1. Introduction
Over the past few centuries, the world has relied on fossil fuels as the main energy source to power vehicles with internal combustion (IC) engines. This has come at a great cost to the environment and human health, as 20% of all carbon dioxide (CO
2) emissions originate from road traffic [
1]. As a result, electric vehicles (EVs) have begun to dominate the automotive industry to reduce CO
2 gas emissions. In 2015, 550,000 units of EVs were sold, and this subsequently increased to 774,000 units in 2016 [
2]. However, EVs have several inherited issues, such as range anxiety [
3], long charging time [
4], and the availability of charging infrastructures [
5]. One of the solutions for these issues is to have different powertrain topologies, such as hybrid electric vehicles (HEVs) [
6].
Rechargeable lithium-ion batteries (LIBs) are the most common type of energy storage for HEVs. However, LIBs generate heat while charging or discharging, and if this heat is not dissipated, it affects the battery’s performance and causes it to have a reduced lifespan, leading to thermal runaway [
7]; thus, the heat generated by LIBs must be adequately thermally managed. For instance, Sato [
8] investigated the effect of LIBs operating above 323 K, which showed a decrease in charging efficiency and LIB longevity.
To mitigate these concerns, a battery thermal management system (BTMS) is introduced to cool LIBs at an optimal operating temperature range of 293–313 K [
9]. The temperature difference within LIBs could also influence their longevity and be kept below 5 K [
10]. Several cooling techniques could be implemented, such as direct and indirect cooling, and with several cooling media, such as air, liquid, phase-change material (PCM), and heat pipes (HPs) [
11].
The heat generation inside LIBs is a complex process that requires knowledge of how the electrochemical reaction rate varies with time, temperature, and current flows. Kim et al. [
12] presented the heat generation of LIBs as
= I (U − V) − I (T
), where
, I, U, and V are the rate of heat generation, the electric current passing the cell, open circuit voltage, and the cell voltage of the LIB, respectively. The resistance loss and reversible entropic heat in the electrochemical reaction are denoted by I (U − V) and (T
), respectively. During charging and discharging processes, LIBs generate three types of heat: (1) the activation of irreversible heat due to the polarisation of the electrochemical reaction, [
13] (2) joule heating due to ohmic losses [
14], and (3) reversible reaction due to the entropy change [
15].
The LIB thermal runaway is a common issue when it reaches the maximal threshold temperature. The thermal runaway is caused by the exothermic reactions among the anode, cathode, and electrolyte [
16]. These exothermic reactions are caused by short circuits, high-speed charging/discharging, and overcharging LIBs [
17]. The thermal runaway of a single LIB may lead to the thermal runaway of the entire LIB pack, which causes severe damage [
18].
BTMSs can be categorised into three strategies: (1) active, (2) passive, and (3) hybrid [
19,
20]. Active BTMSs often use air or a liquid cooling medium where the cooling medium can be in either direct or indirect contact with LIBs. Pumps and fans are used to circulate the cooling medium and transfer the heat to the ambient condition. Passive BTMSs implement PCM and HP on the surface of LIBs to improve the heat transfer with the boundaries [
11]. Lastly, hybrid BTMSs adapt PCM + air, PCM + liquid, or PCM + HP.
Air-cooled is a preferable BTMS solution due to less weight and system complexity compared to those of liquid-cooled solutions. An efficient air-cooled (AC)-BTMS can remove excess heat generated from LIBs during charging/discharging and maintain the optimal operating temperature. However, in forced AC-BTMSs, air cooling capacity is limited due to poor thermal conductivity, which increases the working temperature and causes thermal imbalances within LIBs under harsh operating conditions, in addition to having difficulty in achieving uniform air distribution [
21]. Moreover, forced AC-BTMSs require higher volumetric flow to achieve similar cooling performance to that of other types of BTMS; thus, AC-BTMSs are not efficient in controlling the temperature in LIBs that have fast charging/discharging rates.
For fast charging/discharging applications, direct liquid-cooled (LC)-BTMSs are preferred since they have a higher heat coefficient than AC-BTMSs [
22]. A dielectric cooling medium is used in direct LC-BTMSs due to its electrically nonconductive properties and high thermal stability. Direct LC-BTMSs require two to three times less energy compared to AC-BTMSs to maintain the same average temperature of LIBs [
23]. One of the dielectric cooling media is oil. The rate of the heat transfer of oil was significantly higher than that of air at the same mass flow rate [
24]. However, a newer experiment found that using mineral oil as a dielectric coolant introduced weight to a vehicle, which worsened its performance and increased its energy consumption [
25]. The 3M Novec-7000 is another type of dielectric coolant, and it is commonly used in direct LC-BTMSs. For instance, Thakur et al. [
22] developed the boiling LIB cooling method using 3M Novec-7000. At a discharge/charge rate (C) of 10C, the performance of direct LC-BTMSs was examined by comparing an AC-BTMS and a boiling LC-BTMS. Using the air-cooled system, the LIB temperature increased to 353–363 K, whereas with total immersion in the 3M Novec-7000, the LIB temperature remained at 308 K.
Implementing indirect LC-BTMSs can be considerably more complicated and expensive than direct LC-BTMSs. Indirect LC-BTMSs are where the cooling media do not get in direct contact with LIBs, and are mainly used to prevent electrical conduction with LIBs whilst dissipating heat. However, there are two issues with indirect LC-BTMSs: (1) leakage at cooling connections that cause LIBs to short circuit, and (2) high thermal resistance due to the electrically insulating coating that reduces the heat transfer performance [
26].
For passive PCM-BTMSs, PCM is designed to surround LIBs. When LIBs heat up, the PCM softens and absorbs the heat. When LIBs cool off, PCM hardens and releases all the heat into the atmosphere. PCM stores thermal energy via the latent heat phase of transitions [
27]. Moreover, it has a large amount of latent heat; and during LIBs discharge, it acts as a heat sink. PCM can effectively lower LIBs’ maximum temperature and reduce the temperature differences at the end of the discharge cycle [
28]. However, there are three disadvantages of PCM-BTMSs, namely (1) low thermal conductivity, (2) leakage problems, and (3) limited capacity of heat absorption after melting [
29].
On the other hand, passive HP-BTMSs have high thermal conductivity and are used to maintain the temperature of LIBs within the optimal working temperature range [
30]. For example, Nasir et al. [
31] implemented HP-BTMS to cool HEVs LIBs, and it was found that the system was able to reduce the maximum LIB temperature by 287.7 K and maintain an average temperature below 323 K [
31]. The advantage of using HP-BTMSs is providing a lightweight system with an extended lifecycle. However, the main concern with HP-BTMSs is the relatively high cost compared to other cooling techniques [
29].
To reduce the overall weight of HEVs and, subsequently, the vehicle energy consumption, the ideal volume and weight of BTMSs should be less than 40% of the LIB module [
32]. These requirements motivate the need to find alternative solutions where fuel (already stored in the tank) has the potential to maintain the LIB’s temperature. The concept of using fuel as the cooling medium for BTMSs, introduced in [
33], can reduce the weight of EVs and improve vehicle performance and energy consumption.
Many hydrocarbon molecules in petroleum are found as good dielectrics. However, dielectric permittivity can be altered by the number of carbon atoms in the fuel composition. For instance, using N-Heptane as a coolant for BTMSs managed to control the LIB’s maximum temperature at discharge rates 1C and 2C by not more than 280.9 K and 290.9 K, respectively, compared to a LIB without BTMS [
33]. Fuel must have good material compatibility with LIBs to protect the elastomer seals, copper, and insulation materials from any harm. High flash point reduces the fire risk in high temperatures and improves safety. Fuel components that have dielectric constants (
) lower than 2 are suitable for direct LC-BTMSs [
34].
From the above literature, the challenges are to identify a suitable cooling medium for LIBs that can maintain the optimal operating temperature range and the low-temperature differences between LIBs. Additionally, this approach can reduce the overall weight of HEVs and minimise system complexity. In this paper, we investigated the cooling strategy of petrol components for direct LC-BTMSs using computational fluid dynamics (CFD) for HEVs application. The approach is the first of its kind and has not been revealed elsewhere before, to the best of our knowledge. A commercial software, ANSYS-Fluent, is used to simulate LIB thermal behaviour at different discharge rates. Dielectric components of petrol molecules are investigated for their suitability in BTMSs.
2. Modelling and Parameterisation Study
A transient 3D modelling of LIBs and their enclosure is conducted using the commercial CFD software ANSYS-Fluent. The system design is produced using CATIA V5. Both thermal and electrochemical interactions are accounted for and coupled in the simulation. The main result indicators are the individual LIB maximum temperature, and the temperature differences across the domain and between LIBs. Five cooling media are considered in this study: air, N-Pentane, N-Hexane, N-Butane, and Cyclo-Pentane, based on a direct LC-BTMS strategy. The dimensions and materials used in LIB sub-components are provided in the following sections. In addition, the governing equations used in the numerical simulations are highlighted, along with the coolant properties.
2.1. LIB Module
The LIB module used in this study consists of four LIBs, as shown in
Figure 1. The configuration of the LIB module is based on 4S1P (four LIBs in series and one LIB in parallel). Each of the LIBs is connected to two aluminium tabs (positive and negative connections). These tabs are connected via copper busbars to form the LIB module. In total, the sub-components of the LIB module are four LIBs, eight tabs, and three busbars. LIBs are spaced by 5 mm gaps for the cooling medium. The properties of the LIB and tab and busbar are shown in
Table 1 and
Table 2, respectively.
The LIB module enclosure has inlet and outlet ports for the cooling medium, as shown in
Figure 2. An inlet port is on one side at the top of the wall, and the outlet port is on the diagonally opposite side, positioned at bottom of the wall. The design of the inlet and outlet ports allows for the cooling medium to flow through the LIB module and to fully immerse the LIB module surfaces. This approach can ensure uniform temperature distribution across the LIB module, which helps maintain the optimal working temperature range between 293 and 313 K.
2.2. Meshing
CFD simulation requires high-quality mesh and appropriate convergence criteria to produce accurate and reliable results without requiring high computational time and power. Hence, mesh convergence is performed to ensure the highest possible accuracy. Two different types of mesh are used for the LIB module enclosure and the LIB module, tetrahedral and structured.
The LIB module enclosure is meshed using tetrahedral elements due to its complex geometry and to avoid any distortion. For the LIB module, hexahedral elements are used with four different element size refinements. Four simulation iterations are run based on the full depletion of the LIB module. The element sizes are set between 2 and 4 mm, with 0.5 mm intervals. The average LIB module maximum temperature is plotted against the number of elements.
As shown in
Figure 3, at an element size of 3 mm, the mesh gave a converged model. The differences between the average LIB module maximum temperatures of element sizes 2 mm, 2.5 mm, and 3 mm are negligible. Element sizes of 2 mm and 2.5 mm will require high computational time, but contribute insignificantly to accuracy. Hence, it is more practical to rely on the 3 mm element size. The total number of elements used for the LIB module is 125,068.
Figure 4 shows the final mesh for the BTMS (LIB module, enclosure, and coolant medium).
2.3. Input Parameters
In this analysis, the dielectric fuel components (molecules) are utilised as the BTMS coolant. Only fuel components of a dielectric constant κ < 2 are considered in this model, avoiding any possible short circuit inside the LIB module. To validate the effectiveness of using fuel components as BTMS coolant, they are compared with the conventional coolant (3M Novec-7000).
The 3M Novec-7000 is a dielectric fluid, used for direct liquid cooling in HEV applications or electronic components, as discussed in the literature [
22]. It has proven its efficiency in keeping LIBs within the optimal working temperature range at various discharge rates, and maintaining a uniform temperature distribution across the LIBs.
The cooling medium is set with inlet port temperature of 298 K and outlet port pressure of 1 atmospheric. The flow is simulated for three different inlet velocities: 0.1 m/s, 0.5 m/s, and 1 m/s, to understand the velocity influence on the thermal performance of the LIB module. Additionally, for each of the inlet velocities, three discharge rates are accounted for, 1C, 1.5C, and 2C. The transient discharge effects are considered in a quasi-steady format, with 30-time iterations until the LIB module is fully depleted. The free convection effect is included with the heat transfer coefficient set to 5 W/m2·K. The turbulence model (K-epsilon) is used for the CFD model, where the cooling medium inside the enclosure is expected to show some turbulent flow.
The properties of the proposed coolants are shown in
Table 3. All fuel components have
and reasonably good thermal conductivities, compared to the 3M Novec-7000. These fuel components will be further investigated for their suitability using ANSYS-fluent. In [
35], the thermo-physical properties of different fuel components are provided with detailed methods of calculation, embedded in our model.
4. Conclusions
LIB module temperature without any cooling at the ambient condition of 298 K and 2C exceeds the maximum threshold temperature and can lead to thermal runaway. Hence, an efficient BTMS is required to maintain the optimal working temperature. Some fuel components, such as N-Pentane, N-Hexane, N-Butane, and Cyclo-Pentane, show the potential for coolants in direct cooling of LIB due to their dielectric constants (< 2). The cooling performance of fuel as LC-BTMS (liquid-coolant in battery thermal management system) in the HEVs application was successfully simulated using commercial CFD modelling, ANSYS-Fluent software. The CFD model shows a good agreement with the literature data and can be extended for different cooling media applications to characterise the thermal management of LIB modules.
The results demonstrate that at all discharge rates, the maximum LIB module temperature is managed at:
301.526 K (1C), 304.545 K (1.5), and 306.534 K (2C) for N-Pentane LC-BTMS at inlet velocity 0.1 m/s, indicating an average reduction of 2.8% compared to FAC-BTMS.
299.497 K (1C), 300.834 K (1.5C), and 301.551 K(2C) for N-Pentane LC-BTMS at inlet velocity 1 m/s, indicating an average reduction of 3.9% compared to FAC-BTMS.;
301.395 K (1C), 304.269 K (1.5), and 306.264 K (2C) for N-Hexane LC-BTMS at inlet velocity 0.1 m/s, indicating an average reduction of 2.8%.
299.93 K (1C), 300.461 K (1.5C), and 301.399 K(2C) for N-Hexane LC-BTMS at inlet velocity 1 m/s, indicating an average reduction of 3.9%.
303.456 K (1C), 303.611 K (1.5), and 305.957 K (2C) for N-Butane LC-BTMS at inlet velocity 0.1 m/s, indicating an average reduction of 2.7%.
300.013 K (1C), 300.631 K (1.5C), and 302.575 K(2C) for N-Butane LC-BTMS at inlet velocity 1 m/s, indicating an average reduction of 3.8%.
301.539 K (1C), 3304.236 K (1.5), and 306.674 K (2C) for Cyclo-Pentane LC-BTMS at inlet velocity 0.1 m/s, indicating an average reduction of 2.8%.
299.736 K (1C), 299.879 K (1.5C), and 300.665 K(2C) for Cyclo-Pentane LC-BTMS at inlet velocity 1 m/s, with an average reduction of 4.1%.compared to FAC-BTMSs.
Also, it is noted that N-Pentane, N-Hexane, N-Butane, and Cyclo-Pentane produce an average temperature difference of −0.101 K, −0.232 K, 0.490 K, and −0.380 K, respectively, below that of 3M Novec-7000 LC-BTMS. Additionally, all fuel component LC-BTMSs are able to maintain the temperature difference between LIB cells below 5 K.
To conclude, the relatively low density of these fuel components can help in reducing the overall mass of the vehicle and improve its energy consumption which makes these fuel components viable alternatives to conventional cooling media. It is important to note the flammability of fuels when used in BTMSs. As such, indirect cooling can be more feasible in practice, when the temperature range exceeds the advised temperature limited in this research finding.