CFD-based viability for safely cooling wing-mounted high-performance batteries in an after-market aircraft electrification Open Access

A

= Area (m²);

$an$

= NTGK polynomial constant;

$bn$

= NTGK polynomial constant;

$β$

= Thermal expansion coefficient (1/K);

$Cp$

= Specific heat (J/kgK);

$CD,0$

= Zero-lift drag coefficient;

$C1$

= Constant correction value;

$C2$

= Constant correction value;

D

= Dimensional parameter (m);

$Dw$

= Cross diffusion term in k-$ω$ turbulence model;

e

= Oswald efficiency;

$F→$

= Force vector (N);

$Γ$

= Effective diffusivity;

$G˜k$

= Generation of turbulent kinetic energy because of mean velocity gradients;

$Γω$

= Generation of $ω$ in k-$ω$ turbulence model;

g

= Gravitational acceleration (m/s²);

h

= Heat transfer coefficient (W/m²K);

I

= Current (A);

j

= Local charge transfer current density (A/m²);

$jECH$

= Electrochemical potential (V);

k

= Thermal conductivity (W/mK);

K

= Lift-induced drag;

L

= Length (m);

$ls$

= Turbulence length scale (m);

$Preq$

= Required power (kW);

Pr

= Prandtl number;

q

= Dynamic pressure (Pa);

r

= Radius (m);

Re

= Reynolds number;

$ρ$

= Density (kg/m³);

S

= Wing span (m);

$σ$

= Electrical conductivity (S/m);

t

= Time (s);

$τ$

= Electrical potential (V);

$τ¯$

= Stress tensor;

T

= Temperature (°C);

$Ti$

= Turbulence intensity (%);

u

= Velocity (m/s);

U

= NTGK fitting parameter;

$v∞$

= Flow velocity (m/s);

$v¯$

= Velocity operator (m/s);

$v→$

= Velocity vector (m/s);

V

= Voltage (V);

W

= Weight (kg);

Y

= NTGK fitting parameter;

$Yk$

= Dissipation of k in k-$ω$ turbulence model;

$Yω$

= Dissipation of $ω$ in k-$ω$ turbulence model; and

$Q˙$

= Heat generation or dissipation rate (W).

Approximately 170 electric aircraft projects were launched in 2020 alone (Schwab et al., 2021). Advancing technologies such as high energy density batteries, high power-to-weight ratio electric motors and capable battery thermal management systems have given traction to these projects. Although extensive research is being conducted in the field of electric vehicles and thermal management of lithium-ion battery systems for land-based electric vehicles has been extensively explored, current literature does not cover thermal management of batteries within electric aircraft. This study presents computational fluid dynamic (CFD)-based analyses for the first implementation of an air-cooling system for battery cells integrated in an electric aircraft. The objective of this research is to demonstrate if it is thermally viable to air-cool lithium-ion batteries in electric aircraft and to provide a CFD modelling approach to analysing large battery systems required in electric aircraft.

Infinitus Aero, an Australian aircraft manufacturer, is developing an electric aircraft, the E22 Spark, intended to meet European Union Aviation Safety Agency regulations for flight training applications. The E22 Spark (Infinitus Aero, 2023) is designed to be powered by a 800V-42kWh battery pack composed of lithium-ion pouch cells. These battery cells will be stored inside the wings of the aircraft, which has an airframe identical to the conventionally petrol-powered AeroJones CTLS model, making the batteries a retrofitted system. This research will conduct simulations to demonstrate the viability of an air-cooling system for this aircraft, which is preferred by Infinitus Aero due to air-cooling' slightweight and minimally complex design compared to liquid cooling.

The present research carries out CFD simulations to investigate an air-cooling system designed for the required 669 lithium-ion cells in each wing of the E22 Spark aircraft, to provide electric aircraft manufacturers with a process to evaluate such battery system designs and to more generally evaluate the thermal feasibility of an air-cooled, wing-mounted battery system suitable for powering training aircraft. The study, to our best knowledge, is the first to investigate the replacement of fuel with batteries in the wing of a training aircraft, and is the first CFD-based investigation of battery thermal management of such application and scale.

Lithium-ion cells are widely considered the best option for electric vehicles because of their high energy density, long cycle-life performance and low self-discharge rate compared to other types of battery cells. However, lithium-ion cells generate heat because of electrochemical reactions within the cell between the anode and cathode reacting to produce electrical current. Cell temperature must stay within minimum and maximum limits, and temperature differences between the cells connected in the pack must be minimised (Wang et al., 2011; Liu et al., 2017). When these conditions are not met, the performance and safety of the battery pack are greatly reduced and thermal runaway can eventually occur (Zhang et al., 2023).

A summary of the most recent published studies on forced air-cooling of land-based battery packs in electric vehicles is presented in the  Appendix in Table A1. Notably, 15 of the 20 referenced studies were published in, or after, 2018. The increased number of published studies in recent years is likely attributed to the rise in sustainable transport research and improvements to computational efficiency and simulation software. The largest cooling system analysed was designed for a battery pack of 60 cells which is 90% smaller than the pack in the E22 Spark. Of the 20 studies referenced, eight analysed prismatic-type batteries, 11 analysed cylindrical-type batteries and one analysed pouch-type batteries. The batteries analysed in this study are pouch-type batteries, as they are lighter than prismatic batteries because of the absence of a rigid casing and offer superior performance over cylindrical batteries. All 20 studies used air as the coolant material, one considered water as an option, another considered a heat pipe in addition to the air-cooling and a third considered a phase change material. Of the 20 investigations, 12 used experimental methods to verify their simulation results, indicating the importance of experimental verification of CFD simulations using the heat generation model and heat transfer simulations. In the simulations, 15 of the 20 studies used a certain heat generation model for the batteries, with the remaining five using either a constant heat flux value or experimentally derived heat generation profiles, illustrating the importance of modelling batteries with a comprehensive heat generation model rather than an estimated constant heat generation rate. Additionally, 11 of the 20 studies considered heat dissipation by conduction and convection, with the remaining nine considering convection only. The investigation parameters of each referenced study are also presented, with the most common design optimisation being inlet position and size and less common optimisations involving baffles, spoilers and cell spacing optimisation.

To approach the design of an efficient and effective cooling system, the key findings of the most recent CFD-based studies in Table A1 are presented to shape the cooling system design. Optimising cell layout was investigated by Chen et al. (2020), Zeng et al. (2020) and Wang et al. (2014). Chen et al. (2020) found that a symmetrical pack layout can achieve much better cooling performance, with the maximum temperature difference within the pack decreasing by 3°C compared to an unsymmetrical pack. Zeng et al. (2020) and Wang et al. (2014) found that compared to inlets and outlets aligned on the sides of the pack, an inlet at the top and outlet at the bottom of the pack increased cooling efficiency. Furthermore, compared to circular, hexagonal and rectangular battery arrangements, a cubic battery arrangement is best for cooling.

Cell spacing is also important when designing a cooling system. The effect of cell spacing has been investigated by Hasan et al. (2023) and Chen et al. (2020). Hasan et al. (2023) found that the average temperature decreased by 5°C, 7°C and 2°C when cell spacing was increased from 2 mm to 6 mm for a battery located near the outlet, middle and inlet of the pack, respectively.

The effect of inlet and outlet position, size and angle has been investigated by Xu et al. (2022), Wang et al. (2021a) and Yue et al. (2018). Xu et al. (2022) studied the effect of changing the inlet position with a fixed outlet position for a battery pack with inlet and outlets on the same side of the pack, creating a “U” shaped channel. It was found that placing the inlet as low as possible allowed for maximum airflow around the batteries.

Another investigation by Wang et al. (2021a) concluded that an inlet and outlet centralised at the bottom and top of the pack resulted in the lowest maximum cell temperature, while an inlet at the bottom left-hand side of the pack and outlet centralised at the top of the pack resulted in optimal temperature uniformity throughout the pack.

The research studies referenced in Table A1 have been reviewed to analyse and understand the simulation parameters used in simulations and have shaped the methodology of this research.

The cells researched in this investigation are lithium-ion pouch cells designed specifically for the E22 Spark. To produce an 800V pack with 42.72 kWh of energy to sustain a flight duration of at least 103 min, including pre- and post-flight ground activities required in flight training, the cells are arranged in six parallel strings of 223 cells in series, with a shorthand notation of ‘223s6p’. The battery manufacturer recommends the batteries stay below 45°C during operation, with the optimal temperature range between 15°C and 35°C (Yue et al., 2021).

A C-rate is a measure of how quickly a battery is discharged, where 1C = one hour of battery life. To determine the required C-rate for each flight phase, the engine power required versus airspeed was calculated with equation (1) using relevant aerodynamic data of the E22 Spark. Furthermore, it was assumed that the power required during take-off is equal to the maximum output power of the electric motor (100 kW). Additionally, although the power required for cruise was calculated to be 20 kW, 40 kW was taken as the power required in transient simulations, to conservatively account for adverse flight conditions such as an aggressive headwind, low air density or increased drag on the aircraft.

The C-rates were calculated by dividing the power required in watts for each phase by the voltage of the battery pack (800 V), then dividing by the capacity of each cell (8.9 Ah) and, finally, dividing by the number of parallel strings (six). The E22 Spark is aimed at flight training and, as such, will only ascend to a necessary safe circuit height of 1,000 ft. The full flight profile analysed in the transient CFD simulations, including the aircraft velocity, C-rate and time for each flight phase, is presented in Figure 1:

The wing of the E22 Spark is designed as a unibody structure, with fuel originally stored ahead of the spar in the leading edge of the wing and structural ribs aft of the spar. In the electric aircraft, the batteries will be mounted to the spar, allowing for 61 stacks of 11 cells, spaced 4.5 mm vertically and 5.5 mm horizontally, shown in Figure 2, modelled in Solidworks. This cell spacing is the absolute maximum allowable within the given space of the wing.

The cell configuration is chosen to allow for maximum cell spacing in a symmetrical layout, in alignment with key findings in the studies described in Table A1. The original fuel weight per wing in the CTLS was 65 L or 47 kg. With the batteries, the wings in the E22 Spark now sustain a payload of 68.7 kg per wing, equivalent to a 46% increase per wing, likely limiting permissable dynamic manoeuvres to some extent.

In this study, inlets and outlets placed on the upper surface of the wing are considered. The position of the inlets and outlets with respect to the aerofoil and battery pack is shown in Figure 3^[1]. This layout was chosen such that low-pressure, higher-velocity air accelerated by the aerofoil curvature on the upper surface is ingested by the intakes, in addition to the free stream air directly ingested by the intakes. Additionally, the high-pressure air below the wing is largely unaffected, helping minimise the impact on aircraft lift.

The inlets and outlets modelled in this investigation are simple rectangular cuts in the wing surface, noting NACA inlets would likely be implemented on any manufactured designs.

In all, eight design cases were considered for analysis based on inlet–outlet spacing outlined in Table 1. With inlets and outlets being 100 mm long (span-wise) and 10 and 25 mm wide respectively, all cases were designed such that an outlet is placed at each end of the pack with intermediate inlets and outlets spaced evenly across the wingspan. In this staggered arrangement of inlets and outlets, the air is forced to travel spanwise across the wing over several cells to an outlet before it must exit, compared to having the outlets directly aft of the inlets.

Due to the repeating pattern along the span, each design case can be reduced to simulate the airflow through an inlet and two outlets and the associated cells within that region rather than the entire wing. Case 4, presented in Figure 3, clearly illustrates the repetitions of the outlet–inlet–outlet configurations for the specified spacing.

For the boundary conditions of the simulation, all battery and pack walls were modelled with the no-slip boundary condition and standard wall surface roughness for aluminium. The velocity inlet was set with the velocity at each flight phase as input, noting for further analysis of any production designs with NACA inlets; the velocity at the inlet could be solved by an in-depth analysis of the air characteristics entering the inlet. The turbulent specification method used turbulence intensity and length scale calculated with equations (2) and (3), respectively. The outlets were set to pressure outlets, with turbulence intensity and turbulent length scale also calculated with equations (2) and (3). The velocity at the outlet used in these equations was obtained through multiple flow simulations to estimate a sufficiently valid outlet velocity value for the turbulence parameters at the outlet:

The governing flow and turbulence equations were solved in the steady-state solver for each flight phase velocity in ANSYS Fluent, using the continuity and momentum equations and the Shear-Stress-Transport (SST) k-$ω$ turbulence model. The SST k-$ω$ model was chosen based on similar simulation setups by authors in Table A1 and the Reynolds number of the flow through the wing. The SST k-$ω$ model was developed by Menter (1994), which is widely accepted to be accurate and reliable for flows involving complex behaviours such as adverse pressure gradients and large turbulent shear stress quantities generated when analysing high-velocity flow and intricate geometries. Most authors in Table A1 used the k-$ϵ$ turbulence model. However, the SST k-$ω$ model effectively blends the k-$ϵ$ model for locations in the flow independent of wall behaviours and the enhanced k-$ω$ model for locations in the flow greatly affected by the geometry.

The discretisation scheme used in solving these equations was second-order for pressure and second-order upwind for momentum, turbulent kinetic energy, specific dissipation rate and energy. A pseudo-time step was used for the steady simulations, set to 0.1 s. Relaxation factors were set to default and convergence criteria were set to 1E^-3. The solutions, assessed to have reached a steady state solution once converged, were saved as constant airflow conditions through the wing for the transient heat generation simulations.

To simulate the heat generation and cooling of the batteries during flight, the unsteady transient solver was used to simulate flight-phase-specific discharge rates and times. The steady-state solution for the flow through the wing was saved for each flow velocity and kept as constant during the transient simulations. The second-order implicit model was used for time, with heat transfer enabled through the coupled energy equations for the solid and fluid domains. The heat dissipation between the cells and cooling air is calculated by Fluent with the energy equation applied to coupled thermal boundary conditions for every battery and busbar with the solid batteries in contact with the air [equations (4)–(7)], ignoring the short circuit and abuse models:

The material for the batteries was set to aluminium and a UDS diffusivity of 1.19E⁶ Siemens/s for the positive cathode and 9.83E⁵ Siemens/s for the negative cathode. The connections between the batteries were modelled as copper busbars with positive and negative tabs merged to the busbars for simplicity.

The material properties for air were set as required for the relevant flight phase altitude. The time step size was set to 3 s based on independence tests described in Section 2.8, and simulations were conducted with ten iterations per time step. The residuals for energy and the functions for the positive and negative terminals of the cells were monitored and observed to reach a constant residual by the end of each time step, indicating the number of iterations per time step is sufficient. Pressure–velocity coupling was used with the flow Courant-Friedrichs-Lewy number set based on the average velocity between the cells, with simulated values between 1 m/s and 4 m/s.

To model the heat generation of the cells, the multi-scale multi-domain battery model was used, and the Newman, Tiedemann, Gu and Kim (NTGK) modelling method by Kwon et al. (2006), Kim et al. (2008) and Kim et al. (2011) was specified. The NTGK method is an empirical-based model in which the heat generation is dependent on the current potential. The NTGK model calculates the battery heat generation rate with equation (8). The electrochemical current potential (⁠$jECH$⁠) is dependent on empirical parameters, U and Y, which are third- and fifth-order polynomial functions of the depth of discharge, respectively. Therefore, three data inputs are required for Fluent to accurately model battery cells:

1. cell voltage versus time for;

2. a given discharge rate; and

3. cell capacity.

This data was obtained through experimental tests by the battery manufacturer for discharge rates of 0.2 C, 0.5 C, 1 C, 2 C and 3 C (compared to estimated C-rates of 0.023 C, 0.23 C, 0.94 C and 2.34 C during flight). Importantly, the experimental data was for slightly different cells than those modelled as the modelled cells are not yet produced and covered slightly different C-rates than analysed in simulations for the flight profile. However, Fluent fits the experimental data such that simulated discharge profiles with a C-rate not equal to the experimental data are calculated based on the known data profiles. The U and Y function is then calculated by Fluent per equations (9) and (10), respectively, once the experimental data is curve-fitted. The electrochemical potential is then calculated per equation (11). In these equations, $C1$ and $C2$ are constants, and $an$ and $bn$ are polynomial coefficients for U and Y derived through curve fitting the experimental data:

The NTGK model in this study predicted a temperature of 67.12°C under natural convection conditions (i.e. inlet velocity = 0$m/s$⁠) at an ambient temperature of 30°C after a 75% discharge at 1C. Li et al. (2024) conducted investigations for pouch cells of the same voltage specifications but a capacity of 26 Ah. Their experimental investigations of a single cell yielded very similar results to their simulations using the electrochemical heat generation model. They found that at an ambient temperature of 40°C and a discharge rate of 1C, the maximum temperature rose to 59.85°C (11% less than NTGK) under natural convection. Behi et al. (2020) found that after 2700 s of discharge at 1.5 C a cylindrical lithium-ion battery cell rose to 60°C in both experimental and simulated investigations (also 11% less than NTGK); however, the ambient temperature was only 26°C.

To compare the NTGK model at a different discharge rate, at 2.34 C the NTGK model predicted a battery temperature increase from 30°C to 41.8°C after 3 min. Shahid and Agelin-Chaab (2018) conducted experimental investigations and saw a 6°C increase in the same time period, although the ambient temperature was quite low at 23°C and the C-rate was slightly lower at 2 C.

In all comparison cases mentioned between this research and relevant studies conducted previously, the NTGK model has predicted a slightly higher temperature than appears to be expected from similar battery cells. Predictions of a higher temperature value are therefore conservative in this study, allowing for a factor of safety; however, experimental validation of the NTGK model should be conducted in the future using the specific cell studied in this investigation.

Case 4 was used to conduct mesh independence tests at five different grid refinement levels, with grid elements from 8 million to 32 million. Twenty boundary layers were used in all mesh refinements for all geometries. The Y+, being a critical measure of quality of the mesh including the selection of the number of boundary layers, was calculated by Fluent for each mesh. In each mesh, the Y+ was less than 1 (with the largest Y+ = 0.154), which is required when using SST k-$ω$ turbulence modelling when the flow and thermal boundary layers must be resolved. Polyhedral type meshing was used with prismatic boundary layer generation, consistent with eight of the 20 mesh types used in studies referenced in Table A1.

The transient case simulated in both mesh and time independence studies was 180 s at 2.34 C, with a time step of 5 s used in the mesh independence studies for 36 time steps to simulate the take-off period of flight. All batteries were patched to the same temperature of the air (30°C) at the beginning of the simulation. The independence results for maximum battery temperature, maximum temperature difference between the batteries, and average battery temperature at the end of the take-off period of flight are presented for each grid refinement in Figure 4. As the grid refinement quadrupled from 8 million to 31 million elements, the maximum battery temperature and maximum battery temperature difference between the batteries increased by 1°C (2.5%). However, the maximum average temperature only increased by 0.18°C (0.5%), while the simulation time almost doubled from 2 minutes and 38 seconds per time step with 13 million elements, to 5 minutes and 9 seconds with 31 million elements. As a result, Mesh 2 (13 million elements, Y+ = 0.13) was chosen for all following simulations for flow through the wing at a steady state and heat transfer between the batteries and air at a transient state.

Time step independence was conducted for Mesh 2 in the transient simulation for the take-off phase of flight, by considering five different time steps, varying between 2 s and 30 s. The maximum battery temperature, average temperature and maximum temperature difference between the cells at the end of the take-off period of flight were monitored for the verification and these are presented in Figure 4 for each time step.

Each time step took approximately 2 minutes and 30 seconds to solve with ten iterations per time step. The time steps used in similar CFD studies [such as (Junsheng et al., 2021) and (Napa et al., 2023), the only authors in the referenced studies in Table A1 to explicitly mention their time step selection] are 1 s. It is widely understood that time steps decreasing towards and below 1 s are usually required to fully capture flow phenomena and behaviour. The time independence study, therefore, analysed larger time steps, starting at 30 s, decreasing towards 1 s. However, as the time step decreased from 5 s to 3 s, the relative error for the maximum and maximum average temperature was within just 1.2% and the maximum temperature difference between the cells was within 4.5%. Considering the computational time, a time step of 3 s was selected in the discussion of the results. A time step of 1 s or smaller would have likely been required if the flow was not solved at a steady state before the simulations.

The maximum cell temperature has been reduced by 37%, from 68°C with no cooling, to a maximum temperature of 43°C with Case 4 inlets and outlets, at an ambient temperature of 30°C. As the optimal internal temperature of battery cells is between 15°C and 45°C during operation, Case 4 provides viable cooling performance for the entire flight profile at an ambient temperature of 30°C. The heat generation of the cells occurs mostly during take-off when the C-rate is 2.34 C, rising to 41°C. After the take-off period, when the power is reduced to cruise power (and an associated C-rate of 0.94C), the temperature slightly rises by 2°C, to a maximum temperature of 43°C, before slowly reducing during the cruise to 37°C. The average battery temperature peaked at 41°C at the end of the take-off phase and quickly reduced to below 37°C after 2 min at the cruise speed of 44 m/s and discharge rate of 0.94 C. There is little heat generation during taxi with or without cooling. These results are represented in Figure 5.

When the inlet velocity increases to 36 m/s for take-off and the cells heat rapidly during this phase, the cells near the inlet are rapidly cooled to close to ambient temperature (30°C), while the cells furthest from the outlets heat towards 41°C. This is represented in Figure 5. The non-uniform temperature data is further represented in Figure 6 as a function of the battery distance from the inlet. As the span-wise distance from the inlet increases, from 0 mm for the cells located where the inlet is up to 426 mm where the outlet is, the temperature increases from an average of 37°C to 41°C, with a maximum temperature difference of 6°C. The maximum temperature difference increased to 12°C at the beginning of the cruise when the C-rate is lowered, slowly returning to a maximum temperature difference of 6°C by the end of the cruise phase. The cooling configuration appears to be cooling the cells quite uniformly if the cells located at the inlet are not considered, with the column of cells slightly away from the inlet having an average temperature of 39.5°C and the cells near the outlet having an average temperature of 41°C.

The velocity of air around the cells is a measure of performance of the cell spacing. The higher velocity around the cells at the inlet leads to a lower temperature for the two stacks within 100 mm of the inlet. However, as the distance from the inlet increases, the velocity around the cells slowly decreases from an average of 3 m/s near the inlet to an average of 1.5 m/s at the outlet location. This consistent velocity distribution is excellent, as the batteries furthest from the inlet are still receiving sufficient volumetric flow rate through the gaps between the cells. A visualisation of the steady-state airflow through the simulated section is presented in Figure 6.

It appears that Case 4 is likely at the upper limit for cooling viability at an ambient temperature of 30°C, considering the maximum temperature with Case 4 cooling rose to 43°C. It is, therefore, not recommended that any less inlets are simulated, as it is expected that maximum and average cell temperatures above 45°C would result, and temperature differences between the cells in the pack would rise accordingly.

Cases 5–8 were not simulated, as it appears that Case 4 is viable, and aerodynamically, it is expected that the most viable cooling system is the case with the largest viable inlet–outlet spacing (and, therefore, the smallest number of inlets and outlets required). However, the hypothesised performance of Cases 5–8 is extracted from the Case 4 temperature data as a function of the battery distance from the inlet. For every 50 mm the cells are further away from the inlet, the average temperature increases by approximately 0.5°C. For the design cases with more inlets, it is hypothesised that Cases 5-8 would result in a maximum temperature of 41.2°C, 40.9°C, 40.7°C and 40.6°C, respectively, compared to 41.5°C for Case 4 after take-off. These small decreases in maximum temperature represent a marginal cooling benefit compared to the possibly large and yet-to-be-assessed effect on aerodynamics of the wing because of the increased number of inlets and outlets.

It is important to note the limitations of electric aircraft because of current battery technology shortfalls. Firstly, a longer climb phase (more than approximately 1,000 ft) with given wing geometry of the E22 Spark is not currently possible within current battery technology, as the energy required is significant and a longer climb would not allow for any considerably long cruise phase. However, should there be enough energy within the pack to sustain a longer climb, it is likely that the battery cell temperature would increase to temperatures outside recommended limits, and cooling design optimisation is recommended in this instance to minimise cell temperature as much as possible.

Battery cells must also be spaced a minimum distance apart for air cooling to be effective. For aircraft that require a large number of batteries which does not allow for at least 3 mm cell spacing within the wing geometry, it is expected that cooling performance would drop significantly. Heavier aircraft (such as heavier recreational passenger aircraft) must be able to fit enough batteries to at least meet the energy density of the E22 Spark design whilst allowing for appropriate battery cell spacing. Furthermore, batteries must also be able to be placed in a grid-like pattern in any aircraft wing to allow for optimal flow between the cells towards the outlets. The cost of not meeting these requirements is a small endurance window or excessively hot batteries after take-off.

While this research focuses on training aircraft, the potential for larger aircraft to use an air-cooled battery system similar to that presented cannot be ignored. With current battery technology, manufacturers of larger aircraft could opt for a hybrid petrol-electric system, where the electrical power could be used to support take-off and climb, greatly reducing the fuel used, demand on the petrol power plant, and aircraft noise when close to the ground. Additionally, at the point in the near future where electric aircraft endurance improves because of inevitable advancements in battery technology, electric aircraft become much more viable for training, civilian transport and light cargo fleets alike. This research provides the foundation for manufacturers to explore implementations of electric propulsion in their fleets to reduce emissions, noise pollution and cost of ownership.

The research provides analytically derived evidence that electric aircraft designers could potentially implement air-cooling for battery packs within the wings. The results indicate that liquid cooling is not required for battery packs of large size in electric aircraft, and air-cooled lithium-ion batteries are thermally viable for electric aircraft. The results showed that the battery pack, consisting of 669 cells in each wing, did not exceed 43°C during a simulated training flight with four 100 mm $×$ 10 mm inlets and five 100 mm $×$ 25 mm outlets placed in a span-wise pattern spaced 426.63 mm apart on the wing. Test cases with fewer inlets were found impractical, as the design case presented was already on the upper limit of performance. The demonstrated viability of air-cooling sufficient high-performance batteries in place of fuel in the wing is important to begin designing and producing training aircraft fleets globally with sustainable propulsion, thus reducing a major contribution to global greenhouse emissions by industrialised nations. However, for electric aircraft to be a fully viable replacement of combustion-powered aircraft, improvements are required in battery energy density to allow for longer take-off and cruise periods of flight.

An aerodynamic analysis of the proposed inlets and outlets is warranted to assess whether the aerodynamics of the wing will be affected significantly by the implementation of the Case 4 cooling system design. If the viable design case is aerodynamically viable, then the aerodynamic effects of cases with more inlets should be tested, and the upper limit for the number of inlets based on aerodynamic performance should be identified. Following those analyses, higher ambient temperatures should be analysed up to 45°C to assess the operating condition limits of different cooling system configurations. Further research should evaluate cooling performance under extreme operating conditions to ensure system reliability across diverse flight scenarios. Furthermore, the energy source payload weight does not decrease during flight as fuel normally would, meaning structural analysis for wing loading during take-off, flight and landing must be undertaken and required structural reinforcements or operational limitations must be implemented. Finally, the structural effects of internal flow on the wing skin must also be carefully analysed. It is possible that high forces pushing outwards on the wing skin caused by the stagnation of the internal air flow on the batteries could cause wing skin damage or delamination of the wing skin from the frame. These effects would have to be considered when safely designing a wing for internal flow.

The effects of inlet/outlet shape, position and dimensions on cooling efficiency should also be analysed in future investigations. This research should be conducted in conjunction with research of the external flow over the wing with the inlets and outlets installed. Such research could potentially present significant improvements to the air-cooling system by optimising components such as internal flow velocity, flow through installed NACA inlets and flow rate between the cells.

The heat generation model for the specific cells used in this investigation should be validated by experimental investigations before finalising the aircraft modifications. Such investigations have been conducted by several authors in the literature referenced in Table A1. The heat transfer between the batteries and the air should also be validated in a similar experimental setup.

This research includes computations using the computational cluster Katana supported by Research Technology Services at UNSW Sydney (Link to Research Technology Services at UNSW SydneyLink to a PDF of the cited article.).