Fuzzy Logic Controller Development of a Hybrid Fuel Cell-Battery Auxiliary Power Unit for Remote Applications Benjamin Blunier∗ , Marcelo G. Sim˜oes†, Abdellatif Miraoui∗ ∗ Transport
and Systems Laboratory (SeT) – EA 3317/UTBM University of Technology of Belfort-Montb´eliard, France Emails: (benjamin.blunier, abdellatif.miraoui)@utbm.fr † Colorado School of Mines, 1610 Illinois St.Golden CO 80401-1887, U.S. e-mail:
[email protected]
Abstract—This paper presents the analysis and design of a hybrid fuel cell battery auxiliary power unit (APU) for remote applications implemented with virtual and rapid prototyping techniques. The fuel cell is the main energy source working in low frequency bandwidth while the battery compensates the transient peak power needs. The control of the fuel cell system using two techniques, the first-order based control where the fuel cell delivers the filtered load current and a fuzzy based control. The simulation show that the last one permits the fuel cell to working most of the time in its optimal working zone as well as maintaining the battery state-of-charge in its optimal windows whatever the initial sate-of-charge.
I. I NTRODUCTION This paper introduces the design and simulations of a hybrid auxiliary power unit (APU) for decentralised energy production such as telecom or remote systems applications. Fuel cells are well adapted for such applications as they provide a very good efficiency, low noise and high energy density to provide continuous power during a long time [1]. It is based on a structure using a 300 watts fuel cell stack connected to a Ni-MH battery through a DC/DC buck converter. The NiMH battery is a good solution in terms of energy and power density but it could be replaced in the future by Lithiumbased batteries. Fuel cells are not reversible and have big time transients which are not fully compatible with any kind of application, that is why fuel cells are always combined with a reversible power or energy source (supercapacitors or battery) which can handle peaks of power. For a first approach, a rather simple but robust structure was chosen to validate the relevance of using a fuel cell system in such a system. The aim is to have the most simple and less expensive system which does not need any expensive components and big computation power. In a second step, structures similar to the ones introduced by the authors of [2] will be investigated. Wang [3], [4] propose a low-cost quasiresonant dc-dc converter for fuel cell that could also be used to improve the system efficiency compared to a simple buck converter keeping the costs relatively low. In the actual structure, the batteries provide the peaks of power and the fuel cell is sized to provide nearly the average
Fig. 1.
Fuel cell battery hybrid Auxiliary Power Unit
power. The paper is organized as follows: in a first part, an overview of the architecture and the components of the APU is given. The second part introduces the virtual and rapid prototyping techniques to design the system. The third part presents the modeling of the system and its control. The last part presents the simulation and experimentation results. II. S YSTEM OVERVIEW The figure 2(a) gives an overview of the APU. The fuel cell used in this application is a Polymer Electrolyte Fuel Cell (PEFC). It works at low temperature and at atmospheric pressure on the cathode side. Its maximum power is about 300 watts and the voltage is between 30 V and 58 V. A DC/DC buck converter between the stack and the DC-link allows the fuel cell current to be controlled. The Ni-MH battery imposes the DC bus voltage to be around 24 volts depending on the state of charge of the battery: the system is designed to allow any batteries with a voltage between 9 V and 36 V to be connected. The signal conditioning and processing and the control are done with a microautobox from dSPACE embedded into the APU (see Fig. 2(b)). This system allows rapid prototyping and implementation in a embedded real time system the control laws developed under Matlab-Simulink.
Schematic view of the real system A
purge vanne + B.M.S
V
-
V
Load
A PWM generator
DC/DC
SoC Iload
PWM out
analog inputs Tstack
Istack
αtemp
αfans Air supply control
(a) Electrical architecture overview
digital PWM serial com. analog in. outputs out Battery α SoC
Dspace board interface
Purge on/off
αtot +
Vstack
Vbatt
Temperature supervision
Purges management ref Vstack
Fan control strategy
Vstack
Fuzzy logic controller
First order filter
Energy management strategy
Ni-Mh Battery 300 w fuel cell stack
Power board
Measurement and signal processing board
Dspace board (compiled from simulink code)
Fig. 3.
Rapid prototyping (control implementation on the real system)
1) A first order filter based controller, where the fuel cell current delivers the load current filtered to reduce the fuel cell current dynamics; 2) A fuzzy logic controller to manage the fuel cell current based on the state-of-charge (SoC) of the battery. Micro-autobox
(b) Overview of the real designed and built system Fig. 2. Architecture including a Ni-MH battery and a 300 W fuel cell stack
III. V IRTUAL RAPID PROTOTYPING APPROACH In order to design the fuel cell system module, the virtual prototyping method is used. This method consists to validate all parts of the system, including control and components design (inductance, battery capacity, power electronics, fuel cell power, etc.) using simulation tools. The system model and the control laws (e.g., energy management) are implemented in Matlab-Simulink. The Matlab-Simulink part, validated in simulation is directly downloaded into the dSPACE microautobox. Although the fuel cell system control, including the purges, the air and the temperature managements, is implemented (see Fig. 3), this part is not in the scope of this article; for this, interested reader can refer to the reference [5]. Finally, the virtual prototype permits the energy management strategy (see sec.V-B) to be tested under various driving cycles. This part allows several strategies and set of parameters to be tuned in order to give the best results and check if the parameters respect the system constraints (dynamics, currents, voltage, etc.). In this case, two control strategies for the energy management are tested:
IV. S YSTEM M ODELING AND C ONTROL A complete model of the electrical part of the system is built to test virtually the control laws and energy management strategies before their implementation on the real system. This section presents the fuel cell stack model. In this systems, the battery does not need to be modeled accurately as the real battery has a serial communication interface which can give in real time the SoC of the battery. For the battery model, a simple current integration taking into account charge and discharge efficiencies. A. Fuel cell model As the power required is low, the fuel cell works at atmospheric pressure. The air is supplied with three fans at a high stoichiometric ratio. The fuel cell is air-cooled using the same fans than the air supply. The high stoechiometric ratio ensures that there is nearly no oxygen depletion and no electrodes flooding. As the fuel cell system (energy source) is hybridized with a battery considered here as the peaking power source (PPS) its dynamic does not need to be high. For this reason, but also to increase the lifetime of the fuel cell, the strategy aims at maintaining the fuel cell current dynamics as low as possible (see sec. V-B). Ideally, the fuel cell should deliver a constant power corresponding to the average load power. Considering this remarks, it can be assumed that the fuel cell is almost always in steady state: in this case, a static
model of the fuel cell can be used. For higher powers, the static characteristic cannot be used to describe the fuel cell system: the compressor and the air dynamic (air pressure inside the channels) [6], [7], as well as the water content of the membrane which affects the membrane resistance have to be considered. The authors presented such a model in [8]–[10]. The static characteristic is therefore described as a current (I)-voltage (V ) function [11]: Vstack (I) = E0 − R Istack − A ln(Istack ) + m exp(n Istack ) (1) where E0 , R, A, m and n are empirical coefficients to be determined for the investigated fuel cell stack.
2) First order filter based controller: The overall system control has to give a current reference to the fuel cell: this control has also been successfully demonstrated in a similar system by the author in [5]. The dynamic of the fuel cell can be low as it acts as a range extender. When sudden changes occur in the load power, the fuel cell still works at its previous working point and the battery provides the difference. The fuel cell will take time to reach the desired power: until that time, the battery supplies the load (in the case of a positive current step) or is charged by the fuel cell (in the case of a negative current step). Buck converter bus Istack
Istack
V. F UZZY LOGIC CONTROL
Vstack
A. Fuzzy Control System
+
DC
Fuel cell stack
PI
Istack,ref
V
Load
-
Istack α
Vbatt
+
V
+ −
Most commercial fuzzy products are rule-based systems that receive current information fed back from the device as it operates and controls the operation of a mechanical or other device. A fuzzy logic system has four blocks as shown in Fig. 4. Crisp input information from the device is converted into fuzzy values for each input fuzzy set by the fuzzification block. The universe of discourse of the input variables determines the required scaling for correct per-unit operation.
Itot
Ibatt
DC
Ibatt Vbat Vstack ηcv
bus Istack,ref
1 1+τ p
Power balance
(a) First order filter based controller [5] Buck converter bus Istack
Istack Vstack
+
+
V
DC
Fuel cell stack
+ −
PI
Vbatt V
Load
-
B.M.S
Istack α
Itot
Ibatt
DC
Istack,ref
Fuzzy Controller
Battery SoC
(b) Fuzzy controller Fig. 5.
Fig. 4.
Block diagram of a fuzzy system.
The normalization and scaling are very important because the fuzzy system can be retrofitted for other devices or ranges of operation by just changing the scaling of the input and output. The decision-making logic determines how the fuzzy logic operations are performed (Sup − M in inference), and together with the knowledge base determine the outputs of each fuzzy if − then rule. These are combined and then converted to crisp values with the defuzzification block.
Control strategies
The time with which the fuel cell takes to reach the desired power depends on the control strategy. In this case, the load current (Iload ) is filtered through a first-order transfer function as shown in Fig. 5: the output of this function is the fuel cell bus ) which gives the stack current reference on the DC bus (Istack stack current reference Istack thanks to the DC/DC converter bus , where ηcv is the power balance: Vstack Istack ηcv = Vbatt Istack converter efficiency. This current Istack reference is obviously limited to the maximum current the fuel cell can supply. Over this current, the battery has to supply the remaining current. C. Fuzzy logic controller
B. Control strategy 1) Objectives of the control strategy: The objectives of the control strategy has to take into account the following components constraints: 1) Limits the fuel cell dynamics and voltage cycling (low voltages) to increase its lifetime; 2) Make the fuel cell working around the best working points (where its efficiency is the highest); 3) Maintain the battery State-of-Charge (SoC) in a given window and limits the cycling; 4) Recover the energy during negative load current.
1) Parameters: The parameters of the fuzzy controller are tunable by the user and are represented by trapezoidal functions as shown in Fig. 6. The parameters of the battery state-of-charge management and fuel cell current are given in TABLE. I and TABLE. II, repectively. 2) Rules: The implemented rules are the followings: 1) If (SOC is toolow) then (IFC is toohigh) 2) If (SOC is low) then (IFC is high) 3) If (SOC is good) then (IFC is opt) 4) If (SOC is high) then (IFC is low)
300
Experimental
Interpolation
20 Bursts of current
250
15
Never reached
Optimal power zone
150
Load current (A)
P (W)
200
100 50 0
0
1 Ze
2
3 Low
4
5 I (A) Optimal
6
8
7
9
10
5
10
0
High
Zero load current −5 0
1
2
3
4
Time (s)
(a) Random current load profile generation
Fuel cell current membership function
200
Working zones of the fuel cell system
Description Emergency case SoC is considered to be low SoC is considered to be good (optimal) SoC is considered to be high (the system can work in charge depleting mode in this case : only the battery is working)
Notation SoClimit SoClowmin SoClowmax SoCgoodmin SoCgoodmax SoChighmin
SoC Value 0.1 0.2 0.3 0.45 0.7 0.8
SoChighmax
1
180 160 Average power
Fig. 6.
5 4
x 10
140 120 100 80
TABLE I BATTERY STATE - OF - CHARGE MEMBERSHIP FUNCTION PARAMETERS ( CAN BE EASILY TUNED BY THE USER )
60 0
1
2
3
4
Time (s)
5 4
x 10
(b) Average power along the cycle
VI. S IMULATION AND E XPERIMENTAL R ESULTS
Fig. 7.
Current and average power of the load along the cycle
A. Random load profile generation The load current profile (see Fig. 7) is randomly generated (uniform distribution) to give a average current of 7 A with some long current bursts corresponding to a power much more higher than the fuel cell power. The current profiles can also contain some zero-current during a long time. The average power long the total profile is 150 W as shown in Fig. 7(b). Description Fuel cell maximum current allowed Fuel cell current is considered to be low Fuel cell current for which the fuel cell has been designed (corresponding to the average power of the load) Fuel cell current is considered to be high (used in emergency cases when the battery SoC is too low)
Notation F Ccurrentmax F Ccurrent lowmin F Ccurrent lowmax F Ccurrent optmin
Current (A) 8 2 3 4
F Ccurrent F Ccurrent
optmax highmin
6 7
F Ccurrent
highmax
7.5
TABLE II F UEL CELL CURRENT MEMBERSHIP FUNCTION PARAMETERS ( CAN BE EASILY TUNED BY THE USER )
B. Influence on the battery-converter efficiency The first order control strategy permits the fuel cell to give the load average power along the cycle and permits the battery SoC to be maintained constant (see Fig. 8(a)). However, it works well if the battery-converter efficiency is 100 % but if this efficiency is below 100 % as shown in Fig. 8(b), the SoC cannot be maintained constant because of the losses between the converter and the battery. C. Influence on the initial battery state-of-charge Fig. 9 shows that the first-order based controller does not permit the SoC to be recovered if it is low at the beginning of the cycle as shown in Fig. 9(b) where the initial SoC is very low. D. Comparison between the two control strategies on the fuel cell power distribution Fig. 10 shows that both of the control strategies permits the fuel cell to work around the optimal working zone of the fuel cell. However, using the first order based control (Fig. 10(a)), the fuel cell shuts are works at full power down very often. This strategy may decrease the fuel cell lifetime.
1
1 High
0.9 0.8
0.8
0.7
0.7 0.6
Good SoC
SoC
0.6 0.5
High
0.9
0.4
Good
0.5 0.4
0.3
0.3 Low
Low
0.2
0.2
0.1
0.1 Too low
0 0
1
2
3
4
time (s)
Too low 0 0
5
2
3
4
time (s)
4
x 10
(a) First order filter based control and 100 % battery-converter efficiency (τ = 30 s)
4
x 10
1 High
0.9
0.8
0.7
0.7
0.6
0.6 SoC
Good
0.5 0.4
High
0.9
0.8
Good
0.5 0.4
0.3
0.3 Low
0.2
Low 0.2
0.1
0.1 Too low
0 0
5
(a) First order filter based control and 90 % battery-converter efficiency,τ = 100 s, and initial SoCinit = 1
1
SoC
1
1
2
3 time (s)
4
5 4
x 10
Too low 0 0
2000
4000
6000 time (s)
8000
10000
(b) First order filter based control and 90 % battery-converter efficiency (τ = 30 s)
(b) First order filter based control and 90 % battery-converter efficiency,τ = 100 s, and initial SoCinit = 0.5
Fig. 8. Influence of the system efficiency on the battery SoC profile along the load cycle for the first-order filter based control strategy. The first order filter control strategy does not permit the state-of-charge to be maintained along the cycle.
Fig. 9. Influence of the initial battery state-of-charge on the SoC profile along the cycle. The first order filter control strategy does not permit the state-of-charge to be recovered and maintained along the cycle.
On the contrary, the fuzzy logic controller permits the fuel cell to work exactly at its optimal power with very few working point at zero power and high power whatever the initial state-of-charge. E. Battery state-of-charge profile with the fuzzy controller Moreover, as shown in Fig. 11 the fuel cell is shut down only one time during a long time because the fuel cell power is not anymore needed: the system can work only on the battery during this time. VII. C ONCLUSION A hybrid fuel cell/battery APU has been modeled and designed. The fuel cell in this system acts as the energy source allowing this component to have a low dynamic. The control of the fuel cell system and the control of the overall system has been presented and validated experimentaly on the system. Tests and simulation results will be given in the final paper: they have shown that a very simple control strategy which does not require a high computation power can give very good results the energy management of a hybrid battery-fuel cell
system. Fuzzy logic control showed to improve the efficiency of the system by concentrating the region of operation in the best performance range and to minimizing the shut-down and over-current conditions. With Fuzzy Logic it was also possible to make the system recover the state-of-charge conditions without any extra state machine decision making procedures. Therefore, a fuzzy logic based APU system is expected to have a longer lifetime and best dynamic performance than traditional APU controllers. R EFERENCES [1] B. Blunier and A. Miraoui, 20 Questions sur la pile combustible. Technip, 2008, (in French). [2] A. Khaligh, A. Rahimi, Y. Lee, J. Cia, A. Emadi, and S. Andrews, “Digital control of an isolated active hybrid fuel cell/li-ion battery power supply,” IEEE Transactions on Vehicular technology, vol. 56, no. 6, november 2007. [3] S. Wang, M. Krishnamurthy, R. Jayabalan, and B. Fahimi, “Quasiresonant dc/dc converter for high power fuel cells systems,” in Power Electronics Specialists Conference, 2006. PESC ’06. 37th IEEE, 18-22 June 2006, pp. 1–7. [4] ——, “Low-cost quasi-resonant dc-dc converter for fuel cells with enhanced efficiency,” in Applied Power Electronics Conference and Exposition, 2006. APEC ’06. Twenty-First Annual IEEE, 19-23 March 2006, p. 6pp.
Battery SoC
1
8 7
Low 0 0
1
2
Too low 5
4
4
x 10
8
4 3
High
6 Optimal 4 Low
2
Ze
2
0 0
1
2
3
4
time (s)
1
50
100 150 200 Fuel cell power (W)
250
300
5 4
x 10
(a) SoC and fuel cell current profiles with the fuzzy logic controller (SoCinit = 1) 1
(a) First order filter based control and 90 % battery-converter efficiency,τ = 30 s, and initial SoCinit = 1 35
Battery SoC
0 0
3 time (s)
Optimal zone
5
Good
0.5
FC current
Pourcent of time (%)
6
Too many working points at maximum power
Too many working points at zero power (on/off cycles)
High
High Good
0.5
Low
Optimal zone 0 0
30
1
2
3
4
x 10
8 FC current
Pourcent of time (%)
25
Too low 5
4
time (s)
20 15
High
6 Optimal 4 Low
2
Ze 0 0
10
2
3 time (s)
4
5 4
x 10
(b) SoC and fuel cell current profiles with the fuzzy logic controller (SoCinit = 0.5)
5 0 0
1
50
100 150 Fuel cell power (W)
200
250
(b) Fuzzy logic controller and 90 % battery-converter efficiency and initial SoCinit = 1
Fig. 11. [SoC and fuel cell current profiles with the fuzzy logic controller. The controller permit the battery SoC to be maintained and recovered whatever the initial stage of charge
22 Optimal zone 20
Pourcent of time (%)
18 16 14 12 10 8 6 4 2 0 0
50
100 150 Fuel cell power (W)
200
250
(c) Fuzzy logic controller and 90 % battery-converter efficiency and initial SoCinit = 0.5 Fig. 10. Comparison between the first-order filter based control strategy and the fuzzy logic controller. The first order filter permits the fuel cell to work mostly around the average power but there are a lot of working point at zero power and maximum power (cycling). On the contrary the fuzzy logic controller makes the fuel cell works around the average power and sometime the fuel cell is turned of (when the SoC is high and the load current is low)
[5] D. Bouquain, B. Blunier, and A. Miraoui, “A hybrid fuel cell/battery wheelchair – modeling, simulation and experimentation,” in Vehicle Power and Propulsion Conference, 2008. VPPC ’08. IEEE, 3-5 Sept. 2008, pp. 1–6. [6] B. Blunier, M. Pucci, G. Cirrincione, and A. Miraoui, “A scroll compressor with a high-performance induction motor drive for the air management of a pemfc system for automotive applications,” IEEE Transactions on Industrial Applications, vol. 44, no. 6, pp. 1966–1976, Nov.–dec. 2008. [7] B. Blunier, M. Pucci, G. Cirrincione, M. Cirrincione, and A. Miraoui, “A scroll compressor with a high-performance sensorless induction motor drive for the air management of a pemfc system for automotive applications,” IEEE Transactions on Vehicular Technology, vol. 57, no. 6, pp. 3413–3427, Nov. 2008. [8] B. Blunier and A. Miraoui, “Modelling of fuel cells using multi-domain vhdl-ams language,” Journal of Power Sources, vol. 177, no. 2, pp. 434– 450, 2007. [9] F. Gao, B. Blunier, A. Miraoui, and A. El-Moudni, “Cell layer level generalized dynamic modeling of a pemfc stack using vhdl-ams language,” International Journal of Hydrogen Energy, vol. 34, no. 13, pp. 5498 – 5521, 2009. [10] F. Gao, B. Blunier, A. Miraoui, and A. Moudni, “A multiphysic dynamic 1d model of a proton exchange membrane fuel cell stack for real time simulation,” Industrial Electronics, IEEE Transactions on, vol. PP, no. 99, pp. 1 –1, 2009. [11] B. Blunier and A. Miraoui, Piles a` combustible, Principe, mod´elisation et applications avec exercices et probl`emes corrig´es, ser. Technosup, Ellipses, Ed., 2007, book in French.
