Optimal Design of The Power Transformer With ANSYS Maxwell For Bidirectional Battery Charger
Optimal Design of The Power Transformer With ANSYS Maxwell For Bidirectional Battery Charger
Optimal Design of The Power Transformer With ANSYS Maxwell For Bidirectional Battery Charger
Summary
The leakage inductance of power transformers, windings’ loss estimation and design
optimisation is accomplished using of ANSYS Maxwell™ software. Analysis of high frequency effects
like skin and proximity losses in the transformer windings is presented in this paper. Stranding level
proximity losses in a Litzwire are obtained with Maxwell Transient and Magnetostatic simulation with
additional mathematical post processing. The optimal number of the Litzwire strands and their
diameter are determined for minimum losses in a given transformer.
The resulting new power transformer with higher efficiency for bidirectional DC-DC converter for
Electrical Vehicle charging system is implemented in a practical design
Keywords
High frequency, Eddy currents, finite-element methods, ANSYS/Maxwell inductors, magnetic devices,
numerical field computation, power conversion, power transformers, proximity effect, skin effect
1. Introduction
It is generally recognised that electrical mobility is the route to save the primary energy
consumed in transportation. The overall system, depicted in Figure 1, comprised of mechanically and
electrically optimized and integrated components like PV cells, foil capacitors, batteries, power drives,
motors, magnetic gears as well as Vehicle to Grid interfacing hardware. In case when the electrical
vehicle plugged–in to the grid to either buying or selling electricity a bi-direction Vehicle to Grid (V2G)
interfacing is necessary.
(A) (B) (C
140 VDC Bus HF Trafo 1- 380V DC Bus
PMSM 5kW / 9.9kWp
AC Grid
Inverter Bridge
F=8kHz
DC/DC converter 30..40kHz
2. Transformer optimisation
2.1 Target topology overview
Generally, the miniaturisation of inductors and transformers used in power electronics
converters is the main challenge for designers. Particularly in the target topology an appropriate
design and optimisation of the high frequency power transformer is a key aspect, which will be
analysed in details. The Dual Active Bridge topology presented in Figure 3 (B) consists of two full
bridges with MOSFETs, which are connected to primary and secondary windings of power
transformer. Despite of apparent simplicity of the topology, this circuit has a unique feature, - it utilises
the leakage inductance of the transformer as an element for power transfer and regulation. With an
appropriate phase - shift between primary and secondary bridges the output voltage as well as the
power flow direction can be determined. Important to note that during the switching times of the
primary/secondary bridges, the current in the transformer is inverting and takes very high slew rates,
which are limited by leakage inductance of the transformer only. Taking into account the
miniaturization roadmap in automotive applications, it becomes apparent that the optimization of the
transformer (in terms of its weight/size and cost), as a key component in this topology, has the highest
priority. Increased switching frequencies can, generally, lead to decrease transformer’s size, however
as frequencies are pushed up, many new problems arise: - the AC losses in magnetic materials as
well as in the windings increase dramatically.
Figure 3 V2G Interface including grid coupling inverter/rectifier (C.) and DCDC isolated converter (B.)
Main parameters, which influence the operation of the Dual Active Bridge converter, are:
1. The leakage inductance of the power transformer, which defines the maximum possible power
transfer from the primary to the secondary side in the topology.
2. The windings ratio ‘n’ which together with voltage conversion ratio VH/VL defines the maximum RMS
current in the switches of the bridges. Consequently it has an impact directly to the efficiency of the
converter. Hence the overall transformer optimisation procedure leads to finding out the local
optimum of the above mentioned parameters for minimum RMS (lowest conduction losses), and
global optimum for the transformer design. The latter leads to the trade-off between minimum volume
and switching frequency, hence between the switching and conduction losses as well as core and
winding losses. This in turn has an impact on the transformer efficiency – thereof transformer losses
and its temperature rise.
2.1.1 Transformer optimisation parameters are:
n – Winding ratio
LS – Leakage inductance – (the average power is for the case of a lossless converter)
IRMS – The RMS current is that of the current in the transformer and therefore also the RMS current
in the switches. This RMS current will divide equally between the two phase arms.
FSW – switching frequency (T- switching period)
Figure 4 Optimization of the DC/DC Transformer (DAB Topology); Transformer design specific boundaries:
Power handling, output power limitation, RMS current levels in the transformer windings
RMS RMS
Transformer optimization at Phi =0.1 RMS
Transformer optimium at Phi =0.02 Current Current Transformer optimization Phi =0.2
Current
Prim. /Sec. Winding Ratio
Prim. /Sec. Winding Ratio
Prim. /Sec. Winding Ratio
(A) Minimum Phase Shift =0.02p.u. Minimum possible Phase shift =0.1p.u. Minimum possible Phase shift =0.2p.u.
(1p.u =100% power) (1p.u =100% power) (1p.u =100% power)
Figure 5 Transformer optimisation results for minimum RMS current under different phase shifting factors
Because of non-standard, spatial arrangement of the primary and secondary windings, the
problem for determination of leakage inductances for a given specific layout of the windings is solved
using numerical finite element field-calculation method e.g. Maxwell™ FEM Simulation tool. The FEM
results are compared with practical implementations, which have shown a good agreement with
hardware realisation Figure 7.
a) b) c)
Figure 7 An example of the obtained magnetizing and leakage inductances with FEM simulation and practical
measurement results with impedance analyser Agilent 4294A
Based on the leakage inductance values obtained from the simulation, a prototype Transformer was
constructed and installed in the converter board for further testing (Figure 8, Figure 9).
3D FEM
Magnetic flux density
distribution in the core
a)
3D FEM
Field strength distribution in
the window area
b)
Figure 8 Simulation results and first prototype of the HF Power Transformer, (1/4 of the transformer model)
Power Transformer
Figure 9 Prototype DCDC Converter with the power transformer installed for testing; Infrared FLIR picture
shows high temperature between primary and secondary windings due to eddy current losses in the windings.
a) Test setup for winding loss measurement c) Choke with solid wire
Figure 10 Measurement setup a) with simple chokes b); c) for winding loss estimation
Total losses in the windings are obtained by means of power measurement with DC current IDCT @ 63°C,
with the same temperature rise as with corresponding triangle current IRMS(triangle). The results provided
in Figure 11 have shown a good matching with hardware test results.
Although the general purpose field analysis methods and commercial software like ANSYS
Maxwell™ can be used to find 2D and 3D solutions of arbitrary problems, including analysis of AC
losses (eddy current), in most practical cases the direct simulation of eddy-current effects in the multi
winding transformers is computationally very expensive. Particularly our target transformer must have
many turns (25-30 turns) of fine stranded Litzwire (AWG 35-44 /30-100µm), to reduce eddy current
losses. Compared to dimensions of the E70 core, which is in order of tens centimetres, the scaling of
the individual strands can vary over two to four orders of magnitude. E.g. in case of AWG 44 we
should account a hundred thousand strands of wire for E70 core window (similar case is described
e.g. in [7]). Even when larger wire is used, the skin depth in the wire can be small, for example 100um
at 400 kHz, creating a similar disparity of length scales. In either case a large number of elements are
needed to perform finite-element analysis, and this leads to slow simulations and large memory
requirements. To circumvent this problem, software vendors recommend modelling a stranded
winding as a region of uniform current density (option ‘Stranded’). In this case the field distributions
can be estimated, but it provides no information on losses in the stranded winding (Eddy effects are
switched off). The second problem with direct FEM analysis is the optimization. In previous section
many different winding possibilities with desired leakage inductance are obtained. As a matter of fact
the optimisation procedure in practical design is mostly done by trial and error. Particularly, when each
iteration takes hours to analyze via finite-element analysis, true optimization is not feasible.
In order to circumvent both the limitations of 1D analytical methods, and the limitations of
existing numerical methods we used the so called squared field derivative approach, which combines
numerical calculation of the overall field geometry with analytical calculation of its interaction with the
winding strands. Based on round wires and assuming that the strands diameter (a) in the Litzwire is
small compared to a skin depth (δ), the average proximity loss is given
1 ˆ2 ⎧⎪αaI 1 (αa)⎫⎪ 1+ j
by Pprox = ⋅ H ext ⋅ 2π Re⎨ ⎬ , where α = [3]. The corresponding squared winding current
ρc ⎪⎩ I o (αa) ⎪⎭ δ
derivatives, have been used by many authors, including [8] - [10], [11]. This avoids the scale problem,
but allows applying the power of FEM tool to quickly obtain a much more accurate solution than would
be available through 1D analysis. A similar approach was described in [2] and was also used in [12]
for gapped single-winding inductors with sinusoidal waveforms.
Temperature (°C) 63 - 43 - -
IPeak (A) 6 6 6 6 6
The total time averaged AC loss in all the conductors of a winding is equal to the spatial average of
the time-averaged squared field derivative (1).In case of Litzwire, we have assumed that bundle-level
effects are negligible, which is normally valid for well-designed Litz constructions. [3]
2
(1) P =
π ⋅ lt , j ⋅ N j ⋅ d c4, j ⎛ dB ⎞ ,where PE , j - Time average ac (eddy-current) loss in winding;
E, j ⎜ ⎟
64 ⋅ ρ c ⎝ dt ⎠ Nj - Number of turns in winding;
j
With the aim to verify the results of the SFD lt , j - Average length of a turn;
approach, second test was carried out using 〈•〉 - Spatial average over the region of winding ;
above mentioned test setup with a same choke
geometry but wound with Litzwire ( 8 x 0.5 mm (•) - Time average
strands) and having slightly lower number of
turns due to lower packing factor compared to solid wire. Obtained comparative results between
measurements, direct Maxwell™ transient simulation and Mixed FEM + analytical calculations are
given in Figure 11 (left side). Between hardware test results and direct Maxwell™ transient simulation
a very good correlation of the total winding losses can be observed. The difference is about 8mW.
The post-processing results obtained good matching as well, showing 498mW AC losses compared to
test results with 428mW. Somewhat higher losses predicted with mixed analytical method are due to
the 2D analysis accounting only the normal magnetic fields. In the 3D case, the field may not always
be perpendicular to the axis of the wire, resulting in slightly lower loss in practice, than the calculation
results presented here.
208
100 7
Original design:
80 LV -AWG32 x 180 strands
326 HV-AWG28 x 25 strands
Total Winding loss = 60.77W
60 17
511
40
41 801
1256 1550 3690 8190 17100 34900
20 101 1969 4842
7593 11907 18671
250 660
0
26 28 30 32 34 36 38 40 42 44 46 48 50
AWG
Figure 12. Transformer winding optimisation, with regards to optimal number of strands and strands
diameter for minimum proximity losses in the Litzwire
From the “full bobbin” design (Figure 12, blue curve) it is visible that with AWG38 (0.1 mm) and 1256
strands the winding losses can be reduced up to 17W (1/3 of original design). Further reduction of the
strands diameter and correspondingly increasing number of strands will only slightly reduce the
losses, whereas the cost and the complexity of the windings will increase dramatically.
Important to note that with thinner strands diameters the full bobbin design comes very close to the
optimum design, however the theoretical optimum design is not always buildable. Therefore trade-offs
between best buildable design and minimum losses should be considered. Particularly for our target
transformer the Litzwire range between AWG38/250 to AWG40/ 660 can be considered as optimum.
5. Conclusions
Two problems have been solved during optimization of the power transformer using Maxwell™ FEM
analysis tool. In the first part an appropriate layout of the windings has been found with 2D simulation
to meet the desired leakage inductance for the optimized design. In the second part further
investigation for proximity loss analysis are carried out. With the FEM Simulation the eddy current
losses are verified with the hardware measurement, afterwards the Lizwire optimization in the winding
is carried out with FEM simulation and analytical post-processing using field calculator. It was shown
that the full bobbin design with maximal copper area in the core window is not always the optimal
design. Obtained results have shown a good match between theoretical and practical results, which
will be used for the optimized transformer design with drastically reduced AC losses in the windings.
6. References