A Multiport Power Conversion System for the More Electric Aircraft

In more electric aircraft (MEA), weight reduction and energy efficiency constitute the key figures. In addition, the safety and continuity of operation of its electrical power distribution system (EPDS) are of critical importance. These sets of desired features are in disagreement with each other because higher redundancy, needed to guarantee the safety of operation, implies additional weight. In fact, EPDS is usually divided into isolated sections that need to be sized for the worst case scenario. Several concepts of EPDS have been investigated, aiming at enabling the power exchange among separate sections, which allows better optimization for power and weight of the whole system. In this article, an approach based on the widespread use of multiport power converters for both dc/dc and dc/ac stages is proposed. System integration of these two is proposed as a multiport power conversion system (MPCS), which allows a ring power distribution, while galvanic isolation is still maintained, even in fault conditions. Thus, the redundancy of MEA is established by no significant weight increase. A machine design analysis shows how the segmented machine could offer superior performance to the traditional one with the same weight. Simulation and experimental verifications show the system feasibility in both normal and fault operations.


I. INTRODUCTION
T HE MEA framework draws a lot of attention among the power electronic experts as it simultaneously belongs to the areas of smart power distribution, microgrids, and 2332-7782 © 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
transportation electrification. In the field of power distribution and microgrids, the increasing energy demand challenged the aircraft power distribution systems with strict requirements of safety and reliability [1]. The MEA concept envisages the substitution of the oil distribution with electromechanical actuators or even fully electrical actuators, further increasing the electric power requirements. In the field of transportation electrification, the development of power generation, distribution, storage, and advanced information communication has laid a great foundation for the next generation of MEAs [2]. In the MEA, electrical generators usually employ threephase ac drives, albeit most of the loads are dc loads, or ac loads with different power/frequency requirements. Therefore, the so-called TRUs are utilized in the MEA for feeding different loads [3]. The TRUs confer galvanic isolation and convert ac to dc by using rectifiers. Besides, ac, dc, or hybrid buses are implemented in the MEA for power distribution.
To improve the weight/volume consumption and efficiency as well as increase the safety and availability of the EPDS, this article proposes MPCSs [4]- [7]. With this concept of multiport power converters, both for the dc/dc stages (i.e., MAB converter) and the dc/ac stages (segmented inverters), a ring distribution of the electrical power system featuring galvanic isolation properties among its redundant compartments is established.
In this article, a system concept of MPCS-based EPDS is proposed, by integrating multiple dc/dc converters together with segmented multiphase machine drives. The goal is to have an interconnected EPDS where the power can be exchanged across the whole system and, at the same time, guaranteeing the resiliency due to the galvanic isolation of the dc/dc converters and of the machines. For the first time, the electrical machines and the converters become part of a complex power distribution structure, which shows resiliency and fault tolerance with little additional power/weight installed.
The concept is validated by both simulations and experiments. Section II describes the possible electrical distribution system paradigms and introduces the concept of MPCSs, covering design and implementation aspects of multiport dc/dc and ac/dc converters. Section III illustrates the segmented multi-three-phase machine design guidance as the key energy conversion component for multiport ac/dc converters. Section IV validates the proposed MPCS-based EPDS by simulations during normal and faulty operation. Section V shows the experimental results obtained on a prototype system. Section VI draws the conclusions of the work, highlighting its most important achievements and contributions to the state of the art of the MEA.

II. MULTIPORT POWER CONVERSION SYSTEM AS AN
ELECTRICAL POWER DISTRIBUTION SYSTEM Different types of load are considered in the MEA: fuel and oil pumps and microturbines are ac loads, whereas actuators, battery storage systems, and avionics are dc loads [8]. The mixture of both ac and dc loads requires a complicated structure of EPDSs where the main generators and APUs supply the main dc bus. Different architectures of hybrid EPDSs have been analyzed in [9], concluding that a hybrid distribution with dc implementation and ac/dc rectifiers for the generators is the architecture with the smallest weight. Fig. 1 compares two schematics of a distribution system. In Fig. 1(a), the generators are feeding a three-phase ac bus with variable frequency. The control of the excitation winding of the generators regulates a constant voltage amplitude. Two TRUs are used to provide the dc buses with an HVdc at 270V, and a switching matrix allows for the connection of the APUs to the ac bus bars. In the case of a fault in one engine, the bus bars can also be connected together to keep the EPDS operational. Conversely, in Fig. 1(b), the generators are directly connected to ac/dc converters that provide the required power to the respective dc bus bars. The ac loads are connected to the dc distribution via suitable dc/ac inverters. The LVdc loads are connected to the main dc bus via isolated dc/dc converters. Fig. 2 depicts an example of a dc-based EPDS with multiport power converters. The dc/dc and dc/ac multiport power converters are highlighted in yellow. Compared with Fig. 1(b), a multiport dc/dc converter (i.e., MAB converter) is replacing the two isolated dc/dc converters, and segmented starter/generators with segmented inverter configurations are placed instead of traditional generators with single inverters.
To implement the MPCSs for the MEA, the configurations of both dc/dc MAB converters and the dc/ac segmented machines and inverters for starter/generators are described, followed by fault tolerant and communication considerations.   high-frequency rectifiers, and a multiwinding transformer is involved. The number of windings of the multiwinding transformer is determined by the requirements of sources and loads. The MAB provides a multiport power-flow path for all the connected units. All windings of the transformer are linked by the same flux; therefore, the power can be transferred naturally by the energy delivery within different windings through the common transformer flux. The major advantages of MAB are as follows.

A. DC/DC Multiple Active Bridge Converters
1) The efficiency and power density of MAB are higher than traditional TRUs according to its high-frequency transformers [14]. 2) The voltage and power balancings are realized by flux linkage so that a load of different windings of the transformer can be totally different from each other, giving flexibility and redundancy properties to the design [12].
3) The power density of MAB is higher than separate dc/dc converters: multiwinding structure could decrease the size and weight of insulation by sharing the transformer flux linkage [15]. 4) The efficiency of MAB is higher than the separated configuration because the sharing of the magnetizing current helps to reduce the related copper loss for every single winding and the related inverter. Also, the power flow only needs to go through three stages of power conversion (i.e., dc/ac, transformer, and ac/dc) instead of six stages (i.e., dc/ac, transformer, ac/dc, dc/ac, transformer, and ac/dc [16], [17]).

B. DC/AC Segmented Inverters for Multiphase Starter/Generators
Historically, multiphase machines and drives became an attractive industrial solution in the last decades for the advantages that they can provide in terms of performance, fault tolerance, and control for high-power applications, such as electric ship propulsion [18]- [22], wind farms [23], turbo compressors [24], and aircraft generating systems [25]- [27]. Nowadays, many efforts are being made to discover and fully exploit all the possibilities that this technology can provide.

C. DC/AC Segmented Inverters for Multiphase Starter/Generators
Historically, multiphase machines and drives became an attractive industrial solution in the last decades for the advantages that they can provide in terms of performance, fault tolerance, and control for high-power applications, such as electric ship propulsion [18]- [22], wind farms [23], turbo compressors [24], and aircraft generating systems [25]- [27]. Nowadays, many efforts are being made to discover and fully exploit all the possibilities that this technology can provide.
From the basic principle, a multiphase machine differs from a standard three-phase one for having a number of phases (m) higher than 3. This is possible and simple when the machine is fed by a power electronic converter with an arbitrary number of converter legs. Therefore, the number of machine phases results in a degree of freedom for the overall drive design.
The converter layout can be defined by choosing the number of independent subconverters, how many phases are fed by each of them, and where the legs are connected to. An interesting solution for industrial applications is the one that exploits a set of three-phase converters to design the multiphase system, leading to the so-called multi-three-phase drive. This concept is well presented in Fig. 4, where the same machine is fed by a 12-phase converter in Fig. 4(a) and a four three-phase one in Fig. 4(b). The neutral points are isolated so that in the multi-three-phase layout, faults can be isolated. Also, multileg layouts or multi-H-bridges can be adopted. However, the clear advantage of a multi-three-phase layout is that it allows the use of well-known converter technology.
Not only the power electronics define the capabilities of a multiphase drive but also the winding topology. Conventional three-phase windings are mainly categorized as follows: concentrated or distributed, full or short-pitched, star or delta connected, and single or double layers. However, a traditional three-phase machine always has three balanced phases (with the same coil geometries), 120 electrical degrees shifted (symmetrical winding), with each phase mirrored under all the pole pairs (the related coils are then just series or parallel connected). These options are still available in a multiphase drive, but there are some new degrees of freedom. For example, an m-phase winding can still be based on equally distributed phases (shifted by 360/m electrical degrees) or can be distributed with a different angle (asymmetrical winding) [28], [29]. In particular, considering a multi-three-phase layout, if all three-phase subsystems (balanced and symmetrical) are placed in the same electrical position, it is said that the winding is without phase shift. Instead, if they are shifted by a certain angle, the new winding is called phase-shifted [27]. Furthermore, the phases can be all identically placed under each machine stator pole pair or can be located in the same electrical positions but under different stator areas (sectors). This last category of multiphase windings is named multisectored [30]. Finally, an interesting design of multiphase winding is the one based on a multiwinding layout. In this particular solution, more windings (generally two) are differently wound around the stator in order to control more harmonics of the air gap flux density. However, the advantages of this solution have been mainly exploited for not standard solutions, such as bearingless applications [31]- [33].
Considering the EPDS of Fig. 2, by replacing the main generators with segmented generators the total capacity, weight and volume of the generators, as well as the associated inverters, are not changed. In addition, performance, reliability, and redundancy of the system are enhanced.

D. Fault-Tolerant Strategies for the Combined Structure
A ring distribution of electric power is established by the combination of dc/dc MAB converters and dc/ac segmented inverters for starter/generators, where different power flow loops can be created to operate the system under different fault conditions. In particular, the main working operations of the system are summarized as follows.
1) The segmented generator and inverters may provide full power to the dc bus in normal conditions, as shown in Fig. 5(a).
2) The configuration may provide 50%-100% of full power if any of the main engines is broken, as shown in Fig. 5(b). 3) The configuration may provide 75%-100% of full power if any of the segmented inverters is broken, as shown in Fig. 5(c). 4) The combination may provide 25%-50% of full power in the condition of any main engine broken together with another segmented inverter broken, as shown in Fig. 5(d). Consequently, in a similar way to MAB transformers, the use of multi-three-phase segmented machine configurations not only provides enhanced performance and additional redundancy to the distribution system but also ensures isolation between the three-phase subsystems.

E. Communication Considerations of the Units
For the MEA power distribution systems, optical fibers are used as an information path. Optical fibers are point-to-point transmission media. A star-connected information network is one solution to communication and synchronization. The advantage of a star-connected network is that the time delay of synchronization is very small, and the module redundancy is easily implemented by disabling the gate signal of the H-bridge and bypassing the target module.
Although the star-connected information network is well performed, the number of optical fibers is large, and it is unfavorable for system modularity. A ring-connected information network is appropriate for the configuration shown in Fig. 2. The number of optical fibers is smaller than in star-connected networks, but the time delay of synchronization must be calculated during system operation. Module redundancy could be obtained by means of dual-ring-connected networks (both clockwise and anticlockwise).

III. EXAMPLE OF SEGMENTED MACHINE DESIGN FOR THE MPCS
This section presents a case study of machine analysis, aiming at highlighting the enhancement in the performance that can be reached transitioning from a single-three-phase system to multiple three-phase ones. The electrical machine adopted in this system is an eight-pole IPMs machine with a V-shaped layout, as shown in Fig. 6. The main advantages of this machine are a high torque and an extended power range capability over the base speed (namely, flux weakening region). A summary of the machine parameters is given in Table I. The following analysis is mainly focused on the back EMF, torque, and efficiency performance, considering two different winding layouts: 1) conventional three-phase winding and 2) double-three-phase winding (2 × 3-phase).

A. Torque Analysis
Multiphase design can give significant benefits in terms of ripple and average torque. Indeed, the control strategy of a multiphase machine can be extended to the control of some high-order space harmonics of the air-gap magnetic field [34]. In addition, the winding factor is higher compared with a three-phase winding [35], leading to a higher average torque. In this section, both torque and torque ripple have been carried out by means of FE simulations, maintaining the same machines' overall volume and supply conditions.
In Fig. 7, the torque ripple resulting from a different winding arrangement is shown: for three-phase and 2×3-phase distribution, respectively. The results highlight that the 2×3-phase machine features an average torque of 81.1 N·m with respect to 78.4 N· m in the three-phase machine, with an overall increase in +3.40%. The multi-three-phase layout also helps in reducing the torque ripple. In the three-phase case, the peak-to-peak torque ripple is 25.58%, whereas it is 13.65% for the 2×3-phase motor, leading to an overall reduction of the oscillations by 11.93%, with respect to the single-three-phase winding machine.

B. Losses Analysis
In the literature, many researchers have been working on the estimation of iron losses with analytical and FE models. In [36], an accurate prediction of iron losses for various frequencies and magnetic flux densities, considering the influence of higher order time harmonics and minor loops, is presented. In [37], an improved model of the iron losses for an induction machine based on a comparison of different equations has been developed. The influence of the PWM generated by inverter on machine iron losses is also described in [38].
In this article, in order to have a preliminary indication of the iron losses in the machine, the classical Steinmetz method is considered with the hypothesis of sinusoidal flux in the iron. This is used as a quantitative method for a comparative estimation of the iron losses expressed with the following equation: where B peak is the maximum value of flux density in the iron, k hy and k ed are the hysteresis and eddy current coefficients, respectively, f is the frequency, and α is the Steinmetz coefficient. However, considering the unique value of B peak in the stator core does not permit to consider the asymmetrical flux density behavior. In fact, for a three-phase machine presenting a number of slots/poles per phase q > 1, the flux density is not symmetrical due to high harmonic order fields generated by currents [39]. In this case, another advantage introduced by the 2 × 3-phase winding is that the number of slots/poles per phase is q = 1 so that the flux density results symmetrical in all stator core parts. The approach proposed takes into account these asymmetries allowing for a better comparison between the two considered winding layouts. The iron losses in the two consecutive teeth and on two consecutive yokes at the top of each area are calculated as follows: where P spec,iron takes into account the specific iron losses in the core, while k t , k y , B t , B y , w t , and w y are the correction coefficients, peak flux density values, weights related to the teeth, and the volumes of back iron around the slots, respectively. Table II lists the parameters used for the FE simulations. Finally, it is possible to obtain the total iron losses in the stator by applying the following equation (where Q s is the total number of teeth of the machine): Table III summarizes the results obtained from the comparison of the two different winding arrangements. It is worth to highlight that, albeit the local saturation in the teeth of

C. Efficiency
The mechanical and rotor losses are neglected for simplicity in this qualitative evaluation exercise, as their contribution will only add an equal offset in the quantities of both analyzed machines. The Joule losses have been calculated implementing the classical equation, the function of the resistance R, and phase current I rms squared. In order to take into account the additional losses, such as eddy currents due to flux leakage in the machine housing and other parasitic losses, an additional loss P add equivalent to 10% of the calculated Joule losses P j is introduced. The efficiency is calculated as where T avg is the average torque and b is the rotor mechanical pulsation corresponding to the machine base speed.

D. Results Comparison
In this section, a summary of the compared results between three-phase and 2 × 3-phase systems is given, as reported in Table IV. It is worth to highlight that for the same current loading in the machine slots, the mechanical output power is higher with a multiphase configuration (+3.4%), consequently showing an improved torque capability. If the application is demanding a minimization of the volume, the machine can produce the same torque with the same current load and reduced volume. In fact, being the torque proportional to the stack length of the machine, the active stack length for the multiphase machine can be reduced from l 3 p = 110.0 mm (the active stack length of the three-phase machine) to l 2×3 p = 106.3 mm. This can lead to a weight reduction from 32.3 to 31.2 kg if a multiphase winding is adopted. Therefore, for the same phase current and output torque, the multiphase machine will result in 1.10 kg lighter (about 3.4%). Furthermore, also, the efficiency results to be slightly increased (about 0.10%) compared with the three-phase solution.

IV. SIMULATIONS OF DIFFERENT FAULT SCENARIOS
The simulation model is developed based on Fig. 5 in MATLAB/Simulink and PLECS Blockset with one segmented electrical machine and one QAB converter. The segmented machine is split into two three-phase systems. Two of the dc ports of the QAB are connected to the segmented machine via two separate three-phase half-bridge inverters, and the other two dc ports are connected to either dc loads or dc power sources. For the analysis, voltage V dc1 and V dc2 will be the Port 1 & Port 2 dc buses, whereas V dc3 and V dc4 will represent the Port 3 & Port 4 dc buses of the QAB converter. Table V summarizes the main simulation parameters.
Considering that both the QAB and the segmented drives feature fully bidirectional power electronics, connected to the same buses, different control strategies can be implemented. When multiple power converters have the capability of  controlling the voltage, decentralized control based on droop control has been widely adopted in the area of dc microgrid and motor control [40]. This allows for the independent design of the control loops but may reduce the dynamic performance.
A cascaded current and voltage control is, therefore, employed in the segmented drives as in a traditional rectifier control for machine. Each voltage control generates the current reference i * q for the respective part of the machine. Because there is more than one regulator that has the dc bus voltage V dc as the control target, virtual resistors R v1 and R v2 are used for the grid stabilization, as shown in Fig. 8(a).
The QAB is controlled with a phase shift-control that has the LVdc voltage as the control target. Virtual resistor (R v1 and R v2 ) control is used for the equalization of the power drawn from the HVdc buses, as shown in Fig. 8(b). A decoupling control as the one described in [12] can be implemented to reduce the influence of unbalanced power transfer among the ports on the dynamic performance. Having the balancing control also in the HV side of the QAB control allows for the voltage control of the HVdc bus even in the case of a double fault of the segments connected to the same bus. For the QAB parameters, the switching frequency is 10 kHz, and the dc-link voltage is 50 V. As shown in Fig. 8(b), the parameters of the balancing controller are K p = 0.005 and K i = 1, whereas, in the dc voltage controller, they are K p = 0.01 and K i = 1, and the virtual resistance is set to 1 . For the decoupling strategy, d 31 = output of PI voltage controller 1 + 0.5 d 21 ; d 41 = output of PI voltage controller 2 + 0.5 d 21 .
This control structure has the advantage that there is complete redundancy in the control system in the case of a segmented machine failure. The independent voltage control  allows for the automatic choice of the torque of each segment to balance the HVdc buses. In this way, the multi-threephase machine operates as a power router, exploiting the galvanic isolation of the windings to transfer power between the HVdc buses. The summary of operating conditions (generating or motoring) together with different fault conditions (IPM machine system cut off or QAB port cut off) is shown in Table VI. For the generating mode, the double-three-phase IPM machines are connected to Port 1 and Port 2 of QAB through voltage-source inverters. The speed of the simulated machine is constant (300 rpm). The Port 1 and Port 2 voltages are controlled by the machine controller via inverters separately. The machine segments are with double closed-loop control, for which the outer loop is with a reference of dc voltage and the inner loop with a reference of d − q current. The QAB is controlled following the diagram in Fig. 8(b) so that Port 1 and Port 2 power are balanced. Port 3 and Port 4 are connected to two resistive loads with the resistance of 25 and 25 , respectively. For the motoring mode, the doublethree-phase IPM machine is connected to Port 3 and Port 4 of QAB through inverters. The Port 1 and Port 2 voltages are connected to the same dc voltage source, controlled at 50 V, by separated inverters. At the same time, the machine is torque controlled, by setting the d − q currents to i d = 0 and i q = 3.3 A, respectively. The QAB is controlled using the diagram in Fig. 8(b) so that Port 1 and Port 2 powers are kept balanced. Fig. 9 illustrates the simulation results during normal operation. The two-segmented inverters provide half of the load power each. Fig. 10 shows the simulation results of one segmented inverter open-circuit fault (at 0.4 s) under the motoring mode and the generating mode, respectively. It is obvious that the dc voltage of the four ports of the QAB is remaining the same (i.e., about 50 V), while, for the motoring mode, one of the machine segments is cut off, and the power of another machine segment is doubled. Therefore, the dc supply power of Ports 1 & 2 keeps the same. The fault of one machine segment does not affect the power supply. For the generating mode, one of the machine segments is cut off, and the power of another machine segment is doubled. Therefore, the dc supply powers of Ports 1 and 2 are unchanged. The fault of one machine segment does not affect the dc loads. Fig. 11 shows the simulation results for disconnection of a QAB port (at 0.4 s). It can be observed that the dc voltages of the four ports of the QAB remain the same (i.e., about 50 V), whereas, for the motoring mode, one port of the QAB is disconnected, and the power of the other QAB port is doubled. Therefore, the power of the machine segments is kept constant. The fault of one QAB port does not affect the machine side. For the generating mode, one of the QAB port is cut off, and the power of the other QAB port is doubled. Therefore, the power of the machine segments is not affected by the fault of one QAB port.

V. EXPERIMENTAL VERIFICATIONS
The experimental verification of the proposed MPCS was implemented by connecting a QAB converter to the segmented multiphase IPM machine, as shown in Fig. 2. The experimental platform is shown in Fig. 12, with the parameters in Table VII.  The QAB is connected to the multiphase IPM, where the 2 × 3-phase machine and inverters work as dc loads with two isolated dc ports (i.e., exactly the same as the motoring mode of the simulation). The two ports of QAB on the secondary side (Ports 3 & 4) are connected to the machine, and the other two ports (Ports 1 & 2) of the QAB are connected to the dc power supply. The experimental results are aiming to demonstrate the power-sharing capabilities offered by the multiport converters. In fact, if the power electronics allows for unbalanced power processing, the scenarios presented in the simulation become possible. Table VIII shows the experimental parameters for the test.
When it comes to the experimental configuration, as for the power supply side, two ports of the QAB converter are connected to two dc power supplies. As for the load side, the other two ports are independently connected to two Waveforms with all ports of QAB converter connected (normal condition). inverters to supply the dual-three-phase machine. The control scheme of the experiment is the same as the one implemented in simulation for the motoring mode. The experiment can be divided into three parts: 1) all ports of the QAB converter are connected, and the whole system is running with balance loads; 2) a supply port of the QAB converter is disconnected from dc power supply, and the whole system is running with a single dc source; and 3) a load port of the QAB converter is disconnected from the inverter, and hence, only one inverter is powered, and the machine is running in single-three-phase mode.

A. Experimental Verification for Balanced Loads and Power Sources
To apply the decoupling control strategy in Fig. 8(b), the virtual resistance and the filter coefficients of the ADC into the DSP are changed. Fig. 13 shows the waveforms of the main quantities (currents and voltages) of the QAB converter under normal motoring operation of the electrical machine. Channel 1 and Channel 2 are the phase current of two segmented IPM machine systems, Channel 3 and Channel 4 are the currents of the two input ports from the power supply, and Channel 5 and Channel 6 are the output voltage of QAB at the terminals of the two inverters. It can be seen that the currents of the two input ports are balanced. Fig. 14 shows the QAB ac voltages and currents under normal conditions. The phases of the square wave voltages are the same; therefore, the ac currents of the four ports are similar, which also indicates the same power flow of the four ports.

B. Experimental Verification for Balanced Loads but Single
Power Source Fig. 15 shows the current and voltage waveforms when a source port of the QAB converter is disconnected and the power of the source at the other port is doubled. It can be observed that at the midpoint of the timeline, the input current in Channel 3 goes up, and the input current in Channel 4 falls down to make the input power constant. Meanwhile, the phase currents of the machine and the output dc voltage of the QAB converter do not change, indicating that the machine and the QAB are working well in the dual-three-phase-mode operation. Fig. 16 shows the QAB ac voltages and currents under fault condition (one dc power source cut off). The phase of the square wave voltages is becoming different; therefore, the ac currents of Port 1 is doubled, Port 2 is almost zero, and the currents of Port 3 and 4 are not changed.

C. Experimental Verification for Balanced Power Sources but Single Load
When a load port of the QAB converter is disconnected from the inverter, the key waveforms are shown in Fig. 17. It can be observed that the phase currents of one inverter diminish, while the phase currents of the other inverter increase, indicating that the operating mode of the machine is transferring from the dual-three-phase mode to the single-three-phase mode. The voltages of source ports of the QAB converter keep the same. Furthermore, it can be seen from Channel 3 and Channel 4 that the input current of load ports has a drop at the midpoint of the timeline and then increases to the new steady state, which can be attributed to the fact that the equilibrium point is changed between the two dc loads under the voltage droop control. Fig. 18 shows the QAB ac voltages and currents under fault condition (one inverter load cut off). The phases of the square wave voltages are becoming different; therefore, the ac current of Port 4 is doubled and at Port 3 is almost zero, and the currents of Ports 1 and 2 do not change.
The results are showing the effectiveness of the proposed MPCS, which has been verified experimentally, as shown in Figs. 13-18. The fault-tolerant operation is obtained, as proposed in Fig. 5. With the normal operating condition, the power of the two ports of the loads (inverters) is balanced, Fig. 16. Key waveforms of the QAB converter with one source port of QAB converter disconnected. Fig. 17. Waveforms with one source port of QAB converter disconnected (one sector of the machine is missing). Fig. 18. Key waveforms of the QAB converter with one load port of QAB converter disconnected. and the power of the two ports from the dc power sources is balanced, respectively. Thus, the power flow of each port is similar. With one source port of the QAB converter disconnected (as an emulation of the source fault), the power of the two systems of the segmented multiphase machine is still balanced (no transient state of power control). With one load port of the QAB converter disconnected (as an emulation of the machine fault), the power of the two sources is always balanced. After a transient, the healthy sector of the machine is carrying twice the power and the dc loads return to their normal operation. The time constant of the control is almost the same in simulations and experimental results, validating the control algorithm and the feasibility of the proposed system architecture.

VI. CONCLUSION
In this article, the complete integration of multiport converters for the EPDS of the MEA is carried out. This article provides a novel concept to the EPDS on the MEA, integrating segmented machines and a multiport MAB dc/dc converter for the bidirectional dc/ac conversion for starter/generators. The proposed MPCS for the MEA allows for the realization of a ring electrical power distribution, where the galvanic separation is guaranteed by the windings of the transformers in the MAB converters and the multi-three-phase windings of the machines. This configuration shows its resiliency to faults in multiple scenarios, and this is demonstrated by simulation results realized on a simplified grid. The experimental verification shows evidence of the effectiveness of the MPCS. The results are focused on the unbalanced operation of a multithree-phase machine and MAB converters, demonstrating how this technology can enable the flexible operation of the EPDS in the MEA.