Carrying out time domain to frequency domain conversion on waveforms of secondary voltage V2 of the transformer, or grounding current IC2 of the capacitor of the voltage transformer, or grounding current IP1 of primary side of the step-down transformer or grounding current Ig of the whole transformer, searching whether a stable non-power frequency signal fr exists in the range of 20Hz-150Hz on frequency domain data, if fr exists, calculating deviation delta f of fr and power frequency signal f0, wherein the power frequency signal f0 takes a frequency value corresponding to the maximum gain of 50Hz or 60Hz or the frequency domain data:

When IC2 deviates by more than 5% from the standard value, ig or V2 deviates by less than 2% from the standard value, the capacitor C2 is considered to deviate; when IC2 is deviating, C2 is reduced, and delta C is smaller than 0; when IC2 deviates negatively, C2 increases, and delta C is more than 0; Δc=0, fr=f0;

first, the standard value f0, the nominal capacitance of C2, Calculating an L value:

then the L value, the C2 value, the Fr value is taken to the following equation,

Calculating delta C, wherein the latest C2' value is C2+delta C;

When IC2 is deviating more than 5% from the standard value, ig or V2 deviates more than 2% from the standard value, it is considered that the reduction of the high voltage capacitor C1 results in the reduction of the transformation ratio N of the voltage transformer, resulting in the rise of the output voltage and the rise of IC 2; it is assumed that the high-voltage capacitor and the low-voltage capacitor do not fail at the same time, and therefore the situation belongs to the failure of the high-voltage capacitor;

firstly, the standard value f0 is brought in, C is the nominal capacitance of the high-voltage power supply C1, The L' value is calculated.

Then the L' value, the C1 value, the Fr value is taken to the following equation,

Δc' is calculated. The latest high-voltage capacitor C1 'has a value of C1+ΔC';

The above standard values refer to design manufacturing values or historical test data or average values of fault-free historical operating data.

When the tested transformer does not have the ground quantity collection or the ground leakage quantity collection fails and the data fluctuation quantity is large, and the data cannot be calculated by individual transformers, a mode of joint analysis of a plurality of transformers with the same voltage level is adopted, namely, the probability distribution calculation error of secondary voltages of the plurality of voltage transformers is adopted, and the method specifically comprises the following steps:

(1) Setting the synchronously acquired transformer signals as Ak, wherein k=1, 2 … N, k is the data of a plurality of transformers acquired at the same time point, and N is the maximum number of acquisition time points; ak is the phase and amplitude of the secondary voltage of the electromagnetic voltage transformer or the secondary voltage of the capacitive voltage transformer and the C2 current waveform signal;

(2) Performing quantization processing on the Ak to obtain Bk, wherein Bk=Ak/Akmax, akmax is the maximum value in the Ak, so that the quantized range of the Bk is 0-1, and then performing windowing processing on the Bk by applying a window function to perform time domain-frequency domain transformation to obtain frequency domain characteristic data Ck; taking the modulus value of Ck to obtain the absolute value Ck, and calculating the sum of the modulus values Searching the maximum value | Ckmax | in the |Ck|, calculating the ratio p= | Ckmax |/SAU, when p > = lambda, considering that the whole data is in a good state, when p < lambda considers that the data fluctuation is increased, and taking p as a confidence interval when abnormal data exists; the lambda value interval is 0.8000-0.9999, and the default value is 0.9600. /(I)

(3) The Ak normal distribution is then calculated.

(4) Calculating Ak data and Ak average value of the corresponding confidence interval p as error offset; the obtained data includes:

Electromagnetic voltage transformer: the magnitude offset and the angle offset of the secondary voltage V2 are denoted as PT [. DELTA.ε1,. DELTA.ε2], and DELTA.ε1 is regarded as the specific difference offset, and DELTA.ε2 is the angular difference offset.

Capacitive voltage transformer: the magnitude offset of the secondary voltage V2 and the magnitude offset of the IC2 are marked as [ delta epsilon 1, [ delta epsilon 2]; v2 angular offset, IC2 angular offset [ Δε3, Δε4], then calculating the ratio and angular difference offsets:

Differential offset: delta epsilon 1-Delta epsilon 2

Angular difference offset: delta epsilon 3-Delta epsilon 4

And when the calculated ratio difference and the calculated angular difference offset exceed the specified value of the precision grade of the voltage transformer, the error is regarded as overrun.

Further, the step S2 includes:

A. Neural network rack for establishing following parameters

A1, when the tested equipment is a universal voltage transformer:

the input quantity is as follows: v2peak, V2rms, V2dc, igpeak, igrms, frg,Mfg,N

The output is: verr, δerr, Z, tan delta, PD, YN

A2. when the tested equipment is a capacitive voltage transformer:

The input quantity is as follows: v2peak, V2rms, V2dc, ic2, ip1, Igpeak,Igrms,frg,Mfg,C10,C20,N

The output is: verr, δerr, Z, C2, C1, tan delta, PD, YN

Wherein V2peak = secondary voltage peak; v2rms=secondary voltage effective value; v2dc=secondary voltage dc voltage; igpeak =voltage transformer ground current peak value, igrms =voltage transformer ground current effective value, frg =mirror frequency of current Ig, is non-system power frequency signal of frequency of second largest gain in 20-150Hz range; mfg is the ratio of the sum of other frequency gains greater than the power frequency signal to the sum of all frequency gains, so 0= < Mfg < = 1;

ic2 is the effective current value or root mean square value or peak value of the low-voltage capacitor of the capacitive voltage transformer;

ip1 is the primary grounding current of the electromagnetic transformer of the capacitive voltage transformer;

Is the phase angle of the secondary voltage V2,/> For the phase angle of the low-voltage current IC2,/>The Ig phase angle of the grounding current;

C10 is nameplate or high voltage power supply value of historical test;

C20 is the nameplate or the historically tested low voltage power value.

A low voltage power output by the C2 neural network;

the high-voltage power output by the C1 neural network;

Whether YN has insulation abnormality or overall state abnormality or not, if so, taking 1, and if not, taking 0;

PD partial discharge capacity, unit pc or mV or dB;

B. Training the samples by adopting a neural network algorithm, and building training sample frames with different input amounts and output amounts under the condition of known output amounts, wherein the number of the training samples is not less than 3; if the input quantity is partially missing, the input quantity is marked as 0 value or takes other uniformly fixed numbers, and if the output quantity is known to be partially missing, the input quantity is marked as 0 value or takes other uniformly fixed numbers;

C. according to the input quantity monitored on line, invoking the trained neural network algorithm in the step B to calculate the output quantity, observing the measurement error data and the insulation indexes of the discharge quantity and capacitance according to the output quantity, comparing the synchronism of the error overrun and the insulation abnormal parameter, and explaining the reason of the error overrun;

The basic classification is as follows:

(c1) When the metering error and the dielectric loss are out of standard at the same time and YN=1, judging that the metering is out of tolerance caused by aging of an insulating medium, wetting of the medium or sealing defect;

(c2) When the metering error and the partial discharge amount PD exceed the standard at the same time, and YN=0, judging that the error exceeds the standard and is caused by insulation discharge or discharge temperature rise;

(c3) When the metering error and the partial discharge amount PD exceed the standard at the same time and YN=1, judging that the error exceeds the standard and is caused by heat loss and aging caused by insulation discharge or accumulation of temperature rise of discharge;

(c4) When the insulation resistance Z is normal and the metering error exceeds the limit, but PD exceeds the limit, YN=0, and the metering error exceeds the limit caused by discharge or vibration gaps caused by dirt on the surface of the transformer or poor contact of wiring terminals of high and low voltage layers of the transformer is judged.

Further, a windowing process is performed on Bk by applying a window function, and any one of a Hamming window, a Hanning window, a Gaussian window, a Blackman window, a Black-Harries window, a Keissen window, a flat roof window, a rectangular window and a triangular window is adopted.

In the embodiment of the application, the neural network algorithm is used as an artificial intelligent processing technology similar to the black box theory, is more and more widely applied, and has the key points that the specific algorithm step for drawing the conclusion is not interpreted, but the training sample is reasonably built, and the input quantity and the output quantity are built. The input quantity refers in this embodiment to a parameter that is known or can be obtained by measurement, and the output quantity is a target quantity, i.e. a desired parameter that needs to be deduced by measurement.

Building a physical and electrical model of the output quantity to be combined with the tested equipment, wherein the calculation of the output quantity and the input quantity have an electrical and physical relationship.

Aiming at a 110kV PT voltage transformer, the built neural network input/output framework is as follows:

A1. When the tested equipment is a universal voltage transformer:

the input quantity is as follows: v2peak, V2rms, V2dc, igpeak, igrms, frg,Mfg,N

The output is: verr, δerr, Z, tan delta, PD, YN

Setting the unit of three parameters of input quantity V2peak, V2rms and V2dc to take volts, igpeak, igrms units to take amperes, frg units Hz and Mfg as a percentage;

Output Verr unit percent, δerr unit percent, Z unit G ohm, tan δ unit percent, PD unit leather library, YN is 0 or 1 binary number.

Let the training samples be the first 4 rows of the table below, train with the backward neural network algorithm, and the fifth row is based on the predicted data of the first four sets of data samples.

As can be seen from the prediction data, verr=0.192, δerr=31, z=49.5g, tan δ=0.018, pd=235 pc, yn=1. The ratio difference is close to the required value of 0.2%, the angle difference is also seriously larger, and the dielectric loss and the partial discharge are both higher, so that the insulation discharge problem of the tested transformer is obtained by the YN column.

When the voltage transformer is a capacitive voltage transformer, the sample input and output structure is

The input quantity is as follows: v2peak, V2rms, V2dc, ic2, ip1,Igpeak,Igrms,frg,Mfg,C10,C20,N

The output is: verr, δerr, Z, C2, C1, tan delta, PD, YN

Compared with the calculation of the electromagnetic transformer, the training sample needs to be added with the nominal values of C1 and C2, the actually measured current value and phase of C2 and the like, and the training method and the prediction method are the same as those of the calculation method of the electromagnetic voltage transformer, and are not repeated here.

In the embodiment of the application, the monitoring method can build a monitoring module which consists of a processor, an acquisition module (analog-to-digital converter), a current sampling module, a voltage sampling module, a memory and a communication module.

The acquisition module only needs to acquire voltage and current values, and when the voltage acquisition module is applied to an electromagnetic voltage transformer and a capacitive voltage transformer, the voltage acquisition channels can be completely shared because the secondary voltage specifications are the same 60V or 100V specifications.

The current channels are mainly used for collecting grounding currents Ig and IC2, wherein the grounding currents Ig are also common to electromagnetic voltage transformers and capacitive voltage transformers.

Besides, the hardware for uniquely distinguishing the electromagnetic voltage transformer from the capacitive voltage transformer is as follows: the capacitive voltage transformer increases the current collection of a low-voltage capacitor C2 and the primary grounding current Ip1 of the exciting transformer.

Of course, the architecture can be further simplified, and in the neural network architecture, the fluctuation performance of Ip1 explains whether the fault or the out-of-tolerance is caused by the exciting unit, but because ip1+ic2=ig, if the fault occurrence position is not required to be further determined or the cost is saved, the monitoring of Ip1 is not required. Only by IC2 and Ig.

Therefore, compared with an electromagnetic voltage transformer, the capacitive voltage transformer is only increased in current channel, the same hardware architecture design can be adopted, and particularly when the capacitive voltage transformer is applied to monitoring of multiple sets of transformers, the current channel can be flexibly selected, or the scheme that the hardware scheme of the capacitive voltage transformer is directly adopted is compatible with the scheme of the electromagnetic voltage transformer.

The hardware function of the device also takes temperature, humidity, vibration and the like into account, and plays a good reference role in evaluating the influence of the environment on errors and insulation parameters.

A professional can quickly construct the hardware and implement the relevant algorithms according to the present patent specification.

The device can be widely applied to portable equipment, on-line monitoring equipment, other on-line monitoring functional modules and the like for voltage transformer live-line error testing and live-line insulation performance evaluation.

Of course, the application of the hardware architecture described in this embodiment is not limited to this, and the explanation of the insulation and the error of the transformer or the equivalent transformation or the conceptual substitution of the transformer are all within the scope of protection of this patent.

The above embodiments are only for illustrating the technical solution of the present invention, and are not limiting; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims (6)

1. A voltage transformer monitoring method is characterized in that: the method comprises the following steps:

s1, data acquisition and analysis are carried out on a tested transformer, and the data acquisition and analysis are divided into the following two cases:

A. when the tested transformer is provided with the grounding quantity collection, calculating the metering error of the transformer and analyzing the insulation state according to the secondary quantity and the grounding quantity;

B. When the tested transformer does not have the ground quantity collection or the ground quantity collection fails and the data fluctuation quantity is large, and the data cannot be calculated by the individual transformer errors, a mode of joint analysis of a plurality of transformers is adopted; namely, calculating errors through probability distribution of secondary voltages of a plurality of voltage transformers, wherein the error calculation method specifically comprises the following steps of:

(1) Setting the synchronously acquired transformer signals as Ak, wherein k=1, 2 … N, k is the data of a plurality of transformers acquired at the same time point, and N is the maximum number of acquisition time points; ak is the phase and amplitude of the waveform signal of the electromagnetic voltage transformer secondary voltage or the capacitive voltage transformer secondary voltage and the capacitive current IC 2;

(2) Performing quantization processing on the Ak to obtain Bk, wherein Bk=Ak/Akmax, akmax is the maximum value in the Ak, so that the quantized range of the Bk is 0-1, and then performing windowing processing on the Bk by applying a window function to perform time domain-frequency domain transformation to obtain frequency domain characteristic data Ck; taking the modulus value of Ck to obtain the absolute value Ck, and calculating the sum of the modulus values Searching the maximum value | Ckmax | in the |Ck|, calculating the ratio p= | Ckmax |/SAU, when p > = lambda, considering that the whole data is in a good state, when p < lambda considers that the data fluctuation is increased, and taking p as a confidence interval when abnormal data exists; lambda value interval is 0.8000-0.9999;

(3) Then calculating the Ak normal distribution;

(4) Calculating the difference between the Ak data and the Ak average value of the corresponding confidence interval p as an error offset; or calculating the difference between the Ak data corresponding to the confidence interval p and the set standard value as an error offset;

the obtained data includes:

electromagnetic voltage transformer: the amplitude offset and the angle offset of the secondary voltage V2 are marked as [ delta epsilon 1, [ delta epsilon 2], and delta epsilon 1 is regarded as a specific difference offset, and delta epsilon 2 is an angle difference offset;

Capacitive voltage transformer: the magnitude offset of the secondary voltage V2 and the magnitude offset of the IC2 are marked as [ delta epsilon 1, [ delta epsilon 2]; v2 angular offset, IC2 angular offset [ Δε3, Δε4], then calculating the ratio and angular difference offsets:

Differential offset: delta epsilon 1-Delta epsilon 2

Angular difference offset: delta epsilon 3-Delta epsilon 4

When the calculated ratio difference and angle difference offset exceeds the accuracy grade specified value of the voltage transformer, the calculated ratio difference and angle difference offset are regarded as error overrun;

s2, comprehensively evaluating the relevance of metering error fluctuation and insulation state: and synchronously evaluating the metering and the insulation state by adopting a neural network algorithm, and evaluating the reason of the metering out-of-tolerance according to the insulation index.

2. The method for monitoring a voltage transformer according to claim 1, wherein: the tested transformer in the step S1 comprises an electromagnetic voltage transformer and a capacitive voltage transformer.

3. The method for monitoring a voltage transformer according to claim 1, wherein: when the tested transformer is provided with the grounding quantity collection, the collection and analysis modes are as follows:

the method comprises the following steps of firstly, collecting secondary quantity and grounding quantity of a measured transformer and auxiliary quantity; the auxiliary quantity comprises temperature, humidity and vibration data;

If the tested transformer is an electromagnetic voltage transformer, collecting at least one path of secondary voltage and one path of overall grounding current of the voltage transformer when collecting secondary quantity and grounding quantity;

If the measurement transformer is a capacitive voltage transformer, collecting at least one path of secondary voltage, one path of integral grounding leakage current and one path of low-voltage capacitance grounding current when collecting secondary quantity and grounding quantity; or at least collecting one path of secondary voltage, one path of low-voltage capacitor grounding current and one path of step-down transformer primary grounding current;

And secondly, respectively calculating the error of the voltage transformer and evaluating the insulation state according to the collected secondary quantity and the ground quantity.

4. A method of monitoring a voltage transformer according to claim 3, wherein: when the tested transformer is an electromagnetic voltage transformer, the acquisition and analysis method comprises the following steps:

(a) The waveform Ig of the secondary voltage V2 and the whole grounding current of the transformer is collected at fixed time, and the amplitude, the phase and the frequency spectrum are calculated; temperature, humidity, vibration data;

(b) Calculating the metering instantaneous error: windowing is carried out on the V2 and Ig waveform data, and the direct current components V2dc and Igdc of the V2 and Ig and the ratio kdc =v2dc/Igdc of the direct current components V2dc and Igdc are calculated by adopting wavelet or short-time Fourier transform STFT; when Igdc is carried out, V2dc is increased by more than 10% compared with the fault-free historical average data or kdc is smaller than 90% of the fault-free historical average data, kdc is marked as an abnormal value, and the line is considered to have operation overvoltage or thunder overvoltage or line resonance, so that the voltage V2 is not used as a basis for calculating errors, and collected V2 data is abandoned; when the ratio V2/Ig of the effective value or root mean square value V2 and Ig is increased by more than 10% compared with the fault-free historical average value, the secondary voltage of the voltage transformer is regarded as open circuit, the voltage V2 is not used as the basis of metering error, the collected data of V2 are abandoned, when Igdc, the absolute value of the difference between V2dc and the historical value which is lower than the historical average value 110% or kdc is larger than the historical average value 95%, kdc is recorded as a normal value, and the phase difference between V2 and Ig is calculated Ratio difference between V2 and ig×m, angular difference, dielectric loss:

the ratio difference Vr 0= (V2 Nr-Ig M)/(Ig M) =v2 Nr/(Ig M) -1=kg Nr/M-1

Angular difference

Dielectric loss

Nr is a rated transformation ratio or a transformation ratio value of a history test, M is a reference insulation resistance value or a history test value, kg=v2/Ig;

When M is indeterminate, completing the error calculation by steps (c) + (d);

(c) Calculating the measurement mean error offset:

Obtaining kdc as normal data, secondary voltages V2N, grounding currents Ign, n=1, 2 … N at a plurality of time points, calculating kgn =v2n/Ign, recording a V2 voltage phase angle as θ2n, recording a phase angle of the grounding current Ig as θgn, and calculating Calculating a normal distribution P (kg 0) of kgn based on kg0 with a reference standard value V20, an Ig0 amplitude ratio of kg 0=v20/Ig 0 as a desired value μ, δkg being a standard deviation of kgn;

Calculating a normal distribution P (kgn) based on kgn based on the average value of the amplitude ratio kgn =V2avg/Igavg of the average value of V2n and δkg as the expected value μ and the standard deviation of kgn;

Respectively with reference value Average values are respectively taken as expected values mu,/>Standard deviation/>Calculation/>Obtain/>

Setting a confidence interval, calculating a specific difference offset DeltaP=P (kgn) -P (kg 0) and recording the specific difference offset as the specific difference offset; offset of angular difference

When a confidence interval cannot be set, carrying out Fourier transformation on the amplitude or phase angle difference of N secondary voltages V2N, N >8 and the grounding current Ign to obtain frequency domain data, calculating the proportion of the maximum absolute value of the frequency domain data to the arithmetic sum of all data, wherein the proportion corresponds to the secondary voltages and the grounding currents and is respectively marked as Sv2 and Sign, and the minimum value in [ Sv2 and Sign ] is taken as the confidence interval;

(d) Carrying in the absolute quantity of the ratio difference and the angular difference, and calculating and analyzing the absolute quantity of the ratio difference and the angular difference of the current running state;

Recording the comparison difference and angle difference data of power failure state verification, namely when the power failure data is missing, adopting the average value of the power failure data of the comparison difference and angle difference of other voltage transformers with the same specification in the same operation transformer substation environment to replace the comparison difference and angle difference data of the obtained power failure verification, angle difference data or (a) calculated comparison difference and angle difference data to be Vr0 and delta 0, and then calculating the comparison difference V2=Vr0+DeltaP and angle difference delta=Delta0+Deltadelta of the current voltage transformer to be tested, wherein DeltaP and Deltadelta are the data obtained in the step (c);

D2: when the ratio difference and the angle difference obtained by the D1 are in the standard value range, continuously increasing vibration data or when the vibration amplitude is larger than the history normal value, considering that the external bolt of the voltage transformer is loosened; when the ratio difference and the angle difference obtained by the D1 exceed the standard value range, and when the vibration data is continuously increased or the vibration amplitude is larger than the history normal value, the voltage transformer is considered to have overvoltage, discharge or insulation abnormality;

(e) Taking environmental influence into consideration, regularly analyzing historical data of the ratio difference, the angle difference, the temperature, the humidity and the vibration of the voltage transformer, and calculating normal distribution;

drawing a normal distribution curve based on the ratio difference, the angle difference, the temperature, the humidity and the vibration;

When the ratio difference and the angle difference are over-limit, the temperature and humidity data are in the historical average value or 95% confidence interval, and the voltage transformer is considered to be over-limit in metering error due to internal reasons; when the ratio difference and the angle difference are over-limited, at least one of the temperature, the humidity and the vibration data exceeds the historical average value or the 95% confidence interval, the voltage transformer is considered to have measurement errors which are over-limited due to the influence of the environment or the over-limitation of the errors caused by external insulation or internal insulation.

5. The method for monitoring a voltage transformer according to claim 1, wherein: the step S2 includes:

A. Neural network rack for establishing following parameters

A1, when the tested equipment is a universal voltage transformer:

the input quantity is as follows: v2peak, V2rms, V2dc, igpeak, igrms, frg,Mfg,N

The output is: verr, δerr, Z, tan delta, PD, YN

A2. when the tested equipment is a capacitive voltage transformer:

The input quantity is as follows: v2peak, V2rms, V2dc, ic2, ip1, Igpeak,Igrms,frg,Mfg,C10,C20,N

The output is: verr, δerr, Z, C2, C1, tan delta, PD, YN

Where Verr represents the ratio difference, δerr represents the angular difference, V2peak = secondary voltage peak; v2rms=secondary voltage effective value; v2dc=secondary voltage dc voltage; igpeak =voltage transformer ground current peak value, igrms =voltage transformer ground current effective value, frg =mirror frequency of current Ig, is non-system power frequency signal of frequency of second largest gain in 20-150Hz range; mfg is the ratio of the sum of other frequency gains greater than the power frequency signal to the sum of all frequency gains, so 0= < Mfg < = 1;

ic2 is the effective current value or root mean square value or peak value of the low-voltage capacitor of the capacitive voltage transformer;

ip1 is the primary grounding current of the electromagnetic transformer of the capacitive voltage transformer;

Is the phase angle of the secondary voltage V2,/> For the phase angle of the low-voltage current IC2,/>The Ig phase angle of the grounding current;

C10 is nameplate or high voltage power supply value of historical test;

C20 is nameplate or historical tested low voltage power value;

a low voltage power output by the C2 neural network;

the high-voltage power output by the C1 neural network;

Whether YN has insulation abnormality or overall state abnormality or not, if so, taking 1, and if not, taking 0;

PD partial discharge capacity, unit pc or mV or dB;

B. Training by adopting a neural network algorithm, and building training sample frames with different input amounts and output amounts under the condition of known output amounts, wherein the number of the training samples is not less than 3; if the input quantity is partially missing, the input quantity is marked as 0 value or takes other uniformly fixed numbers, and if the output quantity is known to be partially missing, the input quantity is marked as 0 value or takes other uniformly fixed numbers;

C. according to the input quantity monitored on line, invoking the trained neural network algorithm in the step B to calculate the output quantity, observing the measurement error data and the insulation indexes of the discharge quantity and capacitance according to the output quantity, comparing the synchronism of the error overrun and the insulation abnormal parameter, and explaining the reason of the error overrun;

The basic classification is as follows:

(c1) When the metering error and the dielectric loss are out of standard at the same time and YN=1, judging that the metering is out of tolerance caused by aging of an insulating medium, wetting of the medium or sealing defect;

(c2) When the metering error and the partial discharge amount PD exceed the standard at the same time, and YN=0, judging that the error exceeds the standard and is caused by insulation discharge or discharge temperature rise;

(c3) When the metering error and the partial discharge amount PD exceed the standard at the same time and YN=1, judging that the error exceeds the standard and is caused by heat loss and aging caused by insulation discharge or accumulation of temperature rise of discharge;

(c4) When the insulation resistance Z is normal and the metering error exceeds the limit, but PD exceeds the limit, YN=0, and the metering error exceeds the limit caused by discharge or vibration gaps caused by dirt on the surface of the transformer or poor contact of wiring terminals of high and low voltage layers of the transformer is judged.

6. The method for monitoring a voltage transformer according to claim 5, wherein: and applying a window function to Bk to perform windowing, wherein any one of a Hamming window, a Hanning window, a Gaussian window, a Blackman window, a Black-Harries window, a Kesepia window, a flat top window, a rectangular window and a triangular window is adopted.