CN104810795A - Improved inrush current identification method - Google Patents

Abstract

The invention discloses an improved inrush current identification method. The method includes a main variable data window identification algorithm and an auxiliary acceleration identification algorithm. The method particularly includes: setting threshold values +/-epsilon above and below a time axis respectively, and utilizing the threshold values and an extreme value of a certain section of sampling current to determine a starting point and a finishing point of a variable data window; simulating a standard sine curve forwards and backwards by the extreme value in the selected variable data window according to a sampling interval and data window length, and comparing normalized area indicators of difference area between the standard sine curve and the sampling current so as to realize locking and starting of protection. By the improved inrush current identification method, inrush current waveform and failure current waveform can be differentiated to greatest extent, and when a transformer generates inrush current, locking differential protection can be realized quickly and reliably; mistaken locking differential protection is avoided during inside failure; differential protection can be started quickly when the transformer is switched on in a failure when being no-load; the improved inrush current identification method has high transformer saturation resistance.

Description

A kind of modified model magnetizing inrush current discrimination method

Technical field

The present invention relates to a kind of modified model magnetizing inrush current discrimination method, belong to technical field of power systems.

Background technology

At present, existing power transformer, as the critical elements of transformation of electrical energy and transmission, is widely used in power plant and transformer station.Differential protection is the main protection of transformer, and ideally, transformer differential protection is when normal operation and external short circuit, and the electric current flowing into differential relay is zero, and protective device is reliably failure to actuate.But in fact when transformer during no-load closing produces magnetizing inrush current, the large young pathbreaker of magnetizing inrush current reaches several times even tens times of transformer rated current; On the other hand, magnetizing inrush current only flows through the mains side of transformer, and load side is because of open circuit not electric current, and magnetizing inrush current will flow into the differential circuit of longitudinal difference protection completely, makes to create a very large unsymmetrical current in differential relay.So, differentiate that magnetizing inrush current and internal fault current become the key issue strengthening protection reliability.

Summary of the invention

Object: in order to overcome the deficiencies in the prior art, the invention provides a kind of modified model magnetizing inrush current discrimination method, by the determination of variable data window, the method, compared with the existing discrimination method that shoves, can distinguish inrush and fault current waveform to a greater degree.

Technical scheme: for solving the problems of the technologies described above, the technical solution used in the present invention is:

Comprise main variable data window distinguished number and auxiliary acceleration distinguished number;

Main variable data window distinguished number: arrange threshold value ± ε respectively on a timeline, utilize the extreme value of threshold value and a certain section of sample rate current, extreme value is maximum value or minimum value, determines starting point and the terminating point of variable data window; And in the variable data window selected, according to sampling interval and data window length, fictionalize forward a standard sine curve backward with this extreme value, calculate the normalization index Z of difference area between standard sine curve and sample rate current; Along with the passing of sampled point, algorithm forms multiple variable data window, and each data window can export a normalization index Z, algorithm real-time by normalization index Z and setting value Z _setcompare, and then realize locking and the opening of protection;

Auxiliary acceleration distinguished number: set after auxiliary acceleration distinguished number starts; if phase current does not have Z value to export during one-period; then judge that this phase current is as shoving immediately, this accelerates distinguished number for auxiliary, also can the latch-up protection when one-period so auxiliary acceleration distinguished number is the slowest.

Described modified model magnetizing inrush current discrimination method, is characterized in that: main variable data window distinguished number specifically refers to:

First according to the virtual out standard sine curve of the extreme point in a certain section, virtual standard sine curve out can be divided into two sections:

First paragraph is the curved section Y in data window between extreme point to data window starting point ₁, be formulated as:

Y ₁(k ₁)＝I _lim*sin[π/2-2π(N ₁-k ₁)/N]，k ₁＝1,2,…N ₁

Wherein k ₁for virtual first paragraph standard sine curve Y out ₁discrete point sequence number; I _limfor extreme value in data window (maximum value or minimum value), N is a cycle total sampling number, N ₁for curved section Y ₁interior total data is counted;

Second segment is the curved section Y in data window between extreme point to data window terminating point ₂, be formulated as:

Y ₂(k ₂)＝I _lim*sin[π/2-2πk ₂/N]，k ₂＝1,2,…N ₂

Wherein k ₂for virtual second segment standard sine curve Y out ₂discrete point sequence number, N ₂for curved section Y ₂interior total data is counted;

Then total standard sine curve can be defined as:

y＝[Y ₁Y ₂]

Make S ₁for the area of sample rate current curve and time shaft, S ₂for the area of virtual standard sine curve out and time shaft, then S ₁with S ₂can be expressed as:

$S_1=\underset{j=1}{\overset{N3}{\mathrm{\Σ}}}i\left(j\right)$

$S_2=\underset{k=1}{\overset{N3}{\mathrm{\Σ}}}y\left(k\right)$

Wherein N ₃for Y ₁with Y ₂data window total length altogether, j is N ₃the sampled point sequence number of actual current in data window, k is Y ₁with Y ₂the discrete point sequence number of total standard sine curve y altogether, i (j) is sample rate current instantaneous value during the sampled point of jth in above-mentioned total data window, total y (k) for the instantaneous value of standard sine curve (its length is exactly total data window length) when kth discrete point.

Make the sampled point sequence number of actual current in this data window corresponding with the discrete point sequence number of standard sine curve y, i.e. j=k, then calculating difference area S between sample rate current and standard sine curve is:

$S=\underset{k=1}{\overset{N3}{\mathrm{\Σ}}}|i\left(k\right)-y\left(k\right)|$

Discriminant criterion is normalized, makes S _min=min (S ₁, S ₂)

Z＝S/S _min

S _minfor S ₁, S ₂middle smaller, Z is the index that variable data window judges magnetizing inrush current or fault current; Z _setfor the setting value of Z;

Along with the passing of sampled point, algorithm forms multiple variable data window, and each data window can export an index Z, algorithm real-time by this index Z and certain certain value Z _setcompare, as Z>Zset, the current waveform in data window is determined as and shoves; Work as Z _≤during Zset, the current waveform in data window is determined as fault current.

Wherein, choosing of variable data window specifically refers to: arrange threshold value ± ε respectively first on a timeline, if protection sample rate f, and sampling interval T=1/f, each cycle sampling number is N, comprises the following steps:

1) determination of maximum value or minimum value: before the normal and idle-loaded switching-on of system, difference stream should within threshold interval; When stream of being on duty exceeds threshold interval, main variable data window distinguished number starts, sampled point when selecting sample rate current to pass threshold interval is as the starting point of first extremum search, and sampled point when selecting sample rate current again to pass this threshold interval is as the terminating point of first extremum search; To the data window of terminating point, first forward maximum I is searched out in the starting point of first extremum search _maxor reverse minimum I _min, record the sampled value I corresponding to this extreme point _limand sampled point sequence number n, in like manner find second extreme point I _lim2;

2) determination of variable data window: with first extreme value I _limfor basic point is searched for forward backward, first starting point as variable data window after passing from threshold region using electric current in regulation straight line process, the instantaneous value of starting point is I ₁, sampled point sequence number is a, and search condition is such as formula (1); If do not leave any sampled point in region, then in straight line process using first after current crosses threshold region as the starting point I of variable data window _1,search condition is such as formula (2), and the algorithm seeking data window starting point is:

When

i(j+1)<＝|ε|&i(j)>|ε| j＝n-1,n-2,n-3… (1)

Or

Time,

Then make a=j, I ₁=i (a)

In formula, i is sample rate current instantaneous value, and j is current sample point sequence number, and the sampled point sequence number of the data window starting point after determining is a, and the instantaneous value of data window starting point is I ₁, n is extreme value sampled point sequence number, i (n)=I _lim;

First terminating point as variable data window after passing from threshold region using electric current in search procedure backward, its instantaneous value is I ₂, sampled point sequence number is b, and search condition is such as formula (3); If do not leave any sampled point in region, then same using the sampled point of first after current crosses threshold region as the terminating point I of variable data window in search procedure backward ₂, search condition such as formula (4), namely

i(j-1)<＝|ε|&i(j)>|ε| j＝n+1,n+2,n+3… (3)

Or

Time,

Make b=j, I ₂=i (b);

In formula, i is sample rate current instantaneous value, and j is current sample point sequence number, and the sampled point sequence number of the data window terminating point after determining is b, and the instantaneous value of data window terminating point is I ₂, n is extreme value sampled point sequence number, i (n)=I _lim;

3) first extreme value I is calculated _limand I ₁between sampling number N ₁, comprise I _limand I ₁2 points, fictionalize the first paragraph of standard sine curve, order

Y ₁(k ₁)＝I _lim*sin[π/2-2π(N ₁-k ₁)/N]，k ₁＝1,2,…N ₁

Calculate I _limand I ₂between sampling number N ₂, do not comprise I _limbut comprise I ₂point, fictionalizes the second segment of standard sine curve, order

Y ₂(k ₂)＝I _lim*sin[π/2-2πk ₂/N]，k ₂＝1,2,…N ₂

By sequence Y ₁with Y ₂splice, order

y＝[Y ₁Y ₂]

4) I is calculated ₁, I ₂between points N ₃, comprise I ₁and I ₂, make the sampled point sequence number of actual current in this data window corresponding with the discrete point sequence number of standard sine curve y, i.e. j=k, then calculating difference area S between sample rate current and standard sine curve is:

$S=\underset{k=1}{\overset{N3}{\mathrm{\&Sigma;}}}|i\left(k\right)-y\left(k\right)|$

Described modified model magnetizing inrush current discrimination method, it is characterized in that: along with the passing of data window, algorithm exports Z value continuously, and the decision rule that under different situations, Z value exports is as table 1:

The decision rule that under table 1 different situations, Z value exports

That considers with transformer differential protection starting current is supporting, arranges threshold value ± ε, setting ε=0.18I _n, I _nfor transformer rated current.

Consider the reliability of protection, setting Z _set=0.55.

Beneficial effect: modified model magnetizing inrush current discrimination method provided by the invention, comprises main variable data window distinguished number and auxiliary acceleration distinguished number.The method main innovate point comprises: 1, the ingenious utilization of the method arranges the method for threshold value at zero crossings, solve side that some inrush of oversize transformer is all in time shaft namely so-called not zero passage problem time variable data window On The Choice, this threshold interval is set in addition and also the selection rule of variable data window Algorithms of Selecting under different situations can be unitized.2, by the analysis to Current Waveform Characteristics under different situations, design variable data window Algorithms of Selecting, this algorithm can make variable data window comprise to shove interval angle part when magnetizing inrush current, can make area folded by inrush and virtual standard sine waveform out as far as possible maximum like this; In case of a fault, make area folded by fault current waveform and virtual standard sine waveform out as far as possible minimum.3, propose to shove that locking is auxiliary accelerates criterion, utilize variable data window access time to accelerate criterion as of inrush current distinguishing is auxiliary, ensureing in situation of shoving that algorithm is the slowest also can the latch-up protection when one-period.4, successively exported the difference of Z value by variable data window, in conjunction with phase-splitting locking principle, to transformer belt fault idle-loaded switching-on situation, can unblock not even locking fast.5, this algorithm has stronger anti-instrument transformer saturability.

Demonstrated validity and the superiority of technical solution of the present invention by dynamic simulation test, when transformer belt 1.7% shorted-turn fault idle-loaded switching-on, be protected in the open differential protection of 20ms after closing a floodgate, faster than secondary harmonic brake open and protection speed 23 times.And when idle-loaded switching-on instrument transformer is saturated, algorithm still can make the reliable locking of protection.

Accompanying drawing explanation

Fig. 1 is the searching figure of variable data window maximum value or minimum value;

Fig. 2 be the asymmetric variable data window that shoves choose figure;

Fig. 3 be symmetry shove variable data window choose figure;

Fig. 4 be fault current variable data window choose figure;

Fig. 5 be tape jam idle-loaded switching-on variable data window choose figure;

Fig. 6 (a) is normal idle-loaded switching-on segregated phase differential current and (b) Z value output map;

Fig. 7 (a) is for idle-loaded switching-on is in 1.7% shorted-turn fault segregated phase differential current and (b) Z value output map;

Fig. 8 (a) is for idle-loaded switching-on is in internal ground fault segregated phase differential current and (b) Z value output map;

Fig. 9 (a) is Transformer internal ground fault segregated phase differential current and (b) Z value output map;

Figure 10 (a) for idle-loaded switching-on instrument transformer saturated time segregated phase differential current and (b) Z value output map.

Embodiment

Below in conjunction with specific embodiment, the present invention is further described.

As shown in Figures 1 to 10, a kind of modified model magnetizing inrush current discrimination method, comprising:

(1) main variable data window distinguished number: arrange threshold value ± ε respectively on a timeline, utilize the extreme value of threshold value and a certain section of sample rate current, extreme value is maximum value or minimum value, determines starting point and the terminating point of variable data window; And in the variable data window selected; according to sampling interval and data window length; fictionalize forward a standard sine curve backward with this extreme value, the normalized area index of difference area between standard of comparison sine curve and sample rate current, and then realize locking and the opening of protection.The method can distinguish inrush and fault current waveform to greatest extent, when transformer generation magnetizing inrush current, and can the Blocking Differential Protection of fast and reliable; Blocking Differential Protection is not missed when internal fault; Can also when transformer belt fault idle-loaded switching-on, open differential protection fast; And there is stronger anti-instrument transformer saturability.

In order to promote the performance of magnetizing inrush current discrimination method further, the present invention also proposes a kind of auxiliary to accelerate distinguished number, combines auxiliary acceleration distinguished number, and algorithm is Blocking Differential Protection during one-period after idle-loaded switching-on at the latest.

(2) auxiliaryly distinguished number is accelerated: owing to shoving and the different wave shape of fault current, the access time long to the window of variable data window own also can be used as the factor of shoving and differentiating, therefore can will choose the assistant criteria of variable data window time used as locking of accelerating to shove;

In order to strengthen the reliability of inrush current distinguishing, interval angle data can as often as possible be included when extracting variable data window by main variable data window distinguished number above, if electric current shoves mutually, interval angle is excessive, the sampling number that then variable data window comprises certainly will increase, although enhance the ability that algorithm is differentiated to shove like this, but the time choosing variable data window owing to waiting for more sampling number and long, can affect the output speed of algorithm.Reality, only has the situation of magnetizing inrush current just likely to make interval angle very large, makes variable data window exceed one-period access time, and cause not having Z value to export during one-period in the period after auxiliary acceleration distinguished number starts; So in this setting after auxiliary acceleration distinguished number starts; if electric current does not have Z value to export mutually during one-period; then immediately Cutoff current this for shoving, this accelerates distinguished number for auxiliary, and accelerating so auxiliary that distinguished number is the slowest also can the latch-up protection when one-period.

Therefore, along with the passing of data window, algorithm exports Z value continuously, can carry out Taxonomic discussion to various situation, provides the decision rule that Z value exports under different situations in table 1:

The decision rule that under table 1 different situations, Z value exports

The present invention is in conjunction with main variable data window distinguished number and auxiliary acceleration distinguished number two kinds of algorithms; can not only when the normal idle-loaded switching-on of transformer fast and reliably Blocking Differential Protection; do not miss Blocking Differential Protection when power transformer interior fault, differential protection can also be opened fast when transformer belt fault idle-loaded switching-on.

Embodiment 1:

A) the choosing of variable data window: process of being chosen by variable data window combines the identification algorithm that shoves and is introduced, and arranges very little threshold value ± ε first on a timeline respectively, if protection sample rate f, sampling interval T=1/f, each cycle sampling number is N; The steps include:

1. the determination of maximum value or minimum value: before the normal and idle-loaded switching-on of system, difference stream should within threshold interval.When stream of being on duty exceeds threshold interval, main variable data window distinguished number starts, as shown in Figure 1, sampled point when selecting sample rate current to pass threshold interval is as the starting point of first extremum search, and sampled point when selecting sample rate current again to pass this threshold interval is as the terminating point of first extremum search; In this segment data window, search out first forward maximum or reverse minimum, record the sampled value I corresponding to this extreme point _lim(the I in Fig. 1 _max) and sampled point sequence number n, in like manner can search out second extreme point I according to above-mentioned rule _lim2(the I in Fig. 1 _min);

2. (this sentences the maximum I that Fig. 1 first variable data window searches out in the determination of variable data window _maxfor example is described): with maximum I _maxfor basic point is searched for forward backward, first starting point as variable data window after passing from threshold region using electric current in regulation straight line process, its instantaneous value is I ₁, sampled point sequence number is a, and search condition is such as formula (1); If do not leave any sampled point in region, then in straight line process using first after current crosses threshold region as the starting point I of variable data window _1,search condition is such as formula (2), and the algorithm seeking data window starting point is:

When

i(j+1)<＝|ε|&i(j)>|ε| j＝n-1,n-2,n-3… (1)

Or

Time,

Then make a=j, I ₁=i (a)

In formula, i is sample rate current instantaneous value, and j is current sample point sequence number, and the sampled point sequence number of the data window starting point after determining is a, and the instantaneous value of data window starting point is I ₁, n is extreme value sampled point sequence number, i (n)=I _max;

First terminating point as variable data window after passing from threshold region using electric current in search procedure backward, its instantaneous value is I ₂, sampled point sequence number is b, and search condition is such as formula (3); If do not leave any sampled point in region, then same using the sampled point of first after current crosses threshold region as the terminating point I of variable data window in search procedure backward ₂, search condition such as formula (4), namely

i(j-1)<＝|ε|&i(j)>|ε| j＝n+1,n+2,n+3… (3)

Or

Time,

Make b=j, I ₂=i (b);

In formula, i is sample rate current instantaneous value, and j is current sample point sequence number, and the sampled point sequence number of the data window terminating point after determining is b, and the instantaneous value of data window terminating point is I ₂, n is extreme value sampled point sequence number,

i(n)＝I _max；

3. calculate first extreme value I _maxand I ₁between sampling number N ₁, comprise I _maxand I ₁2 points,

Fictionalize the first paragraph of standard sine curve, order

Y ₁(k ₁)＝I _max*sin[π/2-2π(N ₁-k ₁)/N]，k ₁＝1,2,…N ₁

Calculate I _maxand I ₂between sampling number N ₂, do not comprise I _maxbut comprise I ₂point, fictionalizes the second segment of standard sine curve, order

Y ₂(k ₂)＝I _max*sin[π/2-2πk ₂/N]，k ₂＝1,2,…N ₂

By sequence Y ₁with Y ₂splice, order

y＝[Y ₁Y ₂]

4. calculate I ₁, I ₂between points N ₃, comprise I ₁and I ₂, as shown in Figure 2, make the sampled point sequence number of actual current in this data window corresponding with the discrete point sequence number of standard sine curve y, i.e. j=k, then calculating difference area S between sample rate current and standard sine curve is:

$S=\underset{k=1}{\overset{N3}{\mathrm{\&Sigma;}}}|i\left(k\right)-y\left(k\right)|.$

As shown in Figures 2 and 3, asymmetricly to shove and symmetry is shoved in situation, by above algorithm picks data window, owing to containing interval angle part, it is very large that each variable data window hatched area S can become, and algorithm exports larger normalized area index Z continuously.And for internal fault, as shown in Figure 4, S is less for each variable data window hatched area, algorithm exports less normalized area index Z continuously.

For tape jam idle-loaded switching-on, from the partial waveform of electric current, although in saturated section of transformer, magnetizing inrush current is comparatively large, masks the feature of fault current, when transformer exit saturated enter non-saturated area time, electric current will show the feature of fault current.And above-mentioned variable data window selection algorithm has the ability automatically choosing this unsaturation section electric current.As shown in Figure 5; variable data window selection algorithm alternately can to select after tape jam idle-loaded switching-on saturated section of electric current and unsaturation section electric current in the period 1; then alternately export larger and less normalized area index Z continuously; therefore algorithm has the ability to identify tape jam combined floodgate situation completely within a short period of time, allows and protects opening as early as possible.

B) the adjusting of normalized area index and criterion

Due in actual state; protection sample rate is likely inconsistent and data window can be caused different in size by above-mentioned data window selection algorithm; the size of different situations electric current is also different in addition, then under measurement different situations, the index of hatched area size will be inconsistent.In order to head it off, discriminant criterion need be normalized, if

$S_1=\underset{j=1}{\overset{N3}{\mathrm{\&Sigma;}}}i\left(j\right)$

$S_2=\underset{k=1}{\overset{N3}{\mathrm{\&Sigma;}}}y\left(k\right)$

Wherein N ₃for Y ₁with Y ₂data window total length altogether, j is N ₃the sampled point sequence number of actual current in data window, k is Y ₁with Y ₂the discrete point sequence number of total standard sine curve y altogether, i (j) is sample rate current instantaneous value during the sampled point of jth in above-mentioned total data window, total y (k) for the instantaneous value of standard sine curve (its length is exactly total data window length) when kth discrete point.

Make the sampled point sequence number of actual current in this data window corresponding with the discrete point sequence number of standard sine curve y, i.e. j=k, then calculating difference area S between sample rate current and standard sine curve is:

$S=\underset{k=1}{\overset{N3}{\mathrm{\&Sigma;}}}|i\left(k\right)-y\left(k\right)|$

Discriminant criterion is normalized, makes S _min=min (S ₁, S ₂)

Z＝S/S _min

S _minfor S ₁, S ₂middle smaller, Z is the index that variable data window judges magnetizing inrush current or fault current; Z _setfor the setting value of Z;

Along with the passing of sampled point, algorithm forms multiple variable data window, and each data window can export a normalized area index Z, algorithm real-time by this normalized area index Z and certain certain value Z _setcompare, as Z>Zset, the current waveform in data window is determined as and shoves; As Z≤Zset, the current waveform in data window is determined as fault current.Differentiate that the criterion of fault current and magnetizing inrush current is:

C) auxiliary acceleration distinguished number

In order to strengthen the reliability of inrush current distinguishing, algorithm extract variable data window time can try one's best more than interval angle data are included (as Fig. 2, Fig. 3), if certain shoves mutually, interval angle is excessive, the sampling number that then variable data window comprises certainly will increase, although enhance the discriminating power shoved like this, but the time choosing variable data window owing to waiting for more sampling number and long, can affect the output speed of algorithm.Reality, after algorithm starts, only has the situation of magnetizing inrush current just likely to make interval angle very large, makes variable data window exceed one-period access time, and cause not having Z value to export during one-period in this period.So in this setting after algorithm starts, do not have Z value to export mutually, then judge that this is as shoving immediately during one-period, this accelerates distinguished number for auxiliary, also can the latch-up protection when one-period so algorithm is the slowest.

Preferably, that considers with transformer differential protection starting current is supporting, arranges threshold value ± ε, wherein ε=0.18I _n, I _nfor transformer rated current.Consider the reliability of protection, setting Z _set=0.55.Embodiment 2:

The mode of the present invention's dynamic simulation test is verified algorithm, and dynamic simulation test demonstrates validity and the superiority of the program, when transformer belt 1.7% shorted-turn fault idle-loaded switching-on, is protected in the open differential protection of 20ms after closing a floodgate.Here is the action characteristic analysis of the present invention under each operating mode:

1, the normal idle-loaded switching-on of transformer

Fig. 6 (a) for segregated phase differential current in the normal idle-loaded switching-on situation of transformer, three-phase be A phase, B phase and C phase, segregated phase differential current I _a, I _band I _crepresent with solid line, dotted line and chain-dotted line respectively.From Fig. 6 (a), the idle-loaded switching-on transformer when 0.02s, I in segregated phase differential current _aand I _bshove for one-sided, have obvious interval angle, and interval angle is larger; I _cfor symmetry is shoved, interval angle is less; I _bdue to the impact not zero crossing of DC component, so the method for place's setting threshold achieves choosing of variable data window.

Three-phase Z value is respectively Z _a, Z _band Z _c; From Fig. 6 (b), due to I _aand I _bshove for asymmetric, I _cfor symmetry is shoved, therefore in 5 cycles, according to variable data window Algorithms of Selecting, Z _aand Z _boutput is counted as Z _cexport the half of counting.C phase 6ms after idle-loaded switching-on exports first Z value and is greater than Z _set, A phase and B phase algorithmically main criterion need 24ms Blocking Differential Protection after idle-loaded switching-on respectively, but according to auxiliary acceleration distinguished number, then respectively can after idle-loaded switching-on 20ms Blocking Differential Protection.2, idle-loaded switching-on is in the slight shorted-turn fault of transformer

Fig. 7 (a) is the segregated phase differential current of idle-loaded switching-on transformer star side A phase 1.7% shorted-turn fault.Because little turn ratio short trouble occurs A phase, fault current exists with shoving simultaneously, and fault current is comparatively slight, so difference stream is form of shoving at transformer saturated section of main manifestations, and in transformer unsaturation section, then poor stream shows as the feature of fault current completely.Utilize variable data window distinguished number to extract the variable data window with shove feature and fault signature respectively, then three-phase Z value exports as shown in Fig. 7 (b).

Exported from Fig. 7 (b), A phase Z value and all present the form fluctuated up and down, this is because A difference stream is alternately divided into the data window showing feature of shoving and the data window showing fault signature continuously.Shove in the data window of feature showing, Z value exports all larger; And in the data window showing fault signature, difference stream waveform and standard sine wave is closely similar, therefore the output of Z value is all less.First Z value point that A phase exports during 13ms after idle-loaded switching-on, be judged to and shove, A phase carries out locking rapidly; But when 20ms, export second Z value point, be less than Z _set, be identified as fault current, according to table 1 decision rule, now A phase open and protection rapidly, owing to adopting phase-splitting locking principle, differential protection can action rapidly.Under identical test current conditions, secondary harmonic brake scheme is about 460ms open and protection after idle-loaded switching-on.

3, idle-loaded switching-on is in inside transformer earth fault

Fig. 8 (a) is idle-loaded switching-on transformer star side A phase internal ground fault.Because comparatively catastrophe failure occurs A phase, then I _aand I _cdifference stream main manifestations is fault current feature, and I _bstill there is larger magnetizing inrush current feature.

From Fig. 8 (b), due to the impact by fault current composition, Z _aand Z _coutput is less than Z _set.And I _bowing to still showing feature of shoving, therefore Z _bexport and be still greater than Z _set.And owing to adopting phase-splitting latching logic, therefore A phase and C phase can not lockings after idle-loaded switching-on, differential protection can action rapidly.

4, inside transformer earth fault in running

Fig. 9 (a) is transformer star side A phase earth fault in operation, and segregated phase differential current all presents sinusoidal waveform feature.

From Fig. 9 (b), owing to there is internal ground fault, three-phase Z value all exports twice in one-period, and output valve is all much smaller than set point Z _set, protection can not locking.

5, CT saturation during idle-loaded switching-on

Figure 10 (a) is segregated phase differential current when instrument transformer under idle-loaded switching-on occurs saturated, shoves less, therefore do not occur saturated because B is symmetrical, and A phase and C phase all occur saturated, shove and to decline rapidly when instrument transformer occurs saturated, and produce reverse charging electric current, interval angle disappears.Under Figure 10 (b) provides this situation, three-phase Z value exports, and visible A phase and C phase are due to saturated, and Z value is less than the output in the unsaturated situation of Figure 10 (b) instrument transformer, but three-phase Z value exports all at Z _setabove, thus protection can reliably locking.

The above is only the preferred embodiment of the present invention; be noted that for those skilled in the art; under the premise without departing from the principles of the invention, can also make some improvements and modifications, these improvements and modifications also should be considered as protection scope of the present invention.

Claims (6)

1. a modified model magnetizing inrush current discrimination method, is characterized in that: comprise main variable data window distinguished number and auxiliary acceleration distinguished number;

Main variable data window distinguished number: arrange threshold value ± ε respectively on a timeline, utilize the extreme value of threshold value and a certain section of sample rate current, extreme value is maximum value or minimum value, determines starting point and the terminating point of variable data window; And in the variable data window selected, according to sampling interval and data window length, fictionalize forward a standard sine curve backward with this extreme value, calculate the normalization index Z of difference area between standard sine curve and sample rate current; Along with the passing of sampled point, algorithm forms multiple variable data window, and each data window can export a normalization index Z, algorithm real-time by normalization index Z and setting value Z _setcompare, and then realize locking and the opening of protection;

Auxiliary acceleration distinguished number: set after auxiliary acceleration distinguished number starts; if phase current does not have Z value to export during one-period; then judge that this phase current is as shoving immediately, this accelerates distinguished number for auxiliary, also can the latch-up protection when one-period so auxiliary acceleration distinguished number is the slowest.

2. modified model magnetizing inrush current discrimination method according to claim 1, is characterized in that: main variable data window distinguished number specifically refers to:

First according to the virtual out standard sine curve of the extreme point in a certain section, virtual standard sine curve out can be divided into two sections:

First paragraph is the curved section Y in data window between extreme point to data window starting point ₁, be formulated as:

Y ₁(k ₁)＝I _lim*sin[π/2-2π(N ₁-k ₁)/N]，k ₁＝1,2,…N ₁

Wherein k ₁for virtual first paragraph standard sine curve Y out ₁discrete point sequence number; I _limfor extreme value in data window (maximum value or minimum value), N is a cycle total sampling number, N ₁for curved section Y ₁interior total data is counted;

Second segment is the curved section Y in data window between extreme point to data window terminating point ₂, be formulated as:

Y ₂(k ₂)＝I _lim*sin[π/2-2πk ₂/N]，k ₂＝1,2,…N ₂

Wherein k ₂for virtual second segment standard sine curve Y out ₂discrete point sequence number, N ₂for curved section Y ₂interior total data is counted;

Then total standard sine curve can be defined as:

y＝[Y ₁Y ₂]

Make S ₁for the area of sample rate current curve and time shaft, S ₂for the area of virtual standard sine curve out and time shaft, then S ₁with S ₂can be expressed as:

$S_1=\underset{j=1}{\overset{N3}{\mathrm{\&Sigma;}}}i\left(j\right)$

$S_2=\underset{k=1}{\overset{N3}{\mathrm{\&Sigma;}}}y\left(k\right)$

Wherein N ₃for Y ₁with Y ₂data window total length altogether, j is N ₃the sampled point sequence number of actual current in data window, k is Y ₁with Y ₂the discrete point sequence number of total standard sine curve y altogether, i (j) is sample rate current instantaneous value during the sampled point of jth in above-mentioned total data window, total y (k) for the instantaneous value of standard sine curve (its length is exactly total data window length) when kth discrete point.

Make the sampled point sequence number of actual current in this data window corresponding with the discrete point sequence number of standard sine curve y, i.e. j=k, then calculating difference area S between sample rate current and standard sine curve is:

$S=\underset{k=1}{\overset{N3}{\mathrm{\&Sigma;}}}|i\left(k\right)-y\left(k\right)|$

Discriminant criterion is normalized, makes S _min=min (S ₁, S ₂)

Z＝S/S _min

S _minfor S ₁, S ₂middle smaller, Z is the index that variable data window judges magnetizing inrush current or fault current; Z _setfor the setting value of Z;

Along with the passing of sampled point, algorithm forms multiple variable data window, and each data window can export a normalized area index Z, algorithm real-time by this normalized area index Z and certain certain value Z _setcompare, as Z>Zset, the current waveform in data window is determined as and shoves; As Z≤Zset, the current waveform in data window is determined as fault current.

3. modified model magnetizing inrush current discrimination method according to claim 2; it is characterized in that: choosing of variable data window specifically refers to: first on a timeline, threshold value ± ε is set respectively, if protection sample rate f, sampling interval T=1/f; each cycle sampling number is N, comprises the following steps:

1) determination of maximum value or minimum value: before the normal and idle-loaded switching-on of system, difference stream should within threshold interval; When stream of being on duty exceeds threshold interval, main variable data window distinguished number starts, sampled point when selecting sample rate current to pass threshold interval is as the starting point of first extremum search, and sampled point when selecting sample rate current again to pass this threshold interval is as the terminating point of first extremum search; To the data window of terminating point, first forward maximum I is searched out in the starting point of first extremum search _maxor reverse minimum I _min, record the sampled value I corresponding to this extreme point _limand sampled point sequence number n, in like manner find second extreme point I _lim2;

2) determination of variable data window: with first extreme value I _limfor basic point is searched for forward backward, first starting point as variable data window after passing from threshold region using electric current in regulation straight line process, the instantaneous value of starting point is I ₁, sampled point sequence number is a, and search condition is such as formula (1); If do not leave any sampled point in region, then in straight line process using first after current crosses threshold region as the starting point I of variable data window ₁, search condition is such as formula (2), and the algorithm seeking data window starting point is:

When

i(j+1)<＝|ε|&i(j)>|ε| j＝n-1,n-2,n-3… (1)

Or

time,

Then make a=j, I ₁=i (a)

In formula, i is sample rate current instantaneous value, and j is current sample point sequence number, and the sampled point sequence number of the data window starting point after determining is a, and the instantaneous value of data window starting point is I ₁, n is extreme value sampled point sequence number, i (n)=I _lim;

First terminating point as variable data window after passing from threshold region using electric current in search procedure backward, its instantaneous value is I ₂, sampled point sequence number is b, and search condition is such as formula (3); If do not leave any sampled point in region, then same using the sampled point of first after current crosses threshold region as the terminating point I of variable data window in search procedure backward ₂, search condition such as formula (4), namely

i(j-1)<＝|ε|&i(j)>|ε| j＝n+1,n+2,n+3… (3)

Or

time,

Make b=j, I ₂=i (b);

In formula, i is sample rate current instantaneous value, and j is current sample point sequence number, and the sampled point sequence number of the data window terminating point after determining is b, and the instantaneous value of data window terminating point is I ₂, n is extreme value sampled point sequence number, i (n)=I _lim;

3) first extreme value I is calculated _limand I ₁between sampling number N ₁, comprise I _limand I ₁2 points, fictionalize the first paragraph of standard sine curve, order

Y ₁(k ₁)＝I _lim*sin[π/2-2π(N ₁-k ₁)/N]，k ₁＝1,2,…N ₁

Calculate I _limand I ₂between sampling number N ₂, do not comprise I _limbut comprise I ₂point, fictionalizes the second segment of standard sine curve, order

Y ₂(k ₂)＝I _lim*sin[π/2-2πk ₂/N]，k ₂＝1,2,…N ₂

By sequence Y ₁with Y ₂splice, order

y＝[Y ₁Y ₂]

4) I is calculated ₁, I ₂between points N ₃, comprise I ₁and I ₂, make the sampled point sequence number of actual current in this data window corresponding with the discrete point sequence number of standard sine curve y, i.e. j=k, then calculating difference area S between sample rate current and standard sine curve is:

$S=\underset{k=1}{\overset{N3}{\mathrm{\&Sigma;}}}|i\left(k\right)-y\left(k\right)|.$

4. modified model magnetizing inrush current discrimination method according to claim 2, it is characterized in that: along with the passing of data window, algorithm exports Z value continuously, and the decision rule that under different situations, Z value exports is as table 1:

The decision rule that under table 1 different situations, Z value exports

5. modified model magnetizing inrush current discrimination method according to claim 2, is characterized in that: that considers with transformer differential protection starting current is supporting, arranges threshold value ± ε, setting ε=0.18I _n, I _nfor transformer rated current.

6. modified model magnetizing inrush current discrimination method according to claim 2, is characterized in that: the reliability considering protection, setting Z _set=0.55.