Decentralized Controller for Energy Storage

Management on MVDC Ship Power System with

Pulsed Loads
Samy Faddel, Student Member IEEE, Tarek A. Youssef, Member IEEE and Osama Mohammed, Fellow, IEEE
Energy Systems Research Laboratory, Florida International University, Miami, Florida USA
mohammed@fiu.edu.

Abstract— Medium voltage DC systems (MVDC) are gaining generators commonly used in AES are not going to play a
more interest in ship power systems. Although, MVDC system has significant role in feeding pulsed loads that requires power in
many advantages for the ship power grid, it is challenging to the range of 1 second [3]. Therefore, proper management and
ensure load-generation balance in the presence of pulsed loads. control is needed to ensure load-generation balance and keep
Therefore, proper energy management algorithm of the energy
the MVDC bus voltage within the standards.
storage devices connected to the system should be developed. This
paper proposes an automatic decentralized controller for the Modelling of shipboard MVDC for dynamic analysis was
energy storage devices connected to the MVDC power system. considered in [4]. The focus of this work was to model the
The controller ensures load-generation balance, maintain the stator transient dynamics of the generators. In [5] source-load
MVDC bus constant and ensures proper power sharing among dynamic interactions were considered. The authors tried to
the storage devices. The simulation results, using MATLAB, consider the behavior of the constant power loads (CPLs) and
prove the adequacy of the proposed controller. their interaction with the generator inductances, dc bus
capacitors and cable effects. In both papers, energy storage
Index Terms— Ship power system, decentralized control, MVDC, devices and their control were not considered. Modelling and
pulsed loads, Energy Storage.
control of modular multi-level DC-DC converter in the
I. INTRODUCTION presence of energy storage for MVDC ship power system was
considered in [6]. A controller based on fundamental period
Next generation of ship power system is adopting more averaging and phase shifting technique was proposed. The
electrical energy that increases complexity of the supply and focus of the controller was to control the different submodules
the control process of the isolated power system. This is mainly of the converter under different operating conditions.
driven by the increasing electrical demand and the nature of However, the output voltage is not constant during the
anticipated new types of loads such as electromagnetic aircraft overloading condition due to the absence of sufficient
launch system (EMALS) [1]. This kind of loads draw generation.
intermittent pulses of power from the system [2]. Due to the Energy storage system with isolated modular multilevel
need of high power supply and flexibility in All Electric Ship DC-DC converter was considered in [7], [8]. The authors
(AES) power system, medium voltage direct current (MVDC) developed centralized control strategy for controlling the
systems are going to be a viable option. MVDC power system modular converter to achieve fault current limiting and fault
has multiple advantages against the MVAC system. These ride through using the converter cell capacitor. However, no
advantages include: 1. the replacement of bulky transformers consideration for the pulsed loads on the ship power system.
with the compact power electronic converters. 2. increased fuel Battery energy management was not considered as well.
efficiency of the generators and elimination of the Centralized controller for energy storage management for AES
synchronization problems. 3. Reducing the risk of systematic power systems with pulsed loads was considered in [9], [10].
disintegration while supporting the emerging pulsed loads [2]. In [9], the author proposed a PI controller based energy
Due to the nature of future pulsed loads on AES power storage management to handle the charging and discharging of
systems, energy storage devices are going to play a key role in a hybrid storage system that consists of a supercapacitor and a
the ship power systems. This comes from the fact that the battery. The energy management is based on a centralized
capacity of the generators for the AES power system is controller that requires the knowledge of the generators and the
designed to feed the loads that are continuously connected to load currents. In [10], Fuzzy logic controller was proposed to
the system. Otherwise, more cost is included to oversize the provide energy management of the hybrid energy storage.
generators. Also, due to its slow response, gas driven However, it still requires the knowledge of the generator and

978-1-5386-3048-8/18/$31.00 ©2018 IEEE 254

Authorized licensed use limited to: Renmin University. Downloaded on May 10,2020 at 06:54:34 UTC from IEEE Xplore. Restrictions apply.
load current. In [11], hierarchical control algorithm that MVDC
bus
ensures that the genset, in MVDC power system, to work in
their most efficient operational points was proposed. It is based on Radar
cooperative asymmetrical droop method to coordinate the G Load
Gas
charging among the hybrid energy storage system and the gensets. Turbine Main AC
However, the problem of supplying big loads within short periods Generator
Port
of time (pulsed loads) was not considered. Controllers based on M Propulsion
virtual impedance droop controller for low voltage DC microgrids Motor
with renewable energy was proposed [12], [13]. The authors
proposed a decentralized control algorithm that can change from LV DC
constant control mode to voltage control mode autonomously and Load
ensures good power sharing between batteries in the energy
storage system. Although the algorithm showed a good LV AC
performance, it is hard to be applied in the MVDC ship power Load
LVDC
system due to the slow response of the gas-driven generators. The
Battery
same applies for the use of supercapacitors in [14]. Banks
Starboard
Due to the anticipated use of MVDC and new types of pulsed M Propulsion
loads on AES, there is a need for automatic control algorithms Motor
that should provide smooth insertion and removal of power G
Pulsed
sources and sharing of loads as desired [2]. Gas Auxiliary Load
Turbine
In this paper, an automatic decentralized controller for fair AC
Generator
power sharing among batteries on MVDC ship power system is
proposed. The controller requires no communication or Fig. 1 Notional MVDC system
knowledge of the generator and load currents. It can ensure load-
generation balance for normal operating conditions and during TABLE 1
feeding the pulsed loads. It also ensures that the MVDC bus MVDC SYSTEM PARAMETERS
voltage is within the IEC 60092-101 standards [2]. The Type Quantity Power (MW)
controller is based on a two function controller that uses the Main Generators 2 36
Auxiliary Generators 2 4
concept of virtual impedance controller [12] and the Propulsion Motors 4 32
exponential SOC controller [15]. To avoid unnecessary Radar System 1 3
discharging of the batteries, the controller uses state machine Service Loads - 5
method to ensure efficient use of the energy storage units. Pulsed Loads 1 2

II. NOTIONAL MVDC SYSTEM DESCRIPTION Due to the large power required to feed the pulsed load, battery
The MVDC ship power system is shown in Fig. 1. The banks will be used. However, to maintain the MVDC bus
system parameters are shown in Table I. To meet the total voltage within the standards and to increase the life time of the
installed demand of the loads, two large capacity “main” batteries, proper energy management of the batteries is
generator sets (e.g., 36 MW) can be supplemented with two or required.
more small capacity “auxiliary” generator sets (e.g., 4 MW) III. CONTROLLER DESIGN
[2]. The generators are connected to a controlled rectifier. This
allows more fuel efficiency since the generators are not To ensure adequate operation of the MVDC AES in the
obligated to operate at a fixed speed anymore. The ship is presence of large pulsed loads, there is a need for proper
driven using a propulsion system that uses induction motors. management of the batteries.
The propulsion system represents 80% of the total ship power The energy management system should ensure load-
system loads [5]. The radar system represents a standalone generation balance and avoid unnecessary
load that draws around 3 MW in its steady state operation. Ship discharging/charging of the batteries to increase the life time
service loads are supplied from the MVDC through DC/DC or of the system. Therefore, there is a need for automatic
DC/AC converters. Pulsed Loads represent load center that decentralized control algorithm that should provide smooth
draws intermittent pulses from the system. It draws power in insertion and removal of the batteries [2]. Decentralized
the range of 2 MW within one second. controllers usually have the ability to provide fast response and
The generators on the ship power systems are designed to they are less expensive than the centralized ones.
supply the continuous loads that are connected to the system. Due to the nature of the decentralized controller, it does not
Also, the response time of the gas driven generators is slow. know about the capability of the generators. For example,
Therefore, sudden load additions or rejections to the MVDC when a large load is added to the system, it will cause a
caused by step changes coming from the pulsed loads are met momentary voltage drop on the MVDC bus. This may prompt
by energy storage devices. the batteries to start discharging regardless of the fact that the
generators can supply this added load. Therefore, the control
algorithm should satisfy the following condition:
1. Ensures load-generation balance.

255

Authorized licensed use limited to: Renmin University. Downloaded on May 10,2020 at 06:54:34 UTC from IEEE Xplore. Restrictions apply.
 2. Ensures proper power sharing among the batteries. Fig. 4 (a) Virtual impedance droop controller (b) Exponential controller
3. Avoids unnecessary discharging/charging of the deciding on the current reference value of the controller. This
batteries. It is only when there is a deficit/surplus, the controller is shown in Fig 3. The current reference value of the
batteries will be used. controller IDE is coming from two parts: The first one is based
The proposed controller that satisfies these requirements is on the value coming from the droop part shown in Fig. 4(a).
shown in Fig. 2. It consists of an outer droop-exponential During the discharging of the batteries, the reference current
controller that tries to ensure load-generation balance and will be managed partially by the droop controller. Therefore,
equal power sharing among the batteries. The droop- the MVDC bus voltage is given by:
exponential controller is followed by a state machine logic that
takes the reference current from the controller IDE and decides = − (1)
if this value will be passed to the PI controller or it will be
manipulated to avoid unnecessary discharging/ charging. Once where Rdc is the virtual resistance at each droop controller loop,
the final reference current Iref is obtained, it will be compared ID is the portion of the output current that is coming from the
to the converter current and the error will be passed to a PI droop control part. Vdc is the voltage at the common MVDC
controller that will force the converter to follow the reference bus and Vref is the voltage reference for the MVDC bus.
value of the current. The details of the various parts of the To take into consideration the state of charge of the battery, the
controller are as follows: battery current is measured and the state of charge of the
battery is estimated according to relation (2):
A. Droop Exponential Controller (2)
= (0) −
This part of the controller is a combination of the virtual
resistance droop control that is used for equal power sharing in
the DC micro-grids and the exponential controller that tries to where SOC(0) is the initial state of charge of the battery, Ibat is
take into consideration the state of charge of the battery when the battery current and Cbat is the battery capacity. Once the
SOC is estimated, it will be used in the second exponential part
of the controller. It is desired that the battery with the highest
DC/DC
Battery MVDC SOC is discharged faster than the others to ensure the balance
converter Bus among the batteries and increase the life time of the overall the
Ibat
storage system. In case of charging, it is required that the
SOC
State Machine battery with the lowest SOC to be charged faster than the
Logic
Estimator PWM others. Therefore, the controller will decide on part of the IDE
SOC Charge I
ref current based on the SOC based on the following equation:
Droop- I DE PI
V dc Exponential
I ∙
Controller
Discharge
L
= . (3)
Delay
Del_V
where exp(.) stands for the exponential function and SOCpu,i
Fig. 2 Schematic diagram of the proposed controller = SOCi/Cbat. This relation is show in Fig 4(b). This relation
will bias the effective discharging rate toward the highest
charged battery. In case of battery charging, this current will
Ir be manipulated by the state machine logic.
V dc
The final reference current coming from the droop-
Ir − V dc exponential controller is as follows:
Irbase I = ∙ ∙ (4)
×

DE

Ir
SOC Where K is a constant value that increases/decreases the
reference value based on the battery type and rating. ID is the
Ir − SOC
part of the controller current that is coming from the droop
relation and IE is the part of the controller current that is coming
Fig. 3 Droop Exponential Controller
from the exponential relation.
Once the reference current IDE is obtained, this value will be
I D Vdc=Vref-ID Rdc I
E
passed to the state machine logic that will generate the final
reference value Iref to the PI controller.

B. State Machine Logic

The state machine logic is responsible of generating the final
charge/discharge reference current Iref to the PI controller. It
(a) V dc (b) SOC is shown in Fig. 5.

256

Authorized licensed use limited to: Renmin University. Downloaded on May 10,2020 at 06:54:34 UTC from IEEE Xplore. Restrictions apply.
 Initialization

(Soc>=20) Discharging (Boost mode):

Flag2=dV If SOC<20 Iref=0
Else Iref =IDE

(Soc<20)
((abs(Flag1)< abs(del_V))
Discharge_decrementing && Soc>20)
Flag1=del_V

charge_incrementing
(abs(Flag2)<
abs(del_V)) Charging (Buck mode):
Decrement Discharging:
If SOC>80 Iref=0
Iref =Iref -(0.1*Iref)
Else Iref=Iref+((80-SOC)*Imax)
(abs(Flag2)> (delay; (abs(Flag1)> (delay;
abs(del_V)) abs(Iref)<Imax&&Del_V abs(Del_V)) Iref<Imax&&Del_V<Vth)
<Vth)
Hold on Hold on

Fig. 5 State machine Logic

IV. RESULTS
The inputs to the state machine logic are the discharge
To validate the proposed controller, it was implemented and
reference from the droop-exponential controller, the change in
tested in the notional MVDC system given in Fig. 1. To test the
the voltage of the MVDC bus (Vt - Vt-1) and the SOC of the
ability of the controller for proper power sharing, it is assumed
battery. The main function of the energy storage is to maintain
that the storage system consists of two large storage batteries.
the MVDC bus constant and to ensure load-generation balance.
Each one of them has a rated capacity of 800 AH and nominal
Mainly, the energy storage system will be used when there is a
voltage of 800 V. The validation is performed under different
deficit in the generation, especially while feeding the
loading conditions and different state of charge of the batteries.
intermittent pulsed loads. The charging of the energy storage
devices will occur when there is a surplus of energy that will
be detected by the change of the bus voltage. A. Controller Performance With Equal SOCs of the Batteries
Once there is a change in the MVDC bus voltage, the state First the controller performance is tested when the two
machine logic will be activated and receive the discharge batteries have equal state of charge of 50% of their capacities.
reference current IDE. To avoid unnecessary discharging of the Fig. 6 (b) shows the loading connection/disconnection process
batteries, the reference current IDE will go to the discharge where there is a propulsion system load of 6400A is connected
decrementing block which will reduce the reference current to the system at the beginning. The system continues to start
and wait for few microseconds. The change in voltage will be operation. At t=0.7 sec, service loads of 1000A are added to
detected after that waiting time (the delay). If the voltage has the system. It is worth mentioning that the generators can feed
decreased by decreasing the discharge reference current, this loads up to 8000A when the generators reach their maximum
indicates that the system needs the support from the batteries. capacities. Therefore, as long as the current is less than 8000A,
Therefore, the controller will stop decrementing the the batteries should not supply any current. At t=1.5 sec, radar
discharging current. In case of the voltage did not change or it system load of 600A is added to the system. From 1.5 to 2
increased, this means that the generators are supporting this seconds, the generators are working at their rated power. Fig.7
extra load. So, the discharging current will be decremented till and Fig. 8 show that during this period (0-2 seconds), the
it reaches zero. batteries are idle. At t=2 sec, a pulsed load of (400A) is added
If there is an increase in the MVDC voltage, the state for one second. Since the generators are already running at
machine will switch to the charging mode where it will their rated power, the batteries should supply the extra load to
increase the charging reference current through the charge ensure load-generation balance. This is shown in Fig. 7(a)
incrementing block. This block will increase the charging where the two batteries start to discharge. This also confirmed
current and check the change of the MVDC voltage after each by the decrease in the SOCs in Fig. 8. Since both batteries have
increment to ensure that the charging of the batteries does not the same initial SOC of charge, Fig 7(a) shows that both
negatively impact the MVDC bus. batteries feed the same amount of current. This ensures equal
In case the SOC of the battery is below 20%, this battery will load sharing among the batteries and increases the system life
not participate in the discharge process. Similarly, if the SOC time. It is worth mentioning that high current (1000A) is drawn
of the battery is higher than 80%, this battery will not from the batteries because there are connected at the low
participate in the charging process. voltage side of the converter and the discharge of the battery is
associated by a decrease in the battery voltage as shown in Fig
7(b). After the pulsed load is disconnected at t=3 sec, the

257

Authorized licensed use limited to: Renmin University. Downloaded on May 10,2020 at 06:54:34 UTC from IEEE Xplore. Restrictions apply.
batteries are smoothly disconnected as well since the
generators can feed the existing loads. At t=3.5 sec, another

Voltage (V)
load is disconnected which means that there is a surplus of
energy that can be used to charge the batteries. Therefore, the
batteries are smoothly connected again but at the charging
mode in this case. Fig. 7(a) shows that the current of the
batteries becomes negative (which means charging) after t =3.5
sec. Both batteries charge with the same current which is
desirable. Fig 7(b) shows that the voltages of the batteries
increase after t= 3.5 sec. The charging process is confirmed by

Current (A)
the increase in the SOC as depicted in Fig. 8. Finally, Fig. 6(a)
shows that regardless of the different loading condition on the
ship power system, the MVDC bus voltage is kept constant and
within the standards.

B. Controller Performance With Different SOCs of the

Batteries Fig. 6 Case of equal SOCs
a) The MVDC bus voltage. b) Total load current.
In this test, the same loading conditions are applied but the
initial SOC of the two batteries are different where battery1 has
higher initial SOC of 75% of its capacity while battery 2 has
Current (A)

only 45% of its capacity. Fig. 9(a) shows that the MVDC bus
voltage remains constant with the different loading conditions.
At t= 2 sec, when the pulsed load is connected and the
generators can no longer support this extra load, the two
batteries start to discharge to maintain load-generation
balance. Since battery 1 has higher SOC, its contribution in
supporting the system is higher as shown in Fig. 10(a) where
Voltage (V)

the discharge current of battery 1 is 1500A while the discharge

current of battery 2 is 400A. This is also shown by the
difference in the drop of the battery voltage in Fig 10(b) where
the drop of the voltage is higher for battery 1. The decrease in
the SOC of the batteries during the period 2-3 second is shown
in Fig. 11. At t=3 sec, the pulsed load is disconnected.
Therefore, the batteries are disconnected as well. At t=3.5 sec,
Fig. 7 Case of equal SOCs
when another load is disconnected, both batteries start to a) Battery current. b) Battery voltage.
charge. However, because battery 2 has lower SOC, its
charging current is higher than that of battery 1 which is
desirable. Fig. 11 show the increase of the SOC of the batteries
SOC (%)

after t=3.5 sec.

V. CONCLUSION
In this paper, an automated decentralized controller for
energy storage management for ship power system with pulsed
loads is proposed. The controller is based on a combination of
virtual impedance droop control and SOC exponential control
SOC (%)

to support the system and ensure proper power sharing. State

machine logic was used to avoid unnecessary discharging of the
batteries. The controller was tested in Simulink under different
loading conditions and different SOCs of the batteries. The
results showed that the controller ensures load-generation
balance and the voltage of the MVDC is kept constant. It also
ensures proper power sharing among the batteries which
increase the system life time. Fig. 8 Case of equal SOCs
a) SOC of battery1. b) SOC of battery 2.

258

Authorized licensed use limited to: Renmin University. Downloaded on May 10,2020 at 06:54:34 UTC from IEEE Xplore. Restrictions apply.
 REFERENCES
[1] Z. Jin, G. Sulligoi, R. Cuzner, L. Meng, J. C. Vasquez, and J. M.
Voltage (V)

Guerrero, “Next-Generation Shipboard DC Power System:

Introduction Smart Grid and dc Microgrid Technologies into Maritime
Electrical Netowrks,” IEEE Electrification Mag., vol. 4, no. 2, pp. 45–
57, Jun. 2016.
[2] Institute of Electrical and Electronics Engineers and IEEE-SA
Standards Board, IEEE recommended practice for 1 kV to 35 kV
medium-voltage DC power systems on ships. New York: Institute of
Electrical and Electronics Engineers, 2010.
[3] S. Pish, J. Herbst, D. Wardell, A. Gattozzi, and M. Flynn, “Power
Current (A)

management and energy storage experiments on a MW-scale naval

power system test-bed,” in Electric Ship Technologies Symposium
(ESTS), 2015 IEEE, 2015, pp. 453–458.
[4] J. Shi, R. Amgai, and S. Abdelwahed, “Modelling of shipboard
medium-voltage direct current system for system level dynamic
analysis,” IET Electr. Syst. Transp., vol. 5, no. 4, pp. 156–165, Dec.
2015.
[5] U. Javaid, F. D. Freijedo, D. Dujic, and W. van der Merwe, “Dynamic
Assessment of Source–Load Interactions in Marine MVDC
Fig. 9 Case of different SOCs Distribution,” IEEE Trans. Ind. Electron., vol. 64, no. 6, pp. 4372–
a) The MVDC bus voltage. b) Total load current. 4381, Jun. 2017.
[6] Y. Chen, S. Zhao, Z. Li, X. Wei, and Y. Kang, “Modeling and Control
of the Isolated DC–DC Modular Multilevel Converter for Electric Ship
Current (A)

Medium Voltage Direct Current Power System,” IEEE J. Emerg. Sel.

Top. Power Electron., vol. 5, no. 1, pp. 124–139, Mar. 2017.
[7] R. Mo and H. Li, “Hybrid Energy Storage System With Active Filter
Function for Shipboard MVDC System Applications Based on
Isolated Modular Multilevel DC/DC Converter,” IEEE J. Emerg. Sel.
Top. Power Electron., vol. 5, no. 1, pp. 79–87, Mar. 2017.
[8] R. Mo, M. Steurer, and H. Li, “Controller hardware-in-the-loop
(CHIL) simulation of a multi-functional energy storage system based
on modular multilevel DC/DC converter (M2DC) for MVDC grid,” in
Power Electronics and Motion Control Conference (IPEMC-ECCE
Voltage (V)

Asia), 2016 IEEE 8th International, 2016, pp. 1980–1984.

[9] M. M. S. Khan and M. O. Faruque, “Management of hybrid energy
storage systems for MVDC power system of all electric ship,” in North
American Power Symposium (NAPS), 2016, 2016, pp. 1–6.
[10] M. M. S. Khan, M. O. Faruque, and A. Newaz, “Fuzzy Logic Based
Energy Storage Management System for MVDC Power System of All
Electric Ship,” IEEE Trans. Energy Convers., vol. 32, no. 2, pp. 798–
809, Jun. 2017.
Fig. 10 Case of different SOCs [11] Z. Jin, L. Meng, J. C. Vasquez, and J. M. Guerrero, “Frequency-
a) Battery current. b) Battery voltage. division power sharing and hierarchical control design for DC
shipboard microgrids with hybrid energy storage systems,” in Applied
Power Electronics Conference and Exposition (APEC), 2017 IEEE,
2017, pp. 3661–3668.
SOC (%)

[12] N. L. Diaz, D. Wu, T. Dragicevic, J. C. Vasquez, and J. M. Guerrero,

“Fuzzy droop control loops adjustment for stored energy balance in
distributed energy storage system,” 2015, pp. 728–735.
[13] N. L. Diaz, T. Dragicevic, J. C. Vasquez, and J. M. Guerrero,
“Intelligent Distributed Generation and Storage Units for DC
Microgrids&#x2014;A New Concept on Cooperative Control Without
Communications Beyond Droop Control,” IEEE Trans. Smart Grid,
vol. 5, no. 5, pp. 2476–2485, Sep. 2014.
[14] H. F. Habib, A. A. S. Mohamed, M. El Hariri, and O. A. Mohammed,
“Utilizing supercapacitors for resiliency enhancements and adaptive
SOC (%)

microgrid protection against communication failures,” Electr. Power

Syst. Res., vol. 145, pp. 223–233, Apr. 2017.
[15] S. Faddel, A. T. Al-Awami, and M. A. Abido, “Real time digital
simulation of voltage-based controller for electric vehicle charging,”
in 2016 Clemson University Power Systems Conference (PSC), 2016,
pp. 1–5.

Fig. 11 Case of different SOCs

a) SOC of battery1. b) SOC of battery 2.

259

Authorized licensed use limited to: Renmin University. Downloaded on May 10,2020 at 06:54:34 UTC from IEEE Xplore. Restrictions apply.