Decentralized Power Management for Electrical Power ... - MDPI

electronics Article

Decentralized Power Management for Electrical Power Systems in More Electric Aircrafts Myungchin Kim 1 , Sung Gu Lee 2 and Sungwoo Bae 3, * 1 2 3

*

School of Electrical Engineering, Chungbuk National University, Cheongju 28644, Korea; [email protected] Electronic Robot Engineering, Busan University of Foreign Studies, Busan 46234, Korea; [email protected] Department of Electrical Engineering, Hanyang University, Seoul 04763, Korea Correspondence: [email protected]; Tel.: +82-2-2220-2309

Received: 17 July 2018; Accepted: 7 September 2018; Published: 10 September 2018

Abstract: In order to implement reliable and flexible power management among energy sources, a decentralized power management approach for electrical power systems (EPSs) in the more electric aircraft (MEA) is studied. Considering the increased use of electrical power for various functions, the performance of MEA would be determined by the design and operation of the EPS. By using a virtual impedance that includes both a resistive term and an inductive term, autonomous power sharing is realized. Because of the frequency dependence in the virtual impedance, different power sharing ratios between steady state and transient state can be considered. Not only the operation of various power sources is coordinated without supervision of a centralized controller, but also the operation profile of each source can be adjusted to meet output characteristics of each source. To demonstrate the effectiveness of the proposed approach, a series of simulations that consider various virtual impedance configurations were conducted. The proposed approach contributes to a higher level of operational flexibility, while enabling reliable and cost-effective management of MEAs. Keywords: DC micro-grid; virtual impedance; virtual inductance; more electric aircraft

1. Introduction This paper studies a decentralized approach for electrical power systems (EPS) management in more electric aircraft (MEA) platforms. Similar to vehicular platforms for ground [1] and marine [2] transportation, research on highly electrified aircrafts has received continuous interest. As a result of aircraft electrification, the roles of EPSs have broadened from relatively simple functions to essential functions such as propulsion and flight control [3]. For instance, electrical motors are used for actuation of flight control surfaces [4,5] and aircrafts that use a pure electric or a hybrid electric propulsion architecture have been developed [6,7]. Major motivations for aircraft electrification include social demand toward sustainable transportation and improved flight performance. While combustion engines have been used as the primary source for propulsion in conventional aircrafts [8], platforms that utilize renewable sources to generate propulsion force have been developed [6,9]. By replacing massive hydraulic/mechanical actuation systems with electrical actuators, it is possible to improve performance by weight reduction and efficiency improvement [10]. Thanks to the recent development of electrical motors and power electronic interfaces with high power density, platform electrification is being considered not only for small sized quad-copters (drones) but also for manned aircrafts. In Figure 1, examples of electrical loads used in conventional aircrafts and the MEA are illustrated [3,4,8]. Although the use of electrical interfaces seems to introduce potential advantages compared to conventional aircrafts, electrified aircrafts should also satisfy safety and reliability regulations as in current platforms.

Electronics 2018, 7, 187; doi:10.3390/electronics7090187

www.mdpi.com/journal/electronics

Electronics 2018, 7, x FOR PEER REVIEW

2 of 19

interfaces seems Electronics 2018, 7, 187 to

introduce potential advantages compared to conventional aircrafts, electrified 2 of 19 aircrafts should also satisfy safety and reliability regulations as in current platforms.

More Electric

Conventionial · · · ·

Lighting : Internal/External Heating & Cooling In-flight Entertainment Avionics Equipment : Communication : Navigation

·

+

· · ·

Electric Actuation : Control Surface Electric Taxi Electric Propulsion Integrated Control Systems (Suite) : Flight Control : Mission Operation

Electrification of Aircraft Platforms · Wider Range of Functions · Higher Criticality of Functions more electric electric aircraft aircraft (MEA). (MEA). Figure 1. Increased use of electrical power in a more

Figure Figure 22 shows shows aa generalized generalized architecture architecture of of an an EPS EPS in MEA applications. As shown in Figure 2, the EPS consists of various type of sources, loads and energy storage. Considering the growing interest the EPS consists of various type of sources, loads and energy storage. Considering the growing for the MEA, research both the and itself the EPS components is required.isWhile 115 V/400 Hz interest for the MEA, on research onEPS bothitself the EPS and the EPS components required. While 115 ac and 28 V dc system buses have been considered for electrical distribution in conventional aircrafts [8], V/400 Hz ac and 28 V dc system buses have been considered for electrical distribution in conventional 270 V dc-based havesystems been developed in recent aircrafts [8,9]. aircrafts In addition to the voltage levels, aircrafts [8], 270systems V dc-based have been developed in recent [8,9]. In addition to the research on EPSresearch architecture has also been performed. of architectures include ac systems voltage levels, on EPS architecture has alsoExamples been performed. Examples of architectures with constant [8] or with variable frequency dc systems [9] and ac/dc[9] [11] systems. the include ac systems constant [8] or[10], variable frequency [10],hybrid dc systems and hybrid While ac/dc [11] EPS architecture should be selected based comprehensive review of various factors, been systems. While the EPS architecture shouldonbea selected based on a comprehensive reviewitofhas various reported that selecting 270 V dc for power distribution seems to be a promising option from the factors, it has been reported that selecting 270 V dc for power distribution seems to be a promising perspective weight and stability [9]. In such an EPS, transferred to the using dc option fromofthe perspective of weight and stability [9].the In power such anis EPS, the power is load transferred to distribution, while interface requirements of 28 Vrequirements dc or ac loadsof are28met by or using powerare electronic the load using dc the distribution, while the interface V dc ac loads met by interfaces. In order to design an EPS that meets to various system levelthat requirements (e.g., power quality, using power electronic interfaces. In order design an EPS meets various system level efficiency, and(e.g., weight), optimization studies and on the overall EPS architecture [9], requirements power quality, efficiency, weight), optimization studies onsubsystems the overall and EPS components [12]subsystems have also been architecture [9], andperformed. components [12] have also been performed. Increased Increased use use of of power power electronic electronic interfaces interfaces and and various various power power sources sources in in aircraft aircraft platforms platforms has has raised raised needs needs for for research researchon oncontrol controlof ofdc dcand andac/dc ac/dc hybrid power networks. While power electronic electronic interfaces interfaces enable flexible design and operation of power systems, the performance (e.g., stability, dynamic dynamic response and effective effective protection) of the the overall overall system system is highly highly affected affected by by the the system system control control approach approach [13–15]. [13–15]. In order to perform perform power sharing in a decentralized manner, manner, itit is is general general to droop-based approaches approaches[15]. [15].InIncase case droop is implemented a virtual resistance, to apply droop-based thethe droop is implemented by by a virtual resistance, the the resistance value determines the power sharing ratio among a resistance value determines the power sharing ratio among powerpower sourcessources [14,15].[14,15]. ThroughThrough a stability stability it has been reported that the stability characteristics of droop-controlled dc microgrids analysis,analysis, it has been reported that the stability characteristics of droop-controlled dc microgrids are are highly affected system parameters, such thecable cableresistance resistanceand andload load type type [16]. Although the highly affected by by system parameters, such asasthe virtual virtual resistance resistance enables enables autonomous autonomous control, control, the the desired desired power power sharing sharing ratio ratio can can be be achieved only at at steady steady state. state. Instead, droop-based approaches that that manage manage power power sharing sharing during during transients transients have have been pass filters and high pass filters cancan be been studied. studied. For Forexample, example,Reference Reference[17] [17]introduces introduceshow howlow low pass filters and high pass filters incorporated with thethe virtual resistance toto design be incorporated with virtual resistance designthe thepower powersharing sharingperformance performanceduring duringtransient transient periods. periods. In [18,19], aa virtual virtual capacitance capacitance is is used used to to manage manage the the power power sharing sharing performance performance during during transients. transients. It was demonstrated that the power output of sources during during transients transients is affected affected by by the the filter filter characteristics characteristics or or the the type type of of virtual impedance impedance (e.g., (e.g., resistive resistive or capacitive). capacitive). Control approaches approaches that modular operation [13][13] or minimized powerpower oscillations [20] in hybrid systems that contribute contributetoto modular operation or minimized oscillations [20] inac/dc hybrid ac/dc have also been studied. While the approach of [13] performs system control without requiring full

Electronics 2018, 7, x FOR PEER REVIEW Electronics 2018, 7, 187

3 of 19 3 of 19

systems have also been studied. While the approach of [13] performs system control without requiring fullofinformation of athe system observer, by a nonlinear observer, the study the of [20] improves the information the system by nonlinear the study of [20] improves transient response transient response bybased a neural network based damping controller. characteristics by acharacteristics neural network damping controller. HV Bus HV Load Source

LV

New Source

Battery

LV Load Energy Storage HV Load

LV

LV Load

Source : Power Conversion Interface

: Disconnect

HV: High Voltage LV: Low Voltage Figure 2. 2. Configuration Configuration of of aa generalized generalized EPS EPS in in an an MEA. MEA. Figure

With the increased use of renewable power sources and electrical loads, power management With the increased use of renewable power sources and electrical loads, power management of of EPS has become critical to ensure reliable power supply for loads in MEAs. Similar to stationary EPS has become critical to ensure reliable power supply for loads in MEAs. Similar to stationary microgrids that consist of various renewable sources [21], energy management systems for EPS in microgrids that consist of various renewable sources [21], energy management systems for EPS in airborne platforms have also been studied [22,23]. In order to maintain instantaneous power balance airborne platforms have also been studied [22,23]. In order to maintain instantaneous power balance between sources and loads, not only the output of each source should be effectively controlled [21] but between sources and loads, not only the output of each source should be effectively controlled [21] coordination among power sources is also required [7]. For example, an EPS that consists of proton but coordination among power sources is also required [7]. For example, an EPS that consists of exchange membrane fuel cell (PEMFC), photovoltaics (PV) and battery have been considered for an proton exchange membrane fuel cell (PEMFC), photovoltaics (PV) and battery have been considered unmanned aircraft [24]. As the power demand varies depending on the flight maneuver pattern, for an unmanned aircraft [24]. As the power demand varies depending on the flight maneuver the output of the PEMFC and PV cells are adjusted according to the instantaneous load demand such pattern, the output of the PEMFC and PV cells are adjusted according to the instantaneous load that various flight patterns could be performed and the battery state-of-charge (SOC) level could be demand such that various flight patterns could be performed and the battery state-of-charge (SOC) maintained above a pre-defined level throughout the flight. It is worth noting that energy storage level could be maintained above a pre-defined level throughout the flight. It is worth noting that (e.g., battery) is a critical component in aircrafts as it operates as a back-up source when primary energy storage (e.g., battery) is a critical component in aircrafts as it operates as a back-up source power sources are not operational. When the main propulsion has failed, the battery SOC level would when primary power sources are not operational. When the main propulsion has failed, the battery determine the available flight period for an aircraft to make necessary maneuvers for landing or SOC level would determine the available flight period for an aircraft to make necessary maneuvers recovery in emergency situations. Even during normal flight conditions, the battery can be discharged for landing or recovery in emergency situations. Even during normal flight conditions, the battery to handle rapid increase of power demand that can occur during dynamic maneuvers. can be discharged to handle rapid increase of power demand that can occur during dynamic As the operation mode of each load and the connection status of each disconnect of Figure 2 maneuvers. could be different depending on each mission phase of the overall flight, the system control should As the operation mode of each load and the connection status of each disconnect of Figure 2 effectively respond to configuration changes. In particular, the EPS should be able to handle various could be different depending on each mission phase of the overall flight, the system control should abnormal conditions that could be caused by faults or load conditions. In case a fault occurs, the power effectively respond to configuration changes. In particular, the EPS should be able to handle various flow of the EPS should be reconfigured so that the fault is locally isolated, and the rest of the EPS is not abnormal conditions that could be caused by faults or load conditions. In case a fault occurs, the affected. As the system response differs for each fault case, a comprehensive fault analysis should be power flow of the EPS should be reconfigured so that the fault is locally isolated, and the rest of the conducted. Fault analysis approaches that provide insights on the system behaviors to unsymmetrical EPS is not affected. As the system response differs for each fault case, a comprehensive fault analysis faults have been proposed and the response characteristics for various scenarios have been studied should be conducted. Fault analysis approaches that provide insights on the system behaviors to using fault models of various distributed energy sources [25,26]. Considering the increased connection unsymmetrical faults have been proposed and the response characteristics for various scenarios have of loads in MEA, the effect of constant power loads to dc systems [27] and unbalanced loads to ac been studied using fault models of various distributed energy sources [25,26]. Considering the microgrids [28] have been studied.

Electronics 2018, 7, 187

4 of 19

Meanwhile, researches on design approaches for improved performance of power electronic interfaces in MEA applications have been performed. While researches on control approaches [29] and fault tolerant drive systems [30] have been conducted, design for highly reliable operation [31] and increased power density [32] have received attention for aerospace applications. In order to address needs for weight reduction and meeting requirements of design standards, in particular, studies on performance analysis [33] and optimization [34] of electromagnetic interference (EMI) filters have been actively performed. The main direction for power converters is to achieve a high-power density so that the overall aircraft weight could be reduced [4,35]. In this manuscript, an approach that realizes decentralized power management of EPS in the MEA is introduced. Instead of receiving power commands from a central control unit [36], the command for each source is generated based on the data that are measured from the terminal. Despite the growing interest for MEA, there seems to be limited research on decentralized control approaches of EPS in aircrafts [22,23]. In fact, EPS in most conventional aircrafts has been designed such that each distribution bus is connected to a single generator instead of connecting multiple generators in parallel [37]. With the increased use of multiple sources in MEA, however, it is necessary to study approaches that enable decentralized modular control without violating conventional reliability requirements of the aircraft. Moreover, this paper demonstrates how the virtual impedance could be effectively used to achieve a desired power sharing ratio. The power sharing ratios at steady state and transient state are adjusted separately by designing the virtual resistance and the virtual inductance. Hence, the power flow can be controlled in a more flexible manner. Detailed explanation on the contributions of the presented work are introduced in the following section. The rest of the paper is as follows: Section 2 highlights the detailed contributions of the manuscript, while Section 3 introduces the decentralized power control approach for MEAs. Section 4 demonstrates the performance of the considered control approach and the final section concludes the paper. 2. Contributions of Present Work Detailed explanation on how the current research contributes to effective decentralized control of EPS in MEA platforms are as follows. First, the proposed approach enables autonomous control of EPS in MEA platforms. The application of decentralized control could achieve reliable operation compared to centralized control architectures [38,39]. Because of the decentralized nature, for example, system failures caused by faults in the communication interface could be prevented. Hence, the EPS can be controlled in a fault tolerant manner [38]. Considering the catastrophic consequences of system failure, the importance of system reliability should not be underestimated for aircrafts. Especially, EPS reliability would be a critical factor that determines the overall platform safety because of the increased use of electrical power in highly electrified aircrafts. Although system reliability depends not only on the component reliability but also the system control architecture, prior researches on EPSs of MEAs have been limited to topics such as electrical actuator performance [4] and stability analysis [11]. While approaches for improved reliability have been studied [31,40], such studies have focused on the power interface and not on the overall EPS. Although control laws for MEA EPSs have been discussed, the work of this paper differs by considering a decentralized architecture [22] or considering performance during transient periods [41]. Second, this paper demonstrates that the power sharing ratio among various sources can be controlled to be different between steady state and transient periods. Such increased flexibility in power flow control would contribute to effective operation of MEA so that each power source could be operated according to their inherent output characteristics (i.e., energy density and power density). Figure 3 compares the typical values of energy density and power density for common power sources [42,43]. For example, the output of sources with high power density could be operated to supply high peak power during short transients and limited power at steady state. In case of sources that have high energy densities, a reversed operation profile could be considered.


5 of 19


5 of 19

While the supercapacitor has been used asprofile a representative energy storage thatsupercapacitor handles transient high energy densities, a reversed operation could be considered. While the has power imbalances, the fuel cell has been considered in aircraft applications as a source with been used as a representative energy storage that handles transient power imbalances, therelatively fuel cell large energy densityin [7]. Whileapplications the application PEMFC solid oxide cellsdensity (SOFC)[7]. hasWhile been has been considered aircraft as a of source withand relatively largefuel energy reported for aircrafts [7], PEMFC have been mostly used thanks to its superior properties in power the application of PEMFC and solid oxide fuel cells (SOFC) has been reported for aircrafts [7], PEMFC density, operation temperature efficiency [7,44]. As source could be controlled in a more have been mostly used thanks to and its superior properties in each power density, operation temperature and flexible manner, the output of each source could be coordinated to achieve optimized operation of efficiency [7,44]. As each source could be controlled in a more flexible manner, the output of each the overall EPS. While most previous studies on decentralized power control (i.e., droop control) source could be coordinated to achieve optimized operation of the overall EPS. While most previous have focused on improving the control power sharing performance at steady state of studies on decentralized power (i.e., droop control) have focused on[38,45,46], improvingthe thefocus power this paper is on using the virtual impedance to enable flexible power sharing for both steady state sharing performance at steady state [38,45,46], the focus of this paper is on using the virtual and transients throughout thepower overallsharing flight profile. improving dc microgrid impedance to enable flexible for bothIndeed, steady research state andontransients throughout the performance periods on have been performed [17–19,47]. Instead during of using virtual overall flight during profile. transient Indeed, research improving dc microgrid performance transient capacitors [18,19] the first order filters Instead [17], however, paper studies the application of virtual periods have beenor performed [17–19,47]. of usingthis virtual capacitors [18,19] or the first order inductors. While a virtual inductance was used in [47], the focus of such research was on decreasing filters [17], however, this paper studies the application of virtual inductors. While a virtual power oscillations instead ofthe realizing flexible power sharing. virtual impedances that inductance was used in [47], focus of such research was on Compared decreasing to power oscillations instead are either purely inductive [47] or capacitive [18,19], in addition, this research explores the effective of realizing flexible power sharing. Compared to virtual impedances that are either purely inductive usage virtual impedances that include a resistive term an inductive Byimpedances including a [47] or of capacitive [18,19], in addition, this both research explores theand effective usage ofterm. virtual resistive component, the corresponding source can participate in power sharing for both steady that include both a resistive term and an inductive term. By including a resistive component,state the and transient conditions. corresponding source can participate in power sharing for both steady state and transient conditions.

Figure Figure 3. 3. Comparison Comparison of of power power source source characteristics characteristics (Redrawn (Redrawn using using data data of of [42,43]).

Third, practical practical challenges challenges that that occur occur during during realistic realistic operation operationof of aircraft aircraft EPS EPS are are addressed. addressed. Third, Especially, the needs for minimal minimal impact impact caused caused by by changes changes in inthe theEPS EPSconfiguration configurationare areconsidered. considered. Especially, Thanks to the decentralized and addition of decentralized control controlarchitecture architectureofofthe theconsidered consideredapproach, approach,removal removal and addition components could bebeperformed approaches [13]. [13]. of components could performedininaamodular modularfashion fashion compared compared to centralized approaches Such modularity modularity enables enables aircraft aircraft retrofit tasks to be performed With the the Such performed with with minimal minimal efforts. efforts. With continuous development of advanced electronic payloads [8], system control approaches should continuous development of advanced electronic payloads [8], system control approaches should support activities andand equipment replacement in a simpler way. As the As burden support activitiesofofsystem systemmodification modification equipment replacement in a simpler way. the that each power source should handle at steady state and transients could be set to be different, burden that each power source should handle at steady state and transients could be set to be in addition, the considered virtual impedance-based approachapproach enables system so that the different, in addition, the considered virtual impedance-based enablesoperation system operation so lifetime of the power could becould maximized. As reported in detailed lifetimelifetime studies studies [48,49], that the lifetime of thesources power sources be maximized. As reported in detailed the aging ofaging fuel cells and cells batteries determined by the operation of each source. For example, [48,49], the of fuel and are batteries are determined by theprofiles operation profiles of each source.

For example, fuel cells show expedited aging when it experiences large power transients [48] and the operation profile has direct impact on the thermal stress of batteries [49].


6 of 19

Electronics 7, xexpedited FOR PEER REVIEW 6 of 19 fuel cells2018, show aging when it experiences large power transients [48] and the operation profile has direct impact on the thermal stress of batteries [49]. 3. Decentralized Power Management for an MEA EPS 3. Decentralized Power Management for an MEA EPS 3.1. Limitations of Centralized Power Management 3.1. Limitations of Centralized Power Management For a MEA EPS, system control can be performed by either a centralized [36] or a decentralized For a MEA EPS, system control can be performed a decentralized [38] [38] architecture. In case of the centralized approach,by as either showna centralized in Figure 4,[36] the or central power control architecture. In case of the centralized approach, as shown in Figure 4, the central power control unit unit (PCU) generates command values for each power source. As the detailed information of the (PCU) command for integrated each power source. As can the be detailed information of theoperator overall overallgenerates EPS is transferred tovalues the PCU, information provided to the system EPS is transferred to the PCU, integrated information can be provided to the system operator (i.e., pilot (i.e., pilot of manned aircrafts or ground station control personal of unmanned aircrafts) for increased of manned aircrafts or ground station control personal of unmanned aircrafts) for increased situational situational awareness. Based on the monitored data of the EPS, necessary corrective action could be awareness. the monitored of the EPS, necessary corrective action could be emergency performed by performed Based by theon system operatordata to handle abnormal operation situations such as or the system operator to handle abnormal operation situations such as emergency or faults. Furthermore, faults. Furthermore, optimized operation of the EPS could be performed using the feedback optimized operation of the EPS. EPS could be performed using the feedback information of the overall EPS. information of the overall

Centralized Control Unit

Power Control Unit

Control Algorithm

Communication Links

Information Transfer …

Source

Source

Electrical Bus

Energy Storage

Power Sources

Load

…

Action & Monitoring

Load

Power System Components

Electrical Loads

Connection of EPS Components Figure 4. Centralized control control architecture Figure 4. Centralized architecture for for an an aircraft aircraft EPS. EPS.

Although Although the the centralized centralized approach approach realizes realizes integrated integrated power power management, management, the the operation operation of of the the EPS link reliability. In case feedback of the information from EPS depends dependson onthe thecommunication communication link reliability. In case feedback ofnecessary the necessary information components is unavailable because of faults in thein communication path, path, the performance of theof PCU from components is unavailable because of faults the communication the performance the would also be affected. In case it is assumed that the PCU and EPS components are fully operational, PCU would also be affected. In case it is assumed that the PCU and EPS components are fully the reliabilitythe of the overallofEPS be expressed as expressed as operational, reliability thecan overall EPS can be 𝑀 M

=∏ (t) = ) ∏ R𝑅i (t(𝑡) 𝑅R EPS(𝑡) 𝐸𝑃𝑆

i =1

𝑖

(1) (1)

𝑖=1

where reliability of where M M is is the the total total number number of of EPS EPS components components and and R Rii(t) (t) is is the the reliability of the the communication communication link link between the PCU and the ith EPS component. Assuming a constant failure rate, the between the PCU and the ith EPS component. Assuming a constant failure rate, the reliability reliability of of each each communication communication link link can can be be written written as as [50] [50] λt = e𝑒−−𝜆𝑡 R𝑅(𝑡) (t) =

(2) (2)

where λ is the failure rate of the communication link. Figure 5 shows how the reliability of the EPS where λ is the failure rate of the communication link. Figure 5 shows how the reliability of the EPS at at 1000 h varies as the number of EPS component (M) increases. In order to study the effect of different 1000 h varies as the number of EPS component (M) increases. In order to study the effect of different failure rates, the failure rate of 𝜆1 = 2 failures/1066h, 𝜆2 = 5 failures/106 6h, and 𝜆3 = 10 failures/106 hrs failure rates, the failure rate of λ1 = 2 failures/10 h, λ2 = 5 failures/10 h, and λ3 = 10 failures/106 h were considered. As shown in Figure 5, the reliability of the overall EPS decreases as the number of were considered. As shown in Figure 5, the reliability of the overall EPS decreases as the number of EPS components increase, regardless of the failure rate value. That is, the reliability of the centralized EPS components increase, regardless of the failure rate value. That is, the reliability of the centralized power management approach depends not only on the reliability of the component itself but also on power management approach depends not only on the reliability of the component itself but also the status of the communication link. Hence, it is necessary to develop a decentralized control on the status of the communication link. Hence, it is necessary to develop a decentralized control architecture that performs effective power management in an EPS for the MEA to achieve higher system reliability.


7 of 19

architecture that performs effective power management in an EPS for the MEA to achieve higher Electronics 2018, 7, x FOR PEER REVIEW 7 of 19 system reliability. Electronics 2018, 7, x FOR 1 PEER REVIEW

7 of 19

0.99 1 0.98 0.99 0.97

Reliability Reliability

0.98 0.96 0.97 0.95 0.96 0.94 0.95 0.93 0.94 0.92

 

0.93 0.91



0.92 0.9 1 0.91

1 2 3



1

 3 2

2



3

4

5

6

7

8

9

10

Number of Components (M)

0.9 Figure 5. System with different failure and number of9 components. 1 reliability 2 3 with 4 5 failure6rates 7 number 8 10 Figure 5. System reliability different rates and of components.

Number of Components (M)

3.2. Decentralized Power Management

3.2. Decentralized Power Management

Figure 5. System reliability with different failure rates and number of components.

In order to realize decentralized power management, this impedance to In order to realize decentralized power management, thisresearch researchuses uses the the virtual virtual impedance determine the power sharing ratio among different power sources. Figure 6 shows the block diagram to3.2. determine the power sharing ratio among different power sources. Figure 6 shows the block Decentralized Power Management of the diagram considered approach. As shown in Figure 6, the controller has a cascaded configuration of the considered approach. As shown in Figure 6, the controller has a cascadedthat In order to realize decentralized power management, this research uses the virtual impedance relies on the feedback ofrelies both on thethe voltage andof theboth current. The control is converted a sequence configuration that feedback the voltage and theoutput current. The controltooutput is to determine the power sharing ratio among different power sources. Figure 6 shows the block converted to a sequence of switchingmodulation signals using(PWM) pulse-width [51]. The virtual of switching signals using pulse-width [51]. modulation The virtual(PWM) impedance is used when diagram of the considered approach. As shown in Figure 6, the controller has a cascaded impedance is used when the reference value for the voltage is generated as the reference value for the voltage controller as controller configuration that relies on the feedbackisofgenerated both the voltage and the current. The control output is

(3) 𝑣 ∗ =using 𝑉0 − 𝑍pulse-width converted to a sequence of switching signals modulation (PWM) [51]. The virtual 𝑣 ∙ 𝑖𝑜 ∗ 𝑖 v = V − Z · i (3) v o 0 i impedance is used when the reference value for the voltage controller is generated as where V0 is the no-load voltage, Zv is the virtual impedance, and io is the output current of the power As the command value for the voltage controller by subtracting the product of (3) 𝑣𝑖∗ impedance, = 𝑉0 − 𝑍𝑣 ∙is𝑖𝑜generated where source. V 0 is the no-load voltage, Zv is the virtual and io is the output current of the power the virtual impedance and the output current from the no-load voltage, the virtual impedance source. As theVcommand value for theZvoltage controller is generated subtracting the product of the where 0 is the no-load voltage, v is the virtual impedance, and io isbythe output current of the power performs a role equivalent to an actual impedance. The distinct feature of control law (3) is that the source. As the command value for the voltage controller is generated by subtracting the product of a virtualcommand impedance and the output current from the no-load voltage, the virtual impedance performs is generated using only local information that is easily accessible from each power source. the virtual impedance and the output current from the no-load voltage, the virtual impedance role equivalent to an actual impedance. The distinct featurecan of control law (3) is that the command In other words, the output of different power sources be coordinated without requiring performs a role equivalent to an actualthat impedance. The distinct feature of control law (3) is that the is generated using only local information is easily accessible from each power source. In other communication with the central PCU. command is generated using only local information that is easily accessible from each power source. words, the output of different power sources can be coordinated without requiring communication + In other words, the output of different power sources can be coordinated without requiring + Switch with the central+PCU. Voltage Current Vcommunication PWM with the central PCU. o

_

Vo

Controller

_

V*

+

+ _

V*

_

V

ZV(s) Virtual Impedance ZV(s)

V

Voltage Controller

I*

+ I*

Signal

Controller

_

Current Controller

_

Pulse Width PWM Modulation Pulse Width Modulation

I I

Virtual Impedance

Figure 6. Block diagram of a virtual impedance-based power management.

Figure 6. Block diagram of a virtual impedance-based power management. Figure 6. Block diagram of a virtual impedance-based power management.

Switch Signal


8 of 19

The effect of virtual impedance to the power sharing ratio can be studied using the equivalent circuit of a dc EPS as shown in Figure 7. As shown in Figure 7, the EPS consists of two sources, a common load, the cable impedance and the virtual impedance. The relationship between the output current of each source and the load current can be written as Iload (s) = Io1 (s) + Io2 (s)

(4)

where Iload is the total load current, Io1 is the output current of source 1 and Io2 is the output current of source 2. Because the output current of each source is determined by the impedance between each source and the load, the output current of each source can be written as [52] Io1 (s) =

[ R L + Zv2 (s) + Zc2 (s)]V01 (s) − R L V2 (s) [ Zv1 (s) + Zc1 (s)][ Zv2 (s) + Zc2 (s)] + R L [ Zv1 (s) + Zc1 (s) + Zv1 (s) + Zc1 (s)]

(5)

Io2 (s) =

[ R L + Zv1 (s) + Zc1 (s)]V02 (s) − R L V1 (s) [ Zv1 (s) + Zc1 (s)][ Zv2 (s) + Zc2 (s)] + R L [ Zv1 (s) + Zc1 (s) + Zv1 (s) + Zc1 (s)]

(6)

where RL is the load resistance, Vk , Zvk , Zck , is the no load output voltage, virtual impedance and cable impedance of power source k (k = 1, 2), respectively. As the no-load voltages of power sources that are connected in parallel to a common bus are typically set to an identical value (i.e., Vo1 = Vo2 ), the power sharing ratio between the two sources can be written as Io1 (s) Z (s) + Zc2 (s) Z (s) + Rc2 + sLc2 = v2 = v2 (7) Io2 (s) Zv1 (s) + Zc1 (s) Zv1 (s) + Rc1 + sLc1 where Rck and Lck are the cable resistance and inductance of power source k, respectively. As shown in Equation (7), the current sharing ratio between the power sources is determined by the parameter values of the virtual impedance and the cable impedance. When the virtual impedance is designed to be purely resistive and the values are sufficiently larger than the cable resistance (Rv1 , Rv2 >> Rc1 , Rc2 ), the power sharing ratio between the two sources at steady state is determined by the virtual resistance value. Although selecting a large virtual resistance value would contribute to achieving satisfactory power sharing performance at steady state by decreasing the effect of cable resistance [53], a larger bus voltage deviation would occur as a result of a larger decrease in the voltage reference command. In case the wiring has a large inductance value, furthermore, the load sharing ratio during transients would be highly affected by the cable characteristics rather than the designed virtual resistance value. Hence, the virtual resistance approach would show limited effectiveness for achieving satisfactory load sharing performance during transients. In case of aircrafts, power transients could occur because of dynamic flight patterns, operation mode transitions and activation of pulsed power loads. In order to address such limitations of the purely resistive virtual impedance configuration and the occurrence of frequent power transients in aircrafts, this research considers an alternative virtual impedance configuration by using a virtual inductance as follows. 3.3. Virtual Inductance Based Power Management In this research, the virtual impedance that includes not only a resistive term but also an inductive term is considered. That is, the overall command of the power source is generated as Vi∗ (s) = V0 (s) − ( Rv + sLv )· I (s)

(8)

where Rv is the virtual resistance and Lv is the virtual inductance. Since the virtual impedance includes a frequency dependent term (sLv ), the dynamic response of the voltage command is expected to be different from the case of a purely resistive virtual impedance (Lv = 0) during transients. In order to study the effect of the virtual inductance, the response characteristics of power sources with different types of virtual impedances are considered. For the network of Figure 7, it is assumed that the virtual


9 of 19

impedances of the two sources are configured to be different so that a virtual inductance is included 9 of 19 for power source #2 only. In other words, the virtual impedances of the two power sources were configured as (9) 𝑍v1 =R 𝑅𝑣v Z s) = (9) 𝑣1((𝑠)


Z𝑍v2 ((𝑠) s) = + 𝑠𝐿 sLv =R 𝑅𝑣v + 𝑣2 𝑣

Io1

Io2

Zc1

(10) (10)

Iload

Zc2 RL Zv2

Zv1

Vo1

+ _

+ _

Vo2

: Virtual Impedance : Cable Impedance

Figure 7. Equivalent circuit of an EPS with two sources connected in parallel. Figure 7. Equivalent circuit of an EPS with two sources connected in parallel.

For the considered virtual impedance configuration, the output current response of the two power Fortothe considered configuration, the output current response of the two sources a load current virtual change impedance can be described by the transfer functions written as power sources to a load current change can be described by the transfer functions written as Zv2 (s) + Zc2 (s) I (s) ( Lv + Lc )s + ( Rv + Rc ) (11) M1 (s) = o1 𝐼𝑜1 (𝑠)= 𝑍𝑣2 (𝑠) + 𝑍𝑐2 (𝑠) = (𝐿𝑣 + 𝐿𝑐 )𝑠 + (𝑅𝑣 + 𝑅𝑐 ) I ∆ s ∆(s) s ( ) ( ) 𝑀1 (𝑠) = o = = (11) 𝐼𝑜 (𝑠) Δ(𝑠) Δ(𝑠) Io2 (s) Lc s + ( Rv + Rc ) Z (s) + Zc1 (s) M2 (s) = 𝐼𝑜2 (𝑠)= 𝑍v1 (12) + 𝑍𝑐1 (𝑠) = 𝐿𝑐 𝑠 + (𝑅 𝑣1 (𝑠) I ∆ s ∆(𝑣s+ s ( ) ) 𝑅𝑐 ) ( ) 𝑀2 (𝑠) = o = = (12) 𝐼𝑜 (𝑠) Δ(𝑠) Δ(𝑠) where ∆(s) = ( Lv + 2Lc )s + 2Rv + 2Rc , Rc = Rc1 = Rc2 is the cable resistance, Lc = Lc1 = Lc2 is the cable where Δ(s) M = 1(𝐿 + 2𝐿 , Rc = Rc1the = Rload c2 is current the cable resistance, Lc = Lof c1 = Lc2 is 1, the cable 𝑐 )𝑠 + 2𝑅 𝑣 + 2𝑅𝑐between inductance, is𝑣the transfer function and output current source and M2 inductance, M 1 is the transfer function between the load current and output current of source 1, and is the transfer function between the load current and output current of source 2. Since the output M 2 is theof transfer the load outputaccording current oftosource 2. Since current the twofunction sourcesbetween for a given load current change and responds (11) and (12), the the output power current thebetween two sources for sources a given at load change according to (11) (12), the power sharing of ratio the two steady stateresponds and during transients can and be determined from sharing ratio between the two sources at steady state and during transients can be determined from M1 and M2 . Figure 8 shows how the magnitude of the M1 and M2 varies at different frequency ranges M 1 and M2parameter . Figure 8 shows magnitude the M 2 varies different using the valueshow of Rvthe = 50 mΩ, Lv = of 1 mH, L1c and = 50M µH, and Ratc = 10 mΩ.frequency ranges usingAs theshown parameter values of Rresponse v = 50 mΩ, Lv = 1 mH, Lc =of50 μH, and R c = 10 mΩ. in Figure 8, the characteristics the power sources changes depending on the frequency While of the two transfer are similar at low frequencies, As shownrange. in Figure 8, the the magnitudes response characteristics of the functions power sources changes depending on an apparent difference can be observed at higher Such a difference explained by the frequency range. While the magnitudes offrequencies. the two transfer functions can are be similar at low the increasedan impedance at higher frequencies which is caused by the avirtual inductance. frequencies, apparent magnitude difference can be observed at higher frequencies. Such difference can be In theby considered case,impedance for instance, most of the would beishandled bythe thevirtual power explained the increased magnitude at power higher transients frequencies which caused by source which has a purely resistive virtual impedance. While the virtual resistance value affects the inductance. steady state power sharing ratio, the virtual inductance determines thehandled performance In the considered case, for instance, most of the powervalue transients would be by the during power transient states. source which has a purely resistive virtual impedance. While the virtual resistance value affects the The effect of a sharing virtual inductance to the inductance power sharing ratio could also explained from the steady state power ratio, the virtual value determines thebeperformance during fact that the output current of each source is determined by the effective impedance between each transient states. source and the common load [54]. In case an identical virtualratio resistance is considered for both The effect of a virtual inductance to the power sharing could value also be explained from the sources only source 2 includes a virtual inductance, by it isthe evident thatimpedance the effective impedance fact that and the output current of each source is determined effective between each source and the common load [54]. In case an identical virtual resistance value is considered for both sources and only source 2 includes a virtual inductance, it is evident that the effective impedance seen from the load at non-zero frequencies would be larger for source 2 than source 1. Hence, most of the transient power demand would be supplied by source 1 that is connected to the load through a smaller impedance.


10 of 19

seen from the load at non-zero frequencies would be larger for source 2 than source 1. Hence, most of the transient power demand would be supplied by source 1 that is connected to the load through a smaller2018, impedance. Electronics 7, x FOR PEER REVIEW 10 of 19 Transfer Function Magnitdue 10 Source 1 (M1) Source 2 (M2)

5

Magnitude (dB)

0 -5 -10 -15 -20 -25 -30 -4 10

10

-3

-2

-1

10 10 Frequency (kHz)

10

0

10

1

Figure Frequency response transfer functions source 1 and source Figure 8. 8. Frequency response of of transfer functions forfor source 1 and source 2. 2.

Similar otherdroop-based droop-based approaches [15,38], deviation themain mainbus busvoltage voltagewould would Similar totoother approaches [15,38], a adeviation ininthe occur a result the voltage drop caused thevirtual virtualresistance. resistance. While several approaches that occur asas a result ofof the voltage drop caused bybythe While several approaches that apply an additional voltage restoration control loop have been reported for stationary microgrid apply an additional voltage restoration control loop have been reported for stationary microgrid applications [38,39], is worth reviewing alternative approaches to handle a voltage drop applications [38,39], it isitworth reviewing alternative approaches to handle such asuch voltage drop issue in aircrafts. For example, direct connection of the battery to the electrical bus [27] (which has inissue aircrafts. For example, direct connection of the battery to the electrical bus [27] (which has also also been applied in telecommunication systems [55]) would not only address the voltage drop but been applied in telecommunication systems [55]) would not only address the voltage drop but also also enable immediate response of the battery when all primary power sources are lost. Such enable immediate response of the battery when all primary power sources are lost. Such a solution a solution handle stability issues thatwith could occur with increased connection of constant would alsowould handlealso stability issues that could occur increased connection of constant power loads power loads [27]. Selecting the virtual resistance value according to the range of a normal voltage [27]. Selecting the virtual resistance value according to the range of a normal voltage according to according to design could also beprevious considered as in studies for aircrafts [41]. design standards could standards also be considered as in studies forprevious aircrafts [41]. According to MILAccording to MIL-STD-704F [56], which specifies EPS characteristics for aircrafts, the steady-state STD-704F [56], which specifies EPS characteristics for aircrafts, the steady-state voltage range during voltage range during normal operation is given V for V dc system 22–29 V for normal operation is given as 250–280 V for a 270 as V 250–280 dc system anda 270 22–29 V for a 28and V dc system. a 28 V dc system. Connecting more droop-controlled in parallel the electrical bus would Connecting more droop-controlled sources in parallel tosources the electrical bus to would also contribute to also contribute to voltage drop minimization [37]. Since past research has focused on development voltage drop minimization [37]. Since past research has focused on development of controlof control approaches to mitigate the droop originated voltage research focuses realizing approaches to mitigate the droop originated voltage drop,drop, this this research focuses onon realizing decentralized power sharing using the virtual inductance with a particular focus transient periods. decentralized power sharing using the virtual inductance with a particular focus onon transient periods. Details on implementation examples of voltage restoration controllers can be referred from [38,39]. Details on implementation examples of voltage restoration controllers can be referred from [38,39]. 3.4. Effectivenss Proposed Approach MEA Applications 3.4. Effectivenss of of thethe Proposed Approach forfor MEA Applications ForMEA MEAplatforms, platforms,thetheconsidered considered control approach can usedtotorealize realizeflexible flexiblepower power For control approach can bebeused management among different power sources. The power sharing ratio can be set to be different management among different power sources. The power sharing ratio can be set to be different betweensteady steadystate state and and transient useuse of aofvirtual inductance. Such Such a feature between transient periods periodsthrough througheffective effective a virtual inductance. a contributes to stable operation of EPSs in aircrafts by enabling power management of various sources feature contributes to stable operation of EPSs in aircrafts by enabling power management of various according to their density and power density. Considering the increased use ofuse multi-source sources according toenergy their energy density and power density. Considering the increased of multiEPSs in MEA it is necessary that the that power different properlyiscoordinated source EPSs in [7,24], MEA [7,24], it is necessary theoutput powerofoutput of sources differentis sources properly to meet thetoload overall profile. In particular, the power flow should coordinated meetdemand the load during demandthe during theflight overall flight profile. In particular, the power flow be controlled so that each source operates according to its own power density and energy density should be controlled so that each source operates according to its own power density and energy characteristics. For instance, the EPS the could be could managed such that the source a large with power density characteristics. For instance, EPS be managed such that with the source a density large

power density handles most of the power transients, while the source with superior energy density supplies most of the power during steady state. According to such an operation scenario, the power sharing ratio should be different depending on the load demand profile (i.e., steady state and transient). The importance of managing the output of power sources during transients could also be highlighted from aging characteristics of power sources. In case of a fuel cell, which has been


11 of 19

handles most of the power transients, while the source with superior energy density supplies most of 11 of 19 the power during steady state. According to such an operation scenario, the power sharing ratio should be differentasdepending on the load demand profile (i.e., steady state[7,24], and transient). The importance considered a main power source for various vehicular platforms it has been reported that of managing the output of power sources during transients could also be highlighted from the lifetime is highly affected by the frequency and level of power transients that the fuelaging cell characteristics of power sources.[48]. In case of a fuel cell, which has been considered as a main power experiences during its operation source variousthe vehicular platforms [7,24], it has been reported lifetime is highly affected by By for applying considered power management approach,that thethe power sharing ratio can be the frequency and level of power transients that the fuel cell experiences during its operation [48]. adjusted in a flexible manner by shaping the virtual impedance of each source. In details, the virtual By applying the considered power approach, power sharing ratio can be resistance is determined according to themanagement desired power sharingthe ratio at steady state, while theadjusted virtual in a flexible manner by shaping the virtual impedance of each source. In details, the virtual resistance inductance can be used to adjust the output characteristics during transients only. For the EPS of is determined accordingconsider to the desired power sharing at steady state, while the cable virtualimpedance inductance Figure 7, for example, the case when the ratio virtual impedance and the can be used adjust characteristics during the LEPS of Figure 7, parameters aretoset as Rv1the = 50output mΩ, Lv1 = 0, Rv2 = 25 mΩ, Lv2 =transients 1 mH, Rc1 =only. Rc2 = For 10 mΩ, c1 = Lc2 = 50 μH. for example, consider the case when the virtual impedance and the cable impedance parameters are According to such parameter values, the output impedance frequency characteristics of the two set as R = 50 mΩ, L = 0, R = 25 mΩ, L = 1 mH, R = R = 10 mΩ, L = L = 50 µH. According v2 9. It should v2 be noted that c1 the c2 virtual resistance c1 c2value of source 2 is sources v1 can be plottedv1as Figure to such parameter values, the output impedance frequency characteristics of the be smaller than that of source 1. Hence, source 2 supplies more power than sourcetwo 1 atsources steady can state. plotted as Figure 9. It should be noted that the virtual resistance value of source 2 is smaller than that However, a virtual inductance of 1mH is implemented only in source 2, and the magnitude of the of source 1. Hence,atsource supplies more power1 than sourcethan 1 atthat steady state. However, a virtual overall impedance higher2frequencies of source is smaller of source 2. Hence, source 2 inductance of 1mH is implemented only in source 2, and the magnitude of the overall impedance supplies more power than source 1 at steady state and source 1 handles most of the transient powerat higher frequencies of source 1 is smaller than that of source 2. Hence, source 2 supplies more power than source 2. than source 1 at steady state and source 1 handles most of the transient power than source 2. Electronics 2018, 7, x FOR PEER REVIEW

Imedance Magnitude 30 20

Magnitude (dB)

10 0 -10 -20 -30 -40 -4 10

Source 1 (Z1) Source 2 (Z2) 10

-3

-2

-1

10 10 Frequency (kHz)

10

0

10

1

Figure Figure9.9.Frequency Frequencycharacteristics characteristicsofofimpedances impedancesfor forsource source11and andsource source2.2.

Although Although aa similar similar effect effect can can be be achieved achieved by by aa centralized centralized control control approach, approach, itit should should be be highlighted highlightedthat thatthe thepower powersharing sharingatatboth bothtransient transientand andsteady steadystate statecan canbe beaccomplished accomplishedby bysimply simply designing designingthe thevirtual virtualimpedance impedanceof ofeach eachpower powersource sourcecontrol controllaw lawin inaadecentralized decentralizedmanner. manner.For Foran an EPS EPSthat thatconsists consistsof ofsources sourcesthat thathave havedifferent differentpower powerdensity densitylevels, levels,for forexample, example,ititisispossible possibleto tolet let the thesource sourcewith withthe thehighest highestpower powerdensity densityto tohandle handlemost mostof ofthe thepower powertransients transientsby bydesigning designingthe the virtual virtualimpedance. impedance.While Whilethe thesource sourcewith withthe thelargest largestpower powerdensity densitywould wouldbe beconfigured configuredtotohave haveaa purely purelyresistive resistivevirtual virtualimpedance, impedance,the theresponse responseofofthe therest restofofthe thesources sourcesto toaaload loadchange changecould couldbe be delayed adoptinga avirtual virtual impedance includes a resistive an inductive term. delayed by adopting impedance thatthat includes bothboth a resistive and anand inductive term. Through Through suchitcontrol, it is to possible prevent operation cases fails to meet sudden such control, is possible preventtooperation cases where the where source the failssource to meet sudden load power load power changes because of its inherent slow dynamics. changes because of its inherent slow dynamics. 4. Verification Results The performance of the considered power control approach was verified by a simulation study. Figure 10 shows the configuration of the simulated EPS. As shown in Figure 10, the system uses 270 V dc as the main bus and consists of two sources, cable wiring, and loads. In order to perform power conversion, each power source is connected to the system bus through a buck converter. Detailed parameter values of the simulated system are given in Table 1.


12 of 19

4. Verification Results The performance of the considered power control approach was verified by a simulation study. Figure 10 shows the configuration of the simulated EPS. As shown in Figure 10, the system uses 270 V dc as the main bus and consists of two sources, cable wiring, and loads. In order to perform power conversion, each power source is connected to the system bus through a buck converter. Detailed parameter values of PEER the simulated Electronics 2018, 7, x FOR REVIEW system are given in Table 1. 12 of 19

Source 1

System Bus (270V DC)

Power Conversion

Wiring SW 1

RC1

LC1

Load 1

I O1

Wiring

Load 2 SW 2

RC 2 Source 2

LC 2

Power Conversion

IO 2

Figure10. 10.Configuration Configurationof ofthe thesimulated simulatedEPS. EPS. Figure Table 1. Parameter values used for simulation. Table 1. Parameter values used for simulation. System Parameter

System Parameter

Converter Switching Switching Frequency Frequency Converter Converter Inductanace Converter Inductanace Converter Capacitance Converter Capacitance Source Output Voltage Output Voltage Cable Source Resistance Cable Inductance Cable Resistance

Cable Inductance

Value

Value 20 kHz kHz 20 1 mH 11mH mF 1500 mF V Rc 500 = 10VmΩ 50mΩ µH RLcc==10 Lc = 50 μH

In order to study the effect of a virtual inductance to the power sharing performance, the virtual In orderconfiguration to study the effect of asource virtualwas inductance the power sharing impedance of each set to betodifferent from each performance, other. While the the virtual virtual impedance configuration of each source was set to be different from each other. While the virtual impedance of source 1 was configured to be purely resistance (Rv1 = 50 mΩ and Lv1 = 0), the virtual impedance purelyboth resistance (Rresistance v1 = 50 mΩand andaLvirtual v1 = 0), inductance the virtual impedanceof ofsource source12 was was configured configured to to be include a virtual impedance of source 2 was configured to include both a virtual resistance and a virtual inductance (Rv2 = 50 mΩ and Lv2 = 1 mH). Figure 11 shows the output current of the two sources when the load (R v2 = 50was mΩincreased and Lv2 = from 1 mH). Figure 1114.58 shows the current of the twobetween sources the when the load power 7.29 kW to kW at output t = 0.5 s. The difference two-output power increased 7.29 kW topurposes. 14.58 kWAs at the t = virtual 0.5 s. The difference between currentwas is also plotted from for comparison resistance value is set tothe thetwo-output same value, current is also plotted for comparison purposes. As the virtual resistance value is set to same the differences of the output current between both sources are negligible at steady state. the However, value, the differences of the output current between both sources are negligible at steady state. a noticeable difference is observed during the transient period. It could be seen that the source 1 However, noticeable difference is observed duringsupplies the transient period. It could be seen the2. (i.e., sourcea with a purely resistive virtual impedance) most of the power compared to that source source 1 (i.e., source with a purely resistive virtual impedance) supplies most of the power compared During transients, as shown in Figure 11, instants that the power output of source 1 is 1.8 times larger to source as shown in Figure 11, instants thatsources the power output of source 1 is 1.8 than that2.ofDuring sourcetransients, 2 is observed. Hence, the output of power during transients could be times larger than that of source 2 is observed. Hence, the output of power sources during transients managed by the virtual impedance design. When the transient output of a certain source should be could be for managed bythe thevirtual virtualinductance impedancecould design. When the used transient output a certain source limited, example, be effectively to shape theof transient response. should be load limited, for example, the voltage virtual inductance be effectively usedintoFigure shape 12. the While transient The total current and the bus for the casecould of Figure 11 are shown the response. The total load current and the bus voltage for the case of Figure 11 are shown in Figure 12. bus voltage experiences a transient undershoot as soon as the load is increased at t = 0.5 s, it could While thethat busthe voltage experiences a transient undershoot as soonthat as the load is increased at t =design 0.5 s, be seen bus voltage satisfies the normal voltage range is defined in the aircraft itstandard could be[56]. seen Because that the bus voltage satisfies the normal voltage range that is defined in the aircraft of the virtual resistance, as mentioned in Section 3.4, a voltage drop that design standard [56]. Because of the virtual resistance, as mentioned in Section 3.4, a voltage drop corresponds to the load power level is observed. As the amount of deviation is affected by various that corresponds to the load power level is observed. As the amount of deviation is affected by various system parameters (e.g., load current, virtual resistance, number of droop sources that are connected in parallel), the voltage regulation performance could be managed by adjusting such design factors. Figure 13 shows the waveform when a larger virtual inductance value is considered for the same

Electronics 2018, 7, x FOR PEER REVIEW Electronics 2018, 7, x FOR PEER REVIEW Electronics 2018, 7, 187

13 of 19 13 of 19 13 of 19

current currentduring duringtransients transientsisishighly highlyaffected affectedby bythe thevirtual virtualinductance inductancevalue. value.Because Becauseofofthe thelarger larger virtual inductance used in source 2, the difference of the response dynamics between the two virtual inductance used in source 2, the difference of the response dynamics between the twosources sources has As the a apurely resistive virtual impedance would more system parameters (e.g., load current, virtual resistance, number of droop sources that supply are connected hasincreased. increased. Asa aresult, result, thesource sourcewith with purely resistive virtual impedance would supply more power during the transient period. in parallel), the voltage regulation performance could be managed by adjusting such design factors. power during the transient period.

Figure v2 = = 1 mH). Current, (b): Figure 11. Simulated output current waveform (L 1 mH). mH).(a): (a):Output Output Current, (b):Difference. Difference. v2 Figure11. 11.Simulated Simulatedoutput outputcurrent currentwaveform waveform(L (L v2 = 1 (a): Output Current, (b): Difference.

Figure v2v2 == 1 mH). Current, (b): Voltage. Figure 12. Simulated load current and bus voltage (L 1 mH).(a): (a):Load Load Current, (b):Bus Bus Voltage. Figure12. 12.Simulated Simulatedload loadcurrent currentand andbus busvoltage voltage(L (L v2 = 1 mH). (a): Load Current, (b): Bus Voltage.


14 of 19

Figure 13 shows the waveform when a larger virtual inductance value is considered for the same EPS configuration. In this case, a virtual inductance value that is three times larger than the previous case was considered (i.e., Lv2 = 3 mH). Regarding the output current at steady state, it can be seen that the output of both sources is similar. Such results can be explained by the identical virtual resistance value. By comparing the waveforms of Figures 11 and 13, it can be seen that the behavior of output current during transients is highly affected by the virtual inductance value. Because of the larger virtual inductance used in source 2, the difference of the response dynamics between the two sources has increased. As a result, the source with a purely resistive virtual impedance would supply more power during the transient period. Electronics 2018, 7, x FOR PEER REVIEW 14 of 19

Figure = 3 mH). (a): Output Current, (b): Difference. Figure 13. 13. Simulated Simulated output output current current waveform waveform (L (Lv2 v2 = 3 mH). (a): Output Current, (b): Difference.

Figure 14 shows that the proposed control approach can coordinate the output of power sources Figure 14 shows that the proposed control approach can coordinate the output of power sources such that that the the power power sharing sharing ratio ratio at at steady steady state state and and during during transients transients could could be be set set to to be be different. different. such Such a difference in the power sharing ratio is realized by the parameter values of the virtual impedance. Such a difference in the power sharing ratio is realized by the parameter values of the virtual In case of Figure 14,ofthe virtual values of the two sources were selected asselected Rv1 = 50asmΩ, impedance. In case Figure 14, impedance the virtual impedance values of the two sources were Rv1 0, Rv2 Lv2 =and 0.1 LmH. Because of the smaller virtual resistance value, it is expected =Lv1 50=mΩ, Lv1==25 0, mΩ, Rv2 =and 25 mΩ, v2 = 0.1 mH. Because of the smaller virtual resistance value, it is that source 2 would more power than sourcethan 1 atsource steady 1state. Because of Because the virtual expected that sourcesupply 2 would supply more power at steady state. of inductance the virtual in source 2, however, source 1 is expected to supply most of the power during transients. In fact, inductance in source 2, however, source 1 is expected to supply most of the power during transients. Figure shows response characteristics of the output a load increase at In fact, 14 Figure 14such shows such response characteristics of thecurrent outputwhen current when a loadoccurs increase toccurs = 0.5 at s. tWhile source 1 supplies most of the transient power, the output of source 2 is larger than = 0.5 s. While source 1 supplies most of the transient power, the output of source 2 is larger that of source 1 at steady In case considered decentralized than that of source 1 atstate. steady state.only In the casevirtual only resistance the virtualwas resistance wasforconsidered for power control, it is evident that the output of source 2 (i.e., the source with smaller virtual resistance) decentralized power control, it is evident that the output of source 2 (i.e., the source with smaller would be larger than that be of source 1 at both state and transient However, 14 virtual resistance) would larger than that steady of source 1 at both steadyperiods. state and transientFigure periods. demonstrates that14 andemonstrates additional degree of additional freedom that enables separatethat control between steady state However, Figure that an degree of freedom enables separate control and transient performance is introduced by the virtual inductance. between steady state and transient performance is introduced by the virtual inductance. Figure 15 current response when a step-up/step-down change in the in load Figure 15shows showsthe theoutput output current response when a step-up/step-down change thepower load is introduced. In orderIn to order observe powerthe sharing dependingdepending on the virtual power is introduced. to the observe powerdifference sharing difference on impedance the virtual configuration, it was assumed the virtual resistance value of value sourceof1source is smaller than that of impedance configuration, it wasthat assumed that the virtual resistance 1 is smaller than source (Rv1 =250 Rv2 = 25 mΩ) In addition, a virtual inductance of 1 mH was considered only for that of 2source (RmΩ, v1 = 50 mΩ, Rv2 = 25 mΩ) In addition, a virtual inductance of 1mH was considered only for source 2. When a step increase in the load power is introduced at t = 0.5 s, source 1 handles most of the load demand during transients. Once the transients decay, source 2 supplies more power than source 1 according to the virtual resistance value. In addition, a similar pattern in the transient power sharing was observed when a step decrease in the load power is introduced at t = 0.75 s. Similar to load step up, source 1 shows a larger steeper current change rate than the output current of source

Electronics 2018, 7, 187 Electronics 2018, 7, x FOR PEER REVIEW

15 of 19 15 of 19

source 2. When a step increase in the load power is introduced at t = 0.5 s, source 1 handles most of the Electronics 2018, 7, x FOR PEER REVIEW 15 of 19 load demand during transients. Once the transients decay, source 2 supplies more power than source 1 according to the profile of source 2, while high power density sources (e.g., supercapacitor) operate similar the profile of source 1. That is, In letting power are located ontransient the lowerpower right side according to virtual the profile of source 2, while high powerasources density sources (e.g., supercapacitor) operate according totothe resistance value. addition, similarthat pattern in the sharing of Figure toprofile handle of the load duringsources assigning sources that step similar to 3the ofmost source 1. That is,demand letting power that and are located on power the lower right side was observed when a step decrease in the load power istransients introduced at t = 0.75 s. Similar to load are located thea upper left Figure 3change to supply more power at steady state. Itof is worth noting of Figure 3 on to handle most ofside theof load demand during transients assigning power sources up, source 1 shows larger steeper current rate than theand output current source 2.that For the that determining the optimal power sharing ratio atmore steady stateatand transients would be case are located on the upper left side of Figure 3 to supply power steady state. It is worth noting considered operation scenario, source 1 experiences a power change rate that is twice larger than that dependent. Duringthe theoptimal design power stage ofsharing an actual aircraft, in fact, detailed data would of eachbe power that determining ratio at steady statetheand transients case of source 2. Simultaneously, the output of source 1 at steady state is limited so that source 2 could source couldDuring be considered to determine source should handle more power dependent. the design stage of anwhich actualpower aircraft, in fact, the detailed data of each during power supply more time power for(i.e., a longer period. From thesteady waveforms of Figures 14 and 15, it could be seen different scales shorttotransients and long power state). of suchmore data include power source could be considered determine which sourceExamples should handle power during that the power sharing characteristics in a multi-source EPS can be controlled with higher flexibility source type, energy density, sourcestate). capacity and mission different timepower scalesdensity, (i.e., short transients andpower long steady Examples of suchrequirements. data include power by adjusting both the virtual resistance and the virtual inductance. source type, power density, energy density, power source capacity and mission requirements.

Figure 14. Simulation results with different virtual resistance values. (a): Output Current, (b): Difference. Figure 14. Simulation results with different resistance (a): Output Current, (b): Figure 14. Simulation results with different virtualvirtual resistance values.values. (a): Output Current, (b): Difference. Difference.

Figure 15. Simulation results proposedapproach approachwith with step-up/step-down step-up/step-down changes. (a):(a): Source Figure 15. Simulation results of of thethe proposed changes. Source 1 1 Output Current, (b): Source 2 Output Current. Figure 15. Simulation results of the proposed approach with step-up/step-down changes. (a): Source Output Current, (b): Source 2 Output Current. 1 Output Current, (b): Source 2 Output Current.


16 of 19

From the perspective of energy density and power density of sources, it is worth considering an EPS control approach so that high energy density sources (e.g., fuel cell and battery) operate according to the profile of source 2, while high power density sources (e.g., supercapacitor) operate similar to the profile of source 1. That is, letting power sources that are located on the lower right side of Figure 3 to handle most of the load demand during transients and assigning power sources that are located on the upper left side of Figure 3 to supply more power at steady state. It is worth noting that determining the optimal power sharing ratio at steady state and transients would be case dependent. During the design stage of an actual aircraft, in fact, the detailed data of each power source could be considered to determine which power source should handle more power during different time scales (i.e., short transients and long steady state). Examples of such data include power source type, power density, energy density, power source capacity and mission requirements. 5. Conclusions With the increased use of electrical power to perform critical functions in the MEA, development of reliable control architectures for EPSs is necessary. This paper introduced a decentralized control approach for EPS control in MEA platforms. By using the virtual impedance, the output of different power sources can be coordinated without relying on extensive communication among EPS components. In particular, the effective use of virtual inductance for managing power sharing during transients was studied. As the virtual impedance consists of both virtual resistance and virtual inductance, the power sharing ratio could be adjusted to be different between steady state and transient periods. The effectiveness and performance of the considered control approach were also demonstrated by simulating an aircraft electrical system with different virtual impedance configurations. It was shown that the power sharing ratio during transients can be adjusted by changing the difference of the virtual inductance value. The proposed approach contributes to higher flexibility of configuration and power flow. Not only the needs for EPS configuration modification in aircrafts could be effectively addressed by the decentralized nature of the control architecture, but also the output of each power source could be managed to meet output characteristics of each source. The design of the virtual resistance and the virtual inductance could be effectively used to coordinate the power sharing ratio at both steady state and transient periods in an independent manner. Such increased flexibility would enable optimized operation of MEA by managing the output of various sources that have different power density and energy density characteristics. Author Contributions: Formal analysis, M.K.; Literature Review, S.L.; Verification, S.B.; Writing: Draft Preparation, M.K., S.L. and S.B.; Writing: Review and Editing, S.B. Funding: This research was supported by Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education (Grant Number: 2017R1D1A1B03031723). This work was supported by the Korea Institute of Energy Technology Evaluation and Planning (KETEP) and the Ministry of Trade, Industry & Energy (MOTIE) of the Republic of Korea (No. 20161210200560). Conflicts of Interest: The authors declare no conflict of interest.

Nomenclature M REPS Ri λ Vok ; iok Zvk ; Rvk ; Lvk Iload Zck ; Rck ; Lck RL Vk *

Total number of EPS components Reliability of the overall EPS Reliability of the communication link between the PCU and the ith EPS component Failure rate of the communication link No-load voltage of the kth power source ; Output current of the kth power source Virtual impedance; Virtual resistance; Virtual inductance of the kth power source Total load current Cable impedance; Cable resistance; Cable inductance of the kth power source Load resistance Voltage command of the kth power source


Mk EPS EMI MEA PCU PEMFC PV PWM SOC SOFC

17 of 19

Transfer function between the load current and output current the kth power source Electrical Power System Electromagnetic Interference More Electric Aircraft Power Control Unit Proton Exchange Membrane Fuel Cell Photovoltaics Pulse Width Modulation State-of-Charge Solid Oxide Fuel Cell

References 1.

2.

3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18.

Un-Noor, F.; Padmanaban, S.; Mihet-Popa, L.; Mollah, M.N.; Hossain, E. A comprehensive study of key electric vehicle (EV) components, technologies, challenges, impacts, and future direction of development. Energies 2017, 10, 1217. [CrossRef] Jin, Z.; Meng, L.; Guerrero, J.M.; Han, R. Hierarchical control design for a shipboard power system with DC distribution and energy storage aboard future more-electric ships. IEEE Trans. Ind. Inform. 2018, 14, 703–714. [CrossRef] Sarlioglu, B.; Morris, C.T. More electric aircraft: Review, challenges, and opportunities for commercial transport aircraft. IEEE Trans. Transp. Electrif. 2015, 1, 54–64. [CrossRef] Ganev, E. Selecting the best electric machines for electrical power-generation systems: High-performance solutions for aerospace more electric architectures. IEEE Electrif. Mag. 2014, 2, 13–22. [CrossRef] Bennet, J.W.; Mecrow, B.C.; Atkinson, D.J.; Atkinson, G.J. Safety-critical design of electromechanical actuation systems in commercial aircraft. IET Electr. Power Appl. 2011, 5, 37–47. [CrossRef] Neuman, T. Fly the electrics skies. IEEE Spectr. 2016, 53, 44–48. [CrossRef] Gong, A.; Verstraete, D. Fuel cell propulsion in small fixed-wing unmanned aerial vehicles: Current status and research needs. Int. J. Hydrogen Energy 2017, 42, 21311–21333. [CrossRef] Moir, I.; Seabridge, A. Aircraft Systems: Mechanical, Electrical, and Avionics Subsystems Integration, 3rd ed.; Wiley: New York, NY, USA, 2008. Chen, J.; Wang, C.; Chen, J. Investigation on the selection of electric power system architecture for future more electric aircraft. IEEE Trans. Transp. Electrif. 2018, 4, 563–576. [CrossRef] Chang, J.; Wang, A. New VF-power system architecture and evaluation for future aircraft. IEEE Trans. Aerosp. Electron. Syst. 2006, 42, 527–539. [CrossRef] Areerak, K.N.; Bozhko, S.V.; Asher, G.M.; De Lillo, L.; Thomas, D.W.P. Stability study for a hybrid AC-DC more-electric aircraft power system. IEEE Trans. Aerosp. Electron. Syst. 2012, 48, 329–347. [CrossRef] Roboam, X.; Sareni, B.; Andrade, A. More electricity in the air: Toward optimized electrical networks embedded in more-electrical aircraft. IEEE Ind. Electron. Mag. 2012, 6, 6–17. [CrossRef] Wang, C.; Li, X.; Guo, L.; Li, Y.W. A nonlinear-disturbance-observer-based DC-bus voltage control for a hybrid AC/DC microgrid. IEEE Trans. Power Electron. 2014, 29, 6162–6177. [CrossRef] Dragicevic, T.; Guerrero, J.M.; Vasquez, J.C.; Skrlec, D. Supervisory control of an adaptive-droop regulated DC microgrid with battery management capability. IEEE Trans. Power Electron. 2014, 29, 695–706. [CrossRef] Dragicevic, T.; Lu, X.; Vasquez, J.C.; Guerrero, J.M. DC microgrids-Part 1: A review of control strategies and stabilization techniques. IEEE Trans. Power Electron. 2016, 31, 4876–4891. [CrossRef] Tahim, A.P.N.; Pagano, D.J.; Lenz, E.; Stramosk, V. Modeling and stability analysis of islanded DC microgrids under droop control. IEEE Trans. Power Electron. 2015, 30, 4597–4607. [CrossRef] Gu, Y.; Li, W.; He, X. Frequency-coordinating virtual impedance for autonomous power management of DC microgrid. IEEE Trans. Power Electron. 2015, 30, 2328–2337. [CrossRef] Xu, Q.; Hu, X.; Wang, P.; Xiao, J.; Tu, P.; Wen, C.; Lee, M.Y. A decentralized dynamic power sharing strategy for hybrid energy storage system in autonomous DC microgrid. IEEE Trans. Ind. Electron. 2017, 64, 5930–5941. [CrossRef]


19.

20. 21. 22.

23.

24.

25. 26. 27. 28. 29. 30. 31.

32.

33.

34. 35.

36.

37.

38. 39.

18 of 19

Zhao, X.; Li, Y.W.; Tian, H.; Wu, X. Energy management strategy of multiple supercapacitors in a DC microgrid using adaptive virtual impedance. IEEE J. Emerg. Sel. Top. Power Electron. 2016, 4, 1174–1185. [CrossRef] Ou, T.; Li, K.; Huang, C. Improvement of transient stability in a hybrid power multi-system using a designed NIDC (Novel Intelligent Damping Controller). Energies 2017, 10, 488. [CrossRef] Ou, T.; Hong, C. Dynamic operation and control of microgrid hybrid power systems. Energy 2014, 66, 314–323. [CrossRef] Motapon, S.N.; Dessaint, L.A.; Al-Haddad, K. A comprehensive study of energy management schemes for a fuel-cell hybrid emergency power system of more-electric aircraft. IEEE Trans. Ind. Electron. 2014, 61, 1320–1334. [CrossRef] Zhang, F.; Mollet, F.; Saudemont, C.; Robyns, B. Experimental validation of energy storage system management strategies for a local DC distribution system of more electric aircraft. IEEE Trans. Ind. Electron. 2010, 57, 3905–3916. [CrossRef] Lee, B.; Kwon, S.; Park, P.; Kim, K. Active power management system for an unmanned aerial vehicle powered by solar cells, a fuel cell, and batteries. IEEE Trans. Aerosp. Electron. Syst. 2014, 50, 3167–3177. [CrossRef] Ou, T. A novel unsymmetrical fault analysis for microgrid distribution systems. Electr. Power Energy Syst. 2012, 43, 1017–1024. [CrossRef] Ou, T. Ground fault current analysis with a direct building algorithm for microgrid distribution. Electr. Power Energy Syst. 2013, 53, 867–875. [CrossRef] Kwasinski, A.; Onwuchekwa, C.N. Dynamic behavior and stabilization of DC microgrids with instantaneous constant-power loads. IEEE Trans. Power Electron. 2011, 26, 822–834. [CrossRef] Lim, J.; Kim, H.; Cho, K.; Bae, J. Stand-alone microgrid inverter controller design for nonlinear, unbalanced load with output transformer. Electronics 2018, 7, 55. [CrossRef] Lei, T.; Wu, C.; Liu, X. Multi-objective optimization control for the aerospace dual-active bridge power converter. Energies 2018, 11, 1168. [CrossRef] Onambele, C.; Elsied, M.; Mpanda Mabwe, A.; El Hajjaji, A. Multi-phase modular drive system: A case study in electrical aircraft applications. Energies 2018, 11, 5. [CrossRef] Burgos, R.; Chen, G.; Wang, F.; Boroyevich, D.; Odendaal, W.G.; Van Wyk, J.D. Reliability-oriented design of three-phase converters for aircraft applications. IEEE Trans. Aerosp. Electron. Syst. 2012, 48, 1249–1263. [CrossRef] Muller, J.-K.; Mertens, A. Power electronics design for a direct-driven turbo compressor used as advanced high-lift system in future aircraft. In Proceedings of the 43rd Annual Conference of the IEEE Industrial Electronics Society, Beijing, China, 29 October–1 November 2017; pp. 1–6. Zhai, L.; Zhang, T.; Cao, Y.; Yang, S.; Kavuma, S.; Feng, H. Conducted EMI prediction and mitigation strategy based on transfer function for a high-low voltage DC-DC converter in electric vehicle. Energies 2018, 11, 1028. [CrossRef] Giglia, G.; Ala, G.; Piazza, M.C.D.; Giaconia, G.C.; Luna, M.; Vitale, G.; Zanchetta, P. Automatic EMI filter design for power electronic converters oriented to high power density. Electronics 2018, 7, 9. [CrossRef] Henke, M.; Narjes, G.; Hoffmann, J.; Wohlers, C.; Urbanek, S.; Heister, C.; Steinbrink, J.; Canders, W.R.; Ponick, B. Challenges and opportunities of very light high-performance electric drives for aviation. Energies 2018, 11, 344. [CrossRef] Diaz, N.L.; Luna, A.C.; Vasquez, J.C.; Guerrero, J.M. Centralized control architecture for coordination of distributed renewable generation and energy storage in islanded AC microgrids. IEEE Trans. Power Electron. 2017, 32, 5202–5213. [CrossRef] Gao, F.; Bozhko, S.; Costabeber, A.; Asher, G.; Wheeler, P. Control design and voltage stability analysis of a droop-controlled electrical power system for more electric aircraft. IEEE Trans. Ind. Electron. 2017, 64, 9271–9281. [CrossRef] Kim, M.; Kwasinski, A. Decentralized hierarchical control of active power distribution nodes. IEEE Trans. Energy Convers. 2014, 24, 934–943. [CrossRef] Guerrero, J.M.; Vasquez, J.C.; Matas, J.; de Vicuna, L.G.; Castilla, M. Hierarchical control of droop-controlled AC and DC microgrids-A general approach toward standardization. IEEE Trans. Ind. Electron. 2011, 58, 158–172. [CrossRef]


40. 41.

42. 43. 44. 45. 46. 47. 48. 49. 50. 51. 52.

53.

54. 55.

56.

19 of 19

Aten, M.; Towers, G.; Whitley, C.; Wheeler, P.; Clare, J.; Bradley, K. Reliability comparison of matrix and other converter topologies. IEEE Trans. Aerosp. Electron. Syst. 2006, 42, 867–875. [CrossRef] Gao, F.; Bozhko, S.; Asher, G.; Wheeler, P.; Patel, C. An improved voltage compensation approach in a droop-controlled DC power system for the more electric aircraft. IEEE Trans. Power Electron. 2016, 31, 7369–7383. [CrossRef] Yassine, M.; Fabris, D. Performance of commercially available supercapacitors. Energies 2017, 10, 1340. [CrossRef] Jayasinghe, S.G.; Meegahapola, L.; Fernando, N.; Jin, Z.; Guerrero, J.M. Review of ship microgrids: System architectures, storage technologies and power quality aspects. Inventions 2017, 2, 4. [CrossRef] Gonzalez-Espasandin, O.; Leo, T.J.; Navarro-Arevalo, E. Fuel cells: A real option for unmanned aerial vehicles propulsion. Sci. World J. 2014, 2014, 497642. [CrossRef] [PubMed] Gao, L.; Liu, Y.; Ren, H.; Guerrero, J.M. A DC microgrid coordinated control strategy based on integrator current-sharing. Energies 2017, 10, 1116. [CrossRef] Liu, Y.; Wang, J.; Li, N.; Fu, Y.; Ji, Y. Enhanced load power sharing accuracy in droop-controlled DC microgrids with both mesh and radial configurations. Energies 2015, 8, 3591–3605. [CrossRef] Hamzeh, M.; Ghafouri, M.; Karimi, H.; Sheshyekani, K.; Guerrero, J.M. Power oscillations damping in DC microgrids. IEEE Trans. Energy Convers. 2016, 31, 970–980. [CrossRef] Fletcher, T.; Thring, R.; Watkinson, M. An energy management strategy to concurrently optimize fuel consumption & PEM fuel cell lifetime in a hybrid vehicle. Int. J. Hydrogen Energy 2016, 41, 21503–21515. Bolborici, V.; Dawson, F.P.; Lian, K.K. Hybrid energy storage systems: Connecting batteries in parallel with ultracapacitors for high higher power density. IEEE Ind. Appl. Mag. 2014, 20, 31–40. [CrossRef] Rausand, R.; Holyland, A. System Reliability Theory: Models, Statistical Methods, and Applications, 2nd ed.; Wiley-Interscience: Hoboken, NJ, USA, 2004. Erickson, R.W.; Maksimovic, D. Fundamentals of Power Electronics, 2nd ed.; Springer: New York, NY, USA, 2001. Augustine, S.; Lakshminarasamma, N.; Mishra, M.K. Control of photovoltaic-based low-voltage DC microgrid system for power sharing with modified droop algorithm. IET Power Electron. 2016, 9, 1132–1143. [CrossRef] Lu, X.; Guerrero, J.M.; Sun, K.; Vasquez, J.C. An improved droop control for DC microgrids based on low bandwidth communication with DC bus voltage restoration and enhanced current sharing accuracy. IEEE Trans. Power Electron. 2014, 29, 1800–1812. [CrossRef] Paquette, A.D.; Reno, M.J.; Harley, R.G.; Divan, D.M. Sharing transient loads: Causes of unequal transient load sharing in islanded microgrid operation. IEEE Ind. Appl. Mag. 2014, 20, 23–34. [CrossRef] Gadoura, I.; Grigore, V.; Hatonen, J.; Kyyra, J.; Vallittu, P.; Suntio, T. Stabilizing a telecom power supply feeding a constant power load. In Proceedings of the 20th International Telecommunications Energy Conference (INTELEC), San Francisco, CA, USA, 4–8 October 1998; pp. 243–248. Aircraft Electric Power Characteristics. Available online: http://u.dianyuan.com/upload/space/2012/04/ 06/1333690328-628697.pdf (accessed on 16 July 2018). © 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).