You are on page 1of 13

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/285581158

Development of a dynamic propulsion model for electric UAVs

Conference Paper · November 2015

CITATIONS READS

11 8,880

2 authors:

Andrew Gong Dries Verstraete


The University of Sydney The University of Sydney
18 PUBLICATIONS 593 CITATIONS 101 PUBLICATIONS 1,580 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Andrew Gong on 03 December 2015.

The user has requested enhancement of the downloaded file.


Development of a dynamic propulsion model for electric
UAVs
Andrew Gong 1, Dries Verstraete 1
1
School of Aerospace, Mechanical and Mechatronic Engineering, The University of Sydney, NSW, 2006,
Australia

Abstract

Electric unmanned aerial vehicles (UAVs) are finding increasing use in a range of civilian and military
applications. For many of these applications, an accurate prediction of remaining flight time and ability of the
UAV to fulfil its mission objectives is of paramount importance. This paper outlines a dynamic propulsion model
to accurately predict the remaining battery state of charge, range and endurance based on a given aircraft
platform and flight profile. Detailed empirical models of the battery, electronic speed controller, motor and
propeller are presented and coupled with an aerodynamic model. This allows simulation of the dynamic response
of the propulsion system and estimation of remaining battery charge. The dynamic response of the battery and
the effect of the battery voltage variation on the overall propulsion system efficiency is shown. Estimates of
range and endurance remaining at each stage of the mission are provided and the ability to update these estimates
with variations in the flight profile is demonstrated.

Keywords: electric aircraft, modelling, propulsion, RPAS, UAV, unmanned aircraft

Introduction

Electric unmanned aerial vehicles (UAVs) are being used for a growing variety of missions. In these missions, it
is imperative that the UAV mission planned is within the range and endurance capabilities of the platform. Most
missions contain segments of varying power and duration associated with flight phases such as climb, cruise and
descent, and this reduces the ease of predicting the expected range and endurance. A more detailed propulsion
model is required to be able to quantify the impact of each mission segment to determine the battery state of
charge as well as the expected range and endurance remaining for a given aircraft platform and flight profile.

This paper extends on the authors’ previous work on electric propulsion systems [1]. A dynamic battery model
has been developed that can predict the transient response of the battery to dynamic load variations. The motor
and propeller models have been refined to incorporate empirical data for increased accuracy, and the effect of
battery voltage variation on the efficiency of the ESC and motor is explored.

Aerodynamic Flight Model

A small notional electric UAV is designed using a preliminary design tool and used to simulate the thrust
requirement that is to be produced by the propulsion system [1, 2]. The UAV is designed to cruise at 1000 ft and
is sized to carry a 0.5 kg payload that consumes 20 W of power. The specifications of the UAV are presented in
Table 1. The designed UAV is similar in size to the Kahu UAV, which has a wingspan of 2.29 m and a wing area
of 0.47 m2 [3].

Table 1: Simulated UAV specifications

Characteristic Wingspan (m) Wing area (m2) Aspect ratio Empty weight (kg)
Value 2.3 0.5 11 3.05

The drag polar for this generic UAV is calculated using the method detailed in [2] which results in:

CD = 0.02142 + 0.0459(CL – 0.1536)2 (1)


The propulsive thrust required is then calculated using this drag polar, taking into account any changes in altitude
and the commensurate thrust change required.

Propulsive Models
th
7 Asia-Pacific International Symposium on Aerospace Technology, 25 – 27 November 2015, Cairns
The electric UAV powertrain modelled here consists of the battery, electronic speed controller (ESC), brushless
D.C. (BLDC) motor and propeller. In addition to this, the aircraft aerodynamic model is used to simulate the
performance of the propulsion system during flight.

Dynamic battery model

The core stage of the propulsive model presented is a dynamic battery model. This battery model is a modified
equivalent circuit model and uses empirically derived data to predict the dynamic behaviour of the battery to
electrical loads, including the state of charge of the battery [4]. The equivalent circuit model is shown in Figure
1.

Fig. 1: Battery Equivalent Circuit Model

The equivalent circuit model uses the elements Rp, Cb, Rs, Cs, Rcp and Ccp to capture the dynamics of the battery.
Cb represents the storage of battery charge, Rp represents the parasitic resistance, Ri represents internal resistance,
and the remaining RC pairs represent typical battery dynamics. Cb, Ccp and Rcp are parametrised by state of
charge (SOC) according to the following equations [4]:

Cb = CCb0 + Ccb1·SOC + CCb2·SOC2 + CCb3·SOC3 (2)

Ccp=Ccp0 + Ccp1·exp(Ccp2·SOC) (3)

Rcp = Rcp0 + Rcp1·exp(Rcp2·SOC) (4)

where the state of charge is given by the equation:

SOC=1 - (qmax – qb)/Cmax (5)

The coefficients were determined by fitting the data to test results of a 2200 mAh 3S 25C Turnigy NanoTech
lithium polymer battery and then linearly scaled in capacity to represent a 6600 mAh 3S 25C battery. Figure 2
shows the test current applied to the battery, while Figure 3 shows the measured and modelled response of the
battery to the load case. The model parameters are given in Table 2.

Fig. 2: Dynamic Battery Test Case Fig. 3: Battery Measured and Modelled Response

7th Asia-Pacific International Symposium on Aerospace Technology, 25 – 27 November 2015, Cairns


Table 2: Battery Model Parameters

Rs Cs Rp Ccp0 Ccp1 Ccp2 qmax Cmax


0.0077 3.315e3 8.652e5 1.668e4 -9.334e3 -0.418 2.64e4 2.64e4
Ri Rcp0 Rcp1 Rcp2 CCb0 Ccb1 CCb2 CCb3
0.0095 3.298e4 0.0110 -15.81 30.01 2.18e3 349.8 -465.8

Motor

A brushless D.C. motor (BLDC) model is generated based on test data of the Plettenburg series of motors [5].
This data consists of the performance curves for rpm, torque, voltage, current and efficiency. Based on this, the
required motor voltage Vm and current Im to generate a set torque at a particular speed was fitted according to the
following equations:

Vm = v1·n + v2·Q + v3 (6)

Im= i1·n + i2·Q + i3 (7)

where n is the rotation speed of the motor (RPM), Q is the torque produced (Nm), v1, v2 and v3 and i1, i2 and i3 are
the fitting constants for the voltage and current respectively. The values that were determined for a Plettenburg
Orbit 1018 motor are given in Table 3.

Table 3: Orbit 1018 Motor Model Parameters

Parameter v1 v2 v3 i1 i2 i3
Value 7.656e-4 0.0789 -0.854 -4.801e-5 1.440 2.931

Using the voltage and current performance of the motor, the power and efficiency of the motor can be
determined by the following equations:

Pin = V·I (8)

Pout = n·Q·2π /60 (9)

η= Pout / Pin (10)

where Pin is the input power to the motor, Pout is the output power of the motor and η is the efficiency of the
motor. Figure 4 shows the measured efficiency of the motor over a narrow range in blue as well as the modelled
efficiency in black.

Fig. 4: Motor Measured and Modelled Efficiency (%)

ESC
7th Asia-Pacific International Symposium on Aerospace Technology, 25 – 27 November 2015, Cairns
The electronic speed controller (ESC) is an important but tricky component to accurately model [6]. In this work,
an empirical model proposed in literature [7] is used to approximate the performance of the ESC. This model
combines a linear model of the switching losses with the internal resistive losses to yield an overall loss as a
function of the power drawn by the motor. In this model, the highest efficiency is achieved at maximum power
when switching losses are minimised. The ESC efficiency is given by the following equation [7]:

ηesc = [V – Ri,ESC – PTF (1-PWS)]/ V (11)

where V is the ESC input voltage (equivalent to the battery voltage), Ri,ESC is the ESC internal resistance, PWS is
the power setting (ratio of actual power to full power) and PTF (part throttle factor) is the constant that represents
switching losses. A Ri,ESC value of 0.0035 Ω and a PTF value of 2.1 is used based on the literature for a
Kontronik Jive 80 A ESC [7, 8]. The corresponding ESC efficiency map is shown in Figure 5.

Fig. 5: ESC Efficiency map

Propeller

The propeller model is based on test data of an APC 11x7” thin electric propeller [9, 10]. The coefficient of
thrust CT, coefficient of torque CQ and efficiency η are shown in Figure 6 for a range of advance ratios and
rotational speeds. The thrust T and torque Q generated can then be calculated according to the equations:

T = CT·ρ·n2·D4 (12)

Q=CQ·ρ·n2·D5 (13)

where ρ is the air density, n is the rotational speed (in rps) and D is the propeller diameter (in m).

7th Asia-Pacific International Symposium on Aerospace Technology, 25 – 27 November 2015, Cairns


Fig. 6: Propeller Coefficient of Thrust, Torque and Efficiency

Combined Model

The individual models outlined above are combined to form a dynamic electric propulsion model. This complete
model is able to take a simulated mission profile consisting of airspeed, rate of climb and altitude to generate the
response of the individual components as well as the overall propulsion system. A block diagram is presented in
Figure 7, showing the interaction between each of the model subsystems.

Fig. 7: Propulsion Model Block Diagram

Docile Mission

To evaluate the performance of the electric propulsive model, a docile UAV mission is simulated and the
response of each of the components varied. In this mission, the notional UAV climbs at an airspeed of 14 m/s
with a rate of climb of 2 m/s to 400 ft. This airspeed is maintained while the climb rate drops to 1 m/s to 800 ft,
and then 0.5 m/s until cruise altitude of 1,000 ft is reached. The UAV then cruises for 15 minutes to an object of
interest, when it then descents to 300 ft and orbits for 5 minutes. The UAV then returns to cruise altitude and
transits back to base. The mission altitude and thrust profiles are shown in Figures 8 and 9.

Fig. 8: Mission Altitude Fig. 9: Mission Thrust Requirement

Results

The response of each of the propulsion system components to the simulated mission profile is outlined in this
section.

The propeller response to the mission profile is shown in Figure 10. It can be seen that the propeller rotational
speed varies between 5000 and 6500 rpm in response to the thrust demand required. The efficiency can also vary
greatly, which is typically around 70% but drops below 30% for the descent to orbit. This validates the need to
use an accurate propeller model to predict the performance of the propeller and in turn the overall propulsion
system.

7th Asia-Pacific International Symposium on Aerospace Technology, 25 – 27 November 2015, Cairns


Fig. 10: Propeller RPM, Torque and efficiency variation

The propeller is directly driven by the Orbit 1018 BLDC motor, which constrains the motor to provide the
correct rotational speed and torque to the propeller. The motor power input, power output and efficiency are
shown in Figure 11. It can be seen that as the input power varies between 30 and 150 W, the motor efficiency can
vary by almost 10% from 73 % to 81 %. Note also that the relationship between power and efficiency is not
linear and that maximum efficiency occurs at a medium power setting.

Fig. 11: Motor power in, power out and efficiency variation

The electronic speed controller (ESC) is used to drive the electric motor and this is where the effects of the
dynamic variations in battery voltage begin to be visible. As outlined earlier, the efficiency of the ESC is a
function of the input voltage as well as the power setting (throttle setting). Figure 12 shows the variation of the
ESC efficiency with time. It can be seen that the ESC efficiency is initially high at greater than 84 %, dropping
down to just above 81 % by the end of the mission. As the battery is discharged and its voltage decreases, the
ESC efficiency declines. Note also that during high power phases (at 32 min), the efficiency increases slightly as
the ESC becomes more efficient at high power settings.

7th Asia-Pacific International Symposium on Aerospace Technology, 25 – 27 November 2015, Cairns


Fig. 12: ESC efficiency variation

The core of the dynamic propulsion model is the dynamic battery model. Figures 13 and 14 show the voltage,
power and state of charge variation throughout the duration of the mission profile. It can be seen that the battery
voltage exhibits significant dynamic behaviour and does not settle at any particular voltage. Note also the voltage
transients visible during load changes (e.g. 27 minutes), where the battery displays considerable time delays in
response. The battery state of charge decreases steadily in response to the load, finishing the mission with a state
of charge of 20 %. This corresponds to a state of charge that should be maintained to ensure adequate life and
performance of a lithium polymer battery [11].

Fig. 13: Battery voltage variation Fig. 14: Battery power and SOC variation

In terms of the overall contribution of each component to the efficiency of the propulsion system, it was
determined that the propeller has the greatest impact on overall efficiency, with the motor having the second
largest effect. The ESC efficiency is relatively constant compared to the other two over the course of the mission.

The performance of the overall propulsion system model can be analysed in terms of the endurance and range
remaining. This is determined from the state of charge of the battery as it responds to the electrical load applied.
Figure 15 shows the estimated endurance and range remaining, based on the state of charge of the battery and the
power draw at that moment in time. The endurance and range are set to 0 when the state of charge is 20 % to
allow for some error in the endurance and range predictions and to reduce the risk of battery damage [11].

7th Asia-Pacific International Symposium on Aerospace Technology, 25 – 27 November 2015, Cairns


Fig. 15: Estimated endurance and range remaining

It can be seen that the estimated endurance and range is initially quite low due to the high power draw during the
take-off and climb phase. However, once cruise is reached at 5 minutes, the model is reasonably accurate,
predicting a remaining endurance of 50 minutes compared to the actual mission runtime left of 45 minutes.
Range and endurance estimates increase and decrease with the fluctuations in power consumption throughout the
mission.

Dynamic mission

The mission profile outlined above is an example of a relatively static mission. To investigate the performance of
the propulsion model under a more dynamic profile with greater fluctuations, a modified version of this profile is
proposed.

In this dynamic mission, the UAV climbs to 1,000 ft in the same manner as before, then cruises for 5 min before
starting a descent to 500 ft. The UAV then descends to 200 ft for close observation before climbing back to 500
ft. This is repeated three times before the UAV then climbs back to 1,000 ft for a rapid return cruise at 22 m/s
back to base. The attitude profile and associated thrust profile is shown in Figure 16 and 17.

Fig. 16: Dynamic mission altitude profile Fig. 17: Dynamic mission thrust profile

Figures 18 and 19 show the voltage, power and state of charge of the battery throughout the dynamic mission.
The battery voltage displays large fluctuations corresponding with large load changes as the battery is
discharged. Final state of charge is estimated as 22% at a voltage of 11.05 V.

7th Asia-Pacific International Symposium on Aerospace Technology, 25 – 27 November 2015, Cairns


Fig. 18: Dynamic mission battery voltage Fig. 19: Dynamic mission battery power and SOC

The predicted efficiency of each component in the propulsion system is seen in Figure 20. Once again, the
propeller has the greatest impact on the overall efficiency, varying from 70% to 27%.

Fig. 20: Dynamic mission efficiency evolution

The predicted endurance and range remaining are shown in Figure 21, and it can be seen that the estimate
responds well to the variations in the mission profile. At 5 minutes, the predicted endurance is approximately 50
minutes, which corresponds to the low power cruise being used at that point. When a rapid return to base is
selected at 33 minutes, the estimated endurance drops to 6 minutes to account for the greater power drawn and
periodically updates as the flight profile changes. The UAV operates for a further 8 minutes before it has
returned with a final estimated state of charge of 22 %, close to the 6 minutes predicted.

Fig. 21: Dynamic mission estimated endurance and range remaining

7th Asia-Pacific International Symposium on Aerospace Technology, 25 – 27 November 2015, Cairns


To validate the performance of the dynamic battery model, the electrical load on the battery was simulated and
compared with the predicted voltage expected. The results are presented in Figure 22 and demonstrate the
agreement between model and measurements with a root mean square error value of 0.059. Also shown in green
are the results using an older model outlined in Ref. [1]. The newer dynamic model is able to predict the transient
response in much more detail that the previous static model.

Fig. 22: Dynamic mission modelled and measured battery voltage

Effect of battery voltage on overall efficiency

The results that have been presented so far detail the performance of the individual components under particular
mission scenarios. The overall efficiency can be generalised to explore the effect of voltage and advance ratio on
the overall efficiency of the propulsion system.

Figure 23 shows the overall efficiency as a function of the input voltage to the ESC and the advance ratio J that
the propeller is operating at. The advance ratio J is a non-dimensional parameter to represent the propeller speed
given by the equation:

where v is the aircraft velocity, n is the rotational speed in revolutions per second and D is the propeller diameter.
It can be seen that overall efficiency always increases with increasing voltage, and that a maximum efficiency of
45 % occurs at high voltages and between advance ratios of 0.4 to 0.5.

Fig. 23: Overall propulsive efficiency operating map (%)

Conclusion
The ability to accurately predict the performance of electric UAV propulsion systems is of immense benefit for
their increasing use. This work has developed a dynamic propulsion system model incorporating the battery,
7th Asia-Pacific International Symposium on Aerospace Technology, 25 – 27 November 2015, Cairns
electronic speed controller, motor and propeller. The dynamic battery model is able to predict the time evolution
of voltage and state of charge as the propulsion system responds to the thrust requirement. An aerodynamic
model of a small UAV has been combined with the dynamic propulsion model to simulate the response to a
mission profile and estimate the range and endurance remaining. The results demonstrate that the dynamic
propulsion model accurately predicts the battery voltage response to transients, its effects on the other
components and the estimated range and endurance. Future work will look at integrating the propulsion model
into a flight simulator to analyse the response of the propulsion system to gust inputs during flight and control
inputs from autopilots and other controllers.

7th Asia-Pacific International Symposium on Aerospace Technology, 25 – 27 November 2015, Cairns


References

1. Gong, A and Verstraete, D., “Extending range and endurance estimates of battery powered electric aircraft”, AIAC16:
16yh Australian International Aerospace Congress, Melbourne, Australia, 23-24 February, 2015

2. Verstraete, D., Cazzato, L. and Romeo, G., “Preliminary Design of a Fuel-Cell-Based Hybrid-Electrical UAV”, 28th
International Congress of the Aeronautical Sciences, Brisbane, Australia, September 23-28, 2012

3. Palmer, J.L. et al., “Preliminary flight testing of autonomous soaring with the Kahu UAS”, 14th Australian Aeronautical
Conference, Melbourne, VIC, Australia, February 28 – March 3, 2011

4. Bole, B., Daigle, M. and Gorospe, G., “Online Prediction of Battery Discharge and Estimation of Parasitic Loads for an
Electric Aircraft”, Second European Conference of the Prognostics and Health Management Society, Nantes, France,
July 8-10, 2014

5. Plettenberg High End Motors, Orbit 10/18, Datasheet, available from http://www.plettenberg-
motoren.net/images/stories/datas/Orbit/Orbit10-18.pdf, last accessed 14/07/2015

6. Lundström, D., Amadori, K., Krus, P., “Validation of Models for Small Scale Electric Propulsion Systems”,
Proceedings of the 48th AIAA Aerospace Sciences Meeting including the New Horizons Forum and Aerospace
Exposition, Orlando, Florida, USA, January 4-7, 2010, AIAA Paper 2010-483

7. Roessler, C., Conceptual Design of Unmanned Aircraft with Fuel Cell Propulsion System, Ph.D. Dissertation,
Technische Universitat Munchen, Munich, 2012

8. Mueller, M., eCalc: propCalc – Propeller Calculator, available from http://ecalc.ch/motorcalc.php?ecalc&lang=en, last
accessed 14/07/2015

9. Brandt, J. B., Small-Scale Propeller Performance at Low Speeds, M.S. Thesis, Department of Aerospace Engineering,
University of Illinois at Urbana-Champaign, Illinois, 2005

10. Deters, R.W., Performance and Slipstream Characteristic of Small-Scale Propellers at Low Reynolds Numbers, Ph.D.
Dissertation, Department of Aerospace Engineering, University of Illinois at Urbana-Champaign, Urbana, IL, 2014.

11. Omar, N. et al., “Lithium iron phosphate based battery – Assessment of the aging parameters and development of cycle
life model, Applied Energy, Vol. 113, 2014, pp. 1575-1585.

7th Asia-Pacific International Symposium on Aerospace Technology, 25 – 27 November 2015, Cairns

View publication stats

You might also like