Received 7 February 2023, accepted 27 February 2023, date of publication 6 March 2023, date of current version 10 March 2023.

Digital Object Identifier 10.1109/ACCESS.2023.3253392

Multi-Objective Design Optimization for

HVDC-LCC Converter Transformers: Analytical
and FEA-Based Comparison
MARCOS V. CZERNORUCKI 1,3 ,
MAURICIO BARBOSA DE CAMARGO SALLES 1 , (Member, IEEE), SÉRGIO LUCIANO AVILA 1,2 ,

FELIPE ALVES SOBRINHO3 , WILERSON W. CALIL3 ,

AND JOSÉ ROBERTO CARDOSO1 , (Life Member, IEEE)
1 Laboratory of Advanced Electric Grids (LGRID), Polytechnic School of the University of São Paulo, São Paulo 05508-010, Brazil
2 Laboratory of Applied Scientific Computing (PECCE), Federal Institute of Santa Catarina, Florianópolis 88020-300, Brazil
3 Hitachi Energy Ltd., Guarulhos 07190-904, Brazil
Corresponding author: Sérgio Luciano Avila ([email protected])
This work was supported by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior-Brazil (CAPES) - Finance Code 001.

ABSTRACT The emerging HVDC technology has been used for long-distance power transmission,
increasing flexibility to the power systems, handling asynchronous interconnections, crossing long-distance
submarine cables, unusual loading, and generation profiles, and improving energy market relations. The
HVDC converter transformers are designed based on system parameters, that directly affect the core and
windings geometries, thus the weight and operating losses. In a standard design process, the designer adjusts
the active part dimensions, until the constructive and specified aspects, moreover, the technical standard
restrictions, are satisfied. The main contribution is to formulate analytic equations, in a way that losses and
weight can be obtained for several options of core and windings geometry. However, weight and losses
are opposite objectives in the search for an optimal solution. Understanding this compromise between
opposing goals is relevant to the equipment learning process. In this context, another significant contribution
of the work is to carry out a formal optimization process through the analytical formulation developed.
To minimize the weight and operating losses, and subject to IEC standards and constructive restrictions, the
multi-objective Genetic Algorithm has been used to search for the Pareto Frontier. Far beyond the chosen
solution, the non-dominated frontier obtained for each transformer design, allows the designer to learn about
the equipment and its operation, leading to a continuous improvement of the proposed methodology. The
analytical formulation is validated by an alternative numerical methodology for winding harmonic losses
and short-circuit impedance verification, providing meaningful confidence for the applied method.

INDEX TERMS HVDC, transformer, design, losses, optimization.

I. INTRODUCTION matrix, contributing to the global carbon generation drop [1],

The HVDC systems are implemented worldwide as a reliable and HVDC technology has permitted systems and countries’
solution for bulk energy transportation through long overhead interconnection for power exchange, likewise connecting off-
lines, and underground or submarine cables. It has been shore wind farms, oil and gas platforms for energy generation
recognized as a competitive alternative for AC transmission, and consumption.
integrating different sources of energy, increasing the electri- Thus, designing and manufacturing an HVDC system and
cal system reliability and power quality. Renewable sources its equipment demand special expertise and require great
integrate an important share of the actual global energetic investment for implementation. The converter transformer
is considered a fundamental part of this investment and the
The associate editor coordinating the review of this manuscript and main drive for system efficiency. Hence, optimizing the trans-
approving it for publication was Ali Raza . former design, especially in terms of losses and weight, is a
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License.
23032 For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/ VOLUME 11, 2023
M. V. Czernorucki et al.: Multi-Objective Design Optimization for HVDC-LCC Converter Transformers

TABLE 1. HVDC System Data From the PSCAD Simulations.

valuable engineering task. The design optimization process

must consider several other requirements from the technical
specification, that cannot be neglected. This work proposes
an analytic formulation for converter transformers design
defined for a classic ±800 kV LCC (line-commutated con-
verter) link, for the rectifier and inverter stations. The trans-
formers’ valve windings are connected in delta and wye at
each station. This winding is also directly affected by the
current harmonics since it is physically connected to the
converter valves.
The standard design solution to define an optimal active
FIGURE 1. The simplified generic circuit used for the simulation in
part (core and windings) geometry which fulfils the desired PSCAD [4].
objectives would require carrying out multiple design
options, comparing them individually. Applying optimization
TABLE 2. HVDC Transformers Parameters From the PSCAD Simulations.
algorithms over the analytical formulation, several applicable
solutions are presented at once for each design type and the
most suitable solution is selected. A multi-objective exercise
is performed using the HVDC transformers designs, obtain-
ing a Pareto’s set of solutions, presented graphically, assisting
the designer to select the most suitable option among several
choices.
The work provides a consistent analytic formulation for
HVDC transformers, validated by the valve winding losses
analysis by a numerical alternative methodology. Moreover, The firing angle defines the harmonic content in the load
the short-circuit impedance magnitude is verified providing current. The angle magnitude may assume distinct values
confidence to the method. according to the DC voltage level, overload profile, and line
voltage drop. The simplified generic circuit is shown in Fig. 1
A. GENERIC HVDC LCC SYSTEM and the detailed transformer parameters are given in Table 2.
The HVDC system influences the transformer design through The converter transformers are identified according to the
several effects, intrinsic to the converters switching, or to rectifier (on the left side) and the inverter stations (on the
the system operation, which are widely described in the right side), based on the valve winding connection. The delta
references [2] and [3]. The AC-DC-AC current conversion transformer (YNd) is connected after the lower DC voltage
results in current harmonics that circulate directly through the stage at ± 400 kVdc and the wye transformer (YNy) after the
transformer windings. The selected system for the study was ± 800 kVdc level. The 500 kVac line windings are wye con-
the Belo Monte link located in Brazil rated at Ud ±800 kV nected with an on-load tap changer (OLTC) with 31 electrical
DC voltage, bipole, 2500 km overhead line length, 4000 MW positions [7], for all transformers. The AC system frequency
rated power and 2500 A rated current (Id ) [4]. The 500 kV is 60Hz and the impedances are 15% at rated voltage and
at line side, single-phase, two-winding transformers were power. The transformers of the rectifier station are single-
modeled and connected to the twelve-pulse converters [5]. phase, with a rated power of 393.5 MVA and 333.9 kV voltage
A bank of three single-phase transformers is connected to on the valve side. On the inverter side, the transformers
each six-pulse bridges: one bank has the secondary connected are single-phase, 376.7 MVA and 319.6 kV valve winding
in star and the other in delta. A generic HVDC link model and voltage.
transformers parameters were adjusted according to the Belo
Monte system characteristics. The harmonic current spectrum B. CONTENT OF THE SECTIONS
of this HVDC bipole model was obtained using PSCAD [6] Along the coming sections the authors bring design con-
and considering a 1.5% DC voltage drop along the line. The siderations for the transformer active part dimensions defi-
rectifier converter is modeled with a 15◦ firing angle while the nition (section II) and propose an analytical formulation to
inverter converter had an extinction angle of 17◦ . The detailed calculate the active part weight and converter transformer
system parameters are given in Table 1. losses (section III). Further, the formulation is used to seek

VOLUME 11, 2023 23033

 M. V. Czernorucki et al.: Multi-Objective Design Optimization for HVDC-LCC Converter Transformers

FIGURE 4. Flowchart of a conventional transformer design.

FIGURE 2. Single-phase core design main parameters.

FIGURE 5. Flowchart of a converter transformer design.

FIGURE 6. Cross-section view of one core limb representing the windings’

layout and distances.

FIGURE 3. Winding and conductors’ design, adapted from [9].

A. MAIN DIMENSIONS DEFINITION
The core design defines the geometry of the magnetic circuit,
an optimal solution (section IV) in terms of weight and losses which is given by the diameter (DK ), limb height (LH ),
using multi-objective optimization. Finally, the analytic for- limb pitch (LP), core height (CH ), yoke height (YH ), and
mulation is validated by a numerical model using the Finite core length (LK ). The limb height is defined by the wind-
Element Method to calculate the winding losses and the short- ings height and the distances to the yokes, top, and bottom,
circuit impedance verification (section V). which are a result of the dielectric tests, service voltages, and
mechanical or magnetic requirements. A simplified geometry
II. TRANSFORMER DESIGN CONSIDERATIONS of a single-phase core, with two wound limbs, like the one
The transformer active part design consists basically of the used in the actual study, is given in Fig. 2.
core and windings’ geometrical dimensions definition. Both The winding design is defined by the winding inner diam-
are influenced by specified requirements such as power, volt- eter (Dw ), axial height (H w ), and radial width (RR). The
ages, impedance, dielectric tests, overvoltage, overloads, etc. windings are concentric and assembled around the two core
[3] The core dimension definition will restrict the region limbs. The winding conductor is defined by its height (hc ),
inside the core window, called limb height. The winding width (rr), radius (rd), and paper insulation (insul). The turns
block is placed inside the core window, and the windings’ can be separated by key spacers and the windings by radial
geometry defines the core limb pitch. Thus, the core and ducts filled with insulating material. The number of turns
winding designs are closely linked, and the parameters’ vari- (turn) defines the winding height (H w ), and the number of
ation affects the complete active part design. That interdepen- parallel conductors (paral) defines the radial width (RR).
dence can be visualized in Figs. 2, 3 and 6. Details are shown in Fig. 3.

23034 VOLUME 11, 2023

M. V. Czernorucki et al.: Multi-Objective Design Optimization for HVDC-LCC Converter Transformers

A simplified formulation is presented in section II-D, based TABLE 3. Symbols used in the Formulation.
on [8], to define the geometry of the core, windings, and
conductors. For a specific geometry the core and copper
weight, and the total losses in operation, including current
harmonics are calculated.
The coming sections provide the reader an overview of
the difference between a conventional and a converter trans-
former design.

B. CONVENTIONAL TRANSFORMER DESIGN

The conventional transformer design begins with a technical
specification analysis, where the system voltages are defined
for the high and low voltage sides, frequency, and other
requirements relevant to the transformer design. In the same
specification, the transformer’s main data are provided, such
as the number of phases, rated power, tap changer location,
tap range, short-circuit impedance, losses, dimensions, and
weight limitations. Based on that information combined with parameters were kept fixed since their determination is
the applicable technical standards (IEC, IEEE, etc.), the engi- not related directly to the study purpose. The radial
neer will start the transformer design fixing some parameters ducts (dka, dab and dbc), axial ducts toward the yokes
and leaving others free, to define the core and windings main (dup_A, dup_B, dup_C, dbot_A, dbot_Banddbot_C), dis-
dimensions, given in Fig. 2. tance between winding blocks (ph − ph), conductor insu-
Designing a transformer can be an optimization exercise, lation (insul) and key spacers thickness (spactk), which are
based on restrictions given by the specification and technical directly related to the system insulation coordination [11].
standards, to achieve the most suitable option that fulfills the Those distances are represented in a cross-section of the
given objectives, which can be minimum weight, dimensions, core window in Fig. 6, together with the windings’ layout,
or losses. It can also be a combination of two or more given being illustrated, from the core limb, the regulation (A),
objectives [8]. Figure 4 shows a conventional transformer line (B), and valve (C) windings. The single-phase core has
design basic flowchart. two wounded limbs, and the winding blocks are connected
in parallel. Figure 6 represents the winding block of only
one wounded limb. The conductor corner radius (rr) and
C. CONVERTER TRANSFORMER SPECIAL REQUIREMENTS
the number of parallel conductors (paral) are used for the
The converter transformer design must take into account
conductor’s dimensions definition. The core flux density (φ)
several undesired effects related to the converters’ non-linear
is defined by the overvoltage requirements, saturation, and
operation, HVDC system voltage variation and loading,
core steel material. It certainly affects the core losses, but for
DC bias effect [10], noise level, and electric field distri-
the study, the total losses, no-load, and load losses including
bution [2] and [3]. For losses calculation, the main effect
harmonics and overload profile are analyzed and restricted
to be considered in the design is the load losses increase
together.
due to the harmonic spectrum generated by the converters’
The windings’ cross-section areas (Aw ) are related to the
operation. Such calculation methodology is clearly explained
windings’ current density, which was kept the same. Depend-
in section III-C.
ing on the connection of the delta and wye valve windings,
√ the
According to the technical specification of the Belo Monte
phase voltage will assume different values by 3 relation,
link [4], the transformers shall be designed for a 133% long-
which defines the number of turns and the winding type.
term overload for 30 minutes and 150% short-term emer-
Helical and layer windings are the ones where each round
gency for 5 seconds. Those conditions, in a two-bipole instal-
represents one turn, therefore, the turn per disc (turnd) is
lation, represent the loss of one monopole or one bipole
equal to one. The number of parallel cables (paral) defines
respectively, and the time is defined by the control system
the winding radial width (RR). A disc winding type can have
to rearrange the load, avoiding a forced disconnection. Dur-
several turns in the same disc, and the radial width is defined
ing the optimization in section IV, that loading profile was
by the number of parallel cables (paral) combined with the
represented as a daily cycle, to be applied as an overcurrent
number of turns per disc (ndisc) [12]. The symbols used for
input into the models.
each parameter described are detailed in Table 3.
Since the core flux density is defined, the windings’ num-
D. MAIN DIMENSIONS CALCULATION ber of turns can be determined for each DK input, by (1).
The study presented in this work was performed vary-
ing two core parameters within a certain range: the V
N = (1)
limb height (LH ), and the core diameter (DK). Some 4.44fN ∅Ak

VOLUME 11, 2023 23035

 M. V. Czernorucki et al.: Multi-Objective Design Optimization for HVDC-LCC Converter Transformers

where, TABLE 4. Core steel material magnetic losses [8].

V : phase voltage [V]
f : frequency in [Hz]
N : number of turns
φ : flux density in [T]
Ak : core cross-section area in [m2 ]
The cross-section area Ak (2) is calculated by the core
diameter and the core space factor k s , which was defined as
0.898, considering some fiberglass insulation, core bandage,
and structure material such as flitch plates [8].

2
π DK 10−3
Ak = ks (2)
4
The total winding height in [mm] is given by (3) and the
number of discs, by (4).

Hw = LH − dup − dbot (3)

ndisc = N /turnd (4)

The conductor copper height and width in [mm] are calcu- converter transformer in operation is affected by the cur-
lated by (5) and (6). For layer windings, like the regulating rent harmonics generated by the power electronics ele-
(A), spactk is null. The conductors’ radius is according to the ments switching, increasing specifically the additional load
supplier rules. For this study, the conductors for winding A losses [13]. That design consideration is explained further in
and B used a 1.0mm radius and winding C, 0.5mm. sections III-C and D.

[Hw − spactk (ndisc − 1)] − insul ndisc A. CORE WEIGHT AND LOSSES
hc = (5)
ndisc The magnetic losses related to the core steel material are
(Aw /paral) + rd 2 (4 − π) given by the supplier in terms of W/kg, depending on the flux
rr = (6)
hc density in [T], according to Table 4. The losses due to the
Finally, the winding radial width in [mm] is calculated by Foucault effect were neglected. The core weight GFe in [kg],
(7). is calculated by the mean magnetic path length in (11), based
on the variables defined before, the number of core limbs
RR = (rr + insul) paral turnd (7) (nlimb ), and the steel volumetric density d fe (7650 kg/m3 )
[14].
The windings’ inner diameters Dw in [mm] are given by
(8). GFe = [2LK + nlimb (LH + YH )] Ak dfe 10−3 (11)
DwA = DK + dka Thus, the core losses are a result of the product of the
DwB = DwA + 2RRA + dab weight GFe times the magnetic losses Pmag (12).
DwC = DwB + 2RRB + dbc (8)
P0 = GFe Pmag (12)
The core length in [mm] is defined by (9). The yoke height
For the actual study a core flux density of 1.70 T was
is equal to the core diameter and the limb pitch is given by
chosen, once the specification had a limitation in sound level,
(8).
which would not allow the designer to go close to the material
LK = DK + 2 (DwC + 2RRC ) + ph_ph (9) saturation limit.
LP = 2 (DwC + RRC ) + ph_ph (10) Although, higher or lower flux density values may be
chosen depending on other specification requirements like
This basic formulation defines the geometry of the core, no-load losses, flux density limitations, overvoltage, under-
windings, and conductors for the several DK and LH as input. frequency, or reactive loadings.

III. ANALYTICAL FORMULATION FOR CONVERTER B. WINDING LOSSES AND TOTAL WEIGHT
TRANSFORMER The transformer winding losses are divided into two specific
From the geometry defined by the core and windings’ main components: i) resistive losses, which are the most signifi-
dimensions’ calculation, the magnetic no-load losses, the cant loss component in power transformers, originated from
load losses, and the short circuit impedance can be cal- the windings’ resistance, and also known as ohmic losses;
culated in the coming sections. As mentioned before, the ii) additional losses, resulted by eddy currents circulation in

23036 VOLUME 11, 2023

M. V. Czernorucki et al.: Multi-Objective Design Optimization for HVDC-LCC Converter Transformers

the conductors exposed to the magnetic field [15]. Both loss an enormous amount of labor to achieve a correction factor
components define the winding load losses. which would represent that effect in additional losses and
The converter transformer load current is composed of the impedance calculation. That was performed by dr. Rogowski
fundamental frequency component and different harmonic [20] in the past, who defined a factor that can be applied for
orders [14], which can assume distinct profiles depending such a purpose.
on the HVDC installation and the system operation modes. In a simplified approach, the Rogowski factor Kr for a
That harmonic influence affects the additional losses [16]. transformer that has Hr /π Ḣw < 0.25, where Hr is the geo-
The load losses are also intrinsically related to the transformer metric region of the leakage flux, defined on the ampere-turn
impedance, which represents the part of the energy contained diagram according to Fig. 5, and Ḣw is considered the average
in the leakage magnetic flux. Converter transformers for LCC windings’ height, can be defined in (18) as:
applications normally are specified with considerably high
Hr
impedances, to contribute with the HVDC short-circuit power Kr = 1 − (18)
strength, limiting the fault currents. Hence, the contribution π Ḣw
of the magnetic flux to the total converter transformer load Finally, the winding additional eddy losses, Peddy , in [W]
losses is extremely significant [17]. can be obtained as defined in (19):
From the geometry defined in section II-D, the winding
cross-section area is used to calculate the winding resis- Peddy = π (Dw + RR) 10−6 paral N ac Pe Kr (19)
tance Rw in [] (13), based on the winding inner diameter,
radial dimension, number of turns, and copper resistivity The eddy losses together with the winding ohmic losses
(2.09 10−8 [-m]). (RI2 ) result in the total winding load losses, for a defined
. loading condition [19].
Rw = (Dw + RR)103 πN ρ Aw (13) By the windings’ geometry defined in section II-D, it is
possible to calculate the windings’ weight in [kg] using (20),
And the ohmic losses in [W] are calculated based on the the copper volumetric density is σ = 8.93 [g/cm3 ].
winding load current I w in (14) for a certain reference tem-
perature, T ref . GCu = NAw (Dw + RR) π σ 10−6 (20)


2 234.5 + Tref Thus, the core and copper weight to be analyzed will be the
PRI 2 = Rw Iw (14)
234.5 + 75 sum of (11) and (20), resulting in (21).
The reference [18] contains analytic equations used in this
Gtot = GFe + GCu (21)
work to calculate the theoretical winding load losses. The
peak leakage flux B0 in [T] is generated by the windings
C. IEC/IEEE 60076-57-129 TECHNICAL STANDARD
and its magnitude is defined as a function of the transformer
ampere-turn (NI), the winding height, and the winding radial The HVDC transformers technical standard, IEC/IEEE
dimension, presented in (15). 60076-57-129 [21], provides a formulation for total load
√ losses calculation which has been used over the years to
µo 2NIRR design and test converter transformers with success and
B0 = µ0 H = (15)
Hw extreme reliability, assured by the number of years HVDC
projects have been running worldwide. It defines a multiply-
Being µ0 = 4π 10−7 in the SI system. The winding loss
ing factor F WE , presented in (22), to represent the eddy losses
volumetric density Pe in [W/m3 ] can be calculated by (16),
increase due to the harmonics, also found in [22] and [23],
it is obtained from the frequency, the winding leakage flux,
defined as:
and the conductor radial width for axial losses calculation or
Xn
axial height for radial losses [19]. FWE = kh2 h2
h=1
π
 2 2 2 2
f B0 b Ih fh
Pe = (16) Being kh = and h = . (22)
6 ρ Ir f1
where b is equal to rr for axial losses, and to hc for radial The technical standard [21] recommends that all current
losses. The conductor cross-section area in [m2 ] is given by harmonic orders, h, up to the 49th shall be considered for
(17): the factor’s determination. Different harmonic spectra may
be informed to the manufacturer for several loading or over-
ac = hc rr − rd 2 (4 − π) 10−6 (17)
loading conditions. The currents I h and I r are respectively
The magnetic flux impinging the conductor cross-section the magnitudes of the hth harmonic and the rated in-service
area creates a Foucault current loop. The presence of a flow- current for a specific load condition. The frequencies f h
ing current affects the flux behavior about various points in and f 1 are the harmonic order and rated, which is also the
the section, but the magnetomotive force is found to be can- fundamental frequency. The calculated winding losses for a
celed when calculated around a spot. That fact would demand conventional power transformer are obtained by (23) and for

VOLUME 11, 2023 23037

 M. V. Czernorucki et al.: Multi-Objective Design Optimization for HVDC-LCC Converter Transformers

TABLE 5. Harmonic Spectrum for the Inverter Station.

FIGURE 7. Ampere-turn diagram considered for Rogowski factor

calculation.

E. IMPEDANCE CALCULATION
TABLE 6. Harmonic spectrum for the rectifier station. The transformer impedance is a parameter to be continuously
monitored during the design and especially for converter
transformers, it must be well controlled within a strict tol-
erance. The impedance magnitude is related to the magnetic
leakage flux created by the winding block. Hence, its con-
tribution to the total converter transformer load losses, espe-
cially for eddy losses, is extremely significant [24]. As per
Table 2, the transformers’ impedance for this study was fixed
at 15% at the MVA base. The geometric area is calculated
for each winding and for the radial duct between windings.
In these specific designs, all transformers are composed of
three coils as shown in Fig. 6, and based on the ampere-turn
diagram of Fig. 7, being (A) the regulating winding, (B) the
line winding, and (C) the valve winding. The geometric areas
in [m2 ] sd A , sd 0 , sd B , sd 1 and sd C are calculated in (26):

sd A = (DwA + RRA ) π RRA 3 10−6



sd 0 = (DwA + 2RRA + dab) πdab10−6

sd B = (DwB + RRB ) π RRB 3 10−6


sd 1 = (DwB + 2RRB + dbc) πdbc10−6

a converter transformer are given by (24).
sd C = (DwC + RRC ) π RRC 3 10−6


Pw = Rw Iw2 + Peddy (23) sd = sd A + sd 0 + sd B + sd 1 + sd C (26)

Phw = Rw Iw2 + Peddy FWE (24)
With the weighted area sd calculated, the leakage flux B0
D. TOTAL LOSSES from (15), and the Rogowski factor [20], K r , from (18), the
The total losses including harmonics in [W] are given by (25), percentage impedance [18], U k , can be calculated by (27).
representing the sum of (12) and (24): √
4.44 2B0 sdNfKr
Uk = 100 (27)
Phtot = P0 + Rw Iw2 + Peddy FWE (25) V
The HVDC system described in section I was simulated Thus, by the equations presented in sections II and III, the
using the software PSCAD and the harmonic spectrum was core and windings geometry, losses, weight, and transformer
determined for the rectifier and the inverter wye and delta impedance can be calculated by an analytic methodology.
transformers. The harmonic currents Ih obtained are pre- That formulation was written into a commercial simulation
sented in Tables 5 and 6, and the factor FWE obtained software to perform the optimization exercises presented in
per (22). section IV.

23038 VOLUME 11, 2023

M. V. Czernorucki et al.: Multi-Objective Design Optimization for HVDC-LCC Converter Transformers

IV. DESIGN OPTIMIZATION optimization intends to find a solution that meets both objec-
The parameters from Table 2 were used as input for the ana- tives while respecting the given limits and restrictions. For
lytic formulation of the four converter transformers’ designs. a standard design process that could be achieved by several
The winding type, number of turns, winding height and width, individual runs, comparing the results, until the optimal solu-
conductor cross-section area, and number of parallel cables tion is reached. Using optimization algorithms, the solution
vary according to the three-phase bank connection in wye or can be presented together with other analyses in one single
delta and depend on the power and voltage level of the recti- graphical representation. A possible solution a⃗ ∈ Xf (30) is
fier or inverter station. The core height (CH ), consequently, called non-dominated if:
the limb height (LH ), and the core diameter (DK ), were
∄ x⃗ ∈ Xf : x⃗ ≻ a⃗ (30)
limited by the practical transformer factory experience. That
includes the commercial lifting crane and the vapor-phase That is the main difference between the single-objective
oven height, combined with a maximum transport height approach to multi-objective problems. There is not a single
limitation. The windings’ radial dimensions (RR) follow the optimal solution but an optimal setting in which none can
common manufacturer’s rules to accommodate the electro- be identified as better without a new classification (e.g.,
static ring and the pressboard yoke collars for the windings’ preference for one of the objectives). The group of all
end protection against the core yokes and between windings, non-dominated solutions is called a Pareto-optimal set [31].
as presented in [25] and [26]. Furthermore, the windings’ By correspondence, the set of objective vectors forms the
cross-section areas were defined to result in a current density Pareto-optimal frontier [27]. In this context, optimization
of around 3.0 A/mm2 of the maximum current along the methods deal with a population of possible solutions that have
whole tapping range, which is considered a satisfactory mag- the advantage of obtaining the Pareto-optimal set.
nitude to comply with the overloading cycle, defined in [4].
B. GENETIC ALGORITHM APPLICATION
A. THE PARETO-OPTIMAL SET DEFINITION The Genetic Algorithm (GA) is a stochastic optimization
Optimizing an exercise means searching for the best solution technique based on natural selection and genetics con-
to a given problem, limited by its physical-mathematical cepts [28], [31]. Solutions which generate new solutions are
model [27], [28]. The model for a multi-objective optimiza- selected probabilistically according to their merits (obtained
tion approach for the HVDC transformers designs can be through an objective function that evaluates the quality of the
written as (28): individual for the problem (28)). The better the individual
To minimize ⃗y = f⃗(⃗x ) = (y1 (⃗x ), y2 (⃗x )), is, the greater his chances of reproducing (the greater is the
probability of being selected and suffering genetic operators).
subject to g⃗ (⃗x) and Some advantages of the GA can be highlighted, such as:
⃗e(⃗y) = (e1 (⃗y), e2 (⃗y)) ≤ 0 (28) the generation of a semi-optimal solutions list rather than a
With, single solution, which is of great value for multi-objective
optimization [32]. The evolutionary process and its multiple
⃗y = (Gtot , Phtot ) ∈ Y 2 and x⃗ = (LH , DK ) ∈ X 2 solutions help to understand the compromise between the
conflicting objectives. For the proposed problem, the popu-
Being ⃗y the objectives vector, Y the objectives space (two
lation of possible solutions can be written as (31):
dimensions), x⃗ the parameters vector, and X the parameters
LH n,1 DK n,1
 
space (two dimensions). The parameters’ restrictions g⃗(⃗x) are
the limits, given in (32). The restrictions of the objectives Popn =  .. .. (31)
. .
 

⃗e(⃗y) were defined by the transformer short circuit impedance
tolerance and the total losses including harmonics limitation, LH n,nbind DK n,nbind
given in (29): where each row represents an individual of the nth generation,
e1 (⃗y) → Uk = 15% ± 7.5% nbind is the population size, and LH and DK are the two
optimization parameters. The limb height varied from 1000
e2 (⃗y) → Phtot < Sbase ∗ 0.30% (29)
to 7000 mm, in steps of 10 mm, and the core diameter varied
The impedance restriction comes from the technical stan- from 500 to 2000 mm, in steps of 1 mm. The core diameter
dard IEC 60076-1 [29] and the total losses were established range varies in smaller steps once, in practice, it is con-
based on [30], which is a local Brazilian norm applicable structed by stacking the core steel sheets, which have tenths
for large power transformers. The Pareto optimality concept of millimeters thickness. The limb height is constructed in the
is used for multi-objective modeling, understanding that two longitudinal dimension of the magnetic steel sheet, which has
stated objectives, like Gtot and Phtot , are conflicting with each many meters in length.
other. To reduce core and windings material with the same
core type, winding layout, and material quality, the total C. OPTIMIZATION RESULTS
losses will increase. It means that the improvement of one, The four exercises are large power transformers, that
implies the worsening of the other [27], [28]. Therefore, the justify the limits selected in (32), which have upper

VOLUME 11, 2023 23039

 M. V. Czernorucki et al.: Multi-Objective Design Optimization for HVDC-LCC Converter Transformers

TABLE 7. Inverter Delta exercise. TABLE 10. Rectifier Star exercise.

TABLE 8. Inverter Star exercise.

TABLE 9. Rectifier Delta exercise.

FIGURE 8. The feasible solutions found by the evolutionary constrained
multi-objective optimization process for the inverter delta.

and lower margins from the industry’s common practice.

The evolutionary-constrained multi-objective optimization
methodology [31] has been performed several times. It is
necessary due to its probabilistic nature. The process stop
point happens when it is noticed the stagnation of the Pareto FIGURE 9. The feasible solutions found by the evolutionary constrained
multi-objective optimization process for the inverter star.
frontier.
   
LH 1000 7000
= (32)
DK 500 2000 lowest Gtot is highlighted as diamond cyan. These analyzes
Figures 8 to 11 present all feasible solutions, according to lead to a better understanding of the problem. A piece of bet-
(28), found until the stop point is reached for each exercise. ter knowledge can point to the need of improving the model
Tables 7 to 10 present the solutions found for the lowest and/or to a more enlightened choice of the final solution [33].
Gtot (green) and lowest Ptot (blue). Variation in objectives Still respecting the tolerance reduction around U k ,
implies considerable variation in parameters. The solutions the diamond white solution is the one closest to
in magenta are those that respect a variation not greater than Pareto’s set of solutions. This distance is calculated by
0.25% around U k = 15%. For HVDC converter transformers (dAB = sqrt( (xB − xA)2 + (yB − yA)2 )). It is the choice
it is a valuable strict restriction, once a deviation in the defined as the ‘final solution’, given in the last row of
impedance within an HVDC link may result in unbalances Tables 7 to 10.
and generation of even harmonics that were not considered in Depending on the manufacturer’s factory installation,
the original current spectrum. Among them, the one with the a core with more than 4 meters in limb height would not be

23040 VOLUME 11, 2023

M. V. Czernorucki et al.: Multi-Objective Design Optimization for HVDC-LCC Converter Transformers

FIGURE 10. The feasible solutions found by the evolutionary constrained FIGURE 11. The feasible solutions found by the evolutionary constrained
multi-objective optimization process for the rectifier delta. multi-objective optimization process for the rectifier star.

losses depend on the load current frequency, with the har-

feasible, but the exercise is still valid to provide a view of monics’ spectrum (Tables 5 and 6), each ohmic and eddy
how the different solutions populate the objectives space in losses component, can be obtained for the fundamental and
two dimensions. The final choice was selected for being the all harmonic orders for a specific load profile. The losses
′
closest to the Pareto’s set [31], meaning that it resulted in the with harmonics P hw , (33), are obtained by adding each loss
best relation between total losses including harmonics and the component, proposed by [36], and [37]:
core and windings weight. X49
′
Although, depending on the installation purpose, the objec- Phw = Ph (33)
h=1
tives and restrictions setup may vary. For instance, if the
energy remuneration for a certain system is valued, the opera- The currents for each order are injected in the transformer
tor may prefer a more expensive equipment in terms of mate- valve winding 2D-FEM model. That calculation methodol-
rial investment, that provides lower losses in operation [9]. ogy differs from a standard design, which would use FEM to
On the other hand, if the transformer will be placed on an define the winding losses and apply F WE on the eddy losses,
offshore platform, the total weight and footprint limitations like in (22).
may justify a total loss increase. That choice can result in The core and windings geometry of each type of trans-
lighter core and windings, even though may result in other former was generated into a commercial two-dimensional,
side effects due to the high losses, which can force the axisymmetric model FEM software and utilized a time-
designer to select more expensive high-temperature material harmonic solver. The mesh element size was defined based
for windings insulation and oil. Those examples illustrate on the penetration depth [38], δ, calculated in (34) and fixed
how the transformer application can affect the objectives for the 49th harmonic order frequency (2940Hz) [21]. This
selection and the restrictions definition. resulted in an extremely refined mesh, illustrated in Fig. 12,
used for all simulations, which would not be needed for
lower harmonic frequencies, consuming more computation
V. VALIDATION WITH FINITE ELEMENTS ANALYSIS
resources. For the actual study that was not an issue, but that
The authors considered the winding eddy losses with harmon-
input may be adjusted for other calculations if needed.
ics a leading parameter to validate the analytic formulation
The element size is calculated by.(35) in [mm] and for
of section III once it depends on the leakage flux and can
be a source of deviations in the analysis. Those losses are 2940Hz it is 0.44986mm, for σ 0 =1 2.11410−8 and µ0 in
also frequency dependent and HVDC transformers are sub- the SI system. That element size represents less than 1/3
jected to several current harmonic orders. The valve winding of the winding conductor width, assuring good precision on
is the most stressed one in terms of harmonics, likewise the results. In the 2D simulation, the control region shall be
thermally [34]. That reduces the cellulose material life in defined by a maximum element size of two or three layers of
operation, resulting in possible partial discharges and a future the elements, which means, the size should be between 1/2
dielectric discharge, or an internal short-circuit failure due and 1/3 of the studied dimension [39].
s
to the reduction of the insulation’s mechanical strength [35]. 1
That directly impacts the transformer’s end of life and its δ= (34)
πf µ0 σ0
availability in the power system. The validation was done by
comparing the valve winding losses calculated by the analytic δ
ElemSize = 103 (35)
formulation with a numerical calculation using Finite Ele- 3
ments Method (FEM). Completely distinct methods applied The valve winding, designed as a helical or a disc winding,
over the same transformer geometry (section II). Since eddy had the geometry of each conductor detailed for λ number

VOLUME 11, 2023 23041

 M. V. Czernorucki et al.: Multi-Objective Design Optimization for HVDC-LCC Converter Transformers

FIGURE 12. Mesh geometry inside a single conductor (cross section view).

FIGURE 14. Magnetic field lines and magnetic flux B0 shaded plot.

FIGURE 15. Current density distribution in the refined portion shown in

Figs. 13 and 14.

losses distribution inside every single conductor of the refined

portion show the same behavior in Figs. 15 and 16, stressing
more the conductors placed in the inner diameter than the
ones in the outermost diameter. That is the expected behavior
FIGURE 13. Valve winding geometry showing the refined region.
for that loading analysis [40], illustrating the skin effect
inside the material. Before simulating the harmonic losses,
the transformer percentage impedance was checked based on
of turns or discs respectively, being λ/2 located on the top the magnetic field energy formulation using (36) [41].
and the other half on the winding bottom (Fig. 13). Each
conductor of the refined portion was modeled as a solid ele- 4πf Energy
Uk [%] = 100 (36)
ment and the rest of the winding as a stranded coil, correcting Sbase
the losses by adding the proportional eddy losses from the The Energy in [Joules] is taken directly from the FEM
analytic calculation. simulation and S base is the transformer power base per limb in
The first round of simulations was done setting λ = [VA]. Table 11 presents the calculated percentage impedance
20 and then, for one model, a sensibility study, varying λ was for the analytic calculation and the one resulting from the
performed. The main leakage flux is concentrated between simulation.
the line (B) and the valve (C) windings and the field line That is an important result to assure the simulation is being
close within the core window, as illustrated in Fig. 14. It is performed correctly. If the leakage flux energy results in
possible to verify that the discs or turns located at the top an impedance discrepant from what was expected, it is an
and bottom winding height suffer a more intensive influence indication that the model is not correct.
from the magnetic flux. That justifies the need of refining The four models, rectifier D and Y, and inverter D and Y
the winding in those regions. The current density and load were simulated setting λ = 20 and applying each harmonic

23042 VOLUME 11, 2023

M. V. Czernorucki et al.: Multi-Objective Design Optimization for HVDC-LCC Converter Transformers

TABLE 11. Percentage impedance result.

FIGURE 17. Winding losses of the four designs setting λ = 20.

FIGURE 16. Winding losses distribution in the detailed region shown in

Figs. 13 and 14.

TABLE 12. Losses results and deviation from the analytic calculation. FIGURE 18. Sensibility study for the valve winding losses based on the
portion of refined turns shown in Figs. 13 and 14.

formulation presented in section III is reliable for optimizing

the converter transformers designs.

VI. CONCLUSION
The formulation proposed in sections II and III was vali-
dated in section V by a numerical alternative calculation of
current into the models and obtaining the losses for the the valve winding losses considering harmonics, and by the
respective frequency, up to the 49th , resulting in the total short-circuit impedance verification, for the four transform-
′
winding losses P hw , (33). That result is compared to the ers configurations. The results deviation between the two
analytic calculation P hw , obtained from (24). The final losses methodologies showed a reduction when the winding conduc-
results are presented below in Fig. 17 and the deviation from tors’ geometry was more detailed, validating the formulation
both methods, taking the analytic calculation as the reference, considering the skin effect of the magnetic leakage flux in the
is shown in Table 12. The results show a deviation between extreme winding portions.
the two methods of less than 5%. However, the results from In section IV, the formulation is analyzed by a
Table 9 motivated the authors to investigate if the FEM multi-objective optimization, confronting two antagonistic
losses would converge, maintain, or diverge from the analytic objectives, total losses in operation and active part weight,
calculation when the refined portion increases. That purpose searching for the optimal solution in a 2-D objective space.
is to verify if the analytic calculation was fair enough when Some secondary tasks obtained by the study shall be high-
the conductors of the valve winding are represented in more lighted, such as the simulation of the HVDC link in PSCAD,
detail. defining the parameters of the transformers, and generating
That investigation took place using the rectifier the current harmonics spectrum. Likewise, the creation of an
Y-connected model, varying λ from 6 to 42, in steps of 6 and automatic script to draw the core and windings 2-D geome-
then two more cases considering λ = 60 and λ = 78, with try, input data, and run the FEM program simulation of the
three times the step (Fig. 18). harmonic losses.
The more discretized the detailed portion, the lower the Future improvements can also be foreseen, once the ana-
deviation between the losses calculated by the two different lytical formulation is simplified and can also be developed
methodologies, decreasing close to 2%. That result certifies considering other parameters such as the core structure, leads
the analytic calculation as a realist method, and that the connection, solid and liquid insulation, transformer tank, and

VOLUME 11, 2023 23043

 M. V. Czernorucki et al.: Multi-Objective Design Optimization for HVDC-LCC Converter Transformers

other metallic supports. The losses analysis may also include [17] Q. Ni, L. Luo, J. Fan, and Z. Jin, ‘‘Harmonic loss analysis of converter
other-eddy losses, which are generated by the magnetic flux transformer in LCL-HVDC system,’’ Energy Rep., vol. 6, pp. 352–357,
Dec. 2020.
impinging all metallic parts, which involves a 3-D analysis, [18] L. F. Blume, Transformer Engineering. Hoboken, NJ, USA: Wiley, 1951.
due to the asymmetry of the parts involved in that analysis. [19] R. M. D. Vecchio, B. Poulin, P. T. Feghali, D. M. Shah, and R. Ahuj,
The optimization algorithm may also consider other objec- Transformer Design Principles: With Applications to Core-Form Power
Transformers. Boca Raton, FL, USA: CRC Press, 2010.
tives and restrictions, depending on the HVDC converter [20] W. Rogowski, ‘‘Ueber das streufeld und den streuinduktionskoeffzien-
transformer application. ten eines transformators mit scheibenwicklung und geteilten endspulen,’’
Ph.D. dissertation, VDI, Mitteilung ueber Forschungsarbeiten auf dem
Gebiet des Ingenieurwesens, 1909.
ACKNOWLEDGMENT [21] Power Transformers—Transformers for HVDC Applications,
The authors thank Hitachi Energy for the use of the mentioned Standard IEC/IEEE 60076-57-129, Nov. 2017. [Online]. Available: http://
figures. webstore.iec.ch/publication/28003
[22] Y. Liu, D. Zhang, Z. Li, Q. Huang, B. Li, M. Li, and J. Liu, ‘‘Calculation
method of winding eddy-current losses for high-voltage direct current con-
REFERENCES verter transformers,’’ IET Electr. Power Appl., vol. 10, no. 6, pp. 488–497,
[1] International Energy Agency. (Apr. 2021). Global Energy Review—2021. Jul. 2016.
[Online]. Available: https://www.iea.org/reports/global-energy-review- [23] Y. Wang, W. Zhang, J. Wang, D. Xia, P. Zhang, and J. Li, ‘‘Stray loss cal-
2021 culation of HVDC converter transformer,’’ IEEE Trans. Appl. Supercond.,
[2] M. V. Czernorucki, M. B. C. Salles, A. S. Melo, E. C. M. da Costa, and vol. 22, no. 3, Jun. 2012, Art. no. 5500604.
L. Piegari, ‘‘Effects of the HVDC system on converter transformers,’’ [24] UHV Transmission Technology, Elsevier Inc., Amsterdam,
in Proc. 8th Int. Conf. Renew. Energy Res. Appl. (ICRERA), Nov. 2019, The Netherlands, 2017, doi: 10.1016/C2013-0-15562-6.
pp. 623–630. [25] Y. Shuai, X. Han, L. Zhang, C. Yang, X. Hu, and H. Wu, ‘‘Major
[3] M. V. Czernorucki, M. B. C. Salles, E. C. M. Costa, A. S. Melo, and insulation design consideration of converter transformer,’’ in Proc. Int.
L. Piegari, ‘‘Comprehensive overview on HVDC converter transformer Conf. Condition Monitor. Diagnosis (CMD), Xi’an, China, Sep. 2016,
design: Additional discussions to the IEC/IEEE 60076-57-129 standard,’’ pp. 1004–1007.
IEEE Access, vol. 10, pp. 40165–40180, 2022. [26] M. Yea, K. J. Han, J. Park, S. Lee, and J. Choi, ‘‘Design optimization for
[4] ANEEL. (2015). Bipolo HVDC Xingú—Terminal Rio Conversoras, LT-CC the insulation of HVDC converter transformers under composite electric
±800 kV Xingú—Terminal Rio, LTs E Equipamentos 500 kV Associados. stresses,’’ IEEE Trans. Dielectr. Electr. Insul., vol. 25, no. 1, pp. 253–262,
[Online]. Available: http://aneel.gov.br/transmissão4 Feb. 2018.
[5] W. Sun, L. Yang, F. Zare, Y. Xia, Li Cheng, and K. Zhou, ‘‘3D modeling of [27] F. Werner and N. Vakhania, Multicriteria Optimization Pareto: Optimality
an HVDC converter transformer and its application on the electrical field and Threshold-Optimality. Rijeka, Croatia: InTech, 2020.
of windings subject to voltage harmonics,’’ Int. J. Electr. Power Energy [28] S. L. Avila, ‘‘Otimização paramétrica com computação evolutiva,’’
Syst., vol. 117, pp. 1–10, May 2020. Tech. Rep., 2020.
[6] PSCAD/EMTDC User’s Manual, Manitoba HVDC Res. Centre, Winnipeg, [29] Power Transformers—General, Standard IEC 60076-1, Apr. 2011.
MB, Canada, 2001. [30] Submódulo 2.6. (Jun. 2021). Requisitos Mínimos Para Subestaçães E Seus
[7] M. Sanjo, M. Jeffin, J. Sajan, and A. Deepak, ‘‘Power electronic on-load Equipamentos. ONS. [Online]. Available: http://ons.org.br
tap changer for HVDC converter transformer,’’ Int. J. Innov. Res. Elect., [31] J. Liang, X. Ban, K. Yu, B. Qu, K. Qiao, C. Yue, K. Chen, and
Electron., Instrum. Control Eng., vol. 7, no. 4, pp. 50–58, Apr. 2019. K. C. Tan, ‘‘A survey on evolutionary constrained multi-objective opti-
[8] A. M. Sobrinho, ‘‘Uma contribuição aos projetos de transformadores via mization,’’ IEEE Trans. Evol. Comput., early access, Mar. 1, 2022, doi:
algoritmos naturais e elementos finitos,’’ Ph.D. dissertation, Departamento 10.1109/TEVC.2022.3155533.
de Engenharia Elétrica, Universidade Federal de Uberlândia, Uberlândia, [32] L.-X. Zhang, X.-Y. Kong, S.-S. Zhang, Z. Zhang, W. Lu, and
Brazil, Jun. 2019. X.-M. Zhang, ‘‘Optimization arrangement of angle rings in converter
[9] W. V. Calil, E. C. M. Costa, and M. B. C. Salles, ‘‘Methodology for specifi- transformer valve winding end insulation structure,’’ in Proc. IEEE
cation of generator step-up transformers in photovoltaic power plants based Int. Conf. High Voltage Eng. Appl. (ICHVE), Sep. 2020, pp. 1–4, doi:
on real loading profile,’’ Ph.D. dissertation, Departamento de Engenharia 10.1109/ICHVE49031.2020.9279713.
Elétrica, Univ. São Paulo, São Paulo, Brazil, 2020. [33] S. Yadav and R. K. Mehta, ‘‘Modelling of magnetostrictive vibration and
[10] Y. Zhao and P. Crossley, ‘‘Impact of DC bias on differential protection acoustics in converter transformer,’’ IET Electr. Power Appl., vol. 15, no. 3,
of converter transformers,’’ Int. J. Electr. Power Energy Syst., vol. 115, pp. 332–347, Mar. 2021.
Feb. 2020, Art. no. 105426. [34] HVDC Converter Transformer Failure Survey Results From 2013 to 2020,
[11] Y. Liu, G. Li, L. Guan, and Z. Li, ‘‘The single-active-part struc- CIGRE, Paris, France, Dec. 2021.
ture of the UHVDC converter transformer with the UHVAC power [35] P. Jiang, Z. Zhang, Z. Dong, Y. Wu, R. Xiao, J. Deng, and Z. Pan,
grid,’’ CSEE J. Power Energy Syst., vol. 3, no. 3, pp. 243–252, ‘‘Research on distribution characteristics of vibration signals of ±500 kV
Oct. 2017. HVDC converter transformer winding based on load test,’’ Int. J. Electr.
[12] B. A. Thango and P. N. Bokoro, ‘‘Stray load loss valuation in electrical Power Energy Syst., vol. 132, Nov. 2021, Art. no. 107200.
transformers: A review,’’ Energies, vol. 15, no. 7, p. 2333, Mar. 2022. [36] B. S. Ram, J. A. C. Forrest, and G. W. Swift, ‘‘Effects of harmonics on
[13] Y. Vernay, J. Michel, and J. P. Taisne, ‘‘Study and measurement of converter transformer load losses,’’ IEEE Trans. Power Del., vol. PD-3,
harmonics emission for the HVDC-LCC French station IFA2000,’’ in no. 3, pp. 1059–1066, Jul. 1988.
Proc. Int. Conf. Power Syst. Transients, Perpignan, France, Jun. 2019, [37] Q. Ni, L. Luo, J. Fan, and Z. Jin, ‘‘Harmonic loss analysis of
pp. 1–6. converter transformer in LCL-HVDC system,’’ in Proc. 7th Int.
[14] G. P. Anuoluwapo, ‘‘Power losses in HVDC converter stations,’’ Ph.D. dis- Conf. Power Energy Syst. Eng., vol. 6. Fukuoka, Japan, Sep. 2020,
sertation, Dept. Elect. Eng., Univ. Kwazulu-Natal, Durban, South Africa, pp. 352–357.
Nov. 2018. [38] D. K. Cheng, Field and Wave Electromagnetics, 2nd ed. Reading, MA,
[15] Y. Liu, D. Zhang, Z. Li, Q. Huang, M. Li, B. Li, and J. Liu, ‘‘Eval- USA: Addison-Wesley, 1989.
uation of the calculation method for stray losses in structural parts of [39] D. A. Lowther and P. P. Silvester, Computer—Aided Design in Magnetics.
HVDC converter transformers,’’ in Proc. 5th Int. Conf. Electr. Util- Berlin, Germany: Springer, 1986.
ity Deregulation Restructuring Power Technol. (DRPT), Nov. 2015, [40] M. C. Mgunda, ‘‘Optimization of power transformer design: Losses, volt-
pp. 1780–1785. age regulation and tests,’’ J. Power Energy Eng., vol. 5, no. 2, pp. 45–74,
[16] W. Peng, Q. Liu, Li Liang, W. Jiang, and Z. Zhang, ‘‘Design requirement 2017.
and DC bias analysis on HVDC converter transformer,’’ in Proc. AESEE [41] M. L. Magid, Electromagnetic Fields, Energy and Waves. Hoboken, NJ,
E3S Web Conf., vol. 257, 2021, p. 01038. USA: Wiley, 1972.

23044 VOLUME 11, 2023

M. V. Czernorucki et al.: Multi-Objective Design Optimization for HVDC-LCC Converter Transformers

MARCOS V. CZERNORUCKI was born in São FELIPE ALVES SOBRINHO received the degree
Paulo, Brazil, in 1975. He is currently pursuing in electrical engineering from Brasilia University,
the Ph.D. degree in electrical engineering with the in 2004. In 2002, he started working with power
Polytechnic School of the University of São Paulo. systems analysis and studies as a Trainee with
His research interests include different types of Eletronorte, where he worked as a Planning Engi-
transformers, such as industrial, phase-shifters, neer for the following 12 years. In 2014, he joined
and converter transformers, and their interaction ABB to work with HVDC first in Guarulhos,
with the power system under prospective designs. Brazil, and then in Ludvika, Sweden. Since 2022,
He is a member of Cigré Brazil, actively partic- he has been working as a Principal Consultant with
ipating on the working group A2.09 and on the the Power Consulting Team, Hitachi Energy Ltda.,
standards revision of the Brazilian Technical Standard Association. Raleigh, NC, USA, focusing mainly on new HVDC projects.

WILERSON W. CALIL received the B.Sc. degree

in electrotechnical engineering from the Mauá
School of Engineering, São Paulo, Brazil, in 2005,
and the M.Sc. and Ph.D. degrees in electrical engi-
neering from the University of São Paulo, São
MAURICIO BARBOSA DE CAMARGO SALLES
Paulo, in 2009 and 2020, respectively. He is cur-
(Member, IEEE) received the Ph.D. degree from
rently pursuing the Ph.D. degree in electrical engi-
the University of São Paulo (USP), Brazil, in 2009.
neering with the Polytechnic School of the Univer-
From 2006 to 2008, he was with the Research
sity of São Paulo. Since 2004, he has been working
Team of the Institute of Electrical Machines,
with Hitachi Energy Ltda., where he also holds
RWTH Aachen University. He has been an Assis-
the position of a Principal Project Engineer. His research interests include
tant Professor with the Laboratory of Advanced
applied electromagnetism, magnetic and thermal simulations, power systems
Electric Grids, Polytechnic School of the Univer-
modeling, power transformer design, transformers applied with renewable
sity of São Paulo, São Paulo, Brazil, since 2010.
sources, and high-voltage techniques for experimental analysis. In addition,
From 2014 to 2015, he was a Visiting Scholar with
he is an active member of the International Council on Large Electric
the Harvard John A. Paulson School of Engineering and Applied Sciences.
Systems (Cigre) and has several internationally renowned publications.
His research interests include distributed generation, power system dynam-
ics, control and stability, renewable energy, energy storage, and electricity
markets.

JOSÉ ROBERTO CARDOSO (Life Member,

IEEE) was born in Marilia, São Paulo, Brazil,
in 1949. He received the B.S., M.S., and Ph.D.
degrees in electrical engineering from the Poly-
technic School, University of São Paulo, São
Paulo, in 1974, 1979, and 1986, respectively. Since
1999, he has been a Professor with the Electrical
SÉRGIO LUCIANO AVILA received the dual Energy and Automation Engineering Department,
Ph.D. degrees from the Ecole Centrale de Lyon Polytechnic School, University of São Paulo. He is
(ECL), Lyon, France, and UFSC, in 2009. currently the President of the SBMAG—Brazilian
From 2006 to 2007, he was with the Research Electromagnetics Society and the former Dean of the Polytechnic School
Team of the Grenoble Institute of Technology of the University of São Paulo. He is the Founder of the LMAG—the
(Grenoble INP). He has been an Associate Pro- Electromagnetic Applied Laboratory. He is the author of four books and
fessor with the Laboratory of Applied Scientific more than 70 articles and has guided more than 40 theses/dissertations.
Computing (PECCE), Federal Institute of Santa His research interests include numerical methods for electromagnetics, the-
Catarina (IFSC), Florianópolis, Brazil, since 2010. oretical electromagnetics, electrical machines, electromechanical biomedi-
From 2017 to 2018, he was a Visiting Scholar with cal application, and electrical railway research and engineering education.
the Polytechnic School of the University of São Paulo, São Paulo, Brazil. He was a recipient of the 2013 Emeritus Engineer of the Year in São Paulo.
Currently, he works on solutions for monitoring and predicting behavior in He was awarded Doctor Honoris Causa by the Grenoble-INP Institute of
machines and electrical systems, which involves instrumentation, numerical Engineering, France, in 2019.
modeling, optimization, and computational intelligence algorithms.

VOLUME 11, 2023 23045