Professional Documents
Culture Documents
Load Noise Prediction of A Power Transformer: Journal of Vibration and Control August 2021
Load Noise Prediction of A Power Transformer: Journal of Vibration and Control August 2021
Load Noise Prediction of A Power Transformer: Journal of Vibration and Control August 2021
net/publication/358281951
CITATION READS
1 52
5 authors, including:
Jintai Chung
Hanyang University, Ansan, South Korea
159 PUBLICATIONS 5,299 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
Robust design technology development for the noise and vibration reduction of a radiator cooling fan View project
All content following this page was uploaded by Booyeong Lee on 02 February 2022.
Abstract
In this article, we propose a new regression equation to predict the noise of a power transformer based on the winding
vibration under a loading condition. A regression between load noises and tank vibrations for multiple transformers with
different rated powers was confirmed through measurements and regression analysis. A regression equation for load noise
and winding vibration was derived considering the fact that the winding vibration level is proportional to the tank vibration
level. The electromagnetic force, which is the excitation force of the winding, was obtained using the equivalent magnetic
circuit network method to obtain the winding vibration required for the regression equation. Subsequently, the obtained force
was applied to a finite element model for the winding to achieve the vibration response. The winding vibration obtained
through these methods is closely correlated with the load noise, and the amount of winding vibration transferred to the tank
could be changed according to the distance between the tank and the winding. Accordingly, an equation for predicting the load
noise was established considering the winding vibration and the correlation factors according to the distance of the
transmission path. The proposed prediction equation is considerably more accurate than the previous prediction equation.
Keywords
Winding vibration, load noise prediction, electromagnetic force, power transformer, regression equation
Figure 2. 1/3-octave band frequency spectrum for the measured Figure 3. Load noise level versus transformer rated power.
load noise.
of the winding vibration has a closer correlation with the total number of nodes used was approximately 223,000.
load noise level compared with the rated power, repre- Figure 5 shows the electromagnetic finite element model for
senting the load noise level as a function of the magnitude of the cores and windings of transformer T1, where the finite
the winding vibration makes the prediction of the load noise element of air is not illustrated to show the finite elements of
more accurate. the core and windings.
Numerical analysis of electromagnetic fields was per-
formed using the three-dimensional (3D) equivalent mag-
3. Electromagnetic force analysis
netic circuit network (EMCN) method to calculate the
Before predicting the winding vibration, electromagnetic electromagnetic force generated in the winding (Sim et al.,
analysis models for the three-phase transformer with three 2018). First, the EMCN was constructed by linking all the
limbs were established to calculate the electromagnetic nodes of the electromagnetic finite element model by circuit
force that excites windings. The winding in which several parameters to calculate the magnetic volume force. As it is
layers of copper wires are wound was simplified to a cyl- difficult to show the EMCN for the entire system, we will
inder-shaped solid with a copper material. The tank and discuss the configuration of the EMCN for the part of the
clamps shown in Figure 1 were neglected based on the winding. Figure 6(a) is a partial view of the finite element
assumption that they have a negligible effect on the elec- model for the winding, and Figure 6(b) shows the EMCN for
tromagnetic force of the windings, and the insulators and oil this part. In Figure 6(b), i, j, and k indicate the node posi-
between the core, winding, and tank were modeled as air tioning indexes corresponding to the radial, tangential, and
because their relative permeability is similar to that of air. vertical directions to represent the node number, respectively.
The oil level may have an important influence on the In this figure, the red arrow indicates the direction of the
vibration transmission which is closely related to transformer magnetic flux between the node (i, j, k) and the six adjacent
noise. Jin and Pan (2016) presented a study on the fact that oil nodes. The magnetic flux between the two node (i, j, k) and
affects vibration transmission and they found that oil in- (i+1, j, k) on the straight line in the radial direction and the
creases vibration transmission in the low-frequency range circuit parameters that determine the flux are shown in
and decreases vibration transmission in the high-frequency Figure 7, where the red arrow represents the direction of the
range. However, Kulkarni and Khaparde (2013) claimed the magnetic flux f. In this figure, u is the MSP or the mag-
pressure generated by vibrations of the core and windings in netomotive force of a node, P is the permeance (reciprocal of
a transformer is transmitted to tank surfaces through oil but the magnetic resistance), and W is the magnetomotive force
the vibrations are transmitted without appreciable damping. in the winding current. The radial magnetic flux between the
Since the inside of the transformer tank of this study is always two nodes (i, j, k) and (i+1, j, k) shown in Figure 7 can be
filled with oil, the oil level was not considered in this article. calculated using the following equation (Sim et al., 2018)
Finite element models for the simplified analysis model
were constructed for the finite-element–based electromag- fri,j,k ¼ ui,j,k uiþ1,j,k þ Wi,j,k
r
Pi,j,k
r
, (1)
netic analysis. The finite element mesh for the 32 trans-
formers described in Table 1 was established using Gmsh where the superscript r represents the radial direction.
(Geuzaine and Remacle, 2009), which is a finite element Subsequently, we represented the constructed 3D EMCN
generation program. The finite element used was a hex- as equations in a discretized form and then converted them
ahedral element with a node in the center, and the average into a matrix-vector form to obtain the MSP. When the
Lee et al. 5
Figure 6. (a) Partial view of the magnetic finite element model for the winding and (b) equivalent magnetic circuit network.
Figure 8. Electromagnetic force densities for the variation of the winding angle.
Figure 10. Deformed shape for the B-phase winding of transformer T1: (a) isometric view and (b) x-z plane view.
Figure 12. Load noise level versus the weighted vibration level. Figure 13. Differences between the predicted and measured
load noises for the transformers listed in Table 1.
L ¼ 44 þ 18 logðR=R0 Þ, (11)
c3
L ¼ log Ac1 Dc
n Df
2
þ c4 , (10)
where R is the rated power in MVA and R0 is the reference
where L is the load noise level in dBA, A is the RMS value rated power of 1 MVA. The differences between the pre-
of acceleration calculated at all the nodes when the max- dicted and measured load noises for the 32 transformers
imum current is applied to the B-phase winding, Dn denotes listed in Table 1 are shown in Figure 13, where the red
the distance from the winding surface to the closest tank circular symbol is for the equation of IEC 60076-10 (IEC,
wall, and Df represents the distance from the winding 2016) (i.e., equation (11)), and the blue circular symbol
surface to the farthest tank wall. In addition, the coefficients represents the equation (i.e., equation (10)) proposed in this
c1, c2, c3, and c4 shown in equation (10) are positive article. In this figure, the abscissa represents the model
constants, whose values were determined to minimize the number of the transformer, and the ordinate indicates the
standard deviation of the difference between the measured difference between the predicted and measured load noises.
values and the predicted values using equation (10) for the The vertical bars in Figure 13 are the error bars, the sizes of
32 transformers shown in Table 1 using the generalized which represent twice the standard deviations of the
reduced gradient method (Sharma and Glemmestad, 2013). measured noises. As shown in the figure, the prediction
The R2 and RMSE of the proposed prediction equation equation proposed in this article has a smaller difference
were evaluated to determine the accuracy of the calculation between the predicted and measured load noises than that of
results of the load noise using the equation. After defining IEC 60076-10 (IEC, 2016). Thus, the proposed equation
the weighted vibration level of the winding as can predict load noise more accurately than the previous
c3
Vw ¼ Ac1 Dc n Df
2
in the first term on the right in equation comparable equation.
(10), the weighted vibration levels and the load noise levels
for the 32 transformers shown in Table 1 are plotted in
Figure 12. In this figure, the circular symbol indicates the
7. Conclusions
values for each transformer, and the red solid line is the In this article, we proposed a new equation to predict the
curve corresponding to equation (10). The R2 and RMSE of load noise of three-phase oil-filled power transformers
the predicted equation using the weighted vibration level considering the winding vibration. To this end, the load
illustrated in Figure 12 are 0.94 and 1.94 dBA, respectively. noise of the 32 transformers and the tank vibration were
Compared with the regression equation shown in Figure 11, measured and analyzed to determine that the load noise was
R2 is increased by 0.03, and RMSE is reduced by 0.48 dBA. related to the tank vibration, which is linearly proportional
In conclusion, the prediction using the weighted vibration to the winding vibration. The electromagnetic finite element
level of the winding (Figure 12) is more accurate than the analysis model for the B-phase winding was constructed to
prediction using only the winding vibration (Figure 11). calculate the electromagnetic force, and the acceleration of
The results predicted using equation (10) were compared each node was obtained by imposing the calculated elec-
with those predicted using the previous equation presented tromagnetic force on the structural finite element analysis
by IEC 60076-10 (IEC, 2016) to verify the proposed model. The RMS values of these accelerations were
prediction equation. The equation of IEC 60076-10 (IEC, substituted for the prediction equation proposed in this
2016), which predicts the acoustic power according to the article to predict the load noise of the transformer. It was
rated power of the transformer, is expressed as follows proved by comparing the predicted load noise with the
Lee et al. 9
experimentally measured load noise that the proposed Jin M and Pan J (2016) Vibration transmission from internal
prediction equation was more accurate than the previous structures to the tank of an oil-filled power transformer. Applied
prediction equation. Acoustics 113: 1–6.
Kulkarni SV and Khaparde SA (2013) Transformer Engineering -
Design, Technology, and Diagnostics. 2nd ed. Boca Raton:
Acknowledgment CRC Press.
We would like to acknowledge support provided by Hyosung Mizokami M and Kurosaki Y (2015) Noise variation by com-
Heavy Industries. pressive stress on the model core of power transformers.
Journal of Magnetism and Magnetic Materials 381: 208–214.
Mizokami M and Kurosaki Y (2016) Variation of noise and
Declaration of conflicting interests magnetostriction associated with joint types of transformer
The author(s) declared no potential conflicts of interest with re- core. Electrical Engineering in Japan 194(2): 1–8.
spect to the research, authorship, and/or publication of this article. Moses AJ (1974) Measurement of magnetostriction and vibration
with regard to transformer noise. IEEE Transactions on
Magnetics 10(2): 154–156.
Funding Moses AJ, Anderson PI and Phophongviwat T (2016) Localized
surface vibration and acoustic noise emitted from laboratory-
The author(s) received no financial support for the research, au-
scale transformer cores assembled from grain-oriented elec-
thorship, and/or publication of this article.
trical steel. IEEE Transactions on Magnetics 52(10): 7100615.
Ploetner C and Almeida E (2020) Load sound power levels for
ORCID iD specification purposes of three-phase 50 Hz and 60 Hz liquid-filled
power transformers. CIGRE Working Group Report WG A2 54.
Jintai Chung https://orcid.org/0000-0002-7143-1321
Rausch M, Kaltenbacher M, Landes H, et al. (2002) Combination
of finite and boundary element methods in investigation and
prediction of load-controlled noise of power transformers.
References Journal of Sound and Vibration 250(2): 323–338.
Al-Abadi A (2019) Developing an accurate load noise formula for Reiplinger E (1980) Assessment of grain-oriented transformer
power transformers. 2019 6th international advanced research sheets with respect to transformer noise. Journal of Magnetism
workshop on transformers (ARWtr), Cordoba, Spain, 7–9 and Magnetic Materials 21(3): 257–261.
October 2019, pp. 19–24. Sharma R and Glemmestad B (2013) On Generalized Reduced
Ertl M and Voss S (2014) The role of load harmonics in audible Gradient method with multi-start and self-optimizing control
noise of electrical transformers. Journal of Sound and Vibra- structure for gas lift allocation optimization. Journal of Process
tion 333(8): 2253–2270. Control 23(8): 1129–1140.
Geuzaine C and Remacle JF (2009) Gmsh: a 3-D finite element Sim JH, Ahn DG, Kim DY, et al. (2018) Three-dimensional
mesh generator with built-in pre- and post-processing facilities. equivalent magnetic circuit network method for precise and
International Journal for Numerical Methods in Engineering fast analysis of PM-assisted claw-pole synchronous motor.
79(11): 1309–1331. IEEE Transactions on Industry Applications 54(1): 160–171.
Girgis RS, Bernesjö MS, Thomas S, et al. (2011) Development of Snell D (2008) Measurement of noise associated with model
ultra-low transformer noise technology. IEEE Transactions on transformer cores. Journal of Magnetism and Magnetic Ma-
Power Delivery 26(1): 220–226. terials 320(20): e535–e538.
Hsieh CL and Tuan WH (2005) Elastic properties of ceramic– Tanzer T, Pregartner H, Riedenbauer M, et al. (2018) Magneto-
metal particulate composites. Materials Science and Engi- striction of electrical steel and its relation to the no-load noise of
neering: A 393(1–2): 133–139. power transformers. IEEE Transactions on Industry Applica-
Hsu CH, Huang YM, Hsieh MF, et al. (2017) Transformer sound tions 54(5): 4306–4314.
level caused by core magnetostriction and winding stress Valkovic Z (1994) Effect of electrical steel grade on transformer
displacement variation. AIP Advances 7(5): 056681. core audible noise. Journal of Magnetism and Magnetic Ma-
Hu JZ, Liu DC, Liao QF, et al. (2016) Electromagnetic vibration terials 133(1–3): 607–609.
noise analysis of transformer windings and core. IET Electric Weiser B, Pfützner H and Anger J (2000) Relevance of magne-
Power Applications 10(4): 251–257. tostriction and forces for the generation of audible noise of
IEC (2016) Power Transformers - Part 10: Determination of Sound transformer cores. IEEE Transactions on Magnetics 36(5):
Levels. Geneva: International Electrotechnical Commission. 3759–3777.