Professional Documents
Culture Documents
Transformer Winding Losses Evaluation When Supplying Non Linear Load
Transformer Winding Losses Evaluation When Supplying Non Linear Load
net/publication/224123941
CITATIONS READS
10 229
4 authors, including:
Some of the authors of this publication are also working on these related projects:
The contribution of load control in increasing electric system flexibility View project
All content following this page was uploaded by Marc Petit on 12 May 2015.
Abstract- Nonlinear loads produce harmonic currents, which Posl is stray loss in components other than the windings.
induce additional losses in transformers and cause temperature Rac is AC winding resistance.
rise, especially in the windings. The estimation of winding eddy-
current loss, in the presence of harmonics, is based on the Rdc is DC winding resistance.
knowledge of winding eddy-current loss at power frequency. I is the RMS current
This paper deals with the transformer winding loss estimation
and gives a simplified winding resistance expression that takes When a transformer supplies non linear loads, I2R increases
winding eddy-current into account and allows to estimate with the RMS current variation and winding eddy-current loss
associated losses. The impact of transformer winding
characteristics, as conductor size and layer number, on the increases with the square of the load current and the square of
resistance variation with frequency is shown. the frequency. Equation (2) allows calculating winding eddy-
current loss through a proportionality factor applied to the
hal-00422449, version 1 - 7 Oct 2009
− h1 . sinh[ jk ( x − a )] + h2 . sinh[ jk ( x + a )]
→
J h( x ) =
sinh ( j 2ka ) (7)
∂ 2H z ∂H z ωμσ ⎧ h12 + h2 2 2h h ⎫
= μσ (4) Pds = f1 (ϕ ) − 1 2 f 2 (ϕ )⎬
∂x 2
∂t ⎨ (12)
σ ⎩ 2 2 2 ⎭
A solution of this equation is of the form Hz=Re(h(x)ejωt), With
thus equation (4) becomes: 2
ϕ = eδ and δ=
ωμσ
d 2 h( x )
+ k 2 h( x ) = 0 (5)
dx 2 sinh(2ϕ ) + sin(2ϕ )
With f1 (ϕ ) = (13)
cosh(2ϕ ) − cos(2ϕ )
k 2 = − jωμσ
Where ω is the pulsation and j the complex number defined
by j² = -1.
cosh(ϕ ). sin(ϕ ) + sinh(ϕ ). cos(ϕ )
f 2 (ϕ ) = (14)
The equation (5) solution is cosh(2ϕ ) − cos(2ϕ )
Where e is the thickness of the plate and δ is the depth of
− jk x
h( x) = a1 e jk x
+ a2 e (6) penetration.
This expression can be rewritten as shown in equation (15).
It can be seen that the two functions g1(φ) and g2(φ) give
2
⎧⎪ 2
⎫⎪ good results when the ratio between thickness and depth of
h2 h1 h1
P = ⎨(1 + 2 ). f1 (ϕ ) − 4 . f 2 (ϕ )⎬ (15) penetration is less than 1.4 (errors less than 5%).
ds 2σδ ⎪⎩ h2 h2 ⎪⎭
For conductors with small thickness compared to the depth Thus, the Joule losses in the plate can be expressed
of penetration (φ<<1), functions f1(φ) and f2(φ) can be according to (18) substituting f1(φ) and f2(φ) in equation (15)
approximated by g1(φ) and g2(φ) given in (16) and (17) by g1(φ) and g2(φ).
respectively.
1⎡ 4 ⎤ h
2
⎧⎪ ⎛ e ⎞ ⎛ 4(1 + α ) + 7α ⎞⎫⎪
4 2
f 1 (ϕ ) ≈ g 1 (ϕ ) = ⎢1 + ⋅ ϕ 4 + ε (ϕ 8 ) ⎥ (16) P = 2 ⎨(1 − α ) + ⎜ ⎟ ⎜⎜
2
⎟⎟⎬ (18)
ϕ ⎣ 45 ⎦ ds 2eσ ⎪⎩ ⎝δ ⎠ ⎝ 45 ⎠⎪⎭
⎡
h
1 7 8 ⎤ With α= 1
f 2 (ϕ ) ≈ g 2 (ϕ ) = ⎢1 − 90 ⋅ ϕ + ε (ϕ ) ⎥
4
(17) h2
2ϕ ⎣ ⎦
Fig. 2 shows comparison, between functions f1(φ), f2(φ) III. APPLICATION TO TRANSFORMER WINDINGS
and g1(φ), g2(φ) for different values of the ratio e/δ. Fig. 3
shows the relative error in per cent between the functions and For analysis, the structure of transformer windings can be
their approximations. approximated by a superposition of several conducting plates
[5]. Fig. 4 shows the geometry of MV/LV transformer with
12
the LV and MV windings respectively in the inner and in the
Function f1 outer. The curvature of the conductors, edge and end effects
are neglected.
hal-00422449, version 1 - 7 Oct 2009
Function g1
10
Function f2
Using ampere’s law, magnetic fields on both the left-hand
Function g2
8 side h1 and right-hand side h2 are respectively given by (19)
and (20) for layer m.
Function values
⎛ N ⎞ iˆ
6
h1 = (m − 1) ⋅ ⎜ ⎟ (19)
4
⎝ n ⎠d
2
⎛ N ⎞ iˆ
h2 = m ⋅ ⎜ ⎟ (20)
⎝ n ⎠d
0
-2 With
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Ratio e/ δ
m the layer number from the inner to the outer.
Fig. 2. Functions f1(φ), f2(φ), g1(φ) and g2(φ ) versus ratio e/δ.
N the total number of turns in the windings.
n the total number of layers in the windings.
140
î the peak current in a turn.
100*|f1-g1|/f1 d the height of windings.
100*|f2-g2|/f2
120
Thus, the power per unit length, according to the y
100 direction, dissipated in layer m (21) can be calculated using
Joule losses expression in a plate (18) multiplied by the height
80
of windings and magnetic fields expressions given by (19)
Error in %
and (20).
60
I 2 ⎡ ⎛ e ⎞ ⎛ m 2 m 4 ⎞⎤
2 4
40
⎛N⎞
Pm = ⎜ ⎟ ⋅ ⎢1 + ⎜ ⎟ ⎜⎜ − + ⎟⎟⎥ (21)
20 ⎝ n ⎠ σ ⋅ e ⋅ d ⎢⎣ ⎝ δ ⎠ ⎝ 3 3 45 ⎠⎥⎦
Total loss per unit length in transformer windings is then
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 given by the sum of power dissipated in each layer (22).
Ratio e/ δ
I 2 ⎡ ⎛ e ⎞ 5n3 − n ⎤
Fig. 3. Error in per cent between functions f1(φ)& g1(φ) and f2(φ)& g2(φ) 2 4
n
⎛N⎞
versus ratio e/δ. P = ∑ Pm = ⎜ ⎟ ⎢n + ⎜ ⎟ ⎥
m =1 ⎝ n ⎠ σ ⋅ e ⋅ d ⎢⎣ ⎝ δ ⎠ 45 ⎦⎥ (22)
conductors is less sensitive to frequency than winding
Magnetic core
resistance with rectangular conductors. Thus when a
transformer is supplying non linear loads, winding eddy-
current loss increase is much higher for rectangular
conductors. Moreover (24) calculated with the simplified
functions g1 and g2 gives a conservative result for transformer
d n°1 as soon as e/δ is greater than 1.4 (i.e. frequency greater
than 800 Hz in the present case).
4
hmax
⎛ e ⎞ ⎛ n2 1 ⎞ 2
PEC = ∑ Rdc ⎜⎜ ⎟⎟ ⎜⎜ − ⎟⎟ I h (25)
LV winding layers MV winding layers
h =1 ⎝δh ⎠ ⎝ 9 45 ⎠
by the second term into brackets. Transformer geometry δh is the depth of penetration for harmonic h.
impact can be seen by the presence of layer number – n – and δR is the depth of penetration at the power frequency.
conductor thickness – e – in the expression.
Substituting (26) in (25), eddy-current loss expression can
2 be rewritten and gives the same result that the IEEE Std
n ⎛N⎞
Rdc = ⋅⎜ ⎟ (23) C57.110-1998 (2).
deσ ⎝n⎠
TABLE I
TRANSFORMER CHARACTERISTICS
P = Rac ⋅ I 2
Primary (MV) Secondary (LV) e/δ
⎡ ⎛ e ⎞ 4 ⎛ n 2 1 ⎞⎤ (24)
Transformer Copper, ∅0.8mm Copper, rectangular
2.5
with Rac = Rdc ⎢1 + ⎜ ⎟ ⎜⎜ − ⎟⎟⎥ n°1 21 layers Conductor 3.75*10mm²
at 2.5kHz
⎣⎢ ⎝ δ ⎠ ⎝ 9 45 ⎠⎦⎥
5400 turns 64 turns, 2 layers
Transformer Copper ∅ 0.75mm Copper sheet
0.13
n°2 19 layers 0.2*210mm²
at 2.5kHz
4417 turns 51 turns, 51 layers
IV. NUMERICAL APPLICATION
Generally, LV windings are made with rectangular or sheet 20
Transformer n°1 with g1 and g2 functions
conductors, while MV windings are made with small round 18 Transformer n°1 with f1 and f2 functions
conductors. Table I gives the winding characteristics of two 16
typical 100 kVA transformers which are used for numerical
14
application presented next. The ratio e/δ, given in the last
Rac/Rdc ratio
10
As primary winding characteristics are almost the same, 8
winding ratio evolution is similar. The main difference
between the transformers is the secondary winding geometry 6
(Table I). 4
2
Then we present in Fig. 5 and Fig. 6 the Rac/Rdc ratio 0
evolution as a function of frequency for the secondary 0 500 1000 1500 2000 2500
windings of transformer n°1 and n°2 respectively. The ratio is frequency (Hz)
Fig. 5. Secondary winding Rac/Rdc ratio evolution as a function of frequency
either calculated with (24) using g1 and g2, or with f1 and f2. for transformers n°1 presented in table I, and using functions {g1, g2} -
Theses figures show that winding resistance with sheet equation (24) - or {f1, f2}.
This paper gives a winding resistance expression which
1.12
Transformer n°2 with with g1 and g2 functions takes winding eddy-current into account and permits to
estimate winding losses at power frequency, PEC-0, and also in
1.1
presence of harmonics.
This expression shows the impact of the following winding
1.08
parameters: conductor thickness and layer number. This
Rac/Rdc ratio