Bidirectional operation of the thermoelectric device for active temperature control of fuel cells

Applied Energy 222 (2018) 410–422

Contents lists available at ScienceDirect

Applied Energy journal homepage: www.elsevier.com/locate/apenergy

Bidirectional operation of the thermoelectric device for active temperature control of fuel cells Trevor Hocksun Kwana, Xiaofeng Wub, Qinghe Yaoa, a b

T

⁎

School of Engineering, Sun Yat-Sen University, China School of AMME, University of Sydney, Sydney, Australia

H I GH L IG H T S

TE device’s dual nature is used for temperature control and energy harvesting. • The TEC mode is used to achieve active temperature control. • The TEG mode is used to achieve energy harvesting about the optimal temperature. • The of the proposed system are conducted on MATLAB/Simscape. • Simulations • An experiment using hardware is presented to verify the proposed concept.

A R T I C LE I N FO

A B S T R A C T

Keywords: Thermoelectric device Thermal management system Fuel cell TEG mode TEC mode

The thermoelectric (TE) device enables a conversion interface between the heat transfer and the electricity domain. Specifically, it can operate bi-directionally – Heat can be converted to electricity via the thermoelectric generator (TEG) effect and vice versa via the thermoelectric cooling (TEC) effect. In most state of the art research, the TE device is operated either in the TEG mode or TEC mode but very seldom in both modes for a single control objective. This paper proposes a thermal management system for a fuel cell who exploits the bi-directional characteristics of the TE device to achieve both temperature control and the possibility for energy harvesting when active control is not required. The studied scenarios involve a time-based simulation involving heat generation levels that are typical of a 500 W rated operating proton exchange membrane fuel cell (PEMFC). The overall dynamic system is simulated using Simscape library components in Simulink and the controller itself is implemented using MATLAB s-functions. An experiment involving electric heaters to emulate the fuel cell’s body heat is also conducted to verify the proposed combined TEG-TEC control approach.

1. Introduction The thermoelectric (TE) device is a renewable energy technology who can generate electricity from a temperature gradient via the Seebeck effect (often also known as the thermoelectric generator (TEG) effect). The TE device can also generate a temperature gradient (hence pump heat power from the cold side to the hot side) if an electric current is applied. Operation in such a mode is based on the Peltier effect and is often known as the thermoelectric cooling (TEC) effect. In comparison to other heating or cooling methods, the TE device is especially recognized in literature in that it is a robust, clean and noiseless electric power generator which does not require any active moving parts [1–3]. In terms of the TEG mode, applications include but not limited to the recovery of waste heat from automotive exhaust

⁎

systems [4,5], the solar thermoelectric hybrid power system [6–8], hypersonic engines [9] and fuel cells [10]. Performance via the Seebeck effect is generally characterized by the thermocouple material charα2σ acteristics and is quantified by using the ZT parameter where ZT = λ . Here, α is the Seebeck coefficient, σ is the electrical conductivity and λ is the thermal conductivity. A larger value of ZT represents better Seebeck effect performance, therefore, research in the thermoelectric material field is committed to increase α , σ while minimizing λ [2,11,12]. In addition to materials research, the geometric design of the thermocouples, number of thermocouples and the TE device deign itself is also found to have significant impacts on the Seebeck performance [13–15]. For instance, the multi-objective optimization of the TEG in terms of the aforementioned parameters has been presented in various publications such as Refs. [14,16]. Exergy analysis of the TEG device is

Corresponding author. E-mail addresses: [email protected] (T.H. Kwan), [email protected] (X. Wu), [email protected] (Q. Yao).

https://doi.org/10.1016/j.apenergy.2018.04.016 Received 30 December 2017; Received in revised form 3 April 2018; Accepted 5 April 2018 0306-2619/ © 2018 Published by Elsevier Ltd.

Applied Energy 222 (2018) 410–422

T.H. Kwan et al.

Nomenclature

α σ λ rTEG RTEG n nt Np Ns l A TFC T1 T2

Tamb QH QL QL Qα QK

QJ AIn AOut AModule EFC EHS ITE VTE VTE (OC ) i

seebeck co-efficient (V/K) electrical conductivity (S/m) thermal conductivity (W/(mK)) internal electrical resistance of a single thermocouple of the TEG (Ω) total internal electrical resistance of the TEG (Ω) number of thermocouples in a single TE module total number of thermocouples in the overall TE device number of parallel connected strings of TE modules in the device number of series connected strings of TE modules in the device length of Thermocouple (m) surface Area (m2) fuel cell body temperature (K) temperature of TE device that is at the fuel cell side (K) temperature of TE device end that is at the cooling side (K) temperature of the ambient environment (K) total heat flowing from the fuel cell to the TE device (W) total heat flowing from the TE cold side to the flow water of the liquid cooling system (W) total heat flowing from the TE device to the ambient environment (W) heat flowing because of the Seebeck Effect (W) heat flowing because of Fourier thermal conduction (W)

VOC ISC

heat flowing because of the Joules effect (W) total TE contact area with fuel cell (m2 ) total TE contact area with ambient environment (m2 ) surface area of a single TE module (m2 ) fuel cell heat energy storage buffer (J) heat sink heat energy storage buffer (J) TE device operating current (A) TE device operating voltage (V) TE device open circuit voltage (V) current flowing through one string of TE modules in the device (A) open circuit voltage of the TE device (V) short circuit current of the TE device (A)

Subscripts and superscripts p n

p-type type thermopile n-type type thermopile

Abbreviations TE TEG TEC PEMFC LT-PEMFC HT-PEMFC

thermoelectric thermoelectric Generator thermoelectric Cooler proton exchange membrane fuel cell low temperature proton exchange membrane fuel cell high temperature proton exchange membrane fuel cell

operating current in the TEC mode [30]. Similar to the TEG case, performance in the TEC mode is also sensitive to that of the three thermocouple material characteristics α , σ and λ and also the thermocouple numbers and geometric design [26]. The control of the TE device in the TEC mode from the electronics perspective is also considered in Ref. [27] for active temperature control. Moreover, Ref. [31] proposed using the supercooling effect in a two-stage TE device where the supercooling effect involves using large amplitude current pulses to improve the transient response in the TEC mode. Recently, the study of the TEC as a thermoelectric energy conversion unit (TECU) for both heating and cooling has been conducted in Refs. [32,33] by using steady state models. Here, active air cooling is applied to both the cold and hot sides of the TEC module and the subsequent exhausted air are used separately to achieve the respective cooling and heating functions. In the meantime, the proton exchange membrane fuel cell (PEMFC) is attractive for the automotive application where the key features include low operating temperature and low maintenance requirements [34]. However, the PEMFCs have a narrow operating temperature range where a low temperature (LT) variant typically operates within 60 °C to 95 °C and at a nominal value of ≈80 °C [35]. Analogously, the high temperature (HT) variant operates in the range of 120 to 180 degrees Celsius with nominal values at ≈160 °C [36]. Such prescribed operating temperature ranges typically exist as a trade-off between increased efficiency due to the increasing temperature and issues regarding the membrane properties at high temperatures [36]. In addition to the operating temperatures, PEMFCs operate with conversion efficiencies in the order of 40–55% which means a significant amount of heat is generated during operation. Thus, it is important to provide an effective thermal management system for the PEMFC so that it can retain its optimal output efficiencies and reliability. A common cooling technique for the PEMFC is that of using the oxygen supply fan [37] but this method typically lacks versatility because it couples together the cooling and oxygen supply dynamics. Thus, for higher power applications (> 1 kW), external active liquid or gas cooling systems are more commonly adopted [38]. Recently, the application of the TE device into

also popularly considered with example references being that of Refs. [9,17]. Other publications also deal with the electronics aspect of the TEG where Ref. [18] deals with the side effects of mismatch between multiple TE devices connected in series or parallel. It is worth noting most previously presented publications such as the ones aforementioned analyze the TEG by using steady state models. The small signal model of the TEG was also proposed recently in Ref. [19] where it extended the DC model of the TEG into the dynamic regimes. The TEG is also very popularly considered in energy harvesting applications because the device can virtually be used passively to extract electrical energy from a heat source without the requirement of any moving parts. A popular application is that of energy harvesting from a human body in Refs. [20–23], which are used to power portable waistbands that are typically used for health monitoring purposes. Specifically, Ref. [23] focused on using the human body as a heat source to power an accelerometer whereas Ref. [22] investigated the potentials of using inorganic bulk materials for the same application. The TEG has also been applied in Ref. [24] to power sensors that monitor a gas turbine’s health by energy harvesting from the gas turbine itself. In this context, the requirement for long and numerous power cables to supply power to each individual sensor can be eliminated, thus increasing the reliably of the sensing equipment. TEG based energy harvesting systems to harvest energy from asphalt pavements have even been studied recently in Ref. [25] where it was shown that up to 160kWh energy could be recovered in a day for a 1 km length and 10 m width road. Overall, energy harvesting is very beneficial in that it can reduce the charging requirements on batteries, eliminate wired power connections that are otherwise necessary in conventional implementations or even reduce the size requirements of grid level power plants. In terms of the TEC mode, applications include the active cooling of photovoltaic (PV) cells [26], power electronic switches [27], high powered LEDs [28] and fuel cells [29]. Performance characterization in the TEC mode is often quantified using the co-efficient of performance (COP), the maximum cooling capacity and the maximum allowable 411

Applied Energy 222 (2018) 410–422

T.H. Kwan et al.

the fuel cell application has also been considered, with a relatively early example being that of Ref. [29]. In this aspect which involves TEGs, Ref. [39] considers using the TE device as part of a liquid cooling system where it is used in a heat exchanger to remove heat from the circulating water. A recent study also conducted experimental validations on adopting the TE device for cooling the PEMFC body in Ref. [40]. Besides the PEMFC, the TEG and TEC modules have also been considered for the cooling of other types of fuel cell’s bodies such as the solid oxide fuel cell (SOFC) in Ref. [41], the alkaline fuel cell in Ref. [42] and the direct carbon fuel cell in Ref. [43]. In these research, parametric studies of the various TEG properties were conducted and the results are presented using power versus current density curves. TEG and TEC modules combined cooling systems have also been proposed in recent literature. Typically, such systems involve electrically connecting the TEG module to the TEC module which means the output of the TEG module is used to further pump a temperature differential at the TEC and provide hence an additional cooling effect. An example of where such a system has been applied includes that of Ref. [41] for cooling of a SOFC. Ref. [31] proposed adopting the maximum power point tracking (MPPT) technique on the TEG module so that the power that is delivered to the TEC module in this type of cooling system is maximized. However, literature on adopting the same TE modules in a bidirectional nature is very shallow where the only well-known example may be that of Ref. [44]. Indeed, applying the TE devices in such a manner is not well known in the fuel cell application. Operating the TEG bidirectionally has several major benefits over that of operating only in either modes. For instance, while the TEG mode is advantageous in that it optimally harvests energy from the main device, it is effectively a passive cooling technique and is heavily dependent on the appropriate open loop design of the cooling system. In the meantime, although the TEC mode is advantageous in that active temperature control is made possible, this mode of operation is an active mode and thus requires power from an external source. This paper proposes an innovative TE device based hybrid thermal management system that, with a small increase in system complexity, enables the flexible choice of whether the TE device can be operated in either the TEG or the TEC mode. Here, the key benefits of both modes are better utilized and the impact of their disadvantages are minimized. For instance, the TEG mode is adopted to achieve energy harvesting whenever heating and cooling is not required to maintain the main device’s temperature at the reference value. It is worth noting that the MPPT technique is proposed to be adopted in this mode to maximize the energy harvesting capability. In the contrary, if heating or cooling is required or if the device is operating too far from its optimal temperature range, the TEC mode is adopted. The only well-known example in literature that studied using the same TE device in both the TEG and TEC effect is that Ref. [44]. However, this study bases the analysis of the TEG and TEC performance on separate open loop steady state simulations and does not provide a control method that can actively select whether the TE device should operate in the TEG or TEC mode. Hence, to overcome the limitations of the common previously proposed steady state TEG/TEC models, this paper also proposes a timedependent Simscape model in Simulink to simulate the TE device. The proposed model not only couples together the TE device’s thermal and electrical characteristics in the time domain, but is also highly flexible in that the same model can be used to analyze the device in terms of both the TEG and TEC characteristics. The thermal management system is implemented using MATLAB s-functions where it can switch between the two modes depending on the temperature control requirement and the current state. The presented study will be based on a 500 W rated LT-PEMFC where the LT-PEMFC has a tight operating temperature range (60–95 degrees Celsius) while generating a significant amount of body heat. An experimental verification will be conducted using a power management electronic system that supports the proposed bidirectional control technique and an electric heater that emulates the fuel cell’s heating dynamics.

Fig. 1. The thermal model of the TE device as implemented using Simscape components.

2. System model 2.1. Thermodynamic model The Simscape library is used to simulate the dynamic model because it supports various types of environmental domains and thus enables the capability of a multi-physics simulation. In the context of this paper, the relevant physical environments are that of the thermal, electrical and fluid flow domains. Fig. 1 shows the structure of the thermodynamic model of the TE cooling system as implemented using Simscape components. This model consists of three major components which are respectively the fuel cell thermodynamic model, the TE model and the cooling model. Indeed, a realistic model of the fuel cell would require another multi-physics simulation that additionally involves the fluid dynamics of reactants and products and the production of electricity by the fuel cell itself [41]. On the other hand, the analysis of the TE cooling system only requires the thermodynamic knowledge of the fuel cell. As such, for the sake of brevity, the fuel cell thermodynamic model is simplified and consists of an internally generated body heat QH and a thermal energy storage element (EFC ). The value of QH depends on the fuel cell operating characteristics and thus can be considered as an arbitrary input but indeed, values that correspond to a realistic fuel cell should be chosen. The selection of QH as a function of time for analysis of the proposed combined control method will be discussed in Section 4. The TE device model of Fig. 1 consists of the three most important TE dynamic effects. Specifically, Qα refers to the heat that flows because of the Seebeck effect where α is the Seebeck co-efficient of the thermocouple material. QK refers to the heat flowing because of Fourier thermal conduction and QJ is the Joule heating which is due to the internal electrical resistance of the TE. These expressions are mathematically calculated as according to Eqs. (1), (3) and (4) respectively. The thermal conductance K and electrical resistance rTEG are properties of the thermocouples and are calculated as according to Eqs. (8) and (9) where λ , σ , A , l and n are respectively the thermal conductivity, electrical conductivity, area, length and number of thermocouples. The 412

Applied Energy 222 (2018) 410–422

T.H. Kwan et al.

electronics components such as inductors, capacitors and semiconductor switches which are controlled using very high PWM frequencies (order of 100 kHz to MHz). However, because of the high switching frequencies involved, simulations of power systems at this level will require extremely small-time steps and thus a heavy computational burden. Hence, for the sake of achieving a computationally efficient simulation, a controllable voltage source is adopted. Essentially, the values of Vo are updated in each control time step such that it reflects the steady state characteristics of a real bi-directional converter and load. This model is very computational simple and as such, can be easily integrated with the TE thermal model without impacting the required simulation time. The appropriate value of Vo is dependent on the method of control, thus the details of this is provided in Section 3.

subscripts p and n refer to the positive and negative thermocouples.

Qα = nt αiTAvg

(1)

1 (T1 + T2) 2

(2)

QK = nt K (T1−T2)

(3)

QJ = nt i 2rTEG

(4)

TAvg =

An individual TE module has a surface area AModule , a certain number of thermocouple pairs (n), the thermocouple contact area ( AC ) and length (lc ). Eq. (5) shows the expression to calculate the total surface area of the equivalent TEG device given the area of an individual commercially chosen TEG module AModule . Eq. (6) defines the surface area of the heat sink that is attached to the TEG cold side surface. This value has been assumed to be two times the inlet area based on estimations from the geometry of the connection of the TEG cold side to the flowing water pipes. Eq. (7) calculates the total number of thermocouples in the equivalent TEG module and this parameter will be adopted in the primary thermodynamic model as presented in Eq. (1) to (9).

AIn = NP NS AModule

(5)

AOut = 2AIn

(6)

nt = NP NS n

(7)

K=

λp Ap lp

rTEG =

+

lp σp Ap

λn An ln

+

ln σn An

2.3. Model integration The overall simulation model integrates together the thermodynamic model from Section 2.1 and the electrical circuit model from Section 2.2. This interface is achieved by using a MATLAB s-function whose role is to correctly link together the corresponding parameters of the electrical model and the thermodynamic model. Fig. 3 shows a block diagram that illustrates the interfacing structure where the “Parameters Calculations and Interface” block corresponds to that of the interfacing MATLAB s-function code. Essentially, the inputs and outputs of this block consists of a combination of parameters from both environmental domains. For instance, the inputs include T1 and T2 from the thermodynamic model and ITEG from the electrical circuit model. The outputs of this s-function block are calculated as according to the equations that were presented in Section 2.1 and 2.2. The proposed model has an advantageous benefit where it can simulate the TE device in both the TEG or TEC mode. The TEG mode is engaged whenever the value of VO has the same sign as VTE (OC ) but has a smaller magnitude than it. This occurs because since VTE (OC ) is larger than VO , the resulting current will flow in the direction that charges VO . Otherwise, the TE device will be operating in the TEC mode where the heating mode is applied if VO > 0 and vice versa for the cooling case. The aforementioned relationships have been summarized in Eq. (13).

(8)

(9)

A liquid cooling model is adopted as the cooling system for the cold side of the TEC. The liquid cooling system consists of a long pipe (P1) that runs across the TEC module’s cooling end, a radiator (assumed as another long pipe P2) to reject the heat into the ambient environment (Tamb = 293.5 K) and a water pump to provide the required water flow motion. Referring to Fig. 1, QL denotes the heat that is rejected from the TEC cooling end to the liquid, which involves a combined dynamics of heat conduction from the TEC cooling end to the pipe material and heat convection from the pipe material to the water itself. The term Qamb then denotes the heat that is rejected to the ambient environment via the radiator. τ is applied torque to the water pump and this value is set as a constant in this paper. Its value has been chosen to cause a mass flow rate of approximately 0.15 kg/s where this value is close to that adopted in the experiment of Section 5.

sign (Vo) = sign (VTE (OC ) ) AND |VO | < |VTE (OC ) | ⎧TEG ⎪TEC Heating |V | |V O > TE (OC ) |AND VO > 0 Mode = ⎨TEC Cooling |V | > |V O TE (OC ) |AND VO < 0 ⎪ ⎩ (13)

3. Controller design 2.2. Electrical circuit model The proposed control method operates the TE device in either the TEG or TEC mode and adaptively chooses which mode to use by comparing the current operating state of the system to that of its requirements. The TEC mode is used to actively generate or remove heat

Fig. 2 shows the electrical circuit model which includes both the TE equivalent circuit model and the interface converter. The voltage source VTE (OC ) and resistor RTE both form the standardized equivalent circuit model of the TE device where RTE is calculated as according to Eq. (11). Eq. (10) shows the expression to calculate VTE (OC ) which is based on the standard definition for the Seebeck effect, noting that it also applies to the Peltier effect. The parameter in Eq. (12) denotes the amount of current that flows through a series string of TE modules where ITE is essentially the total current flowing through the entire TE module.

VTE (C ) = NS nα (T1−T2) RTE =

i=

NS × nrTE NP

1 × ITE NP

(10)

(11) (12)

Fig. 2. The TEG and converter electrical model as implemented using Simscape components.

A realistic bi-directional DC/DC converter will consist of power 413

Applied Energy 222 (2018) 410–422

T.H. Kwan et al.

Fig. 3. The block diagram for the parameter interfacing between the thermodynamic model and the electrical circuit model in the Simscape model.

tracker is still climbing the “hill” of the PV curve which means the same direction should be retained.

and should be used whenever the operating PEMFC temperature (TFC ) is outside the required operating range. Conversely, the TEG mode should then be used to extract electrical energy once the PEMFC body is operating within the required temperature range and that heating or cooling is not required to maintain a steady state. The details of the operation of each mode and how they are integrated into the converter model of Fig. 2 are provided in Section 3.1 and 3.2.

3.3. Controller structure Fig. 5 shows the overall structure of the proposed controller to achieve bidirectional energy control of the TE device. The controller contains a Boolean variable WhichMode which registers the value of either zero or one. A value of 0 corresponds to the TEC cooling/heating mode and a value of 1 refers to the TEG power generation mode. The sampling time for the TEC mode is selected as 0.5 s because temperature dynamics are slow in nature and that it takes time to observe significant changes in temperature. On the other hand, the TEG mode’s sampling time is set at a higher value of 0.1 s since the operation of this mode is less reliant on temperature dynamics and more on electrical dynamics. In the TEC mode, the controllable source voltage Vo may be either positive or negative. The former case corresponds to the TE device being in the heating mode where it is attempting to increase the FC body’s temperature while decreasing the coolant liquid’s temperature. The latter corresponds to the TE device’s cooling mode, which indeed, operates oppositely to that of the heating mode. In the meantime, the TEG, the source voltage Vo should be positive for power generation to be possible. It is worth noting that while the simulation model presented in Section 2 involves using controllable voltage source Vo to represent the load (and corresponds to the operating voltage of the TEG), a real hardware system will require power electronic circuits to realize this implementation. First, it is appropriate to separate the operation of the TEC and TEG modes into two separate power circuits which are connected in parallel but are never operated simultaneously. The operation of the appropriate circuit should be decided by the controller based on the WhichMode parameter and a toggle switch should be employed to connect the correct circuit to the TE device from the power source while isolating the other. In the TEC mode, the possibility of a positive and negative value of Vo (to achieve the heating and cooling modes) can be realized by using a converter that enables bi-directional conversion where a common example would be those involving H-Bridge circuits. In the meantime, the TEG mode can be realized by using a MPPT

3.1. TEC mode In this mode, the converter and load model is actively providing electrical power to the TE device which subsequently forces a temperature differential on its two ends. Here, the output voltage source Vo should be regulated using an active control algorithm where the objective is to drive TFC towards its optimal value. A compensator based on classical control theory is adopted to achieve this control purpose. The resulting controller is as per Eq. (14).

Vo 0.005 = 0.05 + 0.02s + T1 s

(14)

On the other hand, when Vo is less than the open circuit voltage of the TE device and has the same sign, the connection of Fig. 3 suggests that the current will flow in the opposite direction and conversely charge the voltage source Vo . This is a sign that the control signal is too weak to support active temperature control and that it is appropriate to switch the operation mode to be the TEG mode. Alternatively, the power consumption can be monitored and if it is observed that the voltage source is being charged then the TEG mode should be engaged to maximize the extractable energy. This characteristic has been implemented in Fig. 5 and will be discussed again in Section 3.3. 3.2. TEG mode The TEG mode should be adopted when TFC is within its optimal range and that heating and cooling is not required for sustainable operation. Here, the converter should act in the MPPT mode so that it can extract the maximum possible power from the TEG for a given thermodynamic condition. Here, any class of MPPT algorithm can be adopted where an extensive survey on this matter is available in example references such as Ref. [45,46]. This paper adopts the perturb and observe (P&O) climbing technique as the example MPPT algorithm which is a well-recognized algorithm in commercial applications for tracking the MPP of solar panels. The pseudocode for this algorithm is illustrated in Fig. 4. Here, Vo is continuously perturbed with small constant voltage steps where the direction is selected based on the change in power between the current and previous time steps. Fig. 4 shows the pseudocode of the P&O algorithm that runs at the nth time step of the sampled loop where parameter Dir should be values of either 1 or −1. The basic concept is that if the change in power since the last time step has become negative, this indicates that the tracker has passed the MPP and that its tracking direction should be reversed (Thus the new value of Dir should be negative of the previous one). Otherwise, the

Begin TimeStep[n] Measure I[n] and V[n] P[n] = I[n] * V[n] If P[n]-P[n-1] < 0 Dir = -Dir End [n] = [n-1] + k * Dir End TimeStep[n] Fig. 4. Pseudocode of the conventional P&O MPPT technique. 414

Applied Energy 222 (2018) 410–422

T.H. Kwan et al.

Fig. 5. The flowchart of the proposed controller for the TE cooling system.

the TEC mode to the TEG mode is that the operating temperature shall be within the prescribed range and the generated power (P) is larger than a small negative tolerance factor (labelled as tol in Fig. 5). Once in the TEG mode, the TE device is now supplying power to the load and the MPPT algorithm should be engaged to extract the maximum possible power. The condition to return to the TEC mode in this case would be when the operating temperature drops outside its required range which, as seen in Fig. 5, is defined as being 340 K and 363 K (or 66.5oC to 89.5oC ). Moreover, the value of TFC (REF ) for TEC control is set as the midpoint of this temperature range which is 351.5 K (77.5oC ) which is assumed to be the nominal operating temperature.

charging circuit, which unlike the TEC mode, can be a unidirectional converter since negative values of Vo is not necessary. It is noted that an example implementation with the aforementioned characteristics is illustrated in the experimental section of this paper (Section 5 and Fig. 11). It is found that by experimentation that the control signal that is required in the TEC mode to achieve the steady state within the required operating temperature range is not often not zero and depends strongly on QH . In a physical sense, the TE device is acting to provide additional heat to the system whenever the generated heat by the PEMFC is insufficient to maintain the prescribed operating temperature. Conversely, the TEG will assist in rejecting the heat towards the heat sink whenever the PEMFC is generating an excessive amount of heat that would otherwise force the PEMFC temperature above its operating range. Based on this discussion, to ensure that the system is stable and does not switch frequently between the two operating modes mode, the TEG mode should only be engaged when the system’s steady state is settled naturally within the optimal temperature range (i.e. heating or cooling is not required). In other words, the control signal that is required by the TEC mode at the steady state should be small enough that its effect on the system is negligible. A condition that indicates this is that, during the TEC mode, when the generated TE electrical power is positive or negatively a small value, this indicates that the voltage control signal Vo is insufficient to overcome the Seebeck effect that is occurring because of the existing TE temperature differential. Subsequently, as shown in Fig. 5, the condition to switch from

4. Simulation 4.1. Simulation parameters In the simulation, a 500 W rated PEMFC is chosen as the study case. As explained in Section 2.1, the fuel cell is modelled as a heat source of QH and a mass thermal energy storage EFC . The specifications of EFC is set based on specifications of a commercial 500 W PEMFC (Ref. [47]) and also tailored slightly to be consistent with the equipment that is used in the experiment of Section 5. Specifically, the mass is set at 2.1Kg and a specific heat capacity of 400 J/(kg K) is adopted. Two different simulation profiles of QH are chosen to analyze the performance of the active controller where the curves are given respectively in Fig. 6(a) and Fig. 9(a). The former simulation profile 415

Applied Energy 222 (2018) 410–422

T.H. Kwan et al.

Fig. 6. The output characteristics of the first simulation (a) PEMFC Operating Temperature (b) The Control Signal Voltage (c) The Load Power (positive is generated electrical power).

involves step changes where the aim here is to assess the response of the proposed control method to large dynamic changes that occur instantly. The latter simulation profile is then used to determine the sensitivity of the proposed control method to slow and gradual dynamic changes.

QH = 300 W). By observing the voltage signal and power curve of Fig. 6(c) and (d), when QH = 500 W, the controller engages the TEC mode in the steady state to provide additional cooling and prevent the PEMFC temperature from reaching values higher than the operating range. Indeed, the most interesting phenomena is when QH = 300 W which is the value that can naturally sustain the PEMFC’s operating temperature within the optimal range for the given cooling system design. Here, heating and cooling of the PEMFC is not required and that the TEC mode is not required. The benefit of using the proposed control method over adopting the TEC mode alone becomes clear during this time. By observing Fig. 8 which is a magnified view of Fig. 6(d) at times 717 s to 737 s, both control methods are outputting a positive power which means power is being generated. It is noted that the reason the TEC mode can generate power is that the control signal voltage is smaller than that of the open circuit voltage of the TE device. Thus, according to model theory, the TE modules can charge the load voltage source. On the other hand, such a scenario is only possible via the implementation of a converter that specifically allows for bi-directional operation. Nevertheless, the electrical power recovered in the TEC only mode is much lower than that of the combined controller because there is no attempt to maximize the electrical energy output of the TE device. In other words, the combined control method is more energy efficient in that it adopts the MPPT method to maximize the electrical energy that

4.2. TE device parameters After a design process, it is determined a TE device that is based on commercial TEC modules (TEC1-12730 [48]) with specifications listed in Table 1 be used in the active cooling system. The design choice was made based on designing the cooling system as a TEG model where the objectives to be optimized are maximum output power when the TEG mode is being operated and ease of achieving the desired temperature control objectives. Another condition that is applied is that the maximum voltage magnitude of Vo is set to be 30 V which was chosen to constrain the maximum amount of power that is given to the TEC. These specifications enable the TEC mode to be operated at power levels of potentially up to 400 W. 4.3. Simulation 1 results Fig. 6 shows the resulting transient responses of both the proposed controller and that utilizing the TEC mode based on the simulation profile of Fig. 6(a). Here, by observing the temperature transient curves shown in Fig. 6(b), both modes are achieving a high performance transient response where the PEMFC operating temperature is almost always kept within the required range of 340 K to 363 K. The required startup time from ambient condition at t = 0 s to the lower temperature value of the operating range is also around 100 s in both cases. On the other hand, the transient response of both modes is different primarily because, in the proposed controller, the TEG mode is engaged at certain times of the simulation. By observing the operating mode profile of Fig. 7 and comparing that to the simulation profile of Fig. 6(a), the TEG mode is engaged when the PEMFC generates a heat of magnitude of 300 W (i.e.

Table 1 TE device specifications.

416

Parameter

Value

NS NP n AIn AC lc

4 1 127 6.2 cm × 6.2 cm 2.4 mm × 2.4 mm 1.3 mm

Applied Energy 222 (2018) 410–422

T.H. Kwan et al.

Table 2 Total effective energy that is supplied by the load to the TEG module. Control method

Energy consumed (kJ)

Combined control TEC mode

126.7 129.1

Table 3 Steady state characteristics for the combined control mode. QH (W)

TFC (K)

Operating mode

PLoad (W)

Control voltage (V)

300 500

347 351.6

TEG TEC

3.55 −117

3.71 −21.5

flexibility for achieving temperature control, if the conditions are not favorable (e.g. when QH < 300 W), a large power consumption will be required. Overall, the output of the TE module is very sensitive to the input QH and that there is a certain combination of QH and thermal resistance to the ambient environment that will enable the TEG mode to be engageable. Subsequently, the recommended design guidelines of the TE cooling system to achieve the best energy efficiency is given as follows:

Fig. 7. The operating mode of the proposed controller in simulation 1.

is recovered during TEG operation. The total energy consumed by the two control methods is provided in Table 2 where the combined control method clearly consumes less energy than that of the TEC mode. While the decrease in energy consumption is not very significant, it is possible to further increase the effectiveness of the combined control method by a careful design consideration of the TE cooling system. Specifically, the TE device design should be such that the TEG mode of the combined control technique can be more frequently adopted. This is achievable by optimizing the TE device design towards the most frequently expected heating power magnitude from the operating PEMFC. Indeed, different steady state values of QH leads to different steady state characteristics of the cooling system. Table 3 shows the steady state characteristics when the combined mode control is used to achieve active temperature control. These results suggest that there is only a small range of QH in which the TEG mode is feasible to operate in the steady state. Furthermore, while the TEC mode has much more

1. Determine the values of QH that are expected to be most frequently used in a practical situation. 2. Design the cooling system for the TE cold side optimally in terms of the TEG characteristics for the given values of QH from step 1. This ensures the TE cooling system can operate in the TEG mode in steady state operation at the nominally selected power levels of QH . 3. Design the TEC control algorithm based on the resulting dynamic model that is obtained by using the cooling system design from step 2. 4.4. Simulation 2 results Fig. 9 shows the resulting transient responses of the combined

Fig. 8. The power generated when the proposed controller and the TEC only mode are applied, zoomed in at times 717 s to 737 s. 417

Applied Energy 222 (2018) 410–422

T.H. Kwan et al.

Fig. 9. The output characteristics of the second simulation (a) PEMFC Operating Temperature (b) The Control Signal Voltage (c) The Load Power (positive is generated electrical power).

output power to the electric heaters. Furthermore, the specific heat capacity of the fuel cell stack is emulated by placing an appropriate number of copper blocks between the electric heaters and the hot side of the TE device. By being able to flexibly set the specific heat capacity and the magnitude of heat generated, the heating emulator solution can be easily designed to be representative to that of a real fuel cell stack. The other side of the TE device is then cooled by a liquid cooling system whose specifications are very close to that used in the simulation. A brushed DC PWM based motor driver, whose DC output characteristics can be repurposed for the TE device application, is used to achieve an adaptively applied voltage signal during the TEC mode. A relay switch is then adopted to switch the operation of the TE between

control and TEC mode only control techniques for the simulation profile of Fig. 9(a) which involved slowly varying values of QH . Clearly, by observing the temperature transient curve of Fig. 9(b) and comparing it to Fig. 6(b), there are no signs of significant instability or degradation of transient performance when a steady change in power is involved. Moreover, the control voltage and power consumed transients and magnitudes shown in Fig. 9(c) and (d) respectively are somewhat similar to that of Fig. 6(c) and (d) which means the control requirement and controller response characteristics are not significantly different between the two simulation cases. According to Fig. 10 which shows the subsequent mode of operation of the combined mode during this simulation, the combined control mode also successfully engages the MPPT mode whenever heating or cooling is not required and extracts electrical energy more efficiently than that of the TEC mode only. 5. Experiment 5.1. Experimental platform setup The proposed bi-directional control concept is verified in this section using the hardware experimental platform that is shown in Fig. 11. Fig. 11(a) shows a graphic illustration of the experimental apparatus and Fig. 11(b) shows a block diagram describing its structure. The fuel cell’s generated body heat (QH in the simulation of this paper) is emulated by using a set of electric heaters that are powered by a DC variable power supply. The emulated thermal solution was chosen over a real fuel cell system because it has a relatively low system complexity, is much more cost effective and the magnitude of the generated heat can be much more flexibly chosen. A similar such solution has previously been adopted in Ref. [49] and thus it has been proven that this method can represent the thermodynamic behaviors of a real fuel cell stack. The magnitude of supplied heat is easily controlled by varying the DC power supply’s operating voltage and monitoring its resulting

Fig. 10. The operating mode of the proposed controller in simulation 2. 418

Applied Energy 222 (2018) 410–422

T.H. Kwan et al.

Fig. 11. The experimental platform architecture (a) A graphic illustration of the apparatus (b) A structural flowchart of the platform.

Fig. 12. Transient response of various parameters (a) Mode of Operation (b) Hot Side Temperature (c) Control Voltage Transient (d) Power Transient. This figure shows the case when the input heat power undergoes a step change from the steady state of the room temperature condition to 300 W at t ≈ 0 s.

419

Applied Energy 222 (2018) 410–422

T.H. Kwan et al.

other hand, this power consumption only occurs during the startup phase and beyond that, electrical power is recovered via the TEG mode. Overall, it can be concluded that with a small sacrifice in power consumption during the startup phase, the proposed controller method can significantly reduce the required time to start up the “fuel cell” and, in addition to this, a partial amount of power that was applied during the TEC mode can even be retrieved in the subsequent TEG mode. Unlike the passive control method, the combined control method can also keep the fuel cell within its optimal temperature range and this characteristic is clearly highlighted in Fig. 14(b) for the second experimental scenario. When the passive control method is adopted, the operating temperature of the “fuel cell” reaches a value of up to 115 degrees Celsius with is clearly outside the acceptable values for a practical LT-PEMFC. On the other hand, the combined control method can maintain the operating temperature to be below 95 degrees Celsius, thus preventing the overheating of the “fuel cell” by adopting the TEC mode in the negative direction. Specifically, as shown in Fig. 14(a), the TEC mode is engaged at the time of around 110 s which corresponds to when the hot side temperature is close to 95 degrees Celsius. Unsurprisingly, similar to that of the startup scenario, the combined control method also has the disadvantage that a power consumption is required during the TEC mode which is a steady state value of −320 W as shown in Fig. 14(d). This also corresponds to the saturated control voltage signal at the lower of −30 V as shown in Fig. 14(c). On the other hand, such a power consumption is necessary in this context as the “fuel cell” would otherwise overheat and thus be damaged if the TEG only mode is adopted.

the TEG and TEC modes. For the sake of simplicity, a manually adjusted variable load resistor is adopted in the TEG mode. This method was chosen because of the lack of available commercial solutions for MPPT control of the typical low output voltages of the TE device (typical values are between 2 V and 4 V in the given implementation). On the other hand, it is found experimentally that the internal resistance of the TE device during the TEG mode does not change significantly for different heating conditions and as such, the variable load resistor solution is sufficient enough for verification purposes. As labelled in Fig. 11(a) and (b), the proposed control algorithm is implemented using an Arduino microcontroller. Temperature, voltage and current measurements are obtained via simple sensors, their relevant sensing circuits and the in-built analog to digital converters (ADC). The switching of the relay switches to use either the TEG and TEC modes is conducted via a digital pin and the control voltage signal that is required for the TEC mode is delivered via a PWM pin and two digital pins as required by the H-bridge of the DC motor driver. The sensed parameters and control signals as collected in the pre-programmed Arduino are then delivered to a laptop via a serial communications link for data monitoring and processing. 5.2. Experimental results The performance of the proposed dual mode controller is verified using two separate experimental conditions. The first is a step change in input heat power from a steady state at room temperature condition to that of 300 W. The second experiment then involves a step change in the supplied heat from a steady state condition of 300 W to 500 W. In both experiments, the step change occurs at the same time as when the data collection process is initiated (i.e. at t ≈ 0 s). It is noted that the power levels of 300 W and 500 W are the same as that adopted in the simulations as presented in Section 4. Moreover, in both experiments, the proposed controller’s performance is compared to that of when only passive control is adopted (i.e. only the TEG mode is engaged). Fig. 12 and Fig. 14 show respectively the resulting transient responses or various parameters when both the proposed control method and only passive control is adopted in respectively the startup to 300 W and 300 W to 500 W scenarios. By observing the temperature transient in Fig. 12(b), the lower optimal temperature limit (60 °C) is reached within 100 s while the passive control method required around 275 s to reach the same value. In other words, the combined control method can very significantly accelerate the “fuel cell” temperature at a much faster rate than that of using only the passive technique. It should be noted this observation and the general shape of the combined control curve is in close agreement with that of the simulation results as presented in Section 4.3. In the meantime, as shown in Fig. 12(a), the combined control method can successfully switch to the TEG mode and as such, begin to extract electrical power from the system in a similar fashion as that of the passive control method. Unsurprisingly, the switching of the operational mode also causes a step change in the control voltage transient and this is illustration via an observation of Fig. 12(c) at the moment the operational mode changed. This characteristic is more clearly highlighted in Fig. 13 which is a zoomed in view of the power transient curve of Fig. 12(d). Here, the initial extracted power using the combined control method (at 160 s) is much higher than that of the passive method but the extracted power value does reduce with time converges at the same values at around 400 s. The reason such a transient behavior exists is because of the initial larger temperature differential at the moment the combined control method switched from the TEC mode to the TEG mode. In other words, a partial amount of power that was applied to the system using the TEC mode before 160 s has been retrieved in the TEG mode between 160 s and 400 s. Indeed, the primary downside of adopting the combined control method is that a power consumption is required during the TEC mode and, and as shown in Fig. 12(d), values of 200 to 400 W have been used. On the

6. Conclusion In this paper, the bi-directional nature of the thermoelectric device has been adopted to achieve temperature control and energy harvesting from an operating PEMFC. The TEG effect is used to harvest electrical energy from thermal energy whenever the PEMFC can be temperature self-sustaining and the TEC effect is used whenever active cooling or heating is required to maintain the PEMFC within its operating temperature range. An active controller that can switch between TEG and TEC model is simulated using the thermodynamic and electrical circuit domains of the Simscape environment. By conducting a time-based simulation involving varying PEMFC body heat magnitudes, it is found that operating by the proposed combined control technique is advantageous over using only either modes of operation. Specifically, while an effective temperature control is achieved, the ability to harvest

Fig. 13. The same power curve as that of Fig. 12(d) but zoomed into the positive range. 420

Applied Energy 222 (2018) 410–422

T.H. Kwan et al.

Fig. 14. Transient response of various parameters (a) Mode of Operation (b) Hot Side Temperature (c) Control Voltage Transient (d) Power Transient. This figure shows the case when the input heat power undergoes a step change from the steady state of 300 W to 500 W at t ≈ 0 s.

References

energy when active control is not required is also acquired, thus combining the benefits of the TEG and TEC modes while eliminating the disadvantages of either modes. It is noted that the performance of the combined control mode is highly sensitive to the TE hardware design and the applied heating load. Specifically, there is a combination of applied heating load and thermal resistance to the ambient environment that will enable the TEG mode to be reliably engaged. Subsequently, a recommended design procedure for the proposed cooling system is to design it with a thermal resistance that matches the most frequently expected heating load conditions so that the TEG mode is more frequently adopted in practice. A future work of this manuscript includes designing a more consolidated power electronics system to practically realize the TE device bi-directional concept. Another potential topic to be explored is the investigation of a more sophisticated control algorithm to further improve the control effort efficiency and provide a more seamless transition between the TEG and TEC modes. Finally, the implementation of the bi-directional concept into a real fuel cell system will also be considered.

[1] Bell LE. Cooling, heating, generating power, and recovering waste heat with thermoelectric systems. Science 2008;321(5895):1457–61. [2] Twaha S, et al. A comprehensive review of thermoelectric technology: materials, applications, modelling and performance improvement. Renew Sustain Energy Rev 2016;65:698–726. [3] He W, et al. Recent development and application of thermoelectric generator and cooler. Appl Energy 2015;143(Supplement C):1–25. [4] Zhang X, Chau KT. An automotive thermoelectric–photovoltaic hybrid energy system using maximum power point tracking. Energy Convers Manage 2011;52(1):641–7. [5] Wang T, et al. Waste heat recovery through plate heat exchanger based thermoelectric generator system. Appl Energy 2014;136:860–5. [6] Li P, et al. Design of a concentration solar thermoelectric generator. J Electron Mater 2010;39(9):1522–30. [7] Xi H, Luo L, Fraisse G. Development and applications of solar-based thermoelectric technologies. Renew Sustain Energy Rev 2007;11(5):923–36. [8] Kraemer D, et al. High-performance flat-panel solar thermoelectric generators with high thermal concentration. Nat Mater 2011;10(7):532–8. [9] Li X, Wang Z. Exergy analysis of integrated TEG and regenerative cooling system for power generation from the scramjet cooling heat. Aerosp Sci Technol 2017;66(Supplement C):12–9. [10] Zhang H, et al. Two-stage thermoelectric generators for waste heat recovery from solid oxide fuel cells. Energy 2017;132(Supplement C):280–8. [11] Puneet P, et al. Preferential scattering by interfacial charged defects for enhanced thermoelectric performance in few-layered n-type Bi2Te3. Sci. Rep.; 2013. p. 3. [12] Goldsmid HJ. Bismuth Telluride and its alloys as materials for thermoelectric generation. Materials 2014;7(4):2577–92. [13] Meng F, Chen L, Sun F. Effects of thermocouples’ physical size on the performance of the TEG–TEH system. Int J Low-Carbon Technol 2016;11(3):375–82. [14] Meng J-H, Zhang X-X, Wang X-D. Multi-objective and multi-parameter optimization of a thermoelectric generator module. Energy 2014;71(Supplement C):367–76. [15] Högblom O, Andersson R. A simulation framework for prediction of thermoelectric generator system performance. Appl Energy 2016;180(Supplement C):472–82. [16] Kwan TH, Wu X. Power and mass optimization of the hybrid solar panel and thermoelectric generators. Appl Energy 2016;165:297–307. [17] Li D, et al. Exergy and energy analysis of photovoltaic-thermoelectric hybrid systems. Energy 2017;126(Supplement C):343–51.

Acknowledgements This work was supported by the National Science Foundation of China (NSFC), grants 11572356 and Special Program for Applied Research on Super Computation of the NSFC-Guangdong Joint Fund (the third phase) under Grant No. U1501501. The project of Guangzhou Science and Technology program (No. 201704030089) also supports this research.

421

Applied Energy 222 (2018) 410–422

T.H. Kwan et al.

[18] Montecucco A, Siviter J, Knox AR. The effect of temperature mismatch on thermoelectric generators electrically connected in series and parallel. Appl Energy 2014;123(Supplement C):47–54. [19] Apertet Y, Ouerdane H. Small-signal model for frequency analysis of thermoelectric systems. Energy Convers Manage 2017;149(Supplement C):564–9. [20] Myers A, Hodges R, Jur JS. Human and environmental analysis of wearable thermal energy harvesting. Energy Convers Manage 2017;143:218–26. [21] Thielen M, et al. Human body heat for powering wearable devices: From thermal energy to application. Energy Convers Manage 2017;131:44–54. [22] Eom Y, et al. Flexible thermoelectric power generation system based on rigid inorganic bulk materials. Appl Energy 2017;206:649–56. [23] Wang Y, et al. Wearable thermoelectric generator to harvest body heat for powering a miniaturized accelerometer. Appl Energy 2018;215:690–8. [24] Wu Y, Zhang H, Zuo L. Thermoelectric energy harvesting for the gas turbine sensing and monitoring system. Energy Convers Manage 2018;157:215–23. [25] Jiang W, et al. Energy harvesting from asphalt pavement using thermoelectric technology. Appl Energy 2017;205:941–50. [26] Najafi H, Woodbury KA. Optimization of a cooling system based on Peltier effect for photovoltaic cells. Sol Energy 2013;91:152–60. [27] Li C, et al. Thermoelectric cooling for power electronics circuits: modeling and active temperature control. IEEE Trans Ind Appl 2014;50(6):3995–4005. [28] Wang J, et al. Experimental study on the thermal management of high-power LED headlight cooling device integrated with thermoelectric cooler package. Energy Convers Manage 2015;101(Supplement C):532–40. [29] Chen M, et al. System modeling and validation of a thermoelectric fluidic power source: proton exchange membrane fuel cell and thermoelectric generator (PEMFCTEG). J Electron Mater 2010;39(9):1593–600. [30] Wang T-H, et al. Parameter analysis and optimal design for two-stage thermoelectric cooler. Appl Energy 2015;154(Supplement C):1–12. [31] Lv H, et al. Enhancement of maximum temperature drop across thermoelectric cooler through two-stage design and transient supercooling effect. Appl Energy 2016;175(Supplement C):285–92. [32] Zhao D, et al. Personal thermal management using portable thermoelectrics for potential building energy saving. Appl Energy 2018;218:282–91. [33] Ibañez-Puy M, et al. Thermoelectric cooling heating unit performance under real conditions. Appl Energy 2017;200:303–14. [34] Mekhilef S, Saidur R, Safari A. Comparative study of different fuel cell technologies. Renew Sustain Energy Rev 2012;16(1):981–9. [35] Chahartaghi M, Kharkeshi BA. Performance analysis of a combined cooling, heating

[36]

[37]

[38] [39]

[40]

[41]

[42] [43] [44]

[45] [46]

[47] [48] [49]

422

and power system with PEM fuel cell as a prime mover. Appl Therm Eng 2018;128(Supplement C):805–17. Abdul Rasheed RK, et al. A review on modelling of high temperature proton exchange membrane fuel cells (HT-PEMFCs). Int J Hydrogen Energy 2017;42(5):3142–65. López-Sabirón AM, et al. Design and development of the cooling system of a 2 kW nominal power open-cathode polymer electrolyte fuel cell stack. Int J Hydrogen Energy 2012;37(8):7289–98. Zhang G, Kandlikar SG. A critical review of cooling techniques in proton exchange membrane fuel cell stacks. Int J Hydrogen Energy 2012;37(3):2412–29. Hasani M, Rahbar N. Application of thermoelectric cooler as a power generator in waste heat recovery from a PEM fuel cell – an experimental study. Int J Hydrogen Energy 2015;40(43):15040–51. Sulaiman MS, et al. Validation of a waste heat recovery model for a 1kW PEM fuel cell using thermoelectric generator. IOP Conf Series: Mater Sci Eng 2017;226(1):012148. Zhang H, et al. Application of cascading thermoelectric generator and cooler for waste heat recovery from solid oxide fuel cells. Energy Convers Manage 2017;148:1382–90. Yang P, et al. Performance evaluation of an alkaline fuel cell/thermoelectric generator hybrid system. Int J Hydrogen Energy 2014;39(22):11756–62. Zhao M, et al. Performance characteristics of a direct carbon fuel cell/thermoelectric generator hybrid system. Energy Convers Manage 2015;89:683–9. Al-Nimr MdA, Tashtoush BM, Jaradat AA. Modeling and simulation of thermoelectric device working as a heat pump and an electric generator under Mediterranean climate. Energy 2015;90(Part 2):1239–50. Ram JP, Babu TS, Rajasekar N. A comprehensive review on solar PV maximum power point tracking techniques. Renew Sustain Energy Rev 2017;67:826–47. Subudhi B, Pradhan R. A comparative study on maximum power point tracking techniques for photovoltaic power systems. IEEE Trans Sustain Energy 2013;4(1):89–98. Store, F.C. Horizon 500W PEM Fuel Cell. 21/09/2017]; Available from: http:// www.fuelcellstore.com/horizon-500watt-fuel-cell-h-500. Hebei I.T. (Shanghai) Co., L. TEC1-12730. 21/12/2017]; Available from: http:// peltiermodules.com/peltier.datasheet/TEC1-12730.pdf. Zakaria I, et al. Thermal analysis of Al2O3–water ethylene glycol mixture nanofluid for single PEM fuel cell cooling plate: An experimental study. Int J Hydrogen Energy 2016;41(9):5096–112.