Hybrid energy storage management in ship power

Electric Power Systems Research 141 (2016) 50–62

Contents lists available at ScienceDirect

Electric Power Systems Research journal homepage: www.elsevier.com/locate/epsr

Hybrid energy storage management in ship power systems with multiple pulsed loads Christopher R. Lashway, Ahmed T. Elsayed, Osama A. Mohammed ∗ Energy Systems Research Laboratory, Department of Electrical and Computer Engineering, Florida International University, Miami, FL 33174, USA

a r t i c l e

i n f o

Article history: Received 6 April 2016 Accepted 16 June 2016 Keywords: Hybrid energy storage Batteries Capacitive energy storage Ship power systems Pulsed loads

a b s t r a c t As various types of energy storage (ES) types continue to penetrate grid, electric vehicle, and Naval applications, a need arises in extending traditional analysis to cover the revised performance metrics associated with a hybrid energy storage system (HESS). Each ES device has its own respective power density, energy density, response time, and voltage stability under load. In some critical applications, such as ship power systems (SPS), it is recommended to combine two or more ES types to overcome the impediments of the other. In this paper, three different series-configured HESS are mathematically modeled, evaluated, and tested experimentally. Lead acid and lithium ion batteries as well as supercapacitor equivalent circuit models are defined as components for each mixed HESS configuration. The impulse response to a constant and pulsed load was used to evaluate each ES model. The charging of mixed ES technologies was then accomplished using a special controller to handle the unique charging constraints of each ES module. Moreover, this same controller was used to apply a “rolling charging” algorithm to extend the operating time of the HESS. The validity of the derived model and controller were validated experimentally through a hardware setup simulating a multi-pulsed load SPS profile. © 2016 Elsevier B.V. All rights reserved.

1. Introduction Research into energy storage (ES) systems continues to flourish to support the future microgrid infrastructure. To serve an everchanging fluctuation in the consumer demand, the grid must rely on the inclusion from a variety of ES sources. The most common solution, an electrochemical battery, can be utilized for a wide range of different tasks including restoring system voltage and frequency following an outage [1–4]. In a utility grid, a wide range of ES can be deployed due to reduced concerns over weight and space. Mobile applications, however, do not have this luxury. The localized microgrid present on a ship, aircraft, or electric vehicle (EV) is susceptible to major operational and logistic challenges. Heavy and frequent pulsed loads, which may present a minimal disturbance to the utility-connected system, can prove to be catastrophic when generation resources are limited. Without the aid of carefully selected ES, the energy must either be available from generators ondemand, or ES units must be prepared and deployed effectively in anticipation of the disturbance.

∗ Corresponding author. Tel.: +1 305 348 3040; fax: +1 305 348 3707. E-mail address: mohammed@fiu.edu (O.A. Mohammed). http://dx.doi.org/10.1016/j.epsr.2016.06.031 0378-7796/© 2016 Elsevier B.V. All rights reserved.

A Naval ship power system (SPS) is composed of a complex isolated power system, typically consisting of 2 main turbine generators (MTG) and 2 auxiliary turbine generators (ATG) [5]. For example, the upcoming DDG1000 Destroyer all electric ship contains 74.8 MW of onboard total shaft power. Critical loads reserve approximately 15% of the available energy, but the next generation of equipment will introduce loads several magnitudes greater than this figure [6]. Energy and power requirements can vary from 100 kW to 10 GW over microseconds to seconds [7,8]. Without proper selection and control, ES units may experience high depth of discharge (DoD) which would reduce their capability of responding quickly to fluctuating demands while significantly reducing their lifespans, or state of health (SoH) [9]. In smart grid applications, ES deployment and control has recently gained increased attention [10]. These cases have been two-fold; providing a method to reduce the intermittency associated with renewable energy sources while offering ancillary backup services. The grid-connected hybrid system in [11] demonstrated a combination of the zinc bromide flow battery (FB) and electrochemical supercapacitors (SC) to reduce the voltage and frequency instabilities as a result of variable wind generation. Several vignettes were tested varying the size of the parallel SC bank where the SC handled short variations and the FB handled

C.R. Lashway et al. / Electric Power Systems Research 141 (2016) 50–62

longer variations. Just as ES has been utilized to handle some of the intermittencies associated with generation, their inclusion has been equally as useful in supporting pulsed loads. Pulsed loads are commonplace in military applications, but are present in a wide range of other applications and fields [12]. The aforementioned grid-connected systems can reduce the impacts following a major disturbance or a variance in generation but under islanded or stand-alone cases, system stability will rely solely on the support of ES when MTGs and ATGs reach their generation capacities. In [13], a battery management system scheme is demonstrated to control the power flow in a lithium ion based battery array. The system is tested under both gridconnected and islanded modes of operation. In islanded mode, a battery with an inverter acts as a synchronous generator providing voltage and frequency support. A number of other control strategies have been demonstrated, but have only focused on one type of ES [14–17]. A SPS presents unique challenges in terms of ES deployment, since they are inherently islanded systems. Pulsed load management and mitigation is an emerging topic in the future all-electric SPS. In [18], a 0.25 Hz 36 MW pulsed load is tested on a notional SPS model where case studies were conducted over the use of a dynamic reactive compensator to maintain bus voltages. However, power demands of multiple pulsed loads present a major challenge in terms of design and implementation. The electromagnetic (EM) railgun and EM catapult were investigated in [19] where short-term pulsed loads were tested, both significantly exceeding the available energy from the MTGs when tested independently. ES was proposed as a solution to support both, but was not demonstrated. An extensive review into the impact of multiple pulsed loads on the electric SPS was performed in [20]. EM railgun and free electron laser firing profiles were tested as connected pulsed loads without electrochemical ES, but employ the railgun launcher rotor as flywheel ES. The system proved the current infrastructure could support at least one important pulsed load, but not both. Investigations have been performed into deploying electrochemical ES as well [21]. In [22], a SC was tested independently with an EM railgun to fill an 800 kA firing pulse. The topology was capable of supplying the pulse but required an enormous 500 F SC. Combinations of ES have offered more realistic solutions [23–26]. In [26], lead acid and sodium sulfide battery banks are simulated in parallel on a SPS to fill a single pulsed load. Each ES bank was installed on a different zonal bus where it was noted that the ES units were able to respond quicker than the MTG to deliver energy. However, hybrid energy storage system (HESS) support has not yet been evaluated on the same bus or in a series configuration. A control topology for a SPS is proposed in [27] where a parallel-configured battery-SC HESS was simulated with respect to a constant and pulsed load. Four operation modes were tested to meet critical and pulsed load demands, but only the voltage recovery period following the pulse was discussed and no investigation was provided into the SC or battery performance. Furthermore, the battery type was not identified. Typical battery and SC HESS have utilized parallel topologies, however, control of these systems is challenging as a result of the wide voltage operating range of the SC. Without a specialized interfacing converter, the SC terminal voltage would follow that of the battery leaving a significant amount of unutilized energy due to a narrowed operating range [28]. Moreover, a mismatch in the equivalent series resistance of each ES would result in unequal, uncontrolled charging or induce internally circulating currents, a phenomenon which parallel-configured lead acid and lithium ion batteries would also be prone to. In [29], a supervisory energy management controller was developed to effectively split EV load demand between a lithium ion battery and SC in

51

a parallel-configured HESS. A multiobjective optimization procedure accounting for both the battery and SC equivalent models and converter topology was solved using dynamic programming. Using these results, a neural network was trained and deployed on the controller with objectives to preserve the battery SoH and enhance total HESS efficiency. Although literature has demonstrated the impacts of pulsed loads on SPS, it has been limited to testing of each pulsed load independently. In practicality, a robust system should have the capability to handle multiple pulsed loads under the same period. Multiple pulsed loads can be seen under a multitude of applications from manufacturing facilities to EVs, but for this focus, this will be realized under a SPS. To the best of the authors’ knowledge, serving multiple pulsed loads on the SPS has not yet been tested and analyzed. In order to overcome this challenge, two novel concepts have been established in this paper. First, several series-configured HESS combinations are proposed and tested through utilizing lead acid and lithium ion batteries as well as a SC bank. The performance of each combination is analyzed. Second, following the selection of each series-connected ES, a specialized dispatch control scheme is demonstrated in an effort to replenish some or all of the energy required to serve one of the pulsed loads considering the SoH trade-offs. Coined as “rolling charging,” a coordination scheme between the load and charging is applied to the heaviest pulsed load in an effort to recover a portion of the discharged energy. The dynamics of each ES is optimized with respect to their operational constraints as well as best practices to preserve their SoH. The remainder of this paper is organized as follows: Section II discusses the concept of multi-chemistry ES and discusses the mathematical models for each type used in this study. Section III discusses the equivalent model and theory of integrating HESS. Section IV describes the proposed coordinated control and rolling charging algorithm in detail. The hardware implementation and experimental results are presented in Section V and Section VI concludes this work. 2. Modeling of multiple energy storage types The following sections describe the model of each electrochemical ES type in-detail. The operational characteristics of each ES play a pivotal role in improving the base case: a traditional seriesconnected lead acid battery system. In order to demonstrate the limitations of each ES, the performance and operational constraints of each are briefly discussed. 2.1. Lead acid batteries The lead acid battery can provide seamless, inexpensive energy to serve a load but suffer from a number of drawbacks. First, their lifespan is heavily dependent on the operational current. Numerous experimental charts have been produced recommending the level of discharge current versus that of the nameplate battery capacity to remain close to the 20-h discharge rate (or C/20), but this can be unfeasible when designated as a primary battery. Second, their lifespan is governed by the operational DoD. Consistent deep discharging of the lead acid battery will exponentially reduce its SoH. Third, their response time to an energy demand is slow as a result of a large double layer capacitance, making it inefficient in high frequency pulsed load applications [30]. The lead acid battery can be represented through a model consisting of two parts. The first part models the energy storage portion through the application of a very large capacitor in

52


2.15

4.2 4

2.05

3.8

2

3.6

1.95

3.4

1.9

3.2

1.85

3

Lead Acid Cell Lithium Ion Cell

1.8 1.75 0

10

20

30

40

50

60

70

80

90

VocLi(t,I,SoC,T)

VocPb(t,I,SoC,T)

2.1

2.8 2.6 100

SoC (%) Fig. 1. Open circuit voltage vs. SoC for lead acid and lithium ion batteries.

parallel with a self-discharge resistor. Fig. 1 provides a graphic representation of the relationship between the open circuit voltage VocPb (t, I, SoC, T ) and SoC for lead acid while also depicting the same relation for lithium ion. The second portion models the unique response for each ES. Fig. 2 shows the equivalent circuits for the three ES types under study where the aforementioned energy model is represented by a nonlinear voltage source. Fig. 2(a) depicts the equivalent circuit model for the lead acid battery. To represent its unique dynamic responses, Electrochemical Impedance Spectroscopy (EIS) lumped parameters are defined based on the Randles circuit. The Randles circuit model depicts the ohmic resistance of the sulfuric acid electrolyte RtPb where electrochemical kinetics of the impulse response are modeled by the polarization resistance RpPb and capacitance CpPb . The transfer function depicting the lead acid battery equivalent impedance is: ZeqPb (s) = RtPb +

RpPb

(1)

CpPb RpPb s + 1

The voltage response VPb (t) at the battery terminals is: VPb (t) = VocPb (t) −

1 CpPb

[e−t/CpPb RpPb + CpPb RtPb ı(t)]I(t)

Rp Li

VocLi[t,SoC,I,T] +

Rt Pb

Rt Li Cp Pb

CpLi

(a)

(b)

Ri SC

Ro SC + Vt SC -

Co SC

RpLi

CpLi RpLi s + 1

(3)

where RtLi is the ohmic resistance of the lithium salt electrolyte and very different polarization resistance RpLi and capacitance CpLi values are used. The voltage response VLi (t) at the battery terminals is: VLi (t) = VocLi (t) −

1 [e−t/CpLi RpLi + CpLi RtLi ı(t)]I(t) CpLi

(4)

2.3. Supercapacitor

One way to begin mitigating the operational risk and battery response time associated with lead acid batteries is to introduce other chemistries with faster reacting times. The lithium ion battery provides an improvement as it is not as sensitive to deep DoD and high discharge currents. Once again, a two part model differentiates

Rp Pb

ZeqLi (s) = RtLi +

(2)

2.2. Lithium ion batteries

Voc Pb[t,SoC,I,T] +

the energy storage and equivalent circuit portions. The energy storage portion is modeled similarly to that of the lead acid except that the behavior of the voltage is very different. Fig. 1 depicts the nonlinear lithium ion open circuit voltage VocLi (t, I, SoC, T ) versus the SoC. The lithium ion equivalent circuit is shown in Fig. 2(b) following the same 1st order Randles circuit form, however, its dynamics reveal drastically different values in terms of its dependence on current and SoC. The transfer function depicting the equivalent impedance is:

Ci SC (c)

Fig. 2. Equivalent circuit models for (a) lead acid (Pb) battery, (b) lithium ion (Li) battery, and (c) supercapacitor (SC).

A SC provides much greater charge storage versus a traditional capacitor due to a highly amplified surface area. Composed of two porous electrodes divided by a separator soaked in a solvent electrolyte [31], no electrochemical reaction takes place thus charge acceptance and delivery is much faster than a battery with a cycle life of over 500,000 cycles even under heavy operation. The drawback, however, is an operating voltage range, from its maximum rated voltage to 0 V. Furthermore, its energy density is significantly less than that of a battery. Despite the SC operating as an electrochemical storage device, its equivalent circuit model does not follow the same form as that of the batteries. An extension beyond the common capacitor and equivalent series resistor is based on a physics-based representation partitioning the SC into parts associated outside and inside of the core material (Fig. 2(c)). In [32], four different SC equivalent models were identified from 2nd to 5th order where the operation frequency dictates the order. Since the pulsed loads in this study are of low frequency, a 2nd order model is sufficient to model the steady-state and transient voltage fluctuation where RoSC and CoSC represent the resistance and capacitance outside the electrode pore and RiSC and CiSC represent the resistance and capacitance inside. The transfer function depicting the SC equivalent impedance is: ZeqSC (s) = RoSC +

RiSC CiSC s + 1 CoSC CiSC RiSC s2 + (CoSC + CiSC )s

(5)


The voltage response VSC (t) across the SC terminals is:

−

iSC e

+

4ZeqPb

4VocPb[t,SoC,I,T]

VSC (t) = VtSC (t)

C

-

53

(a)

− (CiSC + CoSC )t/CoSC CiSC RiSC 1 + + RoSC ı(t) I(t) CoSC + CiSC CoSC (CoSC + CiSC )

(6)

2VocLi[t,SoC,I,T] + -

+

2Zeq Li

2ZeqPb

2VocPb[t,SoC,I,T] (b)

where VtSC (t) is the initial terminal voltage of the SC.

2VocLi[t,SoC,I,T] + -

3. Equivalent models for hybrid energy storage

+

2Zeq Li

Zeq SC

Zeq Pb

Voc Pb [t,SoC,I,T]

In this section, a generalized equivalent models for HESS is obtained. Three distinct cases with different combinations are investigated. In order to validate the obtained HESS models, each equivalent circuit is constructed in the MATLAB/Simulink environment. The response of the derived model is then compared to that obtained from an experimental setup. In the experimental setup, an 85 W constant power and 280 W pulsed power load operating at 0.1 Hz are used. Each 6-cell lead acid battery has a nominal voltage and capacity of 12 V and 21 A h, respectively [33]. The lithium ion battery module is composed of 3 individual cells connected in series where the nominal voltage of each is 3.7 V delivering a similar module voltage to the lead acid at 11.1 V under a matching capacity of 21 A h [34]. The SC used is manufactured by Maxwell and composed of two 58 F modules in parallel each rated at 16.2 VDC [35]. Table 1 shows the equivalent circuit parameters used for each ES type in simulation. Parameters for both lead acid and lithium ion batteries were obtained experimentally and compared to typical values [30,36] while SC parameters were determined using the procedure outlined in [32]. It is assumed that the SoH of batteries of the same type are close thus their parameters are similar. It should be noted that in the case of introducing severely damaged batteries (very low SoH), the parameters will deviate and a higher order model will be needed.

3.1. Case I: 4 lead acid batteries The first case presents the base where a traditional 100% lead acid battery stack in series is tested using 4 modules of matching voltage and capacity. It is anticipated that without the developed controller, all modules serve the pulsed load until a pre-specified DoD. Then, to replenish the lost energy, the modules are decoupled and charged offline. The lumped parameter model of this system is shown in Fig. 3(a). Since a potentially 4th order model is

(c)

Fig. 3. Lumped parameter models for (a) case I, (b) case II, and (c) case III.

condensed to a 1st order similar to (1), the equivalent impedance Zeqcase1 is: Zeqcase1 (s) =

4[RtPb + RpPb + CpPb RpPb RtPb s]

(7)

CpPb RpPb s + 1

The voltage response on the case I system Veqcase1 (t) is then: Veqcase1 (t) = 4VocPb −

4 [e−t/CpPb RpPb + CpPb RtPb ı(t)]I(t) CpPb

(8)

Both the voltage and current responses are shown in Fig. 4(a). It can be seen that the experimental results validate the accuracy of the obtained model. Both voltage responses coincide closely with each other. 3.2. Case II: 2 lithium ion and 2 lead acid batteries In case II, 50% of the lead acid battery modules are replaced with lithium ion modules of matching capacity, introducing a HESS array where half of the modules have less susceptibility to SoH impacts as a result of cycling and discharge currents hereby improving performance. This split system introduces a medium, cost-effective solution to improve the array performance without the need for a total replacement. Using this HESS, the system robustness to heavy pulsed loads is improved while reducing the overall current contribution from each battery where the proposed controller could be used to redistribute the charging pattern. The revised equivalent circuit for case II is shown in Fig. 3(b). The introduction of CpLi increases Zeqcase2 to 2nd order: Zeqcase2 (s) = 2RtLi + 2RtLi + 2RtPb

Table 1 Battery and supercapacitor simulation parameters.

+

Parameter

Name

Value

Lead acid battery initial open circuit voltage Lead acid battery ohmic resistance Lead acid battery polarization resistance Lead acid battery polarization capacitance Lithium ion battery initial open circuit voltage Lithium ion battery ohmic resistance Lithium ion battery polarization resistance Lithium ion battery polarization capacitance Supercapacitor initial terminal voltage Supercapacitor resistance outside pore Supercapacitor capacitance outside pore Supercapacitor resistance inside pore Supercapacitor capacitance inside pore

VocPb RtPb RpPb CpPb VocLi RtLi RpLi CpLi VtSC RoSC CoSC RiSC CiSC

13.17 V 90 m 40 m 160 F 12.80 V 12 m 86 m 0.015 F 16.13 V 35 m 42 F 70 m 18 F

2[(RpPb RpLi CpLi + RpLi RpPb Cppb )s + RpLi + RpPb ]

(9)

CpLi RpLi RpPb CpPb s2 + (RpLi CpLi + RpPb CpPb )s + 1

The voltage response on the case II system Vcase2 (t) is: Vcase2 (t) = 2VocLi + 2VocPb

−2

e−t/CpLi RpLi e−t/CpPb RpPb + + (RtLi + RtPb )ı(t) I(t) CpLi CpPb (10)

Similarly, both the voltage and current responses are shown in Fig. 4(b) validating the accuracy of the developed model.

54


Case I Voltage Voltage (V)

54

50 48

Current (A)

Analytical Model Experimental

52

0

5

10

15

20

0

5

10

15

20

25 30 Time (s) Case I Current

35

40

45

50

30

35

40

45

50

30

35

40

45

50

30

35

40

45

50

35

40

45

50

35

40

45

50

8 6 4 2 25 Time (s)

(a) Case II Voltage Voltage (V)

52

50

48

0

5

10

15

20

25

Current (A)

Time (s) Case II Current 8 6 4 2 0

5

10

15

20

25

Time (s)

(b) Case III Voltage Voltage (V)

56 54 52

Current (A)

50

0

5

10

15

20

0

5

10

15

20

25 30 Time (s) Case III Current

8 6 4 2 25 Time (s)

30

(c) Fig. 4. Voltage and current for (a) case I, (b) case II, and (c) case III.

3.3. Case III: 2 lithium ion batteries, supercapacitor and lead acid battery The final circuit model in case III replaces one of the remaining lead acid battery modules with a SC as shown in Fig. 4(c). In this

HESS, 25% is now served by an ES which can withstand cycling in real-time without major SoH impacts but under a reduced capacity and wider operating voltage range. However, as with case II, using the controllers applied in this paper, dynamic or rolling charging can be used efficiently as discussed in detail in the following section.


The voltage response on the case III system Veqcase3 (t) is:

Power (W)

400 Case I Case II Case III

300

Veqcase3 (t) = 2VocPb + VocPb + VtSC

200 100 0

− 0

5

10

15

20

25

30

35

40

45

1−e

(a)

+

54

−t(CoSC +Ci

SC

)/CoSC Ci

SC

Ri

SC

CoSC + CiSC

50

Time (s)

Voltage (V)

55

2e−t/CpLi RpLi e−t/CpPb RpPb e + + CpLi CpPb

−t(CoSC +Ci

SC

SC

Ri

SC

CoSC

52

)/CoSC Ci

+ (2RtLi + RoSC + RtPb )ı(t)

(12)

50 48 0

5

10

15

20

25

30

35

40

45

50

Time (s)

(b)

The voltage and current responses are shown in Fig. 4(c). It is shown that discharging the SC reduced the overall voltage of the stack. Nonetheless, the model response tracks the voltage reduction introduced by adding the SC.

Current (A)

7.6

3.4. Analytical results discussion

7.4 7.2 7 6.8

0

5

10

15

20

25

30

35

40

45

50

Time (s)

(c) Fig. 5. (a) Power, (b) voltage, and (c) current responses for cases I, II, and III.

The equivalent impedance model Zeqcase3 from the HESS system in case III is given by: 1 (RiSC CoSC CiSC )s2 + (CoSC + CiSC )s

Zeqcase3 = 2RtLi + RoSC + RtPb + +

CiSC RiSC RpPb 2RpLi + + CpLi RpLi s + 1 CpPb RpPb s + 1 (RiSC CoSC CiSC )s + CoSC + CiSC

(11)

In order to demonstrate the differences in response for each HESS configuration, Fig. 5 depicts a comparison of the output power, voltage, and current for all 3 cases. One can see each HESS discharge voltage trend and pulsed load response differs greatly from case to case. Case I follows a steeper discharge voltage trend (higher slope during the pulse) than Case II where the current steadily increases following each pulse to maintain constant power delivery to the load. Following each pulse, the recovery voltage is highly nonlinear due to a high timing constant generated by the large lead acid battery capacitance CpPb . Case II has a more linear voltage behavior outside of the pulse, with a voltage drop less sharp than that of Case I due to reduced ohmic resistance RtLi of the lithium ion modules. The case II voltage ends lower due to a lower terminal voltage VocLi of the lithium ion modules, but the current injected under each pulse is held nearly constant. Case III features the highest initial voltage due to the terminal

+

+

Bypass Relay

Current Measure.

A

V

Voltage Measure.

Charging Relays

Posive Bus Relay

Charging Relays

Posi ve Bus Relay

Charging circuit

DC

Bypass Relay

Current

A Measure. V

DC

Charging circuit

Negave Bus Relay

Negave Bus Rel ay

-

-

(b)

(a)

+

+ Charging Relays

Posive Bus Relay Current Measure.

Bypass Relay

Voltage Measure.

Charging Relays

Posive Bus Relay Current

A

Voltage

V Measure.

Negave Bus Relay

DC

Charging circuit

Bypass Relay

A Measure.

V Voltage

Measure.

Negave Bus Relay

-

(c)

(d)

Fig. 6. ESMS controller – (a) schematic diagram, (b) normal operation, (c) charging mode, and (d) ideal mode.

DC

Charging circuit

56


voltage of the SC VtSC hereby reducing the initial current required. Following each pulsed load (during off-pulse), the stack voltage is nearly flat as shown with the blue line in Fig. 5(b), but due to the reduced storage capacity of the SC, the long term voltage trend declines at the steepest rate hereby causing an increase in the current required from all ES sources to fill the remaining demand (as shown in Fig. 5(c)). These widely varied characteristics add to the importance and highlight the need for developing specialized control for HESS. Otherwise, over-charging or over-discharging of a single ES type within the array could occur leading to an inevitable system failure or permanent damage to the ES device. 4. Coordinated control of HESS In order to handle the diverse charging characteristics for the different ES types within the hybrid stack, a management system is utilized to provide a safe interconnection to each ES. The

following subsections describe the features of the controller deployed for this study as well as how the dynamic rolling charging algorithm functions. Following a discussion over the rolling charging concept, the charging constraints for each ES will be explained in-detail. 4.1. Energy storage management system controllers Modular ESMS controllers, developed initially in [37], are expanded to be utilized to manage each battery and SC module in the HESS. This tool provides individual monitoring and protection for each ES module as well as a number of unique control features. Through a certain topology of relays and measurements interfaced to a LabVIEW-NI DAQ based control platform, the ESMS is able to electrically couple and decouple an ES module in a series configuration while maintaining a path to operate the load. The HESS is then interfaced through a boost converter to maintain a bus voltage

Fig. 7. Configuration of 4 ESMS in series. (a) Schematic and (b) experimental setup.


57

Table 2 Ship power system load profiles. Load type

Equipment

Frequency

Duty cycle

Actual power

Per unit

Scaled power

Test resistance

Constant load Pulsed load 1 Pulsed load 2

Hotel load SPY-1 radar EM railgun

N/A 0.5 Hz 0.05 Hz

N/A 50% 25%

11.22 MW 6.00 MW 38.00 MW

1.000 0.587 3.661

105.0 W 60.7 W 384.4 W

32.0 55.4 8.7

at the required level even with differing voltage ranges associated with each operating scenario. In addition, once decoupled from the load bus, additional terminals provide a connection to a charging bus. A schematic diagram for the ESMS topology is shown in Fig. 6(a). For the sake of paper integrity, operation modes of the ESMS are briefly explained while explicit details can be found in [37]:

1. Normal Operation (Discharging): In order to achieve full isolation, two normally-closed relays connect the positive and negative terminals of the ES to the DC bus. A normally-open relay bypasses the ES to provide an alternative path to maintain continuity of supply. An interlock is provided between the three relays to avoid simultaneous connection. In this mode, the

positive and negative bus relays are closed and the bypass relay is open as shown in Fig. 6(b). 2. Charging Mode: One of the important options featured by ESMS is the capability to charge one ES element while the rest of the stack continues normal operation. During this mode, the positive and negative bus relays are open while the bypass relay is closed to offer an alternate path for the current. After an adjustable short delay, the ES element is connected to the charging circuit through the charging relays. This functionality enables the capability to apply the rolling charging algorithm to recover an ES element or restore a portion of its charge until the next pulse (or load peak) occurs. Fig. 6(c) depicts the current path in this mode. 3. Ideal mode: This mode can be used for maintenance or ES replacement offering complete isolation. The topology is as depicted in Fig. 6(d) where the positive and negative bus relays

Fig. 8. Test I: 2 lead acid and 2 lithium ion battery array under multiple pulsed loads – (a) currents of individual ES modules and (b) 72-s Close-up.

58


are open to isolate the battery from the DC bus. The charging relays are open as well to isolate the battery from the charging circuit, while the bypass relay is closed to provide an alternate path for the current. 4.2. Rolling charging A new concept of rolling charging has been developed to extend the operation time of certain HESS supplying heavy pulsed loads. This concept utilizes capabilities of the ESMS to extract a weak or discharged ES module from the HESS array and place it on charge while maintaining system operation. As an example, in real SPS systems, the EM railgun pulsed power requirement would present a challenge in terms of available energy on the HESS and may drop the SoC of an ES module to dangerously low levels. If the firing frequency and number of shots for the railgun were known, then one

could anticipate when the energy requirement was needed. Using this information, an ES module could then be extracted to charge when the pulse is off and reconnected again only when the pulse is active. Through control and timing, this could be synchronized with the pulsed load. To balance the impact on all ES modules while providing each applicable ES module an opportunity to charge, each ES module is dispatched for a period of time and then “rolled” to the next ES module with the lowest SoC. This tactic can allow the system to sustain for extended operation periods while at the same time permit the maintenance of some ES units while the stack is still under operation. This could be very beneficial for critical loads. 4.3. Charging constraints Charging currents and voltage levels vary based on each type of ES module. Since the utilized topology employs two isolated busses

Fig. 9. Test II: endurance test – (a) without rolling charging and (b) with rolling charging.


namely load and charging bus, another major advantage for the developed ESMS is that it is capable of handling different charging constraints for different types of ES. Consequently, the voltage and current limits of the charging bus can be adjusted dynamically based on the type of ES connected. In more complicated or larger systems, with the aid of the utilized topology, multiple busses can be provided. Moreover, the controller sets the current limits to prespecified values to maintain safe charging currents which will limit SoH degradation. For the 6-cell lead acid battery employed in this experiment, particular care has been taken to regulate the charging current to prolong its SoH. The manufacturer established an absolute maximum charging current at C/4, but to limit thermodynamic and material stress, this has been reduced to a conservative C/6 current under a charging bus voltage of 14.8 V. Lithium ion batteries, however, have much less susceptibility to higher charging currents as long as the charging voltage levels are carefully controlled. For the lithium ion cells deployed in this experiment, a voltage tolerance of 4.20 ± 0.03 V/cell or 12.60 ± 0.09 V is regulated for the 3-cell series module where the maximum charging current can be increased to C/2 [34]. The SC module is an exception as the charging current is not referred to its storage capacity. Theoretically, charging to its full voltage level VmaxSC is only limited by the equivalent series

59

resistance RESR , but the IEC has established a safety limit ImaxIEC for practical applications based on 2.6% of this current [31]. For the SC modules employed in this study, ImaxIEC ∼ = 19 A.

ImaxIEC = 0.026

VmaxSC RESR

(13)

5. Hardware implementation and experimental results In order to investigate the feasibility of the proposed control technique, a hardware setup has been established as depicted in Fig. 7. The lead acid and lithium ion batteries as well as the SC are the same types as listed previously in Section III. The voltage of the load bus is set to 48 V while the voltage of the charging bus is controlled dynamically based on each ES charging requirement. A programmable DC power supply was connected to the charging bus. The load values have been selected closely based on the information gathered from one of the new naval platforms (DDG-1000) [5] where its parameters were scaled down to the per-unit scale as shown in Table 2. In this model, the constant load is used as the base value and the pulsed loads are referenced to it. Under the experimental platform, Pulsed Load 1 is modeled after a radar system presenting the lighter of two considerable pulsed loads installed on the SPS operating at a scanning frequency of 0.5 Hz under a 50%

Fig. 10. Test III: without rolling charging: (a) voltage and (b) power.

60


duty cycle. Pulsed Load 2 introduces the EM railgun, a load that is a major disruption to the SPS. In both tests, the EM railgun is modeled with a 5-s active duration (0.05 Hz frequency) under a duty cycle of 25%. Five different tests are examined on this system. 5.1. Test I: 2 lithium ion and 2 lead acid batteries In this test, the HESS was composed of two lithium ion batteries connected to ESMS 1 and ESMS 2 and two lead acid batteries connected to ESMS 3 and ESMS 4. The test results are shown in Fig. 8 where the test duration was around 20 min (0.33 h), divided into 5 intervals. Looking at Fig. 8(a), it can be seen that in the first interval, all four batteries served the load whereas in the second interval, lithium ion battery ESMS 1 was extracted during the light loading periods (when the large pulsed load was off) and connected to the charging bus. This can be further illustrated by looking at Fig. 8(b), where a close up over a period of 72 s (0.02 h) is shown. The effect of the fast pulsed load (of lower amplitude) is clear and both pulsed loads are overlapped over some periods. It should be noted that the negative current indicates discharging the battery while positive current indicates charging. Then, in the third interval, the second lithium ion battery is extracted to be charged, and so on. This technique helps to replenish some of the lost energy during the high amplitude pulsed load, consequently extending the operation time of the array. 5.2. Test II: endurance test To investigate the effectiveness of the proposed rolling charging technique in expanding the “in operation” time of the HESS, an endurance test was performed utilizing the full potential of the ES modules. Fig. 9(a) and (b) shows the test results without and with the ESMS, respectively. It can be seen from Fig. 9(a) that the HESS was discharged reaching a full discharge voltage cutoff (collapsing point) after 164 min (2.73 h), while Fig. 9(b) shows that with the developed controller, the operation time was extended to 207 min (3.45 h). This system achieved a 26% increase in HESS service time through replenishing some of the lost energy online (while the stack was operating), which would not have been possible in the legacy system. This system could be of significant importance for SPS in critical operating scenarios where it is not possible to shed vital loads [38]. 5.3. Test III: 2 lithium ion, lead acid, and supercapacitor (no rolling charging) In the third test, one of the remaining lead acid batteries was replaced with a SC presenting further complications. To have accurate comparison, this test is used as the base for the case III HESS combination, where no rolling charging was applied. It can be seen from Fig. 10 that the test was constrained to less than 10 min, explained by the vast difference in energy densities between the SC and the batteries. Fig. 10(a) shows that the SC voltage decayed at a much higher rate than that of the batteries, reaching a very low value with which the HESS had to be disconnected. Fig. 10(b) shows that the power absorbed from the SC gradually decreased which had to be compensated by an increase in power injected by the batteries. Continued operation would cause a 33% increase in required current from remaining ES further reducing runtime and impacting the long term SoH. 5.4. Test IV: 2 lithium ion, lead acid, and supercapacitor (SC rolling) Due to the significantly lesser energy density of the SC, it represents the weakest link in the HESS, thus in this test it is elected to be rolling charged. In this scenario, a conservative approach is taken

Fig. 11. Test IV: (a) voltage, (b) voltage – 30 min zoom in, (c) current.

to maximize SoH of the batteries by only dynamically charging the SC. Shown in Fig. 11, the total test duration was 3 h where it can be seen that as the voltage of the capacitor decayed to a pre-specified level, it was decoupled from the stack, charged, and placed back in after reaching full charge. 8 V (50% SoC) was preset in this scenario to avoid a major drop in the HESS array voltage, although this can be set by the operator to any desired value. This process was repeated until one of the remaining batteries reached its full discharge voltage cutoff. The switching operation performed seamlessly without noticeable impacts on the DC bus. Fig. 11(b) shows a close-up of the SC depicting its wide voltage variation as well as how it saturates when approaching a full charge. The impact of the saturation is noted in Fig. 11(c) where the absorbed SC current starts to decrease. An alternative solution to the energy mismatch problem


is to increase the size of the SC, however this would add weight and require more real estate, two factors which are tightly constrained in the modern SPS. The ESMS presents simple, yet effective solution. 5.5. Test V: 2 lithium ion, lead acid, and supercapacitor (charging all) In the final test, the same configuration was utilized except now rolling charging is applied to all 4 ES elements to achieve a maximum runtime and SoC balance amongst ES modules. The test duration was 25 min (0.42 h), divided into 7 intervals. The sequence of the 7 intervals are as follows: no charging of any ES module, charging the SC, charging the first lithium ion module, charging the SC again, charging the second lithium ion module, charging the

61

SC again, and finally charging the lead acid battery. It can be seen from Fig. 12(a) that the varying charging characteristics for all 3 ES types are met and the operation is stable. It is worth to mention that the arrows at the bottom of the figure indicate a current envelope, better understood in Fig. 12(b) and (c). When charging the SC, its voltage quickly increases which was hereby reflected on the voltage of the entire stack. Since constant power loads were used, increasing the stack voltage decreased the required current injection from each ES module. Fig. 12(b) shows the voltage of each individual ES module. Over the entire test, the SC voltage was varied between 16.2 V and 8 V. This test further highlights the importance of individual monitoring and control of each ES module in a HESS. Fig. 12(c) depicts a comparison of the array voltage from the primary side of the boost converter and the output DC bus voltage. The array voltage experiences wide fluctuations due to coupling and decoupling of ES modules, however, these are not reflected on the DC bus voltage due to the converter. Only minor voltage fluctuations are detected on the DC bus and are well within standard limits [39]. It should be noted that during all the performed tests, a simple boost converter is used whose design is beyond the scope of this work. 6. Conclusion In this paper, the modeling and management of multi-chemistry energy storage systems was addressed. The equivalent models for HESS with multiple different combinations were derived and verified experimentally. A coordinated control technique was introduced to handle the charging of different ES types and extend the operating duration of the array under multiple pulsed loads, common in modern SPS. Using the developed management system, a single ES module could be extracted from the array and connected to a charging bus to restore some of the lost energy as a result of heavy loading periods or pulsed loads. This system provides an effective solution to manage multiple ES types to serve multiple pulsed loads on a SPS platform. The novelties of this work are as follows: (1) modeling and evaluation of multiple new series-configured hybrid energy storage architectures composed of lead acid batteries, lithium ion batteries, and SCs, (2) modeling and testing of multiple naval shipboard pulsed loads with varying frequencies and magnitudes via per unit system, and (3) the introduction of a specialized dispatch control scheme to coordinate charging and discharging of individual energy storage units while in operation to extend runtime while acknowledging SoH trade-offs. The effectiveness and seamless operation of the system has been verified through hardware testing. Acknowledgements This work was partially supported by grants from the Office of Naval Research and the U.S. Department of Energy. References

Fig. 12. Test V: (a) current, (b) voltage, (c) DC bus and array voltages.

[1] M. Koller, T. Borsche, A. Ulbig, G. Andersson, Review of grid applications with the Zurich 1 MW battery energy storage system, Electr. Power Syst. Res. 120 (2015) 128–135. [2] M. Singh, L.A.C. Lopes, N.A. Ninad, Grid forming Battery Energy Storage System (BESS) for a highly unbalanced hybrid mini-grid, Electr. Power Syst. Res. 127 (2015) 126–133. [3] P. Denholm, E. Ela, B. Kirby, M. Milligan, The Role of Energy Storage with Renewable Electricity Generation, NREL, Golden, CO, 2010, January, NREL/TP6A2-47187. [4] K.C. Divya, J. Stergaard, Battery energy storage technology for power systems—an overview, Electr. Power Syst. Res. 79 (2009) 511–520. [5] AeroWeb | DDG 1000 Zumwalt Class Destroyer, AeroWeb. [Online]. Available: http://www.bga-aeroweb.com/Defense/DDG-1000-Zumwalt-Destroyer. html (accessed 04.04.16).

62


[6] T. Jaster, A. Rowe, Z. Dong, Modeling and simulation of a hybrid electric propulsion system of a green ship, in: 2014 IEEE/ASME 10th International Conference on Mechatronic and Embedded Systems and Applications (MESA), 2014, pp. 1–6. [7] M. Steurer, M. Andrus, J. Langston, L. Qi, S. Suryanarayanan, S. Woodruff, P.F. Ribeiro, Investigating the impact of pulsed power charging demands on shipboard power quality, in: IEEE Electric Ship Technologies Symposium, ESTS’07, 2007, pp. 315–321. [8] W.A. Walls, W.F. Weldon, S.B. Pratap, M. Palmer, D. Adams, Application of electromagnetic guns to future naval platforms, IEEE Trans. Magn. 35 (January (1)) (1999) 262–267. [9] A. Neffati, M. Guemri, S. Caux, M. Fadel, Energy management strategies for multi source systems, Electr. Power Syst. Res. 102 (September) (2013) 42–49. [10] M. Ibrahim, S. Jemei, G. Wimmer, D. Hissel, Nonlinear autoregressive neural network in an energy management strategy for battery/ultra-capacitor hybrid electrical vehicles, Electr. Power Syst. Res. 136 (July) (2016) 262–269. [11] A. Esmaili, B. Novakovic, A. Nasiri, O. Abdel-baqi, A hybrid system of Li-ion capacitors and flow battery for dynamic wind energy support, IEEE Trans. Ind. Appl. 49 (July (4)) (2013) 1649–1657. [12] A. Mohamed, V. Salehi, O. Mohammed, Real-time energy management algorithm for mitigation of pulse loads in hybrid microgrids, IEEE Trans. Smart Grid 3 (December (4)) (2012) 1911–1922. [13] Z. Miao, L. Xu, V.R. Disfani, L. Fan, An SOC-based battery management system for microgrids, IEEE Trans. Smart Grid 5 (March (2)) (2014) 966–973. [14] J.-Y. Kim, J.-H. Jeon, S.-K. Kim, C. Cho, J.H. Park, H.-M. Kim, K.-Y. Nam, Cooperative control strategy of energy storage system and microsources for stabilizing the microgrid during islanded operation, IEEE Trans. Power Electron. 25 (2010) 3037–3048. [15] H. Karimi, H. Nikkhajoei, R. Iravani, Control of an electronically-coupled distributed resource unit subsequent to an islanding event, IEEE Trans. Power Deliv. 23 (2008) 493–501. [16] G. Diaz, C. Gonzalez-Moran, J. Gomez-Aleixandre, A. Diez, Scheduling of droop coefficients for frequency and voltage regulation in isolated microgrids, IEEE Trans. Power Syst. 25 (2010) 489–496. [17] M. Guerrero, J.C. Vasquez, J. Matas, M. Castilla, L.G. de Vicuna, Control strategy for flexible microgrid based on parallel line-interactive UPS systems, IEEE Trans. Ind. Electron. 56 (2009) 726–736. [18] M. Falahi, K.L. Butler-Purry, M. Ehsani, Reactive power coordination of shipboard power systems in presence of pulsed loads, IEEE Trans. Power Syst. 28 (2013) 3675–3682. [19] E.A. Lewis, M. Butcher, Managing multiple and varying energy demands in naval combatants with integrated electric propulsion, Pulsed Power Symp. (2005), 32/1–32/6. [20] L.N. Domaschk, A. Ouroua, R.E. Hebner, O.E. Bowlin, W.B. Colson, Coordination of large pulsed loads on future electric ships, IEEE Trans. Magn. 43 (2007) 450–455. [21] J.P. Kelley, D.A. Wetz, J.A. Reed, I.J. Cohen, G.K. Turner, W. Lee, The impact of power quality when high power pulsed DC and continuous AC loads are simultaneously operated on a microgrid testbed, in: 2013 IEEE Electric Ship Tech. Symp., 2013, pp. 6–12.

[22] B. Vural, C.S. Edrington, Ultra-capacitor based pulse power management in electrical ships, in: 2012 IEEE Transportation Electrification Conference and Expo (ITEC), 2012. [23] T. Dragicevic, X. Lu, J.C. Vasquez, J.M. Guerrero, DC microgrids—Part II: A review of power architectures, applications and standardization issues, IEEE Trans. Power Electron. 31 (2016) 3528–3549. [24] B.M. Huhman, J.M. Neri, D.A. Wetz, Design of a battery intermediate storage system for rep-rated pulsed power loads, in: 2013 IEEE Electric Ship Technologies Symp. (ESTS), 2013. [25] I. Lahbib, A. Lahyani, A. Sari, P. Venet, Performance analysis of a lead-acid battery/supercapacitors hybrid and a battery stand-alone under pulsed loads, in: 2014 Intl. Conf. on Green Energy, 2014, pp. 273–278. [26] G.L. Kusic, J.M. Heinzel, D.J. Hoffman, Monitoring pulsed power on ship electrical systems, in: 2013 IEEE Electric Ship Technologies Symposium (ESTS), 2013, pp. 13–20. [27] Y. Tang, A. Khaligh, Bidirectional hybrid battery/ultracapacitor energy storage systems for next generation MVDC shipboard power systems, in: 2011 IEEE Vehicle Power and Propulsion Conference (VPPC), 2011. [28] L. Gao, R.A. Dougal, S. Liu, Power enhancement of an actively controlled battery/ultracapacitor hybrid, IEEE Trans. Power Electron. 20 (2005) 236–243. [29] J. Shen, A. Khaligh, A supervisory energy management control strategy in a battery/ultracapacitor hybrid energy storage system, IEEE Trans. Transport. Electrif. 1 (2015) 223–231. [30] C.R. Gould, C.M. Bingham, D.A. Stone, P. Bentley, New battery model and stateof-health determination through subspace parameter estimation and stateobserver techniques, IEEE Trans. Vehicular Technol. 58 (2009) 3905–3916. ˛ [31] F. Béguin, E. Frackowiak, A. Burke, Testing of electrochemical capacitors, in: Supercapacitors, Wiley-VCH Verlag GmbH & Co. KGaA, 2013, pp. 437–471. [32] R.A. Dougal, L. Gao, S. Liu, Ultracapacitor model with automatic order selection and capacity scaling for dynamic system simulation, J. Power Sources 126 (2004) 250–257. [33] PowerSonic, Rechargeable Sealed Lead Acid Battery, Model: PDC-12200 (12V21AH), San Diego, CA, available at: http://www.power-sonic.com/images/ powersonic/sla batteries/PDC-12200.pdf. [34] Batteryspace.com, Lithium Ion Polymer Rechargeable Battery, Model HA055275, Richmond, CA, available at: http://www.batteryspace.com/Prodspecs/A055275-21Ah.pdf. [35] Maxwell, 16V small cell module, Model: BMOD0058 E016 B02 (16.2V-58 F), San Diego, CA, available at: http://www.maxwell.com/images/documents/ datasheet 16v small cell module.pdf. [36] X. Gong, R. Xiong, C.C. Mi, Study of the characteristics of battery packs in electric vehicles with parallel-connected lithium-ion battery cells, IEEE Trans. Ind. Appl. 51 (2015) 1872–1879. [37] A.T. Elsayed, C.R. Lashway, O.A. Mohammed, Advanced battery management and diagnostic system for smart grid infrastructure, IEEE Trans. Smart Grid 7 (2016) 897–905. [38] Y. Levron, D. Shmilovitz, Power systems’ optimal peak-shaving applying secondary storage, Electr. Power Syst. Res. 89 (August) (2012) 80–84. [39] J.S. Sergent, C.D. Coach, R.J. Roux, National Electrical Code Handbook, 12th ed., NFPA, 2011.