Vous êtes sur la page 1sur 10

Electric Power Systems Research 145 (2017) 8998

Contents lists available at ScienceDirect

Electric Power Systems Research


journal homepage: www.elsevier.com/locate/epsr

Approximate method to compute the maximum rms current in a


distribution feeder due to simultaneous inrush currents in
downstream transformers
Elmer Sorrentino
Universidad Simn Bolvar, Caracas, Venezuela

a r t i c l e i n f o a b s t r a c t

Article history: This article shows an approximate method to compute the maximum rms current in a distribution feeder
Received 16 July 2016 due to simultaneous inrush currents in downstream transformers. In the rst part of the article, a method
Received in revised form for the exact calculation of inrush currents in distribution transformers (by solving the differential equa-
14 December 2016
tions) is tested. This exact model was veried with experimental measurements. Subsequently, the exact
Accepted 25 December 2016
model was applied in order to develop the proposed method. The approximate method requires two
approximations, which only use simple algebraic functions. The rst approximation is required to build
Keywords:
an equivalent system for the simultaneous inrush currents of transformers in a distribution feeder. This
Transformer inrush
Total inrush currents in distribution feeders
equivalent system is composed by a single transformer, which is fed by an ideal voltage source con-
Simultaneous inrush of transformers nected in series with an impedance. The second approximation is required to compute the maximum
inrush current of a transformer by assuming that the source impedance is not negligible. These novel
approximations were tested for diverse integration time intervals. The results show that the accuracy of
the developed method can be considered suitable because the average percentage error, in comparison
with the exact model, is in the order of 4% for the rst approximation and 5% for the second approxima-
tion. The developed method allows a straightforward calculation of the maximum rms current in a feeder
due to inrush currents in downstream transformers, and the knowledge of this magnitude is useful for
the proper coordination of overcurrent devices in electric power distribution systems.
2016 Elsevier B.V. All rights reserved.

1. Introduction eter that most inuences the current amplitude. There could be
two positive (or negative) voltage semi-cycles applied consecu-
Transformer inrush currents have been studied for many years tively to a transformer, due to the presence of an uninterruptible
[114], and the literature about this topic is abundant. The oldest power-supply, and the inrush currents during these conditions are
found reference describes some experiments about the phe- computed in Ref. [5]. The equivalent circuit is compared with
nomenon [1], without showing the waveshape of inrush currents, the T equivalent circuit in Ref. [6] for the calculation of transformer
but these waveshapes have been shown for more than 100 years [2]. inrush currents, and an improvement is presented in Ref. [7], in
On the other hand, Refs. [313] are some examples of recent arti- order to correctly describe the air-core inductance from both wind-
cles about this topic. The harmonic content of inrush currents in ings. A reversible model, which is valid independently of which
an off-shore power system is analyzed in Ref. [3], in order to iden- of two windings is excited, is presented in Ref. [8]. Refs. [913]
tify resonance frequencies to avoid problematic scenarios. Some describe some hysteresis models which have been recently applied
limitations in available transformer models are mentioned in Ref. to the calculation of inrush currents, for example: a model which
[4]; consequently, a sensitivity analysis was performed to verify the considers major and minor loop formations, asymmetric core struc-
inuence of each parameter, and the paper indicates that the slope tures, and magnetic cross coupling of the legs is shown in Ref. [9];
of the magnetization curve at extreme saturation is the param- static and dynamic hysteresis models are applied in Ref. [10]; the
inverse JilesAtherton hysteresis model is applied in Ref. [11]; a
vector hysteresis model is applied in Ref. [12], in order to consider
the anisotropy; and a model which includes hysteresis, saturation,
Correspondence to: Dpto. de Conversin y Transporte de Energa, Universidad eddy current losses, and anomalous losses is shown in Ref. [13].
Simn Bolvar, Apdo. Postal 89.000, Caracas 1080, Venezuela.
E-mail address: elmersor@usb.ve

http://dx.doi.org/10.1016/j.epsr.2016.12.025
0378-7796/ 2016 Elsevier B.V. All rights reserved.
90 E. Sorrentino / Electric Power Systems Research 145 (2017) 8998

Transformer inrush currents may reach very large values relay tripping based on waveform of measured currents could be
because the transformer core may be deeply saturated during the ineffective because these relays are not measuring differential cur-
energization. Two random variables determine the magnitude of rents (these relays measure the total currents of the feeders). That
inrush currents: the voltage instantaneous value at the energiza- is, developed rules for transformer differential protection could be
tion instant, and the magnetic remanence. Both variables should ineffective in this case because these relays are measuring the sum
be in their worst case in order to obtain the maximum values of magnetizing inrush currents of downstream transformers plus
of inrush currents. On the other hand, the closing of the main cir- inrush currents of downstream loads, since distribution feeders are
cuit breaker of a distribution feeder implies that all the downstream typically reenergized with load.
transformers are simultaneously energized. Thus, an inrush current
is expected in that feeder, but a direct method to estimate the max- 2. Exact model for inrush currents in transformers
imum inrush currents for this condition was not available. That is,
a trained person could estimate these currents by using an electro- 2.1. Single-phase transformers
magnetic transient program, and subsequently the instantaneous
currents could be integrated to nd the maximum rms currents; For the no-load condition of a single-phase transformer, Ref.
however, these procedures are not usual for the coordination of [25] clearly indicates that the magnetic ux linkages in the wind-
overcurrent devices, and the proper training of engineers to cor- ing connected to the power source have two components: a part is
rectly solve these problems would imply a substantial effort. The nonlinear because it is related to the ferromagnetic core, and the
purpose of this article is to show an approximate method for the other one is linear because it is related to the space between the
straightforward calculation of the maximum expected rms value of core and the winding connected to the power source. Thus, induced
these currents, by using only simple algebraic functions. This fea- voltage (e) in the winding connected to the power source is:
ture is useful for the coordination of overcurrent devices because
e = dH /dt + L T di/dt (1)
the estimation of these currents is only a small part of the main
problem. H is the ux linkage due to the magnetic ux in the ferromagnetic
The rst step was to obtain an exact model for inrush currents core (non-linear), LT is an inductance that represents the effect of
in transformers, by solving the differential equations of the net- ux linkages due to the magnetic ux in the linear region, and i is
works. This exact model was successfully tested by comparison the current.
of its results with experimental measurements. Subsequently, the The energization can be described by a simple equation:
main objective was to obtain a transformer model whose behavior
v = (R S + R T ) i + (L S + L T ) di/dt + (dH /di) di/dt (2)
is representative of the maximum rms currents which are typi-
cally assumed in order to avoid the melting of transformer fuses. v, RS and LS are the voltage, resistance and inductance of the
These values are 25 pu at 0.01 s and 12 pu at 0.1 s [14,15], in per- power source. RT is the winding resistance. Initial conditions are i
unit of transformer rated current. Subsequently, this exact model (t = 0) = 0, and H (t = 0) = R . R is the ux linkage for the remanent
was applied in order to develop the two approximations for the ux condition.
proposed method. The nonlinear component (dH /di) is the derivative of a non-
The rst approximation is required to build an equivalent sys- linear excitation curve of the transformer (H as a function of i). A
tem in order to summarize the effect of simultaneous inrush compact equation is obtained by using RK = (RS + RT ), LL = (LS + LT ),
currents of the transformers of a distribution feeder. No previ- and LK = (LL + dH /di):
ous work was found about this subject in the reviewed literature.
di/dt = (v R K i)/L K (3)
The second approximation is required to compute the maximum
rms inrush current of a transformer by assuming that the source This equation was solved by using Eulers method. Although
impedance is not negligible. As in the previous case, no previous the exact model is based on solving differential equations (which
work was found about this specic issue in the reviewed litera- depends on instantaneous values, resistances and inductances),
ture. Only a paper was found about an estimation of the maximum some values are conveniently shown in this article as reactances,
peak of the transformer inrush current by considering an additional impedances and voltage angles for the sake of simplicity. For exam-
resistance connected in series with the transformer [16], which ple, reactances are at fundamental frequency (X = wL). The voltage
could be useful to estimate the impact of damping resistors on of the power source is assumed to be sinusoidal, at 60 Hz. In case of
transformer inrush current. Therefore, the two proposed approx- voltage angles (), the reference is taken at the energization instant,
imations are novel methods, which have been developed here in and a sine function is assumed (i.e.,  = 0 means that the energiza-
order to obtain a way to estimate the maximum rms currents in a tion is precisely at the zero-crossing instant of the voltage signal,
distribution feeder due to simultaneous inrush currents in down- for the single-phase case).
stream transformers. The use of mathematical models to represent Transformer parameters were estimated in order to obtain the
physical phenomena is a common procedure in different elds of maximum rms currents which are typically assumed for a dis-
the knowledge (some examples can be seen in Refs. [1719]). tribution transformer in the coordination of overcurrent devices
Only the energization of unloaded transformers is considered (25 pu at 0.01 s, and 12 pu at 0.1 s [14,15]; these values are taken
in this article. However, the closing of the main circuit breaker of as reference). For this condition, the transformer is assumed to be
a distribution feeder usually implies the simultaneous energiza- energized by an ideal power source (RS and LS are zero), just when
tion of downstream transformers with their own loads. This fact the voltage is crossing by zero, and R is 0.7 pu. In order to obtain the
implies additional inrush currents due to the dynamic of the loads greatest values of rms current for 0.01 s, the integration for comput-
[2024], which are out of the scope of this article. An estimation ing the rms current was initiated 70 after the energization instant
of these additional inrush currents requires the specic knowledge (i.e., with this procedure, the rst peak of the inrush current is at
about the dynamic behavior of the loads, and accurate informa- the middle of the integration interval).
tion about this point is not typically available. Therefore, a coarse Two excitation curves were applied (Fig. 1). The values of
estimation of these additional inrush currents (based on the avail- current taken as reference can be obtained for both curves (for
able knowledge about the load) is typically necessary. This coarse curve A: RT = 0.0043 pu, XT = 0.0287 pu; for curve B: RT = 0.0038 pu,
estimation would be still necessary in order to complement the XT = 0.0134 pu). Table 1 shows the obtained rms currents for dif-
proposed approximate method. On the other hand, the blocking of ferent source impedances (ZS = RS + j XS ). These results show that
E. Sorrentino / Electric Power Systems Research 145 (2017) 8998 91

Table 2
Comparison of maximum rms currents in star case (impedances and currents are
in pu).

ZN = RN + jLN ZS+ = ZS = ZS0  (IMAX ) IMAX I ( = 0)

0.1 + j 0.1 ZS+ = 0.1 + j 0.1 0 3.42 3.42


0.1 + j 0.1 ZS+ = 0 5 20.45 20.42
0+j 0 ZS+ = 0.1 + j 0.1 0 3.80 3.80
1+j 1 ZS+ = 0.1 + j 0.1 5 3.28 3.27
1+j 1 ZS+ = 0 0 20.17 20.17

Fig. 1. Applied excitation curves.


iA
A
Table 1 AT i1 i[I3CC1 ]AT ICCBT
Computed values of maximum rms currents (pu) in a single-phase transformer, by iB I=0
using two different excitation curves. B
@ 0.01 s @ 0.1 s i
I=0 I=0
C 2
I=0

ZS (pu) 0 0.01/45 0.05/45 0 0.01/45 0.05/45
Curve A 25.0 17.6 8.25 12.0 6.96 2.86
AT iC
Curve B 25.0 17.0 7.70 12.1 6.73 2.68
Fig. 3. Nomenclature for currents in delta case (currents are only in primary wind-
ings).

different excitation curves can be applied in order to obtain the


maximum rms currents taken as reference. Fig. 2 shows the different values of neutral and source impedances, Table 2 shows
obtained currents when curve A was used (curve A is the selected the required angle ((IMAX )) for vA at the energization instant, in
curve for the remaining results of this article, with RT = 0.0043 pu order to obtain IMAX for an integration time interval of 0.01 s. Addi-
and XT = 0.0287 pu). tionally, IMAX is shown in Table 2, as well as the rms current for the
case when energization occurs at vA zero-crossing (I( = 0)). The
2.2. Three-phase banks of single-phase transformers value of IMAX is very similar to I( = 0); therefore, for the sake of
simplicity, the value of I( = 0) was computed in order to obtain the
Three-phase banks of single-phase transformers are analyzed approximate methods.
in this article. Results are also valid for three-phase transform- The neutral impedance (ZN = RN + jLN ) reduces the neutral cur-
ers with ferromagnetic path for zero-sequence magnetic ux (e.g., rents; however, its effect on maximum phase currents is not very
units with 5 or 4 legs). Three-phase three-leg transformers need high. If the ground path has high impedance, the other phases
a different analysis, and they are out of the scope of this article. become saturated (and the return of current is by the other phases
Secondary winding is assumed to be in star connection. Primary instead of the ground path).
winding in star connection and in delta connection are indepen-
dently analyzed here (they are called star case and delta case,
2.2.2. Primary winding in delta connection (delta case)
respectively).
The method is similar to the previous one. In this case, [v] is
composed of line-to-line voltages (vAB , vBC , vCA ), and [i] is composed
2.2.1. Primary winding in star connection (star case) of currents in windings (i1 , i2 , i3 ) as shown in Fig. 3. Therefore, Eq. (5)
The neutral-to-ground connection was considered to be through also describes this case and it can be easily solved. Subsequently, the
a resistance (RN ) connected in series with an inductance (LN ). Thus, line currents can be computed (iA = i1 i3 ; iB = i2 i1 ; iC = i3 i2 ).
the voltage equations can be written in matrix form: The maximum current in a winding is reached when ener-
[v] = [R K ] [i] + [L K ] [di/dt] (4) gization occurs at the zero-crossing instant of the corresponding
line-to-line voltage. It was found that the maximum value of iA
[v] is composed of the phase voltages (vA , vB , vC ), and [i] is composed is reached when the phase of vAB at the energization instant is at
of the line currents (iA , iB , iC ). [RK ] includes all the circuit resistances, +30 from its zero-crossing instant. The maximum value of iA is
as well as [LK ] includes the linear inductances (LL ) and nonlin- obtained when the corresponding winding currents (i1 , i3 ) are rel-
ear elements (dH /di). Therefore, the solving method is similar to atively near to their maximum values (but they are not exactly at
the applied one for the single-phase case, by using the following their maximum values).
equation: At Section 2.1, the transformer parameters were found in order
to obtain a maximum rms current of 25 pu in 0.01 s and 12 pu
[di/dt] = [L K ]1 ([v] [R K ] [i]) (5)
in 0.1 s, for a single-phase transformer. These per-unit values of
The maximum rms current in phase A (IMAX ) is not always found currents are the same for the star case. However, delta case
when energization occurs at the vA zero-crossing instant ( = 0). For is different because two inrush currents (from two single-phase

Fig. 2. Maximum inrush currents for a single-phase transformer, varying the source impedance (: ZS = 0; : ZS = 0.01 pu/45 ; : ZS = 0.05 pu/45 ).
92 E. Sorrentino / Electric Power Systems Research 145 (2017) 8998

R5, L5 R4 , L 4 R3, L3 3. First approximation: inrush current in a feeder, due to


downstream transformers

t=0 T1 T2 T3
R1, L1 R2, L2 3.1. Description of the approximate method

i1 i2 i3
The approximate method consists of using an equivalent trans-
former connected in series with an equivalent impedance. The kVA
Fig. 4. Sample circuit with three transformers which are simultaneously energized.
rating of the equivalent transformer (SEQ ) is the sum of all the kVA
ratings of the installed downstream transformers. Many options
for computing the equivalent impedance were proposed, analyzed
and tested. Finally, only one option was considered suitable; for this
windings) are combined to obtain the line current. In per-unit of option, the equivalent impedance (ZEQ = REQ + jXEQ ) is calculated by
the rated current for the windings, the maximum line current was using the following expression:
found to be 32.6 pu @ 0.01 s and 15.8 pu @ 0.1 s. As the rated line N
  w 
currents are 1.73 times greater than the rated currents for the wind- REQ =  Rj Sj /SEQ ;
j=1
ings, these values are only 18.8 pu @ 0.01 s and 9.1 pu @ 0.1 s, in
per-unit of the rated line currents. N
  w 
XEQ =  Xj Sj /SEQ (6)
j=1

2.3. Simultaneous inrush currents in transformers The branch impedances are Zj (Zj = Rj + j Xj ), and N is the number
of branches. Sj is the sum of the kVA ratings of the transformers
Again, the method is based on solving Eq. (5). Fig. 4 shows a cir- located downstream of the branch impedance Zj . The exponent w
cuit where three transformers (T1 , T2 and T3 ) are simultaneously needs to be found in order to obtain the best approximation. Thus,
energized at t = 0. In the single-phase case, there is a nonlinear ele- the main problem here is to determine the value of the exponent
ment for each transformer. In the three-phase cases, there are three w.
nonlinear elements for each transformer bank. For the sake of sim- The exponent w is found by comparing the solution of the exact
plicity, a simple single-phase case is shown in this section, with model with the solution of the approximate model. The exact model
R2 = R3 = R5 = 0, L2 = L3 = L5 = 0, R1 = R4 , and L1 = L4 . Thus, the short consists in the solution of all the differential equations for the circuit
circuit level is the same at each transformer location. The three (i.e., by considering detailedly the current in each transformer). The
transformers have the same kVA rating in this simple example. approximate model consists of the solution of the differential equa-
Fig. 5 shows the current in each transformer (i1 , i2 , i3 ) and the total tion only for the equivalent transformer with its series impedance
current (iTOT ), for two cases of impedances (Z1 = R1 + j L1 ; Z4 = R4 + j (ZEQ ).
L4 ): ) Z1 = Z4 = 0.01 pu/45 ; ) Z1 = Z4 = 0.05 pu/45 .
3.2. Description of analyzed cases

Single-phase transformers, and three-phase banks (star case


2.4. Experimental verication of these exact models and delta case) were analyzed. Four congurations were ana-
lyzed, by applying combinatorial variations of branch impedances
These models were veried by comparing the computed results and kVA ratings. The number of combinations NC is the same for
with measured values, for many different conditions. Some exam- single-phase units and for delta case. In star case, the number
ples are shown in Appendix A. of combinations NCY is nine times greater than NC because there

Fig. 5. Computed results for circuit of Fig. 4, with R2 = R3 = R5 = 0, L2 = L3 = L5 = 0, R1 = R4 , and L1 = L4 . The three transformers are identical single-phase units. Currents are
shown in per-unit of the rated current of single-phase units. ) Z1 = Z4 = 0.01 pu/45 ; ) Z1 = Z4 = 0.05 pu/45 .
E. Sorrentino / Electric Power Systems Research 145 (2017) 8998 93

Table 5
Combinatorial test cases for conguration C3.
10 11 12
SSC [pu] [ ]
t=0
1 2 3 4 5 6 7 N5 N4 T1 T2 T3 N5 N4 T1 T2 T3

500 400 400 300 300 70 70 50 50 50


8 9 500 400 20 20 20 70 50 50 30 30
500 400 100 100 20 50 50 50 50 50
500 100 20 80 80 50 50 50 30 30
13 14 15 16 17 500 100 20 20 20
500 20 100 15 15
100 80 80 20 20
18 19 20 21
100 20 20 15 15
20 15 15 10 10
Fig. 6. Conguration C4: circuit with 21 transformers.

Table 3 Table 6
Combinatorial test cases for conguration C1. Combinatorial test cases for conguration C4.

SSC [pu] [ ] SSC [pu] Main 100 50 20 100 100 50


Sub-branch 100 50 20 50 20 20
T1 T2 T3 T1 T2 T3
 [ ] Main 75 45 15 75 45
500 400 300 70 50 50 Sub-branch 75 45 15 45 15
500 100 80 70 50 30
500 100 20 50 50 50
125 100 80 50 50 30
30 25 20 50 30 30

Table 4
Combinatorial test cases for conguration C2.

SSC [pu] [ ]

N5 T1 T2 T3 N5 T1 T2 T3

500 400 300 300 70 50 50 50


500 400 80 20 70 50 30 30
500 20 20 20 50 50 50 50
100 80 80 80 50 50 30 30
100 80 20 20
100 20 20 20
20 15 15 15
20 15 10 10

is also a loop for the combinatorial test cases of zero-sequence


impedances. Details of the combinatorial test cases are in the fol-
lowing paragraphs. The congurations are:

a) Conguration C1. Fig. 4 (R1 = R2 = 0, L1 = L2 = 0). NC = 625. Fig. 7. Examples of results for the rst approximation.

b) Conguration C2: Fig. 4 (R4 = 0, L4 = 0). NC = 800.


c) Conguration C3: Fig. 4 (without null elements). NC = 900. shown in Table 6 (SSC is in per-unit of Sj , which is the sum of the
d) Conguration C4: Fig. 6 (21 transformers). NC = 72. kVA ratings of the transformers located downstream of the branch
impedance Zj ); but the combinations of the three rst cases of SSC
Each conguration was analyzed for three integration time with the two last cases of were not used.
intervals: 0.01 s, 0.05 s, 0.1 s.
For C1, C2, and C3: the kVA rating of transformer 1 is the refer- 3.3. Results
ence (1 pu), and the combinations of kVA ratings of transformers
2 and 3 are S = [0.1, 0.4, 1, 4, 10]. Branch impedances are derived Different objective functions could be selected in order to mini-
from short circuit levels, whose module (SSC ) and angle () are in mize the error of the approximation. The maximum absolute value
Tables 35 (e.g., Table 5 shows 9 combinations of SSC and 4 com- of the percentage error (E%-MAX ) was selected here. In order to
binations of , for the 36 combinations of branch impedances of search the optimal value for w, the value of E%-MAX is simply com-
conguration C3). SSC is in per-unit of SEQ . N4 and N5 are points at puted as a function of the exponent w, for all the combinations of
load side of impedances Z4 and Z5, respectively. each analyzed conguration.
For C4, the kVA ratings of transformers were alternated by using Fig. 7 shows two examples of results, for two integration time
the numeration of Fig. 4. Three cases of kVA ratings were applied intervals. For conguration C2 in delta case, the result is the value
(SCASE1 = [1, 1, 1]; SCASE2 = [1, 4, 10]; SCASE3 = [0.1, 1, 10]). For exam- of E%-MAX for the 800 combinations, as a function of the exponent
ple, kVA ratings for SCASE3 are: 0.1 pu for units 1, 4, 7, 10, 13, 16 w. For conguration C4 in star case, the result is the value of E%-MAX
and 19; 1 pu for units 2, 5, 8, 11, 14, 17 and 20; and 10 pu for units for 648 (72 9) combinations.
3, 6, 9, 12, 15, 18 and 21. Transformers with numbers in the range Fig. 8 shows the global result for all the cases, i.e.,: four congu-
17 and 1317 are considered to be in the main feeder, and the rations (C1C4), three transformer types (single-phase case, delta
other ones are in sub-branches. Branch impedances are computed case, star case), and three integration time intervals (0.01 s, 0.05 s,
from the combination of 6 cases of SSC and 5 cases of  which are 0.1 s). The minimum value of the maximum error (23%) is obtained
94 E. Sorrentino / Electric Power Systems Research 145 (2017) 8998

eg . (p.u.)
50 28
28
global

mx (%)

0 .0 1 s(pu)
40 24
24
20
20
30
E%-MAX

@ @0.01s
16
16
20 12
12
10
Error
88

eficaz
0 44

IRMS
00

Corriente
2.0 2.1 2.2 2.3 2.4 2.5

10000
10000
7000
4000
2000
1000
700
400
200

77
7000
4000
2000
1000
700
400
200
100
70
40
20
10

44
22
10
100
70
40
20
Valor de w
m (pu)
Va l ore s de m (p.u .)

Fig. 8. Global results for the rst approximation.

. (p.u.)
14
14

(pu)
12
12
with w = 2.375 (for this condition, the average of the absolute val-

0 .1 seg
10
10
ues of the percentage errors is 4%). E%-MAX does not vary very much

@ @0.1s
88
if w is between 2.175 and 2.425; therefore, a value of w = 2.3 can be
66
also recommended because it is in the middle of the range.

eficaz
44
Actually, there are many possible congurations for distribution

IRMS
22
feeders, as well as for location and sizes of transformers in such

Corriente
00
feeders. Therefore, the recommendation about using w = 2.3 can be

10000

700

200
400
7000
4000
2000
1000
10000

70
40
20
10
77
44
22
100
7000
4000
2000
1000
700
400
200
100
70
40
20
10
seen as a prudent engineering criterion, in order to be relatively far
from the values where obtained errors tended to increase with a m (pu)
Va l ore s de m (p .u .)
higher slope in Fig. 8.

s eg . (p.u.)
55

@ @1s1 (pu)
4. Second approximation: inrush current in a transformer 44
considering the source impedance
33

22
eficaz
The maximum inrush currents were analyzed by varying the
short-circuit level (SSC ) at the transformer location, in per-unit
IRMS
11
Corriente

of the transformer rating (ST ). In this article, the ratio |SSC /ST | is
denoted by m, and  is the angle of SSC . For three-phase cases, SSC 00
10000
10000
7000
4000
2000
1000
700
400
200

77
7000
4000
2000
1000
700
400
200

70
40
20
10

44
22
100
100
70
40
20
10
and ST are three-phase values. Fig. 9 shows examples of maximum
rms inrush currents in a single-phase transformer for the analyzed m (pu)
Va l ore s de m (p.u .)
values of m and , and for three integration time intervals. For each
cell of the horizontal axis, m is constant, and  varies from 0 (left)
Fig. 9. Examples of maximum rms inrush currents in a single-phase transformer,
to 90 (right), in 10 steps.
depending on the source impedance (in the gure, m = |SSC /ST | is the ratio of the
For delta case, the graphs are similar to those of Fig. 9, as a func- short-circuit kVA to the transformer rated kVA).
tion of the three-phase short circuit level. For star case, there is
one graph for each condition of zero-sequence impedance (Z0 ). approximate current (IAPR-1 ) is computed by using the following
Positive- and negative-sequence impedances are assumed to be equations:
identical (Z+ ). Thus, two additional variables are necessary in star
case: F0 is |Z0 /Z+ |, and Z0 is the angle of Z0 . IAPR-1 = f (y, ) = f 1 (y) + g (y, ) = f 1 (y) + g1 (y) g2 () (7)
The approximation consists in obtaining the maximum rms 3 2
f 1 (y) = A (/2 atan (a y + b y + c y + d)) (8)
inrush currents by using algebraic functions. Many options of alge-
2
braic functions were proposed, analyzed and tested, and nally two g1 (y) = B/(e (y y0 ) + 1) (9)
options are recommended here. The maximum absolute value of
g2 () = (1/cos (C ( 0 ))) 1 (10)
the percentage error (E%-MAX ) was selected as the objective function
to be minimized. In the results, the average of the absolute values f, f1 , g1 and g2 are functions, and m0 , A, a, b, c, d, e, B, y0 , C and  0
of the percentage errors (<E% >) is also shown. In order to search are the required constants for the approximation.
the optimal parameters for the algebraic functions, two different For star case, IAPR-1 is divided by a correction function (h1 ).
optimization tools were combined: the fminsearch function of Another logarithmic variable (u0 = ln(F0 )) is useful to compute the
Matlab, and the solver of Excel. The combination of both tools approximate current (IAPR-1-Y ):
was important because this optimization problem is highly nonlin-
IAPR-1-Y = IAPR-1 /h1 (y, , u0 , Z0 ) = f(y, )/h1 (y, , u0 , Z0 )
ear. That is, when a tool was not able to nd a path to move out of
a local optimum, that point was applied as initial point to the other (11)
tool, and this iterative procedure was repeated until the nding
of acceptable solutions. On the other hand, the use of momentary
changes in the objective function was also useful for this purpose h1 (y, u0 ) = 1 + 1/[h2 (u0 ) h3 (y, u0 ) + h4 (u0 )] (12)
(e.g., the use of minimization of <E% >, while E%-MAX was restrained
to be lower than a selected top value). h2 (u0 ) = D (1 (2/) atan (k (u0 0 ))) (13)
2
h3 (y, u0 ) = (y (p u0 + q)) (14)
4.1. Option 1 for the second approximation
h4 (u0 ) = 1/(r u0 + s) (15)

A logarithmic variable (y) is useful for this approximation. y is h2 , h3 and h4 are functions, and D, k, 0 , p, q, r and s are the required
ln(m0 /m), and m0 is a constant which was set to m0 = 1000. The constants for the approximation.
E. Sorrentino / Electric Power Systems Research 145 (2017) 8998 95

Table 7 Table 10
Constants for single-phase transformers (option 1). Constants for single-phase transformers (option 2).

t [s] @ 0.01 @ 0.05 @ 0.10 @ 0.50 @ 1.0 t [s] @ 0.01 @ 0.05 @ 0.10 @ 0.50 @ 1.0

A [p.u.] 8.0257 4.7955 3.7276 1.6649 1.1637 MI [pu] 0.04000 0.06548 0.08333 0.17690 0.25010
a 0.1293 0.1582 0.1276 0.1349 0.1468  I [ ] 49.43668 30.85890 21.89142 20.88289 14.84596
b 1.2890 1.4559 1.0839 1.1756 1.3095 a2 0.05897 0.09057 0.11447 0.12296 0.16006
c 5.0440 5.1259 3.7597 4.1439 4.6137 A2 0.35505 0.16599 0.89066 2.22407 4.40686
d 7.2888 6.4884 4.8556 5.4111 5.9492 B2 1.11080 3.36357 5.35687 12.20415 17.52567
B [p.u.] 5.8074 2.5425 1.1100 0.3315 0.2314  2 [ ] 28.94768 25.56604 26.69278 25.23631 28.98942
e 0.5815 0.3613 0.3433 0.2573 0.2435
<E% > 0.802 1.193 1.657 2.104 1.977
y0 4.0716 3.4522 3.4823 3.5711 3.5421
E%-MAX 2.197 3.807 5.230 6.155 6.457
C [1/ ] 0.8847 1.1245 1.2599 1.3359 1.3119
 0 [ ] 35.9843 30.2310 28.7307 28.7056 27.5500

<E% > 1.3178 2.1281 3.2757 3.8834 3.7839 Table 11


E%-MAX 3.9796 5.5820 7.2714 8.7988 8.8062 Constants for delta case (option 2).

t [s] @ 0.01 @ 0.05 @ 0.10 @ 0.50 @ 1.0


Table 8 MI [pu] 0.04580 0.07525 0.09650 0.20400 0.28900
Constants for delta case (option 1).  I [ ] 48.27341 28.53303 20.68348 18.50267 14.60430
a2 0.08342 0.12947 0.15107 0.18177 0.21961
t [s] @ 0.01 @ 0.05 @ 0.10 @ 0.50 @ 1.0
A2 0.42816 0.23701 0.85895 2.39095 3.79217
A [p.u.] 6.4851 3.9078 2.9140 1.3172 0.9374 B2 1.06802 3.48217 5.40535 12.19268 16.61683
A 0.1253 0.1259 0.1224 0.1361 0.1314  2 [ ] 25.29035 25.26092 25.74452 25.67516 27.94233
B 1.2922 1.1905 1.1157 1.2567 1.2013
C 5.1805 4.4813 4.1379 4.5779 4.3793 <E% > 0.642 1.180 1.520 1.910 1.856
D 7.6101 6.0344 5.6236 6.0710 5.8311 E%-MAX 2.016 3.494 4.922 5.764 5.967
B [p.u.] 7.1154 2.1750 1.0018 0.2529 0.1934
E 0.3612 0.4096 0.3347 0.2110 0.2215
y0 3.7492 3.5460 3.5675 3.5197 3.5503 Table 12
C [1/ ] 0.7733 1.0777 1.2030 1.2065 1.2195 Constants for star case (option 2).
 0 [ ] 33.0595 26.1742 26.1463 21.8986 22.9412 t [s] @ 0.01 @ 0.05 @ 0.10 @ 0.50 @ 1.0
<E% > 1.0644 2.7522 3.1156 3.7409 3.7512 MI [pu] 0.03918 0.06405 0.08585 0.17212 0.23794
E%-MAX 4.3247 6.4461 7.6178 8.7870 8.6835  I [ ] 56.39980 45.77784 36.80111 20.76681 10.00248
a2 0.07535 0.07153 0.09897 0.19286 0.36111
A2 0.47813 0.24500 0.35249 2.01838 2.28233
Table 9 B2 0.89373 2.96156 4.61896 10.01604 9.95437
Constants for star case (option 1).  2 [ ] 21.81001 18.88739 21.75369 25.92622 28.80486
t [s] @ 0.01 @ 0.05 @ 0.10 @ 0.50 @ 1.0 D 1.48000 3.00121 1.27397 0.67735 0.21147
k 2.81727 7.61523 3.05910 2.15009 1.15639
A [p.u.] 8.0828 4.8011 3.6757 1.6498 1.1668 0 0.25960 0.42189 0.68450 0.69627 1.07123
a 0.1617 0.1693 0.1604 0.1541 0.1852 p 0.91239 0.58794 0.49998 0.52411 0.27349
b 1.7296 1.6789 1.5599 1.4499 1.8572 q 7.49885 6.14237 6.71487 6.89841 7.34776
c 6.8443 6.1873 5.6834 5.1548 6.8844 r 0.11305 0.11582 0.11847 0.09768 0.05991
d 9.5746 8.0071 7.2506 6.5463 8.8797 s 0.11024 0.07798 0.10180 0.17246 0.56139
B [p.u.] 4.7875 2.3746 1.2555 0.2470 0.1847
e 0.4181 0.3934 0.3232 0.2189 0.2169 <E% > 3.510 3.728 4.156 4.332 4.600
y0 3.9010 3.5825 3.6571 4.1483 3.6834 E%-MAX 10.967 12.712 14.294 17.063 18.122
C [1/ ] 0.8633 1.0363 1.1365 1.2617 1.3107
 0 [ ] 30.4246 23.6786 22.9356 24.2581 26.5270
D 1.2647 0.6706 0.3315 0.6500 0.5599 g4 (M) = 1 a2 ln(M) (19)
k 1.9100 1.7088 1.3901 1.4000 1.5012
0 0.2223 0.5875 0.9472 0.4700 0.6936 g5 () = A2 + B2 cos( 2 ) (20)
p 0.8898 0.8776 0.5457 0.5318 0.9353
q 7.7589 7.9286 8.1416 7.6849 7.7061 g3 , g4 and g5 are functions, and MI , I , a2 , A2 , B2 and 2 are the
r 0.0439 0.0300 0.0045 0.0395 0.0677 required constants for the approximation.
s 0.2959 0.3640 0.5350 0.3727 0.1773
For star case, IAPR-2 is also divided by h1 in order to compute the
<E% > 2.7959 3.2489 3.5062 4.4927 4.9966 approximate current (IAPR-2-Y ):
E%-MAX 10.256 11.532 12.164 14.069 14.696
IAPR-2-Y = IAPR-2 /h1 (y, , u0 , Z0 ) (21)

Tables 79 show the constants that minimize the error E%-MAX For this option, cases with  or  Z0 equal to 90
were not
(the value of <E% > is also shown). included, because they are not strictly practical and they incre-
ment the error of approximation. Instead of them, values of 
4.2. Option 2 for the second approximation and  Z0 equal to 85 were selected for the combinatory of cases.
Tables 1012 show the constants that minimize the error E%-MAX
This approximation uses a weighted sum of the complex value of (the value of <E% > is also shown).
source impedance (ZS = M/); M = (1/m) and a constant impedance
(ZI , whose module and angle are MI and I , respectively), to com- 4.3. General comparison between both options
pute the approximate current (IAPR-2 ):
Both analyzed options offer a reasonable accuracy for the
IAPR-2 = |1/[g3 (M, ) ZS + ZI ]| (16) approximation. Errors are not comparable in a simple way because
ZI = MI /I (17) the greatest values of  and  Z0 are not identical (90 for option
1, and 85 for option 2). In general, the average error (<E% >, in
g3 (M, ) = g4 (M) g5 () (18) Tables 712) is lower than 5% for all the cases. Therefore, from the
96 E. Sorrentino / Electric Power Systems Research 145 (2017) 8998

point of view of the accuracy of the results, both options can be in downstream transformers, was developed. The approximate
recommended. method is based on two novel approximations, which only use
The number of required constants is lower in option 2. On the simple algebraic functions. The rst approximation is required to
other hand, option 1 does not require the use of complex numbers. build an equivalent system for the simultaneous inrush currents of
transformers in a distribution feeder. This equivalent system con-
5. General analysis of results and summary of the proposed sists of a single transformer, which is fed by an ideal voltage source
approximate method connected in series with an impedance. The second approxima-
tion is required to obtain the maximum inrush rms current of a
5.1. General analysis of results transformer when the source impedance is not negligible. Thus,
the proposed approximate method simply requires two steps: (a)
Initially, a method for the exact calculation of inrush cur- compute the source impedance for the equivalent system, using the
rents in distribution transformers, by using the numerical solution rst approximation; (b) compute the maximum rms inrush current
of the system of differential equations, was successfully imple- in the feeder, using the second approximation.
mented. The results of this method were compared to results of The rst approximation only needs an exponent (w), in order
experimental measurements with a satisfactory degree of accuracy to compute the source impedance for the equivalent system. The
(Appendix A shows several examples of comparison between com- exponent w that minimizes the approximation error is 2.375, but
puted results and experimental measurements). On the other hand, this error does not vary very much if w is between 2.175 and 2.425.
this exact method was used to obtain the transformer parameters A value of w = 2.3 is recommended because it is in the middle of
in order to represent the rms inrush currents which are typically this range.
assumed in coordination of protective devices (i.e., 25 pu at 0.01 s, The second approximation only needs a set of constants, in order
and 12 pu at 0.1 s, when the power source impedance is negligible). to apply algebraic functions, and these constants are shown for
This model was applied in order to develop the two approximations diverse integration time intervals. Two options, based on different
or the proposed approximate method. algebraic functions, are shown because both are suitable for this
The rst approximation needs an exponent (w), in order to approximation.
compute the source impedance for the equivalent system. This
approximation was tested by applying combinatorial variations of Acknowledgements
branch impedances and transformer kVA ratings in four congu-
rations of circuits that were taken as examples. The exponent w Author is grateful to Isaas Latuff, Felipe Garayar, Jorge Melin,
that minimizes the maximum percentage error of the approxima- Alberto Abouganem and Luis Prez, for their valuable help.
tion is 2.375. However, the error does not vary very much if w is
between 2.175 and 2.425, and a value of w = 2.3 can be also rec- Appendix A. Some examples of verication of the exact
ommended because it is in the middle of the range. The average of model which is described in Section 2.
absolute values of percentage errors is in the order of 4% for this
approximation. All the experimental tests were performed with 2.7 kVA single-
The second approximation needs a set of constants, in order to phase transformers (240 V/120 V). Thevenin equivalent circuit of
apply algebraic functions whose results are similar to the results the power source was estimated from experimental measure-
of differential equations. These constants were found for diverse ments. Thevenin voltage is 123.6 V (phase-to-neutral), and 212 V
integration time intervals. Two options, based on different alge- (line-to-line). Thevenin impedances are 0.1237 + j 0.0320 /phase,
braic functions, are considered suitable for this approximation. In and 0.1429 + j 0.0373  for the neutral wire. All the tests were per-
general, the average of absolute values of percentage errors is lower formed by energizing the side of 120 V, except the tests for delta
than 5% for this approximation. case which were performed at the side of 240 V. Very many inrush
The main strategy to obtain certainty about the validity of the currents were recorded for each case, in order to have certainty
results was the test of these approximations for a wide range of about the maximum inrush currents. Curve A (Fig. 1) was selected,
cases. On the other hand, the proper selection of algebraic functions and transformer parameters are not the generic ones (which are
was important for the second approximation, in order to obtain
functions which stabilize to specic values for very high values of
m, and to non-negative values for very low values of m.

5.2. Summary of the proposed method to compute the maximum


rms current in a distribution feeder due to simultaneous inrush
currents in downstream transformers

The proposed approximate method simply requires the follow-


ing steps:

a) Compute the source impedance for the equivalent system, using


Eq. (6).
b) Compute the maximum inrush current in the feeder, using the
source impedance of the previous step and the algebraic func-
tions described in Section 4 (option 1 or 2).

6. Conclusion

An approximate method to compute the maximum rms current Fig. A1. Examples of measured (m) and computed (c) results for a single-phase
in a distribution feeder, due to the simultaneous inrush currents transformer. ) ZAD = 0; ) ZAD = 0.2 /0 .
E. Sorrentino / Electric Power Systems Research 145 (2017) 8998 97

Fig. A2. Examples of measured (m) and computed (c) results for phase currents in Fig. A5. Examples of measured (m) and computed (c) results for simultaneous
star case. ) ZN = 1.935 /0 ; ) ZN = 4.95 /0 . A is the angle of vA . energization of single-phase transformers. ) three units in parallel, without addi-
tional impedance; ) for circuit of Fig. 4: Z5 = ZTH + 0.39 /0 ; Z4 = Z3 = 0.39 /0 ;
Z2 = Z1 = 0 /0 .

0.6 pu; except for delta case, where R is 0.45 pu (because Thevenin
voltage is lower than 1pu in delta case).
Figs. A1A5 show examples of measured and computed results.
In single-phase case and delta case, additional impedances (ZAD )
were included in series with the transformer. In star case, a neutral
impedance was included (ZN ). In case of simultaneous energization
of single-phase units, included impedances are in accordance with
Fig. 4.

References

[1] J. Fleming, Experimental researches on alternate current transformers, J. Inst.


Electr. Eng. 21 (1892) 594686.
[2] T. Yensen, Starting current of transformers, Univ. Ill. Bull. 55 (1912) 143.
[3] R. Turner, K. Smith, Transformer inrush currents, IEEE Ind. Appl. Mag. 16 (5)
(2010) 1419.
[4] N. Chiesa, B. Mork, H. Hidalen, Transformer model for inrush current
calculations: simulations, measurements and sensitivity analysis, IEEE Trans.
Fig. A3. Examples of measured (m) and computed (c) results for neutral current in
Power Deliv. 25 (4) (2010) 25992608.
star case. ) ZN = 0 /0 ; ) ZN = 1.935 /0 . A is the angle of vA .
[5] A. Farazmand, F. De Len, K. Zhang, S. Jazebi, Analysis, modeling, and
simulation of the phase-hop condition in transformers: the largest inrush
currents, IEEE Trans. Power Deliv. 29 (4) (2014) 19181926.
[6] F. De Len, A. Farazmand, P. Joseph, Comparing the T and equivalent circuits
for the calculation of transformer inrush currents, IEEE Trans. Power Deliv. 27
(4) (2012) 23902398.
[7] S. Jazebi, F. De Len, A. Farazmand, D. Deswal, Dual reversible transformer
model for the calculation of low-frequency transients, IEEE Trans. Power
Deliv. 28 (4) (2013) 25092517.
[8] S. Zirka, Y. Moroz, C. Arturi, N. Chiesa, H. Hidalen, Topology-correct reversible
transformer model, IEEE Trans. Power Deliv. 27 (4) (2012) 20372045.
[9] P. Moses, M. Masoum, H. Toliyat, Dynamic modeling of three-phase
asymmetric power transformers with magnetic hysteresis: no-load and
inrush conditions, IEEE Trans. Energy Convers. 25 (4) (2010) 10401047.
[10] S. Zirka, Y. Moroz, A. Moses, C. Arturi, Static and dynamic hysteresis models
for studying transformer transients, IEEE Trans. Power Deliv. 26 (4) (2011)
23522362.
[11] R. Naghizadeh, B. Vahidi, S. Hosseinian, Modelling of inrush current in
transformers using inverse JilesAtherton hysteresis model with a
neuro-shufed frog-leaping algorithm approach, IET Electr. Power Appl. 6 (9)
(2012) 727734.
[12] J. Leite, A. Benabou, N. Sadowski, Transformer inrush currents taking into
account vector hysteresis, IEEE Trans. Magn. 46 (8) (2010) 32373240.
[13] J. Faiz, S. Saffari, Inrush current modeling in a single-phase transformer, IEEE
Trans. Magn. 46 (2) (2010) 578581.
[14] M. Bishop, S. Mendis, J. Witte, K. Leix, Selecting overcurrent protection for
Fig. A4. Examples of measured (m) and computed (c) results for delta case. )
three-phase transformers, IEEE Ind. Appl. Mag. 2 (2) (1996) 3541.
ZAD = 0 /0 ; ) ZAD = 0.39 /0 . AB is the angle of vAB .
[15] IEEE Tutorial Course: Application and Coordination of Reclosers,
Sectionalizers and Fuses, Course Text 80 EH0157-8-PWR.
[16] Y. Wang, S. Abdulsalam, W. Xu, Analytical formula to estimate the maximum
inrush current, IEEE Trans. Power Deliv. 23 (2) (2008) 12661268.
excellent to obtain 25 pu in 0.01 s and 12 pu in 0.1 s, as maximum [17] M. Valipour, Optimization of neural networks for precipitation analysis in a
rms currents for a distribution transformer). Transformer parame- humid region to detect drought and wet year alarms, Meteorol. Appl. 23 (1)
ters for this specic case are RT = 0.0043 pu and XT = 0.016 pu. R is (2016) 91100.
98 E. Sorrentino / Electric Power Systems Research 145 (2017) 8998

[18] M. Valipour, M. Banihabib, S. Behbahani, Comparison of the ARMA, ARIMA, [22] O. Mirza, Usage of CLPU curve to deal with the cold load pickup problem, IEEE
and the autoregressive articial neural network models in forecasting the Trans. Power Deliv. 12 (2) (1997) 660667.
monthly inow of Dez dam reservoir, J. Hydrol. 476 (7) (2013) 433441. [23] D. Fischer, S. Cress, R. Beresh, Characterization of distribution systems
[19] M. Khasraghi, M. Sedkouhi, M. Valipour, Simulation of open- and closed-end overcurrents using expert systems, in: Proceedings of the Power Systems
border irrigation systems using SIRMOD, Arch. Agron. Soil Sci. 61 (7) (2015) Conference: Advanced Metering, Protection, Control, Communication, and
929941. Distributed Resources, Clemson, SC, USA, 2007, pp. 7884.
[20] M. Persson, W. Baig, T. Thiringer, Measurements and modelling of three- and [24] H. Weng, X. Lin, Studies on the unusual maloperation of transformer
ve-limb transformer behaviour during large voltage and frequency differential protection during the nonlinear load switch-in, IEEE Trans. Power
disturbances, IET Gener. Transm. Distrib. 10 (2) (2016) 334340. Deliv. 24 (4) (2009) 18241831.
[21] E. Kurtz, Transformer current and power inrushes under load, Electr. Eng. 56 [25] R. Yacamini, A. Abu-Nasser, Numerical calculation of inrush current in
(8) (1937) 989994. single-phase transformers, IEE Proc. B Electr. Power Appl. 128 (6) (1981)
327334.

Vous aimerez peut-être aussi