- Email: [email protected]
The development of a hybrid PEMFC power system
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 0 ( 2 0 1 5 ) 4 6 3 0 e4 6 4 0
Available online at www.sciencedirect.com
ScienceDirect journal homepage: www.elsevier.com/locate/he
The development of a hybrid PEMFC power system Yi-Fu Guo, Han-Che Chen, Fu-Cheng Wang* Department of Mechanical Engineering, National Taiwan University, No. 1, Roosevelt Road, Sec. 4, Taipei 10617, Taiwan
article info
abstract
Article history:
This paper describes the development of a stationary hybrid power system consisting of a
Received 16 December 2014
3 kW proton exchange membrane fuel cell (PEMFC) module, a battery set, and solar cells.
Received in revised form
We design PEMFC control and power managements for the hybrid system, and construct a
28 January 2015
simulation model based on experimental data. The simulation model is then used to
Accepted 29 January 2015
analyze performance and efficiency of the system, employing different power manage-
Available online 26 February 2015
ment strategies. The study comprises three parts: PEMFC control, power management, and the simulation model. First, we apply robust control to regulate the PEMFC temperature
Keywords:
and current, and achieve root mean square errors of 0.06 C and 0.07 A, respectively.
PEMFC
Second, we integrate the hybrid power system and design power management strategies
Hybrid power
for sustainable operation. Third, we develop a Matlab™ SimPowerSystem model based on
Robust control
experimental data. The simulation results are shown to match the experiments with a
Energy efficiency
deviation of 0.21%. Last, we apply the simulation model to discuss the impacts of different
SimPowerSystem
power management strategies on system efficiency. Based on the results, the designed controllers and power management strategies are deemed effective at improving system performance and efficiency. In addition, the developed simulation model can be applied in the future to develop customized hybrid power system. Copyright © 2015, Hydrogen Energy Publications, LLC. Published by Elsevier Ltd. All rights reserved.
Introduction The proton exchange membrane fuel cell (PEMFC) is a promising alternative power source because of its advantageous properties, such as low operating temperature, high efficiency, and low pollution. A PEMFC system operates on the principles of diffusion dynamics, proton concentration dynamics, cathode kinetics, and the Nernst equation. Many researchers considered these dynamics to develop complicated PEMFC models. For example, Fabio et al. [1] applied Simulink to develop a PEMFC based on experimental data. Abraham et al. [2] used artificial networks to predict the performance of
complex fuel-cell systems. Saddi et al. [3] compared the performance of three PEMFC models by experiments. From a system point of view, the PEMFC dynamics can be simplified for control design. For instance, Wang et al. [4] developed a multivariable linear model with two inputs (hydrogen and oxygen) and two outputs (voltage and current). Based on simplified PEMFC models, advanced control algorithms can be applied to improve system performance. For example, Tirnovan and Giurgea [5] designed optimal air management for improving PEMFC performance. Kurz et al. [6] applied proportional-integral control to regulate the PEMFC temperature and air flow to achieve high power density. Liso et al. [7] proposed a control-oriented dynamic model of a liquid-
* Corresponding author. Tel.: þ886 2 3366 2680; fax: þ886 2 23631755. E-mail address: [email protected] (F.-C. Wang). http://dx.doi.org/10.1016/j.ijhydene.2015.01.169 0360-3199/Copyright © 2015, Hydrogen Energy Publications, LLC. Published by Elsevier Ltd. All rights reserved.
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 0 ( 2 0 1 5 ) 4 6 3 0 e4 6 4 0
4631
Fig. 1 e The hybrid power system.
cooled PEMFC and applied proportional-integral-derivative (PID) control to regulate the stack temperature. Cedric et al. [8] developed a model-based control strategy to improve PEMFC water management. Wang and Chen [9] designed fixed-order robust controllers to improve performance and
Table 1 e Specification of hybrid power system components. Component PEMFC Module [13] Solar module [15]
LiFePO4 Battery [14] DC/DC Converter
DC/AC Inverter [16]
Type M-Field™ LPH8020 ASEC-220G6S
A123™ ANR26650 M-Field™ S/N:00051 MW™ TS-3000-148
Specifications See Ref. [13] Maximum power: 220 W Open circuit voltage: 33.86 V Short circuit current: 8.61 A Nominal voltage:52.8 V Nominal capacity: 23 Ah Input voltage: DC 44e85 V Output voltage: DC 42e57 V Maximum power:3 kW Input voltage: DC 42e60 V Output voltage: AC 110 V Maximum power: 4.5 kW
stability of a PEMFC system. This idea was extended in Ref. [10] to develop robust PID controllers for industry applications. Nowadays, PEMFC has been widely applied in hybrid power systems such as stationary power and electric vehicles. For instance, Dursun et al. [11] developed a hybrid PEMFC power model using Matlab/Simulink to simulate the effects of different power management strategies. Andaloro et al. [12] integrated a 5 kW PEMFC and battery to develop a hybrid powertrain for an urban bus. Wang et al. [13] developed a stationary PEMFC hybrid power system to supply uninterruptible power to telecom stations during power failures. This paper aims to develop a stand-alone hybrid power system for remote areas where the power grid might be unavailable. This paper proposes a hybrid power system that consists of a PEMFC module, a battery set, and solar cells. We develop robust controllers and power management for the system, and build a SimPowerSystem model based on experimental data. The developed simulation model is then used to discuss the effects of different power management strategies. The paper is organized as follows: Section The hybrid power system description integrates the hybrid power system that is composed of a PEMFC, a battery set, and a solar module. Section The PEMFC control introduces the PEMFC
Fig. 2 e The 3 kW PEMFC system.
4632
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 0 ( 2 0 1 5 ) 4 6 3 0 e4 6 4 0
converter, and a DC/AC inverter, as shown in Fig. 1. Because we used laboratory loads for performance verification, we selected the system components (see Table 1) by considering the following issues: (1) The average power load was 1.2 kW, with a peak of 3 kW. Therefore, we chose the 3 kW PEMFC module [13]. (2) The PEMFC has a voltage output of 48e60 V, so that we picked a DC/DC converter with an input of 40e80 V and an output of 42e57 V. (3) Considering the DC/DC converter output range, we integrated a lithiumeion battery set that composes of 16 10 A123 ANR26650M1A cells [14] to provide a nominal voltage of 52.8 V and a nominal capacity of 23 Ah to compensate the rapid-varying loads. (4) We connected a 660 W solar module [15] to the battery set, as the auxiliary power. (5) Because the electric utilities require 110 V AC, we chose an inverter [16] with a DC input of 42e60 V and an AC 110 V output. We will apply this hybrid power system to develop control and power managements in the following sections.
The PEMFC control
Fig. 3 e Identification of the PEMFC thermal model.
This section introduces the 3 kW PEMFC module, and designs robust controllers to regulate the PEMFC temperature and current. The applied PEMFC module consists of a 3 kW PEMFC
system, and designs robust controllers to regulate the PEMFC temperature and current to improve PEMFC performance and efficiency. Section Power management develops power management strategies so that the hybrid system can supply sustainable power at high efficiency. Section SimPowerSystem model builds a SimPowerSystem model based on experimental responses. We apply the hybrid power system to supply electricity to the lab, and compare the simulation and experimental results. In addition, we use the simulation model to analyze the impacts of different power management strategies. Last, we draw conclusions in Section Concluding remarks.
The hybrid power system description The proposed hybrid power system consists of a PEMFC module, a secondary battery set, a solar module, a DC/DC
Table 2 e The PEMFC thermal models. Load 10 20 30 40 50 60
A A A A A A
First GT10;1 GT20;1 GT30;1 GT40;1 GT50;1 GT60;1
0:1154 ¼ sþ0:01019 0:2396 ¼ sþ0:01186 0:3363 ¼ sþ0:01441 0:3594 ¼ sþ0:01617 0:3939 ¼ sþ0:01959 0:4749 ¼ sþ0:02044
Second GT10;2 GT20;2 GT30;2 GT40;2 GT50;2 GT60;2
0:1155 ¼ sþ0:01030 0:2420 ¼ sþ0:01233 0:3567 ¼ sþ0:01770 0:3629 ¼ sþ0:01663 0:4157 ¼ sþ0:01709 0:4798 ¼ sþ0:02114
Fig. 4 e PEMFC responses with temperature tracking (at 60 A load).
4633
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 0 ( 2 0 1 5 ) 4 6 3 0 e4 6 4 0
stack, four 60 W fans for air-cooling and oxygen supply, a purge valve at the stack outlet for dead-end operation, a hydrogen inlet valve at the stack input, a pressure regulator for maintaining appropriate fuel pressure, and an output breaker for emergency shutdown. We apply the current shunt, dc transmitter, thermal couple, and pressure sensor to measure the PEMFC current, voltage, temperature, and hydrogen pressure, respectively. A pressure regulator is implemented to keep the stack operating at appropriate pressure. The PEMFC system is shown in Fig. 2, with the specifications illustrated in ([13], Table 1). We control the PEMFC temperature and current by considering the following issues: (1). The PEMFC should be operated at the optimal temperature. (2). The fans must supply sufficient oxygen for the stack reaction. (3). The PEMFC can provide varying loads for the hybrid system.
PEMFC temperature control We first derive the PEMFC thermal models, and design robust controllers to regulate the PEMFC temperature. For the system point of view, we can regard the PEMFC thermal dynamics as the following single-input single-output (SISO) model: Tstack ¼ GT ðsÞ$Ufan ;
(1)
where we can regulate the duty cycle of the cooling fans (Ufan) to control the stack temperature (Tstack).GT is a transfer function that represents the thermal model of the PEMFC module. We derived the thermal model GT by generating a chirp signal of 0.002e0.02 Hz to control the input Ufan and then measuring the output temperature Tstack, as shown in Fig. 3. We accounted for system variation by repeating the experiments twice at the current loads of 10, 20, 30, 40, 50, and 60 A. Considering the oxygen supply and operating temperature, we set different fan duties for corresponding loads. The fans provide oxygen to generate loads and heat, but also take away heat. Therefore, the PEMFC temperature responses Tstack are delayed reactions of the fan duties Ufan. We applied the subspace state-space system identification (N4SID) algorithms [17,18] to obtain the PEMFC thermal models, as illustrated in Table 2. We then verified the models by applying the same inputs to the models and comparing the simulation and
Table 3 e Performance and energy density of the temperature control. Load
Fig. 5 e PEMFC temperature control.
10 20 30 40 50 60
A A A A A A
RMSE ( C)
Energy density (J/L)
Open-loop
Closed-loop
Open-loop
Closed-loop
0.1114 0.7446 0.7149 1.1356 4.2160 2.9761
0.0495 0.0581 0.0593 0.0550 0.0520 0.0466
6160.44 5863.98 5587.28 5243.39 5159.39 4878.74
6186.13 5892.64 5597.16 5428.18 5195.96 4903.45
4634
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 0 ( 2 0 1 5 ) 4 6 3 0 e4 6 4 0
experimental temperature responses, as shown in Fig. 3, where the maximum error between simulation and experiments is less than 2.1%. That is, the derived transfer functions can represent the PEMFC thermal dynamics. Noting the variation in the PEMFC thermal model GT shown in Table 2, we analyze the system variation and design H∞ robust controllers to regulate the PEMFC temperature. Suppose a nominal plant G0 has a normalized left coprime factorization of G0 ¼ M1N, where M, N2RH∞ and MM* þ NN* ¼ I. Assume that a perturbed system GD can be expressed as GD ¼ (M þ DM)1(N þ DN) with k½ DN DM k∞ < ε, and DM, DN2RH∞. Because the coprime factorization of a system is not unique, we can define the gap between a nominal plant G0 and a perturbed plant GD as [19]: The smallest value of k½ DN DM k∞ < ε which perturbs G0 into GD, is called the gap between G0 andGD, and is denoted as d(G0,GD). Therefore, we can select a nominal plant G0 from Table 2, which minimizes the maximum gap between G0 and the perturbed plantsGD, as follows:
G0 ¼ arg min max dðG0 ; Gi Þ ; G0
Gi
cGi :
(2)
where Gi ¼ GTj,k, for j ¼ 10, 20,/,60 and k ¼ 1, 2. Considering the models in Table 2, we selected the following nominal plants: GT0 ¼ GT20;1 ¼
0:2396 ; s þ 0:01186
(3)
with a gap of d(GT0,GTi,j) 0.3497, for j ¼ 10, 20,/,60 and k ¼ 1, 2. These gaps can be regarded as the module variation, which might be caused by model simplification and the changes in operating conditions, such as the environmental temperature and loads. A closed-loop system with a perturbed system GD and a controller K can be represented as ([20], Fig. 4). Therefore, applying the Small-Gain Theorem [21], the system is internally stable for all perturbations D ¼ ½ DN DM with kDk∞ < ε if and only if:
K
K 1 1 1
KÞ M ¼ ðI G 0
I ðI G0 KÞ ½ I
I ∞
1 G0
ε: ∞
(4)
We can further define the system's stability margin b(G0,K) as [15]:
Table 4 e The PEMFC current models.
49 V 51 V 54 V
1st
2nd
GI49;1 ¼ s2999:5s12030 þ16:86sþ57:39 GI51;1 ¼ s21003s11860 þ17:03sþ58:01 GI54;1 ¼ s2887:5s11020 þ16:14sþ55:15
GI49;2 ¼ s2976:3s11860 þ16:63sþ56:58 GI51;2 ¼ s2993:9s11860 þ17:03sþ58:11 GI54;2 ¼ s21031s12240 þ17:65sþ60:60
1
K 1
bðG0 ; KÞ≡
I ðI G0 KÞ ½ I G0 ; ∞
(5)
so that the closed-loop system is internally stable for all uncertainties D ¼ ½ DN DM with kDk∞ < ε, if and only if its stability margin b(G0,K) ε. Hence, the objective of controller synthesis is to design a controller K for the nominal plant Gi0 such that the stability margins are greater than the system gaps. We applied H∞ loop-shaping techniques [22], and set the following weighting functions ([20], Fig. 5): W1 ¼
ð2Þðs þ 0:04Þ ; sðs þ 5Þ
W2 ¼ 1;
(6)
to shape the plant as Gs ¼ W2GT0W1. The weighting functions were selected to provide a high gain at the low-frequency range to improve system performance, and a low-gain at the high-frequency range for noise attenuation. In addition, we added an integrator to the weighting functions to eliminate the steady-state errors. Following the standard procedures, we designed the following H∞ robust controller: 1:199s2 6:032s 0:1673 : (7) KT∞ ¼ s2 þ 5:23s þ 0:2007 with a stability margin of b(Gs,KT∞) ¼ 0.6404. Because the stability margin is much greater than the system gap of 0.3497, system stability can be guaranteed during operation. We implemented the designed controller on the PEMFC and analyzed the system performance. First, we let the PEMFC track the step temperature commands at different current loads; e.g., Fig. 4 shows the system responses with a load of 60 A. The results show that the PEMFC power generation and balance of plant (BOP) power consumption varied at different operating temperatures. Second, we define the following energy density sstack and PEMFC efficiency hPEMFC to analyze the effects of operating temperature: sstack ¼
Estack;elec Ebop;elec Exp
H2
Fig. 6 e Hardware setup for PEMFC current control.
(8)
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 0 ( 2 0 1 5 ) 4 6 3 0 e4 6 4 0
hPEMFC ¼ 50 49V 51V 54V Command
45 40
Stack Current(A)
35 30 25 20 15 10 5 0
0
5
10
15
20
25
30
Time (sec)
(a) With different loads. 50 45 40
PEMFC current(A)
35 30 25
Estack;elec Ebop;elec Exp
H2
LHV
100%
4635
(9)
where Estack,elec and Ebop,elec represent the energy generation Exp and BOP consumption of the PEMFC, respectively, while H2 is the practical hydrogen consumption. Note that hPEMFC is proportional to the energy density of (8) by a factor of low heating value (LHV ¼ 120 MJ kg1) [23]. Fig. 5(a) illustrates the energy density and efficiency of the PEMFC at different loads and temperatures. We can reduce the fan duties to raise the PEMFC operating temperature, and to reduce the BOP power consumption. However, the temperature control is limited by two factors: (1) the operating temperature should be less than 60 C [24]; and (2) the fans should provide sufficient oxygen for stack reaction. Fig. 5(a) indicates that the energy density increases as the temperature increase, and is improved by about 250 J/L using appropriate temperature settings. Third, based on Fig. 5(a), we set the optimal PEMFC operating temperature as shown in Fig. 5(b), where the PEMFC is controlled to track the reference temperature to achieve the highest energy density. Last, we compare the PEMFC temperature responses obtained with the open-loop and closed-loop control, as shown in Fig. 5(c) and Table 3. The designed robust controller can track the temperature commands with a root mean square error (RMSE) of 0.06 C. As a result, the average energy density is effectively improved by 51.7 J/L. In addition, the BOP power, which is mainly consumed by the fans, is lowered and smoothed by the designed robust controller. That is, the fan power consumption and noises were effectively reduced.
20
PEMFC current control 15
Because the PEMFC must supply power to the hybrid system, we connected the PEMFC to the DC/DC converter, as shown in Fig. 6. This allowed us to control the converter to regulate the PEMFC current. The procedures are similar to the temperature control: we derived the models by experiments, and then designed robust controllers to regulate the PEMFC current. From the system point of view, the system can be depicted as the following SISO model:
10 5 0
0
100
200
300
400
500
600
700
800
Time(sec)
(b) With different current commands. 100 90
Ifc ¼ GI ðsÞ$Uconv ;
80
Efficiency (%)
70 60 50 40 30 PEMFC efficiency Converter efficiency Power efficiency
20 10 0
0
10
20
30 40 Stack current (A)
50
60
70
(c) The efficiencies. Fig. 7 e PEMFC current control. Fig. 8 e PEMFC current settings.
(10)
4636
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 0 ( 2 0 1 5 ) 4 6 3 0 e4 6 4 0
Fig. 9 e The SimPowerSystem model.
where GI represents the transfer function from the control signal (Uconv) to the PEMFC current (Ifc). We generated chirp signals Uconv and measured the corresponding current Ifc, and applied the N4SID method to derived the model GI. Considering system variation, we repeated the experiment twice at different operating loads of 49, 51, and 54 V. The derived models are illustrated in Table 4, from which we selected the nominal plant as: GI0 ¼ G149;2 ¼
976:3 s 11860 ; s2 þ 16:63s þ 56:58
(11)
with the system gap of d(GI0,GIi,j) 0.0476, cGIi,j. We applied H∞ loop-shaping techniques ([20], Fig. 5) with the following weighting functions: W1 ¼
ð0:004Þðs þ 3Þ ; s
W2 ¼ 1;
1:308s2 21:28s 68:12 s2 þ 31:35s þ 89:11
hconv ¼
Econv;out 100%; Econv;in
h ¼ hPEMFC hconv ;
(14)
(15)
where Econv,in and Econv,out represent the input and output energy of the converter. We applied both temperature control and current control, and compared the system efficiencies in Fig. 7(c), where the power efficiency reached maximum at 20 A and dropped to minimum at 60 A. This factor will be considered when designing the power management of the hybrid system in Section Power management.
(12)
That is, we shaped the plant as Gs ¼ W2GI0W1 and designed the robust controller as follows: KI∞ ¼
We further defined the following converter efficiency hconv, and power efficiency h, to analyze system performance:
(13)
which gives a stability margin of b(Gs,KI∞)¼0.6073. Because the stability margin is much greater than the system gap of 0.0476, system stability can be guaranteed during operation. We implemented the weighted controller K ¼ W1KI∞W2 and recorded the system responses, as shown in Fig. 7. First, the designed controller succeeded in providing a steady current with a RMSE of 0.070 A (see Fig. 7(a)). Second, the controlled system can track different current commands (see Fig. 7(b), where the load-meter was set at a constant load of 50 V). That is, the PEMFC can be controlled to provide steady current. This is beneficial because varying loads may cause performance degradation [25,26].
Power management We applied the PEMFC to the hybrid power system shown in Fig. 1, and developed power management strategies by considering the following issues: (1) the PEMFC supplies auxiliary power when the solar and battery cannot provide sufficient electricity to the loads. In addition, we control the PEMFC to supply a steady current of 20 A if possible, because it achieves the maximum energy efficiency (see Fig. 7(c)). (2) High charging current to the battery set should be avoided because it might damage the battery [27]. (3) We limited the SOC operating ranges to 45e80% to prevent over-charging and over-discharging. Referring to Fig. 1, we monitored the current Ico and Ibatt, and calculated the battery SOC using the Coulomb counting method. The following power management strategies were developed:
4637
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 0 ( 2 0 1 5 ) 4 6 3 0 e4 6 4 0
(2) When the SOC continuously drops to 45%, the required load (Iload Isolar ¼ Ico þ Ibatt) is high and the battery discharges rather than charges. Therefore, we increase the PEMFC current Ifc according to Fig. 8 to charge the battery until the SOC reaches 50%. In addition, we set a threshold of 2 A to avoid chattering. (3) The safety protection will be activated by one of the following conditions [13]: (i) Low pressure protection: when the stack input pressure is less than 0.2 barg. (ii) Low voltage protection: when Vfc is less than 40 V. (iii) Over temperature protection: when Tstack is greater than 70 C. (iv) Over load protection: when Ifc is greater than 70 A. (v) Low battery protection: when Vbatt is less than 48 V.
SimPowerSystem model This section develops a simulation model using the Matlab™ SimPowerSystem [28], as shown in Fig. 9, for performance verification. The model can be used in the future to evaluate power management strategies and to select system components for developing customized hybrid power systems.
PEMFC model verification
Fig. 10 e Verification of the PEMFC model.
(1) When the battery SOC drops to 50%, the PEMFC system is started to supply a steady current of 20 A until the SOC reaches 80%.
We first developed the PEMFC sub-system, which consists of a PEMFC stack, the fuel and air supply, and the thermal model. We tuned the corresponding parameters and compared the simulation and experimental responses in Fig. 10 and Table 5. Note that the vertical spikes in the PEMFC flow-rate responses were caused by purges, which are necessary to avoid flooding and to keep suitable humidity in the stack [29]. We activated the purge function in both experiments and simulation, when the PEMFC output reached 2300 As according to the manufacturer's suggestion. First, the simulation and experimental responses are similar. That is, the simulation model correctly represented the PEMFC dynamics. Second, the designed controller can effectively regulate the PEMFC operating temperature according to Fig. 5(b). That is, the PEMFC can be controlled to reach the highest energy density. Last, the simulation model achieves similar hydrogen efficiency and PEMFC efficiency to that obtained experimentally, as illustrated in Table 5. Therefore, we can use this PEMFC sub-system to develop the hybrid power system model.
Table 5 e PEMFC model verification by Fig. 10. Simulation Load(A) o
RMSE ( C) Hydrogen efficiency (%) [30]. PEMFC efficiency (%)
Experiment
20 A
40 A
60 A
20 A
40 A
60 A
0.0136 98.04 57.51
0.0109 98.65 53.37
0.0049 98.88 48.28
0.0153 98.10 56.97
0.0525 98.38 53.67
0.0528 98.77 48.22
4638
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 0 ( 2 0 1 5 ) 4 6 3 0 e4 6 4 0
Fig. 11 e Responses of the integrated system.
Hybrid power model verification We further integrate the hybrid power model (see Fig. 9) which is composed of the PEMFC sub-system, a battery sub-system, a solar sub-system, a DC/DC converter, an electric loader, and an energy management system. We conducted the
experiments when the solar power was high, and obtained the following experimental data: the lab load, solar power, PEMFC responses, and battery responses, as illustrated in Fig. 11. Then we applied the recorded load and solar power to the simulation model, and compared the system responses in Fig. 11(c) and (d) and Table 6. First, the hybrid system
Table 6 e Hybrid model verification by Fig. 11.
Fuel cell
Solar cell Battery Load
Energy production (MJ) H2 consumption (L) Operating time (sec) Average current (A) Energy density (J/L) Energy production (MJ) Initial SOC (%) Final SOC (%) Energy consumption (MJ)
Simulation
Experiment
10.26 2311 10,295 19.98 4439 3.66 100 76.21 10.92
10.25 2306 10,290 19.96 4445 3.66 100 75.53 10.92
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 0 ( 2 0 1 5 ) 4 6 3 0 e4 6 4 0
and solar power for 24 h, as shown in Fig. 12, where the power disruptions were a result of the refrigerator compressor. We applied the following four scenarios to discuss the impacts of charging strategies:
3000
Power (W)
2500 2000 1500 1000 500 0
0
5
10
15
20
15
20
Time (hr) 1000 800 Power (W)
4639
600 400 200 0
0
5
10 Time (Hr)
Fig. 12 e The 24-h loading cycle and solar power.
succeeded in providing sustainable power for the lab loads. And the PEMFC was started up to provide a constant current of 20 A when the battery SOC dropped to 50%, according to the management strategies. Second, the solar power module can supply up to 33% of the auxiliary energy during the test, as shown in Table 6. Therefore, we can implement more solar panels to further reduce the hydrogen consumption in the future. Last, the simulation and experimental responses are similar, with a deviation of 0.21% for hydrogen consumption and 0.13% for energy density. Hence, the simulation model can be applied to predict the system responses in the future, because it is very time-consuming and expensive to conduct experiments at high power. Note that the spikes in the PEMFC flow-rate responses in Fig. 11(c) were caused by purges.
The impacts of power management strategies The developed simulation model can effectively predict the experimental responses; therefore, it can be applied to estimate the impacts of different power management strategies because conducting long-time experiments can be very costly and time-consuming. For example, we recorded the lab loads
(1) Strategy 1 applies the power management strategies described in Section Power management. That is, the PEMFC starts to supply a steady current of 20 A when the battery SOC drops to 50%, and stops when the SOC reaches 80%. If the SOC drops to 45%, we gradually increase the PEMFC current Ifc according to Fig. 8. (2) Strategy 2 is similar to Strategy 1: the PEMFC starts to supply a steady current of 20 A when the battery SOC drops to 50%, but directly increase Ifc to 50 A when the SOC drops to 45% until it reaches 50%. (3) Strategy 3 starts the PEMFC to supply a constant Ifc ¼ 50 A when the battery SOC drops to 45%, and stops until it reaches 80%. (4) Strategy 4 is similar to Strategy 1, but with a smaller SOC operating range of 45%e60%. The simulation results are compared in Table 7. First, all strategies can provide continuous power for the applied loads. Second, strategy 1 can effectively reduce the hydrogen consumption because it operates the PEMFC at the highest efficiency (20 A) most of the time. That is, strategy 1 provides the best energy density, with about 11% hydrogen consumption improvement compared with Strategy 3. Third, comparing Strategies 1 and 4, a larger SOC operating range can slightly increase the energy density. This is because the larger SOC range can extend the operating time and allow the PEMFC to operate at high efficiency (20 A). In addition, Strategy 1 can reduce the number of PEMFC restarts to avoid performance degradation. We can also apply the simulation model to evaluate the impacts of resizing system components, such as the battery set and solar panel. For example, if we increase the battery capacity from 15 Ah to 75 Ah, the hydrogen consumption can be reduced by 125 L by the simulation. On the other hand, the hydrogen consumption can also be reduced by about 10% with a 660 W solar panel, and 20% with a 1320 W solar panel. These model adjustment can be considered in the future when developing customized hybrid power systems.
Table 7 e Comparison of different power management strategies.
Fuel cell
Solar cell Battery Load
Energy production (MJ) H2 consumption (L) Operating time (sec) Number of starts Average current (A) Energy density (J/L) Energy production (MJ) Initial SOC (%) Final SOC (%) Energy consumption (MJ)
Strategy 1
Strategy 2
Strategy 3
Strategy 4
91.28 20,342 74,535 11 24.51 4487 10.33 100 80.05 101.87
91.37 20,415 74,521 11 24.61 4475 10.33 100 80.05 101.87
93.38 22,600 41,681 48 48.74 4132 10.33 100 80.05 101.87
91.23 20,342 74,581 29 24.53 4485 10.33 100 80.05 101.87
4640
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 0 ( 2 0 1 5 ) 4 6 3 0 e4 6 4 0
Concluding remarks [9]
This paper has developed a PEMFC hybrid power system that consisted of the PEMFC, secondary battery set, and solar module. First, we designed robust controllers to regulate the PEMFC temperature and output current, with RMSEs of 0.06 C and 0.07 A, respectively. And the average energy density was improved by 51.7 J/L by the proposed control. Second, we integrated the hybrid power system and designed power management strategies to guarantee sustainable operation. Last, we built a SimPowerSystem model and tuned the parameters based on experimental results. The simulation results were shown to comply with the experiments, with a maximum deviation of 0.21%, so that the simulation model can be used to predict the experimental responses. For example, we discussed the impacts of different power management strategies on system efficiencies. The results indicate that the system performance can be improved by PEMFC control, and the hydrogen consumption was improved by up-to 11% by appropriate power management strategies. In the future, the developed simulation model can be applied to design customized hybrid power systems employing different energy sources.
[10] [11]
[12]
[13]
[14] [15]
[16] [17]
[18]
Acknowledgment
[19]
This work was financially supported in part by the National Science Council Taiwan under Grand NSC 103eETeEe002e002eET.
[20]
references
[1] Fabio M, Fausto T, Luca O, Paola GS, Giovanni D, Stefano L, et al. PEMFC system simulation in MATLAB-Simulink environment. Int J Hydrogen Energy 2011;36:8045e52. [2] Abraham U, Chavez R, Roberto MG, Duron-Torres SM, Ferraro M, Brunaccini G, et al. High power fuel cell simulator based on artificial neural network. Int J Hydrogen Energy 2010;35:12125e33. [3] Saddi A, Becherif M, Aboubou A, Ayad MY. Comparison of proton exchange membrane fuel cell static models. Renew Energy 2013;56:64e71. [4] Wang FC, Chen HT, Yang YP, Yen JY. Multivariable robust control of a proton exchange membrane fuel cell system. J Power Sources 2008;177:393e403. [5] Tirnovan R, Giurgea S. Efficiency improvement of a PEMFC power source by optimization of the air management. Int J Hydrogen Energy 2012;37:7745e56. [6] Kurz T, Keller J. Heat management in a portable high temperature PEM fuel cell module with open cathode. Fuel Cells 2011;11:518e25. [7] Liso V, Nielsen MP, Kær SK, Mortensen HH. Thermal modeling and temperature control of a PEM fuel cell system for forklift applications. Int J Hydrogen Energy 2014;39:8410e20. [8] Cedric D, Michel B, Brigitte GP, Jean PC, Bruno GP. A novel non-linear model-based strategy to improve PEMFC water
[21]
[22]
[23] [24]
[25]
[26]
[27]
[28]
[29]
[30]
management e the flatness-based approach. Int J Hydrogen Energy 2015;40:2371e6. Wang FC, Chen HT. Design and implementation of fixedorder robust controllers for a proton exchange membrane fuel cell system. Int J Hydrogen Energy 2009;34:2705e17. Wang FC, Ko CC. Multivariable robust PID control for a PEMFC system. Int J Hydrogen Energy 2010;35:10437e45. Dursun E, Kilic O. Comparative evaluation of different power management strategies of a stand-alone PV/Wind/PEMFC hybrid power system. Int J Electr Power Energy Syst 2012;34:81e9. Andaloro L, Napoli G. Design of a hybrid electric fuel cell power train for an urban bus. Int J Hydrogen Energy 2013;38:7725e32. Wang FC, Kuo PC. Control design and power management of a stationary PEMFC hybrid power system. Int J Hydrogen Energy 2013;38:5845e56. A123 systems. Available at: http://a123systems.com/. ASEC-220G6S. Available at: http://cdn.photovoltaik-web.de/ media/mod_solarmoduledb/Datenblaetter/Apollo_Solar_ Energy_Asec-200G6S-2BB_Asec-220G6S-2BB.pdf. TS-3000e148A. http://www.meanwell.com/search/ts-3000/ TS-3000-spec.pdf. Goethals I, Pelckmans K, Suykens JA, De Moor B. Subspace identification of Hammerstein systems using least squares support vector machines. IEEE Trans Autom Control 2005;50:1509e19. Van Overschee P, De Moor B. N4SID-Subspace algorithms for the identification of combined deterministic-stochastic system. Automatica 1994;30:75e93. Georgiou TT, Smith MC. Optimal robustness in the gap metric. IEEE Trans Autom Control 1990;35:673e86. Wang FC, Guo YF. Robustness analyses of PEMFC systems on the production line. Int J Hydrogen Energy 2015;40:1959e66. Glover K, McFarlane DC. Robust controller design using normalised coprime factor plant descriptions. SpringerVerlag New York, Inc; 1989. McFarlane DC, Glover K. A loop-shaping design procedure using H∞ synthesis. IEEE Trans Autom Control 1992;37:759e69. Fuel cell formulary. http://www.pemfc.de/FCF_Smart.pdf. Ballard Mark1020 ACSTTM PEMFC stack. Available at: http:// 140.112.14.7/~sic/PaperMaterial/MAN5100192_new% 20guide_mark1020.pdf. Lin R, Li B, Hou YP, Ma JM. Investigation of dynamic driving cycle effect on performance degradation and microstructure change of PEM fuel cell. Int J Hydrogen Energy 2009;34:2369e76. Yan X, Hou M, Sun L, Cheng H, Hong Y, Liang D, et al. The study on transient characteristic of proton exchange membrane fuel cell stack during dynamic loading. J Power Sources 2007;163:966e70. Klein R, Chaturvedi NA, Christensen J, Ahmed J, Findeisen R, Kojic A. Optimal charging strategies in lithium-ion battery. In: Proceedings of the 2011 American control conference; 2011 Jun 29eJuly 1; San Francisco, USA; 2011. p. 382e7. Njoya S, Tremblay O, Dessaint LA. A generic fuel cell model for the simulation of fuel cell vehicles. IEEE Veh Power Propuls Conf VPPC 2009;1e3:1501e8. Chen J, Jason BS, Anna GS, James RW. Optimization of purge cycle for dead-ended anode fuel cell operation. Int J Hydrogen Energy 2013;38:5092e105. Wang FC, Peng CH. The development of an exchangeable PEMFC power module for electric vehicles. Int J Hydrogen Energy 2014;39:3855e67.
Available online at www.sciencedirect.com
ScienceDirect journal homepage: www.elsevier.com/locate/he
The development of a hybrid PEMFC power system Yi-Fu Guo, Han-Che Chen, Fu-Cheng Wang* Department of Mechanical Engineering, National Taiwan University, No. 1, Roosevelt Road, Sec. 4, Taipei 10617, Taiwan
article info
abstract
Article history:
This paper describes the development of a stationary hybrid power system consisting of a
Received 16 December 2014
3 kW proton exchange membrane fuel cell (PEMFC) module, a battery set, and solar cells.
Received in revised form
We design PEMFC control and power managements for the hybrid system, and construct a
28 January 2015
simulation model based on experimental data. The simulation model is then used to
Accepted 29 January 2015
analyze performance and efficiency of the system, employing different power manage-
Available online 26 February 2015
ment strategies. The study comprises three parts: PEMFC control, power management, and the simulation model. First, we apply robust control to regulate the PEMFC temperature
Keywords:
and current, and achieve root mean square errors of 0.06 C and 0.07 A, respectively.
PEMFC
Second, we integrate the hybrid power system and design power management strategies
Hybrid power
for sustainable operation. Third, we develop a Matlab™ SimPowerSystem model based on
Robust control
experimental data. The simulation results are shown to match the experiments with a
Energy efficiency
deviation of 0.21%. Last, we apply the simulation model to discuss the impacts of different
SimPowerSystem
power management strategies on system efficiency. Based on the results, the designed controllers and power management strategies are deemed effective at improving system performance and efficiency. In addition, the developed simulation model can be applied in the future to develop customized hybrid power system. Copyright © 2015, Hydrogen Energy Publications, LLC. Published by Elsevier Ltd. All rights reserved.
Introduction The proton exchange membrane fuel cell (PEMFC) is a promising alternative power source because of its advantageous properties, such as low operating temperature, high efficiency, and low pollution. A PEMFC system operates on the principles of diffusion dynamics, proton concentration dynamics, cathode kinetics, and the Nernst equation. Many researchers considered these dynamics to develop complicated PEMFC models. For example, Fabio et al. [1] applied Simulink to develop a PEMFC based on experimental data. Abraham et al. [2] used artificial networks to predict the performance of
complex fuel-cell systems. Saddi et al. [3] compared the performance of three PEMFC models by experiments. From a system point of view, the PEMFC dynamics can be simplified for control design. For instance, Wang et al. [4] developed a multivariable linear model with two inputs (hydrogen and oxygen) and two outputs (voltage and current). Based on simplified PEMFC models, advanced control algorithms can be applied to improve system performance. For example, Tirnovan and Giurgea [5] designed optimal air management for improving PEMFC performance. Kurz et al. [6] applied proportional-integral control to regulate the PEMFC temperature and air flow to achieve high power density. Liso et al. [7] proposed a control-oriented dynamic model of a liquid-
* Corresponding author. Tel.: þ886 2 3366 2680; fax: þ886 2 23631755. E-mail address: [email protected] (F.-C. Wang). http://dx.doi.org/10.1016/j.ijhydene.2015.01.169 0360-3199/Copyright © 2015, Hydrogen Energy Publications, LLC. Published by Elsevier Ltd. All rights reserved.
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 0 ( 2 0 1 5 ) 4 6 3 0 e4 6 4 0
4631
Fig. 1 e The hybrid power system.
cooled PEMFC and applied proportional-integral-derivative (PID) control to regulate the stack temperature. Cedric et al. [8] developed a model-based control strategy to improve PEMFC water management. Wang and Chen [9] designed fixed-order robust controllers to improve performance and
Table 1 e Specification of hybrid power system components. Component PEMFC Module [13] Solar module [15]
LiFePO4 Battery [14] DC/DC Converter
DC/AC Inverter [16]
Type M-Field™ LPH8020 ASEC-220G6S
A123™ ANR26650 M-Field™ S/N:00051 MW™ TS-3000-148
Specifications See Ref. [13] Maximum power: 220 W Open circuit voltage: 33.86 V Short circuit current: 8.61 A Nominal voltage:52.8 V Nominal capacity: 23 Ah Input voltage: DC 44e85 V Output voltage: DC 42e57 V Maximum power:3 kW Input voltage: DC 42e60 V Output voltage: AC 110 V Maximum power: 4.5 kW
stability of a PEMFC system. This idea was extended in Ref. [10] to develop robust PID controllers for industry applications. Nowadays, PEMFC has been widely applied in hybrid power systems such as stationary power and electric vehicles. For instance, Dursun et al. [11] developed a hybrid PEMFC power model using Matlab/Simulink to simulate the effects of different power management strategies. Andaloro et al. [12] integrated a 5 kW PEMFC and battery to develop a hybrid powertrain for an urban bus. Wang et al. [13] developed a stationary PEMFC hybrid power system to supply uninterruptible power to telecom stations during power failures. This paper aims to develop a stand-alone hybrid power system for remote areas where the power grid might be unavailable. This paper proposes a hybrid power system that consists of a PEMFC module, a battery set, and solar cells. We develop robust controllers and power management for the system, and build a SimPowerSystem model based on experimental data. The developed simulation model is then used to discuss the effects of different power management strategies. The paper is organized as follows: Section The hybrid power system description integrates the hybrid power system that is composed of a PEMFC, a battery set, and a solar module. Section The PEMFC control introduces the PEMFC
Fig. 2 e The 3 kW PEMFC system.
4632
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 0 ( 2 0 1 5 ) 4 6 3 0 e4 6 4 0
converter, and a DC/AC inverter, as shown in Fig. 1. Because we used laboratory loads for performance verification, we selected the system components (see Table 1) by considering the following issues: (1) The average power load was 1.2 kW, with a peak of 3 kW. Therefore, we chose the 3 kW PEMFC module [13]. (2) The PEMFC has a voltage output of 48e60 V, so that we picked a DC/DC converter with an input of 40e80 V and an output of 42e57 V. (3) Considering the DC/DC converter output range, we integrated a lithiumeion battery set that composes of 16 10 A123 ANR26650M1A cells [14] to provide a nominal voltage of 52.8 V and a nominal capacity of 23 Ah to compensate the rapid-varying loads. (4) We connected a 660 W solar module [15] to the battery set, as the auxiliary power. (5) Because the electric utilities require 110 V AC, we chose an inverter [16] with a DC input of 42e60 V and an AC 110 V output. We will apply this hybrid power system to develop control and power managements in the following sections.
The PEMFC control
Fig. 3 e Identification of the PEMFC thermal model.
This section introduces the 3 kW PEMFC module, and designs robust controllers to regulate the PEMFC temperature and current. The applied PEMFC module consists of a 3 kW PEMFC
system, and designs robust controllers to regulate the PEMFC temperature and current to improve PEMFC performance and efficiency. Section Power management develops power management strategies so that the hybrid system can supply sustainable power at high efficiency. Section SimPowerSystem model builds a SimPowerSystem model based on experimental responses. We apply the hybrid power system to supply electricity to the lab, and compare the simulation and experimental results. In addition, we use the simulation model to analyze the impacts of different power management strategies. Last, we draw conclusions in Section Concluding remarks.
The hybrid power system description The proposed hybrid power system consists of a PEMFC module, a secondary battery set, a solar module, a DC/DC
Table 2 e The PEMFC thermal models. Load 10 20 30 40 50 60
A A A A A A
First GT10;1 GT20;1 GT30;1 GT40;1 GT50;1 GT60;1
0:1154 ¼ sþ0:01019 0:2396 ¼ sþ0:01186 0:3363 ¼ sþ0:01441 0:3594 ¼ sþ0:01617 0:3939 ¼ sþ0:01959 0:4749 ¼ sþ0:02044
Second GT10;2 GT20;2 GT30;2 GT40;2 GT50;2 GT60;2
0:1155 ¼ sþ0:01030 0:2420 ¼ sþ0:01233 0:3567 ¼ sþ0:01770 0:3629 ¼ sþ0:01663 0:4157 ¼ sþ0:01709 0:4798 ¼ sþ0:02114
Fig. 4 e PEMFC responses with temperature tracking (at 60 A load).
4633
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 0 ( 2 0 1 5 ) 4 6 3 0 e4 6 4 0
stack, four 60 W fans for air-cooling and oxygen supply, a purge valve at the stack outlet for dead-end operation, a hydrogen inlet valve at the stack input, a pressure regulator for maintaining appropriate fuel pressure, and an output breaker for emergency shutdown. We apply the current shunt, dc transmitter, thermal couple, and pressure sensor to measure the PEMFC current, voltage, temperature, and hydrogen pressure, respectively. A pressure regulator is implemented to keep the stack operating at appropriate pressure. The PEMFC system is shown in Fig. 2, with the specifications illustrated in ([13], Table 1). We control the PEMFC temperature and current by considering the following issues: (1). The PEMFC should be operated at the optimal temperature. (2). The fans must supply sufficient oxygen for the stack reaction. (3). The PEMFC can provide varying loads for the hybrid system.
PEMFC temperature control We first derive the PEMFC thermal models, and design robust controllers to regulate the PEMFC temperature. For the system point of view, we can regard the PEMFC thermal dynamics as the following single-input single-output (SISO) model: Tstack ¼ GT ðsÞ$Ufan ;
(1)
where we can regulate the duty cycle of the cooling fans (Ufan) to control the stack temperature (Tstack).GT is a transfer function that represents the thermal model of the PEMFC module. We derived the thermal model GT by generating a chirp signal of 0.002e0.02 Hz to control the input Ufan and then measuring the output temperature Tstack, as shown in Fig. 3. We accounted for system variation by repeating the experiments twice at the current loads of 10, 20, 30, 40, 50, and 60 A. Considering the oxygen supply and operating temperature, we set different fan duties for corresponding loads. The fans provide oxygen to generate loads and heat, but also take away heat. Therefore, the PEMFC temperature responses Tstack are delayed reactions of the fan duties Ufan. We applied the subspace state-space system identification (N4SID) algorithms [17,18] to obtain the PEMFC thermal models, as illustrated in Table 2. We then verified the models by applying the same inputs to the models and comparing the simulation and
Table 3 e Performance and energy density of the temperature control. Load
Fig. 5 e PEMFC temperature control.
10 20 30 40 50 60
A A A A A A
RMSE ( C)
Energy density (J/L)
Open-loop
Closed-loop
Open-loop
Closed-loop
0.1114 0.7446 0.7149 1.1356 4.2160 2.9761
0.0495 0.0581 0.0593 0.0550 0.0520 0.0466
6160.44 5863.98 5587.28 5243.39 5159.39 4878.74
6186.13 5892.64 5597.16 5428.18 5195.96 4903.45
4634
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 0 ( 2 0 1 5 ) 4 6 3 0 e4 6 4 0
experimental temperature responses, as shown in Fig. 3, where the maximum error between simulation and experiments is less than 2.1%. That is, the derived transfer functions can represent the PEMFC thermal dynamics. Noting the variation in the PEMFC thermal model GT shown in Table 2, we analyze the system variation and design H∞ robust controllers to regulate the PEMFC temperature. Suppose a nominal plant G0 has a normalized left coprime factorization of G0 ¼ M1N, where M, N2RH∞ and MM* þ NN* ¼ I. Assume that a perturbed system GD can be expressed as GD ¼ (M þ DM)1(N þ DN) with k½ DN DM k∞ < ε, and DM, DN2RH∞. Because the coprime factorization of a system is not unique, we can define the gap between a nominal plant G0 and a perturbed plant GD as [19]: The smallest value of k½ DN DM k∞ < ε which perturbs G0 into GD, is called the gap between G0 andGD, and is denoted as d(G0,GD). Therefore, we can select a nominal plant G0 from Table 2, which minimizes the maximum gap between G0 and the perturbed plantsGD, as follows:
G0 ¼ arg min max dðG0 ; Gi Þ ; G0
Gi
cGi :
(2)
where Gi ¼ GTj,k, for j ¼ 10, 20,/,60 and k ¼ 1, 2. Considering the models in Table 2, we selected the following nominal plants: GT0 ¼ GT20;1 ¼
0:2396 ; s þ 0:01186
(3)
with a gap of d(GT0,GTi,j) 0.3497, for j ¼ 10, 20,/,60 and k ¼ 1, 2. These gaps can be regarded as the module variation, which might be caused by model simplification and the changes in operating conditions, such as the environmental temperature and loads. A closed-loop system with a perturbed system GD and a controller K can be represented as ([20], Fig. 4). Therefore, applying the Small-Gain Theorem [21], the system is internally stable for all perturbations D ¼ ½ DN DM with kDk∞ < ε if and only if:
K
K 1 1 1
KÞ M ¼ ðI G 0
I ðI G0 KÞ ½ I
I ∞
1 G0
ε: ∞
(4)
We can further define the system's stability margin b(G0,K) as [15]:
Table 4 e The PEMFC current models.
49 V 51 V 54 V
1st
2nd
GI49;1 ¼ s2999:5s12030 þ16:86sþ57:39 GI51;1 ¼ s21003s11860 þ17:03sþ58:01 GI54;1 ¼ s2887:5s11020 þ16:14sþ55:15
GI49;2 ¼ s2976:3s11860 þ16:63sþ56:58 GI51;2 ¼ s2993:9s11860 þ17:03sþ58:11 GI54;2 ¼ s21031s12240 þ17:65sþ60:60
1
K 1
bðG0 ; KÞ≡
I ðI G0 KÞ ½ I G0 ; ∞
(5)
so that the closed-loop system is internally stable for all uncertainties D ¼ ½ DN DM with kDk∞ < ε, if and only if its stability margin b(G0,K) ε. Hence, the objective of controller synthesis is to design a controller K for the nominal plant Gi0 such that the stability margins are greater than the system gaps. We applied H∞ loop-shaping techniques [22], and set the following weighting functions ([20], Fig. 5): W1 ¼
ð2Þðs þ 0:04Þ ; sðs þ 5Þ
W2 ¼ 1;
(6)
to shape the plant as Gs ¼ W2GT0W1. The weighting functions were selected to provide a high gain at the low-frequency range to improve system performance, and a low-gain at the high-frequency range for noise attenuation. In addition, we added an integrator to the weighting functions to eliminate the steady-state errors. Following the standard procedures, we designed the following H∞ robust controller: 1:199s2 6:032s 0:1673 : (7) KT∞ ¼ s2 þ 5:23s þ 0:2007 with a stability margin of b(Gs,KT∞) ¼ 0.6404. Because the stability margin is much greater than the system gap of 0.3497, system stability can be guaranteed during operation. We implemented the designed controller on the PEMFC and analyzed the system performance. First, we let the PEMFC track the step temperature commands at different current loads; e.g., Fig. 4 shows the system responses with a load of 60 A. The results show that the PEMFC power generation and balance of plant (BOP) power consumption varied at different operating temperatures. Second, we define the following energy density sstack and PEMFC efficiency hPEMFC to analyze the effects of operating temperature: sstack ¼
Estack;elec Ebop;elec Exp
H2
Fig. 6 e Hardware setup for PEMFC current control.
(8)
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 0 ( 2 0 1 5 ) 4 6 3 0 e4 6 4 0
hPEMFC ¼ 50 49V 51V 54V Command
45 40
Stack Current(A)
35 30 25 20 15 10 5 0
0
5
10
15
20
25
30
Time (sec)
(a) With different loads. 50 45 40
PEMFC current(A)
35 30 25
Estack;elec Ebop;elec Exp
H2
LHV
100%
4635
(9)
where Estack,elec and Ebop,elec represent the energy generation Exp and BOP consumption of the PEMFC, respectively, while H2 is the practical hydrogen consumption. Note that hPEMFC is proportional to the energy density of (8) by a factor of low heating value (LHV ¼ 120 MJ kg1) [23]. Fig. 5(a) illustrates the energy density and efficiency of the PEMFC at different loads and temperatures. We can reduce the fan duties to raise the PEMFC operating temperature, and to reduce the BOP power consumption. However, the temperature control is limited by two factors: (1) the operating temperature should be less than 60 C [24]; and (2) the fans should provide sufficient oxygen for stack reaction. Fig. 5(a) indicates that the energy density increases as the temperature increase, and is improved by about 250 J/L using appropriate temperature settings. Third, based on Fig. 5(a), we set the optimal PEMFC operating temperature as shown in Fig. 5(b), where the PEMFC is controlled to track the reference temperature to achieve the highest energy density. Last, we compare the PEMFC temperature responses obtained with the open-loop and closed-loop control, as shown in Fig. 5(c) and Table 3. The designed robust controller can track the temperature commands with a root mean square error (RMSE) of 0.06 C. As a result, the average energy density is effectively improved by 51.7 J/L. In addition, the BOP power, which is mainly consumed by the fans, is lowered and smoothed by the designed robust controller. That is, the fan power consumption and noises were effectively reduced.
20
PEMFC current control 15
Because the PEMFC must supply power to the hybrid system, we connected the PEMFC to the DC/DC converter, as shown in Fig. 6. This allowed us to control the converter to regulate the PEMFC current. The procedures are similar to the temperature control: we derived the models by experiments, and then designed robust controllers to regulate the PEMFC current. From the system point of view, the system can be depicted as the following SISO model:
10 5 0
0
100
200
300
400
500
600
700
800
Time(sec)
(b) With different current commands. 100 90
Ifc ¼ GI ðsÞ$Uconv ;
80
Efficiency (%)
70 60 50 40 30 PEMFC efficiency Converter efficiency Power efficiency
20 10 0
0
10
20
30 40 Stack current (A)
50
60
70
(c) The efficiencies. Fig. 7 e PEMFC current control. Fig. 8 e PEMFC current settings.
(10)
4636
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 0 ( 2 0 1 5 ) 4 6 3 0 e4 6 4 0
Fig. 9 e The SimPowerSystem model.
where GI represents the transfer function from the control signal (Uconv) to the PEMFC current (Ifc). We generated chirp signals Uconv and measured the corresponding current Ifc, and applied the N4SID method to derived the model GI. Considering system variation, we repeated the experiment twice at different operating loads of 49, 51, and 54 V. The derived models are illustrated in Table 4, from which we selected the nominal plant as: GI0 ¼ G149;2 ¼
976:3 s 11860 ; s2 þ 16:63s þ 56:58
(11)
with the system gap of d(GI0,GIi,j) 0.0476, cGIi,j. We applied H∞ loop-shaping techniques ([20], Fig. 5) with the following weighting functions: W1 ¼
ð0:004Þðs þ 3Þ ; s
W2 ¼ 1;
1:308s2 21:28s 68:12 s2 þ 31:35s þ 89:11
hconv ¼
Econv;out 100%; Econv;in
h ¼ hPEMFC hconv ;
(14)
(15)
where Econv,in and Econv,out represent the input and output energy of the converter. We applied both temperature control and current control, and compared the system efficiencies in Fig. 7(c), where the power efficiency reached maximum at 20 A and dropped to minimum at 60 A. This factor will be considered when designing the power management of the hybrid system in Section Power management.
(12)
That is, we shaped the plant as Gs ¼ W2GI0W1 and designed the robust controller as follows: KI∞ ¼
We further defined the following converter efficiency hconv, and power efficiency h, to analyze system performance:
(13)
which gives a stability margin of b(Gs,KI∞)¼0.6073. Because the stability margin is much greater than the system gap of 0.0476, system stability can be guaranteed during operation. We implemented the weighted controller K ¼ W1KI∞W2 and recorded the system responses, as shown in Fig. 7. First, the designed controller succeeded in providing a steady current with a RMSE of 0.070 A (see Fig. 7(a)). Second, the controlled system can track different current commands (see Fig. 7(b), where the load-meter was set at a constant load of 50 V). That is, the PEMFC can be controlled to provide steady current. This is beneficial because varying loads may cause performance degradation [25,26].
Power management We applied the PEMFC to the hybrid power system shown in Fig. 1, and developed power management strategies by considering the following issues: (1) the PEMFC supplies auxiliary power when the solar and battery cannot provide sufficient electricity to the loads. In addition, we control the PEMFC to supply a steady current of 20 A if possible, because it achieves the maximum energy efficiency (see Fig. 7(c)). (2) High charging current to the battery set should be avoided because it might damage the battery [27]. (3) We limited the SOC operating ranges to 45e80% to prevent over-charging and over-discharging. Referring to Fig. 1, we monitored the current Ico and Ibatt, and calculated the battery SOC using the Coulomb counting method. The following power management strategies were developed:
4637
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 0 ( 2 0 1 5 ) 4 6 3 0 e4 6 4 0
(2) When the SOC continuously drops to 45%, the required load (Iload Isolar ¼ Ico þ Ibatt) is high and the battery discharges rather than charges. Therefore, we increase the PEMFC current Ifc according to Fig. 8 to charge the battery until the SOC reaches 50%. In addition, we set a threshold of 2 A to avoid chattering. (3) The safety protection will be activated by one of the following conditions [13]: (i) Low pressure protection: when the stack input pressure is less than 0.2 barg. (ii) Low voltage protection: when Vfc is less than 40 V. (iii) Over temperature protection: when Tstack is greater than 70 C. (iv) Over load protection: when Ifc is greater than 70 A. (v) Low battery protection: when Vbatt is less than 48 V.
SimPowerSystem model This section develops a simulation model using the Matlab™ SimPowerSystem [28], as shown in Fig. 9, for performance verification. The model can be used in the future to evaluate power management strategies and to select system components for developing customized hybrid power systems.
PEMFC model verification
Fig. 10 e Verification of the PEMFC model.
(1) When the battery SOC drops to 50%, the PEMFC system is started to supply a steady current of 20 A until the SOC reaches 80%.
We first developed the PEMFC sub-system, which consists of a PEMFC stack, the fuel and air supply, and the thermal model. We tuned the corresponding parameters and compared the simulation and experimental responses in Fig. 10 and Table 5. Note that the vertical spikes in the PEMFC flow-rate responses were caused by purges, which are necessary to avoid flooding and to keep suitable humidity in the stack [29]. We activated the purge function in both experiments and simulation, when the PEMFC output reached 2300 As according to the manufacturer's suggestion. First, the simulation and experimental responses are similar. That is, the simulation model correctly represented the PEMFC dynamics. Second, the designed controller can effectively regulate the PEMFC operating temperature according to Fig. 5(b). That is, the PEMFC can be controlled to reach the highest energy density. Last, the simulation model achieves similar hydrogen efficiency and PEMFC efficiency to that obtained experimentally, as illustrated in Table 5. Therefore, we can use this PEMFC sub-system to develop the hybrid power system model.
Table 5 e PEMFC model verification by Fig. 10. Simulation Load(A) o
RMSE ( C) Hydrogen efficiency (%) [30]. PEMFC efficiency (%)
Experiment
20 A
40 A
60 A
20 A
40 A
60 A
0.0136 98.04 57.51
0.0109 98.65 53.37
0.0049 98.88 48.28
0.0153 98.10 56.97
0.0525 98.38 53.67
0.0528 98.77 48.22
4638
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 0 ( 2 0 1 5 ) 4 6 3 0 e4 6 4 0
Fig. 11 e Responses of the integrated system.
Hybrid power model verification We further integrate the hybrid power model (see Fig. 9) which is composed of the PEMFC sub-system, a battery sub-system, a solar sub-system, a DC/DC converter, an electric loader, and an energy management system. We conducted the
experiments when the solar power was high, and obtained the following experimental data: the lab load, solar power, PEMFC responses, and battery responses, as illustrated in Fig. 11. Then we applied the recorded load and solar power to the simulation model, and compared the system responses in Fig. 11(c) and (d) and Table 6. First, the hybrid system
Table 6 e Hybrid model verification by Fig. 11.
Fuel cell
Solar cell Battery Load
Energy production (MJ) H2 consumption (L) Operating time (sec) Average current (A) Energy density (J/L) Energy production (MJ) Initial SOC (%) Final SOC (%) Energy consumption (MJ)
Simulation
Experiment
10.26 2311 10,295 19.98 4439 3.66 100 76.21 10.92
10.25 2306 10,290 19.96 4445 3.66 100 75.53 10.92
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 0 ( 2 0 1 5 ) 4 6 3 0 e4 6 4 0
and solar power for 24 h, as shown in Fig. 12, where the power disruptions were a result of the refrigerator compressor. We applied the following four scenarios to discuss the impacts of charging strategies:
3000
Power (W)
2500 2000 1500 1000 500 0
0
5
10
15
20
15
20
Time (hr) 1000 800 Power (W)
4639
600 400 200 0
0
5
10 Time (Hr)
Fig. 12 e The 24-h loading cycle and solar power.
succeeded in providing sustainable power for the lab loads. And the PEMFC was started up to provide a constant current of 20 A when the battery SOC dropped to 50%, according to the management strategies. Second, the solar power module can supply up to 33% of the auxiliary energy during the test, as shown in Table 6. Therefore, we can implement more solar panels to further reduce the hydrogen consumption in the future. Last, the simulation and experimental responses are similar, with a deviation of 0.21% for hydrogen consumption and 0.13% for energy density. Hence, the simulation model can be applied to predict the system responses in the future, because it is very time-consuming and expensive to conduct experiments at high power. Note that the spikes in the PEMFC flow-rate responses in Fig. 11(c) were caused by purges.
The impacts of power management strategies The developed simulation model can effectively predict the experimental responses; therefore, it can be applied to estimate the impacts of different power management strategies because conducting long-time experiments can be very costly and time-consuming. For example, we recorded the lab loads
(1) Strategy 1 applies the power management strategies described in Section Power management. That is, the PEMFC starts to supply a steady current of 20 A when the battery SOC drops to 50%, and stops when the SOC reaches 80%. If the SOC drops to 45%, we gradually increase the PEMFC current Ifc according to Fig. 8. (2) Strategy 2 is similar to Strategy 1: the PEMFC starts to supply a steady current of 20 A when the battery SOC drops to 50%, but directly increase Ifc to 50 A when the SOC drops to 45% until it reaches 50%. (3) Strategy 3 starts the PEMFC to supply a constant Ifc ¼ 50 A when the battery SOC drops to 45%, and stops until it reaches 80%. (4) Strategy 4 is similar to Strategy 1, but with a smaller SOC operating range of 45%e60%. The simulation results are compared in Table 7. First, all strategies can provide continuous power for the applied loads. Second, strategy 1 can effectively reduce the hydrogen consumption because it operates the PEMFC at the highest efficiency (20 A) most of the time. That is, strategy 1 provides the best energy density, with about 11% hydrogen consumption improvement compared with Strategy 3. Third, comparing Strategies 1 and 4, a larger SOC operating range can slightly increase the energy density. This is because the larger SOC range can extend the operating time and allow the PEMFC to operate at high efficiency (20 A). In addition, Strategy 1 can reduce the number of PEMFC restarts to avoid performance degradation. We can also apply the simulation model to evaluate the impacts of resizing system components, such as the battery set and solar panel. For example, if we increase the battery capacity from 15 Ah to 75 Ah, the hydrogen consumption can be reduced by 125 L by the simulation. On the other hand, the hydrogen consumption can also be reduced by about 10% with a 660 W solar panel, and 20% with a 1320 W solar panel. These model adjustment can be considered in the future when developing customized hybrid power systems.
Table 7 e Comparison of different power management strategies.
Fuel cell
Solar cell Battery Load
Energy production (MJ) H2 consumption (L) Operating time (sec) Number of starts Average current (A) Energy density (J/L) Energy production (MJ) Initial SOC (%) Final SOC (%) Energy consumption (MJ)
Strategy 1
Strategy 2
Strategy 3
Strategy 4
91.28 20,342 74,535 11 24.51 4487 10.33 100 80.05 101.87
91.37 20,415 74,521 11 24.61 4475 10.33 100 80.05 101.87
93.38 22,600 41,681 48 48.74 4132 10.33 100 80.05 101.87
91.23 20,342 74,581 29 24.53 4485 10.33 100 80.05 101.87
4640
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 0 ( 2 0 1 5 ) 4 6 3 0 e4 6 4 0
Concluding remarks [9]
This paper has developed a PEMFC hybrid power system that consisted of the PEMFC, secondary battery set, and solar module. First, we designed robust controllers to regulate the PEMFC temperature and output current, with RMSEs of 0.06 C and 0.07 A, respectively. And the average energy density was improved by 51.7 J/L by the proposed control. Second, we integrated the hybrid power system and designed power management strategies to guarantee sustainable operation. Last, we built a SimPowerSystem model and tuned the parameters based on experimental results. The simulation results were shown to comply with the experiments, with a maximum deviation of 0.21%, so that the simulation model can be used to predict the experimental responses. For example, we discussed the impacts of different power management strategies on system efficiencies. The results indicate that the system performance can be improved by PEMFC control, and the hydrogen consumption was improved by up-to 11% by appropriate power management strategies. In the future, the developed simulation model can be applied to design customized hybrid power systems employing different energy sources.
[10] [11]
[12]
[13]
[14] [15]
[16] [17]
[18]
Acknowledgment
[19]
This work was financially supported in part by the National Science Council Taiwan under Grand NSC 103eETeEe002e002eET.
[20]
references
[1] Fabio M, Fausto T, Luca O, Paola GS, Giovanni D, Stefano L, et al. PEMFC system simulation in MATLAB-Simulink environment. Int J Hydrogen Energy 2011;36:8045e52. [2] Abraham U, Chavez R, Roberto MG, Duron-Torres SM, Ferraro M, Brunaccini G, et al. High power fuel cell simulator based on artificial neural network. Int J Hydrogen Energy 2010;35:12125e33. [3] Saddi A, Becherif M, Aboubou A, Ayad MY. Comparison of proton exchange membrane fuel cell static models. Renew Energy 2013;56:64e71. [4] Wang FC, Chen HT, Yang YP, Yen JY. Multivariable robust control of a proton exchange membrane fuel cell system. J Power Sources 2008;177:393e403. [5] Tirnovan R, Giurgea S. Efficiency improvement of a PEMFC power source by optimization of the air management. Int J Hydrogen Energy 2012;37:7745e56. [6] Kurz T, Keller J. Heat management in a portable high temperature PEM fuel cell module with open cathode. Fuel Cells 2011;11:518e25. [7] Liso V, Nielsen MP, Kær SK, Mortensen HH. Thermal modeling and temperature control of a PEM fuel cell system for forklift applications. Int J Hydrogen Energy 2014;39:8410e20. [8] Cedric D, Michel B, Brigitte GP, Jean PC, Bruno GP. A novel non-linear model-based strategy to improve PEMFC water
[21]
[22]
[23] [24]
[25]
[26]
[27]
[28]
[29]
[30]
management e the flatness-based approach. Int J Hydrogen Energy 2015;40:2371e6. Wang FC, Chen HT. Design and implementation of fixedorder robust controllers for a proton exchange membrane fuel cell system. Int J Hydrogen Energy 2009;34:2705e17. Wang FC, Ko CC. Multivariable robust PID control for a PEMFC system. Int J Hydrogen Energy 2010;35:10437e45. Dursun E, Kilic O. Comparative evaluation of different power management strategies of a stand-alone PV/Wind/PEMFC hybrid power system. Int J Electr Power Energy Syst 2012;34:81e9. Andaloro L, Napoli G. Design of a hybrid electric fuel cell power train for an urban bus. Int J Hydrogen Energy 2013;38:7725e32. Wang FC, Kuo PC. Control design and power management of a stationary PEMFC hybrid power system. Int J Hydrogen Energy 2013;38:5845e56. A123 systems. Available at: http://a123systems.com/. ASEC-220G6S. Available at: http://cdn.photovoltaik-web.de/ media/mod_solarmoduledb/Datenblaetter/Apollo_Solar_ Energy_Asec-200G6S-2BB_Asec-220G6S-2BB.pdf. TS-3000e148A. http://www.meanwell.com/search/ts-3000/ TS-3000-spec.pdf. Goethals I, Pelckmans K, Suykens JA, De Moor B. Subspace identification of Hammerstein systems using least squares support vector machines. IEEE Trans Autom Control 2005;50:1509e19. Van Overschee P, De Moor B. N4SID-Subspace algorithms for the identification of combined deterministic-stochastic system. Automatica 1994;30:75e93. Georgiou TT, Smith MC. Optimal robustness in the gap metric. IEEE Trans Autom Control 1990;35:673e86. Wang FC, Guo YF. Robustness analyses of PEMFC systems on the production line. Int J Hydrogen Energy 2015;40:1959e66. Glover K, McFarlane DC. Robust controller design using normalised coprime factor plant descriptions. SpringerVerlag New York, Inc; 1989. McFarlane DC, Glover K. A loop-shaping design procedure using H∞ synthesis. IEEE Trans Autom Control 1992;37:759e69. Fuel cell formulary. http://www.pemfc.de/FCF_Smart.pdf. Ballard Mark1020 ACSTTM PEMFC stack. Available at: http:// 140.112.14.7/~sic/PaperMaterial/MAN5100192_new% 20guide_mark1020.pdf. Lin R, Li B, Hou YP, Ma JM. Investigation of dynamic driving cycle effect on performance degradation and microstructure change of PEM fuel cell. Int J Hydrogen Energy 2009;34:2369e76. Yan X, Hou M, Sun L, Cheng H, Hong Y, Liang D, et al. The study on transient characteristic of proton exchange membrane fuel cell stack during dynamic loading. J Power Sources 2007;163:966e70. Klein R, Chaturvedi NA, Christensen J, Ahmed J, Findeisen R, Kojic A. Optimal charging strategies in lithium-ion battery. In: Proceedings of the 2011 American control conference; 2011 Jun 29eJuly 1; San Francisco, USA; 2011. p. 382e7. Njoya S, Tremblay O, Dessaint LA. A generic fuel cell model for the simulation of fuel cell vehicles. IEEE Veh Power Propuls Conf VPPC 2009;1e3:1501e8. Chen J, Jason BS, Anna GS, James RW. Optimization of purge cycle for dead-ended anode fuel cell operation. Int J Hydrogen Energy 2013;38:5092e105. Wang FC, Peng CH. The development of an exchangeable PEMFC power module for electric vehicles. Int J Hydrogen Energy 2014;39:3855e67.