Comprehensive Overview on HVDC Converter Transformer Design: Additional Discussions to the IEC/IEEE 60076-57-129 Standard

HVDC is an economical and technical advantageous technology for power transmission through long distances, asynchronous interconnections and long submarine cables crossing. Despite DC transmission benefits to power systems, the converters non-linearity produces undesirable effects to the converter transformer in service, mainly listed in the technical standard IEC/IEEE 60076-57-129. However, additional discussions and complementary information can be found in a plurality of references, which are brought in the article under a comprehensive overview perspective. Some design solutions deal with these effects increasing the technical margins, which have direct influence on manufacturing costs and transformer reliability and availability levels. This article goes through the main topics pointed by the standard and the references, investigating their consequences in the converter transformer operation, in order to provide a comprehensive tutorial on design solutions and considerations to deal with those undesirable effects.


I. INTRODUCTION
HVDC system technology became commercially and practical feasible with the advent of the mercury-arc valves in the 1950's, developed to a solid-state thyristor in late 1960's and introduced new technologies, allowing DC transmission implementation on a wider number of applications in the late of 1990's [1]. It has been considered a technical and economical solution for bulk energy transmission in long distances, to interconnect asynchronous systems and provide power delivery through long submarine cables crossing. The global interest in the HVDC system solutions is evident by the number of projects within the last years [2].
Nowadays, renewable energy generation represents an important share of the global energetic matrix, as a reliable, environmental and economical solution, contributing in a strategic way to the global carbon generation drop policies [3]. HVDC technology can integrate different renewable energy sources, promoting enough reliability and power quality for energy transportation, even under uncertainty wind flow and sunshine intensity [4]. It also provides a higher system efficiency when compared to an equivalent high-voltage AC system [5]- [7], connecting energy generation plants in different regions, countries or even continents. Thus, HVDC increases the electricity marketing relation [8] and helps to strength networks shortcircuit power, increasing reliability and energy availability, under a smart grid prospective [4], [9]. In this context, some worldwide projects can be mentioned, such as BritNed project connecting Great Britain to the Netherlands, Fenno-Skan connecting Finland to Sweden [10], or EuroAsia and EuroAfrica interconnections between Israel-Greece and Egypt -Greece respectively. Some reasons that make HVDC more competitive than HVAC for power transmission through very long distances are the number of conductors of the overhead line, what also reduces the environmental impact and terrain expropriation costs [11], reflecting in more than 50% rightof-way shrinkage, as illustrated in Fig. 1. The costs are further reduced (around 25 to 35%) once the HVDC does not require reactive compensation along overhead lines, underground or submarine cables. Furthermore, losses in The HVDC systems represent an efficient solution for interconnecting asynchronous systems, since DC installation performs as a buffer, isolating the systems electrically [14]. It can also connect 50Hz and 60Hz electrical system, e.g., the Itaipu link, which transmits 6300MW over 800km by a ±600kVdc line [15]. Finally, DC links can operate synchronized with AC systems, improving power flow control, avoiding cascade failures and blackouts [16]. Figure 3 illustrates the break-even point where an HVDC installation becomes commercially competitive. The break-even distance differs between overhead transmission lines and underground/submarine cables due to the base material cost [17]. The HVDC system main costs refer to the converter station, including the valve hall, smoothing reactor, transformers and the converter valves, and not essentially to the overhead lines. The HVAC initial costs are lower, increasing while the transmission reaches longer distances, due to the reactive compensation, represented by the dashed lines [18]. The total system cost is a combination of the terminal installation and the energy losses, represented by the continuous curves in Fig. 3 [15].

A. WORLDWIDE APPLICATION
AC technology is being successfully applied in the power generation, transmission and distribution systems worldwide. Nevertheless, it has technical and economical limitations on implementation and operation, that make DC transmission a realistic and, often, an even unique alternative [20].
The greatest HVDC installations are located mainly in China and India. In the references [21]- [26] the reader can find HVDC links worldwide, listed by the commissioning date, country of installation, transmission line length, DC system voltage, transmission power and DC current. The references [27]- [33] provide an overview on the HVDC projects in operation categorizing reliability, performance, energy availability, energy utilization, forced and scheduled outages and other data taken from different utilities along the years.
Nowadays, there are two basic converter topologies commonly applied: the line-commutated Current Source Converter (LCC or CSC) and the self-commutated Voltage Source Converter (VSC) [34].

B. HVDC SYSTEM TOPOLOGIES 1) LINE-COMMUTATED CURRENT SOURCE CONVERTER
The conventional line-commutated LCC (Fig. 4) is built using a six-pulse or Graetz bridge. It is common to have two six-pulse bridges connected in series, resulting in a twelve-pulse bridge in cascade. This is obtained by using two transformers connected to each 6-pulse bridge with different phase displacement in the valve terminal. One transformer is built with a delta connection and the other is wye connected. A thirty-degree displacement is applied in the AC voltage source, establishing a distinct thyristor firing sequence within the period. This combined operation does not eliminate fifth and seventh current harmonic generation, but despite they still flow through the valves and transformer   The LCC technology requires a strong network connection to operate successfully, since the thyristors commutation are synchronized with the AC network. That avoids the voltage instability, but also does not permit LCC to be connected to an isolated load, like an offshore platform. The network strength is measured by its short-circuit ratio (SCR) which is the relation between the system short-circuit power by the converter station rated power [35].
Nowadays LCC installations are reaching DC voltage rates of ±1100kV and transmission power up to 12GW in projects like ChangJi-GuQuan [36], [37] and XinJiang-AnHui in China [38], [39]. Figure 5 displays how HVDC developed along the years, by rated voltage and power, and its rapid growth in the last 25 years, picking some main projects in each period. 2) SELF-COMMUTATED VOLTAGE SOURCE CONVERTER Self-commutated VSC ( Fig. 6) with pulse-width modulation (PWM) was introduced as a system solution in the 1990's. The increase in power and voltage ratings has enabled VSC to operate where, in the past, only LCC would be used. The VSC converters are self-commutated by an insulated-gate bipolar transistor (IGBT), connected to soliddielectric extruded HVDC cables.

FIGURE 6. VSC valve connection inside the valve hall, authorized by Hitachi Energy
VSC permits a rapid control of the active and reactive power independently, controlled at each terminal, no mattering the DC transmission voltage level. This control flexibility allows VSC converters to be placed anywhere in the system, without restriction such as minimum network short circuit capacity. Self-commutation allows the converter to synthetize a balanced three-phase voltage which simulates the operation of a synchronous generator, using the control capability. The dynamic support of the AC voltage in each converter improves the voltage stability and permits a transferred power increase between sending and receiving ends, thereby extending the transfer capability of the DC link [41].
Projects like EuroAsia, interconnecting Israel and Greece, EuroAfrica, interconnecting Egypt and Greece, and Western Link in UK, certify VSC as suitable to transmit energy in DC voltages higher than ±500kV and rated powers over 2000MW. Due to the self-commutating characteristic, it has been widely used to connect offshore links to wind farms, supply oil and gas isolated platforms.
The table below describes the HVDC implementation in each application according to the system topology. The Capacitor Commutated Converter (CCC) was not mentioned before. It is similar to the LCC topology, with a commutated capacitor placed between the converter transformer and the thyristor bridge. It gives a contribution for the valve voltage, allowing the operation on smaller firing angles, reducing the reactive power requirement, appearing to be less dependent on the AC network.

3) THE STANDARD FOR HVDC APPLICATION
The IEC/IEEE 60076-57-129 [42], titled Transformers for HVDC Application, has the first edition published on November 2017. It is a combination of the normative experiences brought into the former IEEE C.57.129 and the IEC 61378-2. The standard brings specific requirements related to losses, tests, sound level, components such as bushings and tap changer, some informative annexes related to overloading, transformers for VSC converters, design review, audible sound and service load losses determination.
The scope of the standard [42] includes single or three-phase HVDC transformers of two, three or multiple windings and back-to-back applications. Industrial and traction transformers are excluded, being normalized by IEC 61378-1 / IEEE C57.18.10 and IEC 60310, respectively.
This article intends to point the main matters addressed by the standard and provide additional discussions found in an extensive search on the literature of each topic, focusing in the effects of the HVDC system operation over the converter transformer design. It has not the presumption to approach all subjects the technical standard presents, but it purposes to give the reader a comprehensive overview on the six main effects found in the standard, which are transformer loading, load losses, sound level, thermal design, DC bias and electric field distribution, under an HVDC transformer design perspective. They are fully part of LCC operation, however VSC may also be susceptible to some, like for an asymmetric topology or when the AC filters are not located between the converters and the transformers.

III. HVDC SYSTEM EFFECTS ON TRANSFORMERS
The converter transformer makes the electrical link between the AC and DC transmission systems, responsible for part of the DC voltage regulation and permits the reactive power control. The converter valves non-linearity operation causes some undesirable effects that demand a special design of the converter transformers [43].
Increasing technical margins not necessarily enlarge safety margins or life expectance, but it will surely reflect in manufacturing costs. However, enough margins and a suitable design should be verified by a consistent design review [44]. Guidelines for such a revision can be found in [45] detailing checking points for specification requirements, system data, environmental conditions, transformer design, fabrication, inspection, test plan, transport and installation.

A. HVDC TRANSFORMER LOADING
Since HVDC systems, specially LCC, transport high bulk energy and interconnect large power systems, the converter transformers normally operate under stressful load profiles. During normal operation, the valve switching provide a considerable harmonic content, that increases transformer losses and noise level. Also, overload cycles are usually specified in the technical documentation. Details of those effects and their undesirable consequences are given in the coming sections.
A fault in the DC line or in the DC yard may require the HVDC system to operate under a reduced voltage profile. That operational mode requires a larger thyristor firing angle 2, increasing losses and localized hots-spots, when compared to normal operation, angle 1, consuming more reactive power (Fig. 7).

1) IEC/IEEE 60076-57-129 HIGHLIGHTS
The converter transformer load current is composed of the fundamental frequency component and the harmonic orders, whose magnitudes depend on the converter station, and include some residual DC components.
The HVDC system consumption of reactive power will penetrate the transformer windings, likewise the current harmonic content and any wave form distortion, which must be considered in the design. That information should be analyzed in combination with the AC voltage and tap changer position [42].
Overloading the transformer in service is a result of the following events:  Planned overloads;  Emergency overloads;  Failure of auxiliary equipment of the unit or of the converter station.
Any overload may result in significant reduction of the insulation life, increasing the risk of a transformer failure. Planned overload condition can be combined to a low ambient temperature profile or limit duration capability.
The overload condition shall be informed with additional data as ambient temperature, duration, overload factor [p.u.], preload [p.u.], number of coolers in service, permissible overtemperatures for oil, windings and hotspot, harmonic spectrum and tap changer position.

2) ADDITIONAL DISCUSSIONS
Transformers for LCC installations are normally specified with on-load tap changer. Despite the AC line voltage variation, which is limited by the standards between 5 and 10%, the tap changer is used to control the transformer valve side RMS voltage [46]. In combination with the thyristors firing angle, the transformer valve side voltage is used to control the amount and direction of the DC link power flow.
During normal operation the converter  angle can be set for instance in 15° and the inverter  angle in 17°. The transformer tap changer in the converter station transformer is set in a lower tap, resulting in a higher valve terminal voltage than the inverter station transformer. That defines a normal power flow from converter to inverter station and the power magnitude is a result of those parameters. During reverse operation the pole voltage polarity is changed, in order to force the power flow to reverse [47], what stresses the transformer insulation by a combination of AC and DC stresses. That is tested by the polarity reversal test during the FAT (Factory Acceptance Tests). In reverse operation, normally the transmitted power is reduced due to inverter's capacity limitations.

FIGURE 7. Single-phase full bridge wave form of an in-phase voltage and current considering a small (a) and large (b) thyristor firing angle
The HVDC transformer voltage regulating tap range is far wider than for power transformers. It is common to require more than 30% range above the nominal voltage, reflecting in a large regulating winding number of turns. This is used during reduced voltage profile, when the system DC voltage is required to be dropped around 70 to 80% of the nominal pole voltage. The thyristor's firing angle is set around 30°, resulting in a much higher current harmonic distortion, the transformer losses and reactive power consumption [48]. During normal operation, the reactive power represents about 50% of the active system power [49], [50]. During reduced voltage operation this amount can reach more than 70%. A high reactive loading is relevant when the AC system requires to improve stability, transmission efficiency and HVDC conversion performance, controlling the steady-state and fast voltages transients [51]- [53]. The reactive loading can be seen by two perspectives: load compensation and voltage support [54].
During the overload, the firing angles are kept close to the ones practiced during normal operation or even lower, like 12°. That results in a less severe harmonic spectrum during overloads, but the current magnitudes can reach 133% for planned and 150% for emergency cycles. For a twobipole installation those overloads represent the loss of one monopole or one bipole.

B. LOAD LOSSES CALCULATION
The converter transformer leakage flux contains the same harmonic orders than the load current, increasing additional load losses in the windings and in the internal metallic parts, such as core clamps, tank, flitch-plates and active part or tap-changer supports. This augmentation can be around 25 to 30% of the total load losses, resulting in local hotspots higher than in a conventional transformer operation, what may generate dissolved gases in oil [48], [55].

1) TECHICAL STANDARD HIGHLIGHTS
The winding rated current combines the fundamental with different harmonic components, which may change significantly from one converter installation to another.
The eddy losses are generated by the leakage flux impinging the winding conductors, creating a circulating current in the material. The same situation happens in other metallic parts, resulting the stray other-eddy losses [56]. The standard [42] gives a formulation for the total load losses calculation: (1) Being, P N : total load losses P WE1r : additional eddy losses P SE1r : additional stray other-eddy losses The factors F WE and F SE are defined as: Being and ℎ .
From (2) it is possible to verify the eddy losses dependency of the harmonic frequency by an exponent 2 and from (3), the stray losses by an exponent 0.8. That is an important information when measuring the load losses in different frequencies.
Considering all the current harmonic orders up to the 49 th , it is possible to calculate the rated current, which consists the transformer load current in nominal condition: Different harmonic spectra can be informed to the manufacturer for several loading or overloading conditions. The one to be applied in the temperature rise test, performed in the highest equipment current tap position, may not be the same than the one used for load losses evaluation. During temperature rise test, an equivalent load current can be applied: , No-load losses can be affected by voltage harmonics, but it is normally negligible when compared to the total losses. Nevertheless, DC-bias can affect significantly the noload losses and it shall be considered in the design stage and the losses increase during temperature rise test.

2) ADDITIONAL DISCUSSIONS
The standard method has been applied for factory testing, but it may not represent completely the reality in service [57]. The load current in operation generates a nonlinear and non-sinusoidal leakage flux, which results in additional losses and located overheating [58].
The references [59]- [62] propose different methods for additional load losses calculation considering the winding inductance variation, material nonlinearity and hysteresis characteristics, frequency domain behavior and electromagnetic circuit coupling. According to the authors, those are sources of errors when calculated and measured losses are compared. For instance, the load current and the leakage flux should be represented for each harmonic order and simulated in a 2-D FEM (Finite Element Method) program to calculate winding eddy losses. The other-eddy losses should be calculated by a 3-D FEM program due to the active part and tank structure asymmetry [63]. Magnetic and nonmagnetic materials do not behave similarly on losses. That may also be a source of error, when a more detailed calculation is made, considering material hysteresis, nonlinearity behavior and frequency characteristics.
Some corrections and alternative formulations are proposed by [64]- [66] to adjust the technical standard factors. In fact, losses calculation in HVDC transformers is being an issue and it has encouraged many authors to seek alternative methods to do it more accurate, considering effects such as skin and proximity, which are not covered by the traditional calculation [67].

C. THERMAL DESIGN
The transformer thermal behavior is defined by the losses' distribution inside the active part. In LCC installations, the harmonic filters are connected to the transformer AC terminal, so all converters nonlinearity components are driven directly to the valve winding.

1) TECHICAL STANDARD HIGHLIGHTS
Due to the frequency harmonic and DC current effects, the losses may not be evenly distributed along the winding height, like expected during temperature rise test [42]. So, in operation, hotspot locations and magnitudes may be significantly different [69]. The intense radial leakage flux generates high temperatures in the extreme winding discs or turns, represented by a multiplying constant called hotspot factor. The winding losses are calculated normally by a 2D-FEM software [70], considering the harmonic content, following the method described in section 3.1.1.
The standard [71] proposes an indirect calculation method based on the mean winding-oil temperature gradient and the tank oil temperatures, using the hotspot factor to define the winding hotspot from calculated and measured temperatures. Higher hotspot factors are expected in converter transformers than in power transformers.
The hotspot temperature can be measured directly by placing fiber optic probes inside the transformer windings, especially at the main windings top discs or turns. The hotspots are located at a discrete and normally restricted region, so even defining a loss distribution calculation and an oil flow circulation, the designer cannot fully assure the fiber optic location will be placed exactly at the same region where the highest temperatures occur. It is not insured the installation of one or two fiber optics sensors will detect relevant hotspot temperatures. Up to 10K difference were found on experiences with sensors located at windings top discs.
Additionally, probes can be placed in the core, between adjacent sheets and in other active part and tank locations where the hottest spots are expected. Those actions increase the effectiveness of the measurements taken by the direct method.
Load beyond nameplate shall be specified giving its duration, redundant cooling in operation (on/off), ambient temperature, DC and RMS currents and specific harmonic spectrum. The harmonic spectrum composition changes significantly depending on the different load profiles and may submit the transformer insulation to severe thermal stresses.
The standard [42] reports some side effects by loading a transformer beyond nameplate like, the insulation material gassing evolution, active part cellulose structure mechanical strength reduction, gaskets leaking, tap changer contacts premature wear, auxiliary equipment aging and excessive oil expansion, increasing tank internal pressure.
Also, metallic structures may suffer localized saturation due to the high leakage flux magnitude, such as tank, tank shields, bushing turrets, core laminations, core clamps, tie-plates, yoke bolts, etc. Installing thermocouples to monitor metallic parts not directly in contact to the insulation, may be a good practice. The thermal behavior is verified externally by a thermal-scan during the heat-run test to check possible hotspot regions (Fig. 8).

2) ADDITIONAL DISCUSSIONS
All specific load harmonic content shall be informed, even those related specific load conditions where the commutation angle is expected to be higher than the nominal. Overload and special applications might result in transformer and its components life expectance reduction, since hotspots can reach temperatures beyond cellulose thermal limits, reducing its mechanical strength, causing a premature dielectric failure due to the insulation degradation along the operating life [72]- [74]. A complete failure report involving thermal, dielectric and other failure causes is found in [75] which is an HVDC transformer, including back-to-back, failure survey with data collected along the years.

FIGURE 8. Converter transformer external thermal scan on tank during heat-run test, without (a) and with (b) thermal scan use of image authorized by Hitachi Energy
The DC voltage of a back-to-back installation is much lower than in a classic HVDC. Therefore, the valve windings are normally located next to the core. Furthermore, in a threephase unit, both valve windings, delta and wye connected, are manufactured in the same winding shell, like in GSU (Generator Step Up) transformer axial split winding. But, due to the 30° phase displacement and the low reluctance, there is a considerable leakage flux radial component different than zero in region "A" of Fig. 9, between the two DC windings, directed to the core, resulting an overheating of the winding turns located in "B".

FIGURE 9. Back-to-back converter transformer winding arrangement per limb
The investigation [76] verified that the delta winding insulation was more thermally stressed than the wye, since its strand dimension was 18% higher, suffering a greater overheating due to the higher losses. The reference also presents a calculation method fixing the leakage flux in wye winding in 1.0pu, showing the leakage flux in the delta reached /√6=1.28pu at fundamental 60Hz frequency, Fig.  10(b). The resulted RMS leakage flux in region "A" was 44.8% at the fundamental frequency, added to other harmonics of higher frequency orders, showed in Fig. 10(c).
The thermally damaged winding insulation had the mechanical strength dramatically reduced, resulting in two 240MVA transformers dielectrically failed in service, followed by the spare unit gas generation, which was in operation just for 18 months, 15 of them under a reduced load of 180MVA. Simulations showed the hotspot reached 159C on the damaged region, against the 120C limit for thermally upgraded paper.
The failed units valve windings were redesigned with a smaller conductor height and removing the extra insulation used at region "A". The paper thickness was not a dielectric requirement, once the rated RMS voltage was only 23.3kV and there was enough distance between the two windings and from windings to core. In fact, it was reducing the copper-oil heat exchange capacity, collaborating with the overheating in that region.

D. SOUND LEVEL
The converter transformer sound level under operation is a combination of many components divided in:  Core excitation at rated frequency;  Winding load current;  Harmonics in the load current and service voltage;  DC bias effect;  Cooling equipment.
The complete sound level will be a combination of the different sound powers, obtained by a logarithm sum of all those components.

1) TECHICAL STANDARD HIGHLIGHTS
The sound level is expressed as the A-weighted of the total sound power, including all frequency bands or as the Aweighted sound level in 1/3 octave band. It is used to assure the total specified converter station sound level [42].
The core sound frequency spectrum consists mainly of the power of frequency even harmonics. For 60Hz, it is represented by 120, 240, 360 and 480Hz [77].
The harmonic influence in load audible sound occurs only in service and cannot be reproduced during the FAT due to unavailability of sufficient power supply in the manufactures facility to generate distorted wave shapes [42]. For a conventional transformer it consists mainly of twice the power frequency, i.e., 120Hz for 60Hz [69]. For converter transformers it is associated to the converter harmonic spectrum in different frequencies.

FIGURE 11. Oil to air cooler bank installed in an inverter station
The cooling system sound level shall be measured separately and added to the total sound calculation by a logarithmic sum. The frequency spectrum is defined by the motors' broadband, plus some discrete tones and the harmonic content [77].

2) ADDITIONAL DISCUSSIONS
The core excitation refers to the magnetic circuit vibration, excited usually at rated voltage. Even though, the guarantee can be specified in other excitation condition, but always at rated frequency. It is defined by the core geometry (limb height, limb pitch and core diameter), the actual flux density and the core steel type. The tank design can also influence if magnetic tank shunts and sand inside the stiffeners are used.
The windings vibration is defined by the load current and, in case of converter transformers, the current harmonic content [78]. It is defined by the winding geometry (diameter, radial width and total height), the conductor dimensions, the number of single strands or within a CTC (continuous transposed cable) cables in radial and axial directions, the number of spacers along the winding circumference and the winding block distance to tank.
Comparing a converter transformer to an equivalent power transformer, the sound level may vary from 2 to

FIGURE 12. Sound panels (a) and enclosure (b) installation, use of image authorized by Mecart Transformer Screens and Enclosures [80]
The cooling equipment is composed by fans, pumps, oil to air (Fig. 11) or oil to water coolers, depending on the transformer application, installation requirements or design choice.
The core, load and cooling system audible sound can be optimized by the designer, by choosing suitable materials or dimensions and distances, depending on the specified and environmental requirements. Some actions to minimize the total noise level are the use of high oriented grain or domain refined core steel, use low noise cooling fans, increase windings design space factor, increase windings to tank distances, consider sand inside tank stiffeners, install external sound panels (Fig. 12a) or sound enclosures (Fig.  12b). These actions can drop down the sound level in more than 10 dB(A). The installation of sound panels or enclosures may increase the oil temperature rise, for the tank cooling capacity reduction.

E. DC BIA EFFECT
A DC current flowing through the transformer winding is denominated DC bias current. It will create an offset of the rated core magnetization current showed in Fig.  13. Every electronic amplifier has some bias effect and HVDC valves are not an exception.

1) TECHICAL STANDARD HIGHLIGHTS
An unsymmetrical converter switching causes an uncompensated DC bias current, resulting in a half-cycle core saturation, showed in Fig. 13. DC bias affects the noload losses, noise level and cooling design, however, the standard [42] mentions the losses increase can be neglected compared to the transformer total losses. DC bias effect cannot be verified by FAT, only in-service operation.

2) ADDITIONAL DISCUSSIONS
The returned DC current magnitude is typically assumed to be up to 10A in the neutral [42], [45]. It is created by a combination of the ground potential increase, called ground potential rise (GPR), an electromagnetic coupling effect and an unsymmetrical thyristor's firing operation. In this section, the first two causes will be detailed once the third one tends to be minimized by the operational control.

A) GPR -GROUND POTENTIAL RISE
The DC bias current sourced by the GPR results in a current return by the neutral terminal during monopolar, bipolar or homopolar operation, under normal, emergency or fault conditions [82]. Any AC transformer or metallic utilities located in the surrounding area of the electrode or the converter stations, may be affected by the GPR. Returning currents can be measured in cooling pipes, cooling equipment, cabinets and other accessories, causing premature corrosion.
The DC current magnitude is a function of the electrode configuration, ground geological composition [83] and the installation distance from the HVDC electrode. It is simulated using different soil models, as horizontally layered, vertically layered or exponentially varying [84]. It is verified by two-electrode voltage difference (voltage source) and the analyzed AC system zero sequence resistance. With fewer substation transformers in operation and minimum AC system zero sequence resistance, which means, a maximum number of AC lines connected between the electrode stations, the maximum DC bias current will be sourced, and the highest magnitude will be reached. So, it will increase proportionally to the electrode surrounding AC system expansion.
The electrodes are usually constructed in two different configurations: Shallow Horizontal like ring (Fig.  14), n-pointed star, branched star and linear electrode arrangements, and Deep Vertical which consists in the interconnection of several deep conductors to reach low resistance soil layers. The vertical electrode is used when the land dimensions do not allow a horizontal electrode installation. In [84], the maximum earth gradients given by the touch and step voltages are analyzed. It is also possible to verify the measured GPR for ring and 4-arm star electrode, for a 2000A ground return current and a soil layered in three layers of 50 ohm-m and 4m thickness in the top, 11 ohm-m and 18m in the middle, and a bottom layer of 260 ohm-m resistivity.

FIGURE 14. HVDC ground electrode and soil layers division
The maximum voltages in the tables 2 and 3 are achieved in distances below 800m for the ring pattern and below 1200m for the 4-arm star.
Currents up to 6.6Adc were measured along a 600mile (966 kilometers) Pacific Gas Transmission and Electric Company pipeline located 35 miles (56.3 kilometers) far from Celilo electrode, which is part of Dalles-Los Angeles link in USA. The ground return transmitted current was 2000A [84].
The HVDC link connecting Talcher-II to Kolar  station, through a 1370km 2000MW DC transmission line  (±500kVdc, 2000A), had in 2003 an unexpected GPR registered right after a monopolar operation with ground return [85]. The cause was assigned to the geological composition and not directly the electrode design.
GPR measurements in Hydro-Quebec's Radisson-Sandy Pond link showed magnitudes around 30Adc per phase in the Radisson substation 315kV converter transformers and 25Adc per phase in the 735kV autotransformers.
The reference [86] shows results of a DC current injection test in two single-phase autotransformers of 370MVA and 550MVA respectively, both 735kV at HV. The current was injected in the tertiary winding in steps of 425, 850, 1700 and 2550A, resulting in 12.5, 25, 50 and 75A on the HV terminal. A voltage source varying between 0-55V, connected to a diode bridge, was capable for supplying a current up to 3000Adc. The measurements detect significant increase in core losses and magnetizing current peaks. It also verified the core structure flitch-plates losses and temperature rises. Excessive current magnitudes shall be monitored and avoided for more than few minutes. The transformer design influences considerably the DC bias effects in the core. Three-limbed three-phase cores are less affected than a threephase five-limbed or single-phase cores. That happens because a three-limb core design provides a higher zero sequence reluctance, dropping down the zero-sequence flux generated by a DC current. A smoother magnetization curve withstands higher DC current than a stepped one. This behavior is related to the selected steel material. The reference [87] simulates and analyzes the DC current effect in a three-phase five-legged core (three main and two return limbs) model in terms of magnetizing current, low harmonic orders increase and magnetic field intensity.
The GPR can be verified by the installation distance to the electrode, considering some parameters such as station GPR in volts, station GPR as a percentage of the DC GPR and magnetizing current peak as a percentage of the nominal value [88]. The parametric study presented in that reference considered a typical system under monopolar operation with ground return and all those parameters are analyzed in detail.

B) ELECTROMAGNETIC COUPLING EFFECT
The electrical system expansion may result in AC and DC lines physical proximity and parallel length in the same right of way, causing an electromagnetic coupling [89]. That interaction results in an overvoltage on the DC system, higher than expected for a conventional DC link [90], promoting low-order harmonic interaction [91], steady-state induction of fundamental frequency voltages [92] and zero sequence fault currents induction [93].
The electromagnetic coupling of an AC line in normal operation induces a fundamental frequency AC circulating current along the DC coupled line. That current, driven into the thyristors of the converter or inverter station, is naturally switched to a DC current; which flowing through the HVDC transformer valve winding, causes an increment in the core saturation [94] and promotes a core saturation instability [95].
The reference [96] presents a study of a typical AC line coupling effect over an LCC-MMC (multilevel modular converter) hybrid system (Fig. 15). The power was transmitted from Yunnan to Shanghai, in China, over a 1018km overhead line. The study set the parallel length between 20 and 200km and separation distance from 40 to 200m. The DC link transmits 3000MW at ±500kV and the AC system, 5000MW at 1000kV.

FIGURE 15. Parallel installation of HVDC and HVAC systems
The line induced voltage was simulated varying the separation distance between 40 and 200 meters for a fixed line length of 100 kilometers, and the parallel length from 40 to 200 kilometers for a fixed separate distance of 100 meters (Fig. 16). The positive pole is closer to the AC line than the negative one. That is why a higher voltage is induced in the positive line, more than 30kV for a separation distance of 40 meters, decreasing significantly for distances over 100 meters. The induced voltage does not show significant variation for distances over 140 meters.
The induced voltage is not proportional to the parallel length, since the voltage is a function of the line inductance (self and mutual) and the coupling capacitance. The aim of the study is to verify the magnitude of the circulating current induced along the DC line connecting the rectifier LCC and the inverter MMC stations as a function of the separation distance and the parallel length (Fig. 17). The shorter the separation distance, the greater the circulating current will be. For separation distances over 140m, the current magnitude does not change significantly.
The reference [97] does not recommend a circulating current higher than 30A. Considering that, the separation distance should be kept over 40 meters.

F. ELECTRIC FIELD DISTRIBUTION IN THE INSULATION SYSTEM
Besides the DC electric field distribution considerations, the dielectric design must consider transient effects such as:  Polarity reversal for power flow inversion, causing an electric field disturbance for a short period, stressing the insulation system;  Valve misfire in the converter station, promoting oscillatory voltage surges in the valve windings;  Converter station opened line switching, causing voltage surges with magnitudes up to 1.5pu of the rated voltage;  Lightning impulses in the overhead lines reaching magnitudes around 2 or 3 times the system rated voltage but being a function of the protective devices.

1) TECHICAL STANDARD HIGHLIGHTS
The transformer insulation system is mainly dielectrically stressed in service by the sources described above, extracted from [42] and [98], and FAT is the standard method to check its integrity and robustness [99]. The details of the tests listed below are found in [42]  Applied tests are performed during a long duration period, combined with partial discharge (PD) measurement, intending to simulate the DC and AC stresses on the valve side during converting operation. In a similar way, Polarity Reversal simulates LCC system reversal operation by changing the DC voltage polarity, combining direct and alternating stresses.

2) ADDITIONAL DISCUSSIONS
The valve windings are exposed to the converter operational DC voltage component [101]. By using a 2D-FEM program and analytical equations, the Laplacian AC field is precisely obtained, once there are no free chargers, resulting in a quasi-stationary situation, from a capacitive distribution, where rotE=0 [102] and [103]. Furthermore, that behavior is similar to the conductivity, stressing the solid insulation more than the oil [104], differently than the AC fields.
Temperature and moisture cause little variation in the main material parameters such as permittivity and conductivity in AC field analysis. It is considered negligible for practical purposes, once the insulation materials are properly stored, well dried and oil impregnated. Furthermore, the relative insulation permittivity varies between 2.0 and 7.0.
On the other hand, the insulation performance for DC fields is considerably affected by temperature and moisture, which shall be well controlled during manufacturing process, once permittivity and conductivity are exponentially influenced by those variables [105]- [107]. The materials breakdown characteristics under different voltage stress cases are analyzed by the following electric field equations. Considering the applied voltage V(t): Applying the divergence in both sides of (6) and introducing constitutive relations to the current and electric density vectors, the equation can be re-written as: Replacing the electric field H by minus gradient of V(t) and the gradient divergence is a Laplacian operator. The derivative in time and a constant portion established by the conductivity are defined in (8).
The DC voltage can be defined as a long-time energization in steady state [108]- [110] and a simplification is done taken 0, whereas derivatives in time are null.
For a sinusoidal excitation sin , the angular frequency times the permittivity becomes considerably large when compared to the conductivity.
, , sin 0 and . 0 From (10), it is noted that permittivity rules the potential distribution in an AC field configuration, while the insulation materials excitation voltage defines the DC field distribution.

3) DC X AC FIELD DISTRIBUTION PLOT
In this section, the authors used the same insulation structure, Fig. 18, to apply 680kVdc and 520kVac test voltages, using a 2D-FEM software. The electric field strength is represented by the shaded color and the equipotential voltage lines by the contours. The electric field distribution is influenced by the space charge established on the insulation, which defines the insulation dielectric strength in the windings, from the active part to the grounded parts, in the connections and between windings [111]. The simulation confirms the behavior mentioned previously.

4) VOLTAGE HARMONICS INFLUENCE
The technical standard [42] states that the voltage harmonics can affect the no-load losses, but it is negligible compared to total losses magnitude. Looking by an electrical field perspective, the reference [112] warns of a combination of voltage harmonics with partial discharges and other dielectrics stresses superimposed, which not only distort the valve winding voltage waveshape, but also contribute with the insulation degradation [113]- [119]. VOLUME XX, 2022 13

FIGURE 18. DC and AC field distribution inside an insulation structure
In [112], a voltage harmonic spectrum is analyzed considering 3k, 6k, 6k±1, 12k and 12k±1 harmonic orders, where k is an integer from 1…n, for the Huainan Converter Station in China which is rated in ±800kVdc and 6400MW. The THD (Total Harmonic Distortion) in percentage is given in Table 4, for the simulated and measured results, considering Y/Y and Y/D transformer winding connections.
A 3-D model was developed to analyze the electrical field inhomogeneity into the minor insulation, by a FEM simulation, especially along the valve winding electrostatic shield ring. It is enhanced by the voltage harmonics and the insulation regions closer to the winding presented higher voltage changing rate ( ⁄ ). Such rate increase can enlarge partial discharges magnitude and accelerate the insulation degradation. Furthermore, defects found into the insulating material presented higher PD levels under voltage harmonics influence. Voltage harmonics can also be a source for commutation failure and outages during bipole reversal of polarity [120], [121]. The Cigré report [122] informs that 14 out of 22 failures in HVDC transformers occur in the valve winding. State Grid and China South Power Grid have registered thirty failures in operation from 2016 until now, due to severe operational conditions [112]. Thus, voltage harmonics is an issue that may compromise the converter transformer insulation withstand along the operation, requiring a consistent design of the solid insulation between windings and toward the core yokes, combined with satisfactory quality control to avoid undesired material defects.

IV. CONCLUSION
This paper provides the reader a comprehensive overview on how HVDC system operation affects the converter transformer design, covering the technical standard IEC/IEEE 60076-57-129 highlights and including additional discussions found in an extensive search on the literature. The authors believe this article shades light and amplifies the view on the consequences of each effect and how to mitigate then through practical design solutions. The authors also included a comparison of the electrical field behavior under DC and AC stresses, via a finite element simulation, that shows the most stressed parts on the insulation where a reinforcement should be verified. Those main effects must be considered during technical specification conception and checked by a design review audit. The way the design solutions deal with those effects can be determinant to validate the most optimized proposal and to assure the transformer will perform satisfactorily. The power systems expansion and interconnection worldwide are increasing the distances, power and transported energy, reaching continental scales. Thus, the electrical, thermal, magnetic and mechanical stresses tend to increase, and the transformer design to become more complex, requiring a great technical knowledge to specify and verify those requirements.