You are on page 1of 13

Aerospace Science and Technology 59 (2016) 132–144

Contents lists available at ScienceDirect

Aerospace Science and Technology


www.elsevier.com/locate/aescte

Optimal performance and sizing of a battery-powered aircraft


Giulio Avanzini a,1 , Emanuele L. de Angelis b,∗,2 , Fabrizio Giulietti b,3
a
University of Salento, Lecce, 73100, Italy
b
University of Bologna, Forlì, 47121, Italy

a r t i c l e i n f o a b s t r a c t

Article history: This paper presents an analytical and experimental framework for investigating the performance of fixed-
Received 7 June 2016 wing electrically-driven propeller aircraft. After the experimental derivation of a novel constant power
Received in revised form 29 July 2016 discharge model for lithium-polymer (Li-Po) batteries, closed-form expressions for flight endurance and
Accepted 15 October 2016
range are derived by equating power required in steady level flight with power supplied by the battery
Available online 24 October 2016
pack. Best endurance and best range airspeed are obtained. A methodology is also described to optimally
size the battery capacity for a given set of battery and aircraft characteristics. The proposed approach is
validated by application to a test case relative to a small-size unmanned aerial vehicle.
© 2016 Elsevier Masson SAS. All rights reserved.

1. Introduction the power dissipated by aerodynamic drag, accounting for losses


due to the propeller, motor, and motor controller. If the same
The interest in Remotely Piloted Aerial Systems has been battery pack also supplies energy to payload and avionic system,
steadily and rapidly increasing over the last 15 years, starting from a further share of electrical power needs to be considered in the
applications in academic research to nowadays widespread opera- energy balance equation. Note that, if on one side such equation
tive scenarios in both civil and military applications. In particular, is straightforward, on the other hand modeling of the battery be-
such platforms represent a great potentiality in the field of aerial havior is far from trivial, when effective battery capacity depends
photography and 3D reconstruction, environmental monitoring [1], on temperature, aging, cycling, and current drawn. The so-called
disaster prevention and relief [2], and search and rescue opera- Peukert law suggests a power relationship between discharge cur-
tions [3]. According to the particular requirement, different kinds rent and delivered capacity over some specified range of discharge
of vehicles and propulsive systems are currently employed. Rotary- currents [7]. Peukert equation was empirically proven for lead-acid
wing configurations gained a particular relevance, thanks to their batteries, but its validity was recently tested for different types of
hovering, vertical take-off and landing capabilities, and to the abil- batteries (e.g, lithium-ion batteries), in the case of constant dis-
ity of flying in confined spaces. On the other hand, when higher charge current [8].
velocity and/or higher altitude and/or longer range are required, In Ref. [9] endurance and range equations are derived using
fixed-wing aircraft represent the most suitable solution. Peukert law for the determination of discharge time as a func-
Expressions to estimate the endurance and range of piston- tion of the current drawn, which in turn is obtained from the
propeller and jet aircraft are long-standing results (e.g., Breguet expression of necessary power. The maximum endurance and the
equations), where the aircraft is subject to a decrease of weight corresponding best endurance condition are evaluated and later
caused by fuel consumption [4]. Conversely, the analysis of perfor- validated by means of an experimental investigation [10]. An ap-
mance for battery-powered aircraft represents a relatively recent proximation for maximum range and best range airspeed is also
topic of discussion [5,6]. In steady rectilinear level flight, the bat- derived, which provides the exact result only in the case of an ideal
tery pack is required to deliver an electrical power which equates battery, when Peukert exponent n is equal to 1 and the delivered
capacity is thus independent of the discharge current. In Ref. [11]
the exact analytical expression for the best range condition is de-
*Corresponding author. rived for the general case (n = 1). It is important to underline that
E-mail addresses: giulio.avanzini@unisalento.it (G. Avanzini), the results in [9] and [11] are valid under the hypothesis of con-
emanuele.deangelis4@unibo.it (E.L. de Angelis), fabrizio.giulietti@unibo.it
stant discharge current and constant battery voltage. The latter is a
(F. Giulietti).
1
Professor, Department of Engineering for Innovation (DII). major simplifying assumption, provided that battery voltage slowly,
2
Research Fellow, Department of Industrial Engineering (DIN). yet steadily, decreases during the discharge process and higher cur-
3
Associate Professor, Department of Industrial Engineering (DIN). rents are thus required to provide the same power to the electric

http://dx.doi.org/10.1016/j.ast.2016.10.015
1270-9638/© 2016 Elsevier Masson SAS. All rights reserved.
G. Avanzini et al. / Aerospace Science and Technology 59 (2016) 132–144 133

motor [12]. On the other hand, a closed-form solution is no longer tronic load. In Section 3 the power required for steady level flight
available when the variation of voltage with battery charge is then is evaluated and endurance and range equations are solved. The
taken into account, even in the case of an elementary discharge best airspeed values are obtained as a function of battery features.
model like the one proposed in [12] and [13]. In Section 4 the sizing of battery capacity is performed for both
For many applications of interest, a constant power discharge endurance and range optimization. Numerical simulations and a
process is more representative of the actual battery loading, as it laboratory-test campaign, validating the effectiveness of the pro-
happens for a steady-level flight of a fixed-wing aircraft or a hov- posed approach, are then presented in Section 5. A section of
ering condition of a multi-rotor vehicle. In Ref. [14] an open-circuit concluding remarks ends this paper.
voltage discharge model for the constant-power case is proposed,
provided battery specifications are known (including internal re- 2. Battery model
sistance, Peukert coefficient, etc...). The effect of discharge rate on
available capacity is taken into account by modifying Peukert equa-
tion for variable current discharges. 2.1. Experimental setup
In the present paper, a novel integral formulation for constant
power battery discharge process is proposed, where time is di- In this Section, a battery discharge model is derived on the ba-
rectly expressed as a function of discharged capacity and absorbed sis of experimental results. All tests are performed on Li-Po battery
power. The model is based on an extensive experimental campaign packs at room temperature, T 0 = 23 ◦ C, by means of a Maynuo®
performed on Li-Po batteries, one of the most popular choices for M9715 DC electronic load, which allows discharge programs at
unmanned aerial vehicles, due to the high energy-to-weight ra- constant voltage, current, load resistance, or power. In particular,
tio, the capability of high-rate discharge, and the variety of shapes each experiment is conducted at constant battery power, defined
and sizes. As an advantage with respect to existing models for as
constant-power discharge processes, the integral model here pro-
posed does not require the knowledge of voltage variations as a Pb = V I (1)
function of current and residual capacity. This in turn allows for a
closed-form solution to both the endurance and range equations of where V is the voltage across the terminals of the battery when
electrically-driven propeller aircraft, without the need for simplify- the discharge current I is delivered. Since V decreases during the
ing assumptions or numerical discretization required for the use of discharge process, the experiments are conventionally terminated
constant voltage models or models derived from Peukert law. Re- when the open-loop battery voltage, V O L , reaches the nominal
sults are analyzed and compared to the models introduced in [9] voltage, V0 = N · Vcell , where N ∈ N0 is the number of series-
and [11]. Some considerations on the effect of ambient tempera- connected cells and Vcell = 3.7 V is the nominal open-loop voltage
ture on battery performance are also included in the discussion. of a single Li-Po cell. In such a way, the battery is prevented to op-
Finally, an optimal sizing approach for the battery pack is de- erate in the over-discharging zone and to be damaged. The open-
rived, extending the method proposed in [15] for rotary-wing un- loop voltage can be estimated by means of Ohm’s Law, namely
manned aerial vehicles to the case of fixed-wing battery-operated
aircraft. Excluding the possibility of extending range and/or en- VO L = V + R I (2)
durance by dumping exhausted batteries out of the aircraft in
where R I is the voltage loss caused by battery internal resistance
flight [16], and starting from the balance equation of required and
and polarization of the active materials during discharge. The re-
available power, a model is derived in [15], where the endurance of
sistance, R, which is largely independent of the state of charge,
a multi-rotor vehicle in hovering condition is expressed as a func-
increases as the battery ages and cycles, and it depends on battery
tion of airframe features, rotor figure of merit, payload required
temperature, T b = T 0 +  T joule , where  T joule =  T joule ( I ) > 0 is
power, and on-board battery capacity, based on Peukert discharge
model. In such a way, a condition for maximum endurance is de- the temperature increase due to Joule effect and T 0 represents bat-
termined in terms of optimum capacity and weight, thus allowing tery temperature at the beginning of each experiment. Different
to properly size the battery pack with respect to the vehicle take- methods are available in the literature to measure R. In this study,
off weight. a sample test is performed on an aged N = 6 cells Li-Po battery
In this paper, a similar approach is derived for the battery sizing manufactured by Robbe® , with a nominal capacity C 0 = 5 Ah and
of a fixed-wing aircraft for both endurance and range optimization, a nominal voltage V0 = 6 · 3.7 = 22.2 V. Internal resistance is eval-
based on the novel constant-power discharge model. The model is uated by means of the electronic load according to the current-off
here modified in order to express discharge time as a function of method [19], performed at different discharge currents. Results are
battery nominal capacity. The conditions for battery pack size that reported in Fig. 1, where the circle markers represent the values of
provide maximum endurance or maximum range are derived for R obtained experimentally as a function of I . A polynomial regres-
the most general case, by means of a numerical root-finding proce- sion is proposed in order to fit the experimental data, namely
dure. A particular analytical solution is also obtained for maximum
endurance, when systems power P s is neglected or provided by a 
3

dedicated battery. These results represent a valuable instrument R(I ) = rj I j (3)


when aircraft design process is at a conceptual stage. j =0
All the derivations carried out in this paper are applied to a
practical case relevant for current technologies, which limit the where r0 = 6.23 · 10−2 , r1 = −4.17 · 10−4  A−1 , r2 = 3.11 ·
use of wholly-electrically powered aircraft to the domain of small- 10−6  A−2 , and r3 = −1.16 · 10−8  A−3 . Internal resistance de-
size unmanned aerial vehicles. Nonetheless, it is worth highlight- creases for higher discharge currents. Since battery temperature T b
ing that the principles developed for the determination of best- increases as a consequence of Joule effect, internal chemical reac-
endurance and best-range conditions, as well as the sizing proce- tions are accelerated. Battery pack sealing, external surface area,
dure, are of general validity and, consequently, they could be easily and housing also affect heat transfer to the environment, thus in-
extended to future electrically-driven manned aircraft [17,18]. fluencing T b . High temperatures thus improve performance at the
In the next section the battery model is presented, based on cost of increasing unwanted parasitic chemical reactions, resulting
experimental results conducted by means of a programmable elec- in a corresponding loss of battery life [20].
134 G. Avanzini et al. / Aerospace Science and Technology 59 (2016) 132–144

Fig. 1. Battery internal resistance (N = 6, T 0 = 23 ◦ C).

2.2. Constant power discharge model of power, P b , and approximately proportional to N. On the other
hand, no trend is identifiable for β  , and a constant value equal to
In the first set of experiments, the battery is discharged at 4 the mean value, β = 0.9664, calculated from Table 1, can be as-
different power levels, namely 50 W, 100 W, 200 W, and 300 W, sumed for the model in Eq. (5). A least squares algorithm is used
starting from a fully charged condition. The results are described in to seek the updated values of α that best fit the model in Eq. (5),
Figs. 2 and 3, where the measured closed-loop voltage and current when β  = β . The new values of α are reported in Table 1 with
are plotted as a function of the discharged capacity, C , obtained as the indication of the algorithm Root Mean Square Error (RMSE).
A test–retest procedure is also performed in order to evalu-
t ate the measurement repeatability under the same experimental
C (t ) = I (s) ds (4) conditions. In particular, for each battery pack and each constant-
0
power value, the test–retest variability of α proves to be very small
over four equal experiments, provided β  = β . The case repre-
and reported in terms of Depth of Discharge (DOD), which repre- sented by the discharge at P b = 200 W of the pack with N = 6,
sents the percentage of discharged capacity with respect to the for example, shows a dispersion of the obtained α values with
nominal capacity, C 0 . Battery voltage decreases during the dis- R M S E = 0.0003 about the reference value reported in Table 1,
charge process, and higher currents become necessary to provide namely α = 0.119.
the same power required by the electrical load. Hence, for a given For each number of series-connected cells, α ( P b , N ) is plotted
time t, discharged capacity increases with the requested power in Fig. 5 as a function of P b . For each battery type, α decreases
(see Fig. 4, where discharge time is reported as a function of exponentially with P b (see the circle markers). A curve-fitting is
discharged capacity). Taking into account the results reported in thus performed in order to characterize the model
Fig. 4, a battery model is presented where time, t, is a function of
discharged capacity, C , and requested power, P b . Assume α ( P b , N ) = δ( N ) P b (N ) (6)
β in terms of two coefficients, δ( N ) > 0 and  ( N ) < 0, reported in
t (C , P b ) = α C (5)
Table 2, and used to plot the curves in Fig. 5 (see the solid lines).
where α and β  are positive functions of power and battery pa- In Fig. 6 the coefficients δ and  are plotted as a function of cells
rameters (for example, the number of cells, N), to be identified number, N (see the circle markers). In order to characterize battery
experimentally. configurations other than the ones given in Table 2, the following
In order to characterize the model proposed in Eq. (5) for dif- models are used to identify the functions δ = δ( N ) and  =  ( N )
ferent battery configurations, a second set of experiments is con- for a generic value N ∈ N0 (see the solid lines in Fig. 6):
sidered. Different Li-Po battery packs are tested with N = 1, 2, 4, 6
cells, each one produced by a different manufacturer (Nano-Tech® , δ( N ) = −0.1067 N 3 + 0.8960 N 2 + 2.488 N + 0.6299 (7)
FightPower® , Turnigy® , Robbe® , respectively). For each pack, 4 dif- −4 3 −3 2
ferent power levels are set, for a total of 16 experiments. Note that
 ( N ) = 2.917 · 10 N − 1.375 · 10 N
−3
each battery pack was previously characterized in terms of internal + 3.083 · 10 N − 1.041 (8)
resistance at room temperature, in order to stop the experiments
when V O L ≡ V0 , disregarding the characteristics of discharge pro- 2.3. Temperature effects and other considerations on battery
cess used for the particular test. After each constant-power dis- performance
charge test, a least squares algorithm is used for evaluating the
coefficients α and β  that best fit the model proposed in Eq. (5) As for any empirically derived model, based on experimental
to the experimental data points. In Table 1 the experimental re- results and interpolation techniques, the values of model param-
sults are reported, with the desired power, P b , set on the electronic eters are related to the particular class of investigated battery
load, the mean value, P b , of the actual power discharged, and the and operating conditions (such as battery age, number of dis-
values of α and β  resulting from the selected nonlinear curve- charge/recharge cycles, temperature, etc.). As an example, temper-
fitting algorithm. Note that α shows to be a decreasing function ature at which a battery is discharged affects its service life and
G. Avanzini et al. / Aerospace Science and Technology 59 (2016) 132–144 135

Fig. 2. Battery voltage during constant power discharges (N = 6, T 0 = 23 ◦ C).

Fig. 3. Battery current during constant power discharges (N = 6, T 0 = 23 ◦ C).

Fig. 4. Discharge time during constant power discharges (N = 6, T 0 = 23 ◦ C).


136 G. Avanzini et al. / Aerospace Science and Technology 59 (2016) 132–144

Table 1
Experimental results (Eq. (5) characterization, T 0 = 23 ◦ C).

N V0 [V] C 0 [Ah] Pb (P b , R M S E) [W] α β α (β  = β ) R M S E [h]


2.5 (2.475, 0.002) 1.5045 0.9627 1.5080 0.0025
5 (4.980, 0.001) 0.7348 0.9613 0.7377 0.0007
1 3.7 0.75
10 (9.977, 0.003) 0.3521 0.9578 0.3545 0.0005
15 (14.979, 0.004) 0.2221 0.9562 0.2241 0.0003

10 (9.958, 0.003) 0.7782 0.9670 0.7782 0.0012


20 (19.952, 0.006) 0.3815 0.9650 0.3814 0.0007
2 7.4 2.5
40 (39.959, 0.008) 0.1830 0.9665 0.1830 0.0003
60 (59.954, 0.009) 0.1184 0.9769 0.1186 0.0004

50 (49.904, 0.008) 0.3205 0.9670 0.3206 0.0012


100 (99.891, 0.011) 0.1581 0.9664 0.1581 0.0006
4 14.8 5.4
200 (199.904, 0.041) 0.0762 0.9828 0.0757 0.0002
300 (299.932, 0.046) 0.0496 0.9717 0.0498 0.0002

50 (49.853, 0.055) 0.4836 0.9659 0.4834 0.0016


100 (99.855, 0.011) 0.2404 0.9662 0.2404 0.0008
6 22.2 5
200 (199.850, 0.026) 0.1193 0.9640 0.1190 0.0005
300 (299.864, 0.062) 0.0787 0.9656 0.0787 0.0003

Fig. 5. Experimental results (Eq. (6) characterization, β  = β = 0.9664, T 0 = 23 ◦ C).

Fig. 6. Experimental results (α0 , α P -characterization, β  = β = 0.9664, T 0 = 23 ◦ C).


G. Avanzini et al. / Aerospace Science and Technology 59 (2016) 132–144 137

Table 2 formance, when the battery pack is new, and degraded conditions,
Experimental results (Eq. (6) characterization, β  = β = 0.9664, T 0 = 23 ◦ C). when close to the end of their use. An accurate estimate of dis-
N V0 [V] δ( N )  (N ) charge time would thus require to repeat the whole experimental
1 3.7 3.872 −1.039 campaign at various stages of battery life. Anyway, to the authors’
2 7.4 8.471 −1.038 experience with Li-Po batteries, the trends of δ( N ) and  ( N ) are
4 14.8 18.18 −1.032 correctly evaluated, with only minor variations. For larger battery
6 22.2 24.96 −1.009 packs an ad hoc experimental campaign is also required, to extend
the model based on larger values of cells (N > 6).
voltage characteristics, due to the reduction in chemical activity
and the increase in the internal resistance of the battery at lower 3. Performance optimization
temperatures [20]. In particular, at lower ambient temperature, T 0 ,
and the same discharge power, P b , a reduction of usable capacity 3.1. Required and available power for cruise flight
is expected, as well as an increase in the slope of the discharge
curve (see [21] for the effect of ambient temperature on valve- The total power required by an electric aircraft in steady-level
regulated lead-acid batteries). In what follows it is investigated flight is
how the effects described above of temperature on battery per-
P tot = P s + P r (9)
formance influence the coefficients of the model in Eq. (5).
Consider the battery pack tested in Section 2.2, with N = 2. where P s = P a + P p is the total power required for avionics and
A set of constant power discharges is performed with the same possible payload, and P r is the power required to overcome drag
power levels (10 W, 20 W, 40 W, and 60 W) at a constant room D at a given flight velocity V , namely
temperature T 0 = 17 ◦ C. In particular, each test is stopped at the
Pr = D V (10)
same value of the closed-loop voltage which was reached at the
end of the corresponding test performed at higher temperature. A parabolic drag polar in the form C D = C D 0 + kC L2
is assumed,
In Fig. 7 a comparison is reported for the 60 W discharge test where C D 0 is the zero-lift drag coefficient and kC L2
is the term
between the closed-loop voltage time histories. The detrimental  
related to the induced drag. Remembering that C L = 2W / ρ S V 2 ,
effect of lower temperature can be quantified at a given value of
where ρ is air density, S is wing reference area, and W is aircraft
discharged capacity. Assuming C = 1.41 Ah, the experiment con-
weight [9], the power required for cruise flight in Eq. (10) can be
ducted at P b = 60 W shows a decrease of discharge time of the
written in the form
colder battery about 1.92% with respect to the same experiment
performed at 23 ◦ C. For the sake of brevity, the experimental re- P r = D V = 0.5ρ V 3 SC D = A V 3 + B V −1 (11)
sults are not discussed in detail. The procedure used to estimate
the coefficients α and β  that best fit the model in Eq. (5) to the where
experimental data points provides a mean value to β  given by 2kW 2
β = 0.9728, calculated over four experiments at T 0 = 17 ◦ C. At this A = 0.5ρ SC D 0 ; B= . (12)
value of β , the curve-fitting technique that allows to characterize
ρS
the model in Eq. (6) gives δ = 8.709 and  = −1.053 (respectively In a condition of steady flight, the power required to over-
2.8% bigger and 1.5% smaller than the corresponding values re- come drag, P r , and the available power, P a = T V , related to the
ported in Table 2 for the case when T 0 = 23 ◦ C), whose variation thrust, T , generated by aircraft propeller, are equal. Note however
denotes the expected decrease of endurance performance. that the power output of the battery dedicated to the propul-
In general, the parameters that define the variation of δ and  sion system is reduced by losses within the electric driving system
as a function of the number of cells N depend on battery condition made of a speed regulator, an electric motor, and a propeller. Al-
and type. For this reason, the experiments were performed on bat- though each element has its own efficiency, ηr , ηm , and η p , respec-
tery packs with exactly the same technology, at approximately half tively, for the purpose of the present work, they will be combined
of their operational lifespan, as a compromise between better per- into an overall efficiency, ηtot = ηr ηm η p , such that

Fig. 7. Battery voltage during constant power discharges (N = 2, P b = 60 W, different room temperatures).
138 G. Avanzini et al. / Aerospace Science and Technology 59 (2016) 132–144

ηtot ( P b − P s ) = P r dx
(13) = δ V − Ā (1 + 3 ) V 4 + P s V + B̄ (1 −  )
dV
where ηtot is assumed to be a constant. Taking into account   − 1
Eqs. (11), (12), and (13), the total power requested from the bat- × Ā V 4 + P s V + B̄ Cβ = 0
tery for cruise flight becomes
This yields the condition
P b = Ā V 3 + B̄ V −1 + P s (14)
Ā (1 + 3 ) V 4 + P s V + B̄ (1 −  ) = 0 (21)
where
that can be analytically solved.
0.5ρ SC D 0 2kW 2 In the case when the power absorbed by aircraft avionics and
Ā = ; B̄ = (15)
ηtot ρ S ηtot payload is negligible ( P s ≈ 0 W), Eq. (21) gives

3.2. Endurance optimization  −1 4 B̄
V br =
4
(22)
1 + 3 Ā
For a given usable capacity, battery technology, and ambient
temperature, cruise endurance is estimated by means of Eqs. (5) and the best airspeed ratio is
and (6), provided the discharge power is given by Eq. (14). It fol-
lows that  −1
bbr =
4
(23)
  1 + 3
t = δ P  C β = δ Ā V 3 + B̄ V −1 + P Cβ s (16)
b Note also that, if  = −1, the ideal situation in which discharge
where battery power is assumed to be nonzero. Since  ≤ −1, time is simply inversely proportional to the absorbed power is ob-
according to the proposed discharge model, one obtains that the tained. In such a particular case, the ideal best range condition
best endurance is achieved by flying at minimum-power condition, identified in [9] for ideal discharge process with unity Peukert
which occurs at the airspeed [9] coefficient, n = 1, is recovered by Eqs. (22) and (23), which give
 V br ≡ V E max and bbr ≡ 1, respectively.
Consider now the general case represented by Eq. (21). Let
1 4 B̄
V be = √ (17)
4
3 Ā 0 = 12 Ā B̄ (1 + 3 ) (1 −  ) ; 1 = 27 Ā P s2 (1 + 3 ) (24)
Let E = L / D be the aircraft aerodynamic efficiency in the selected and
cruise condition. A non-dimensional speed ratio b = V / V E max is
 
adopted as in [11], for representing the actual cruise condition 
3 1 + 21 − 4 30
with respect to the maximum efficiency condition, E = E max ,

Q = ;
where V = V E max . It is easy to express the lift coefficient and the 2
aerodynamic efficiency as a function of b in the form [11]   
1 1 0
R= Q + (25)
1 2b2 2 3 Ā (1 + 3 ) Q
CL = C L Emax ; E= E max (18)
b2 1 + b4
Provided  ≤ −1, it is possible to prove that the fourth order poly-
where both the values of C L Emax = C D 0 /k and E max = 1/ 4C D 0 k nomial in Eq. (21) always has one pair of complex conjugate roots.
depend on the coefficients of the drag polar only and may be thus One of the remaining real solutions is always negative. The only
assumed as fixed, for a given vehicle aerodynamic configuration. real positive solution provides the value of the best range airspeed
Note, in particular, that V E max = B̄ / Ā, such that Eq. (17) provides
4
in the form [22]
the optimal value of the airspeed ratio for endurance optimiza- 
tion: 1 Ps
√ V br = R + −4R 2 − (26)
bbe = 1/ 3
4
(19)
2 R Ā (1 + 3 )

Note that for higher value of P s , the endurance is reduced and the best airspeed ratio
(a greater share of battery power is exploited by on-board systems    
and payload), but the best endurance flight velocity is unaffected 4 Ā 1 Ps
bbr = R+ −4R 2 − (27)
by P s , being always equal to the minimum-power airspeed. B̄ 2 R Ā (1 + 3 )

3.3. Range optimization for range optimization. For higher values of P s , the range obtained
from Eq. (20) for V = V br derived in Eq. (26) steadily decreases,
The range of an electrically-driven propeller aircraft is obtained whereas the best range airspeed increases, as it will be shown nu-
by multiplying the endurance t in Eq. (16) by the airspeed V , merically in Section 5.
namely
  4. Sizing of battery capacity
x = t V = δ V P b C β = δ V Ā V 3 + B̄ V −1 + P s Cβ (20)
4.1. Preliminary considerations
Note that, according to the battery model coefficients obtained
in Section 2, the endurance in Eq. (16) is expressed in hours, h. In Section 3 the best endurance and best range conditions were
Hence, a conversion factor equal to 3.6 should be taken into ac- determined on the basis of the proposed battery discharge model
count if a range expressed in kilometers is desired. in terms of airspeed in steady level flight for a given aircraft con-
The optimal value for the best range airspeed is obtained by figuration (total weight, geometric and aerodynamic characteristics,
equating to zero the derivative of range x with respect to V in battery size, etc.). Contrary to what happens in conventional fuel-
Eq. (20), namely powered aircraft, for which weight steadily decreases during cruise
G. Avanzini et al. / Aerospace Science and Technology 59 (2016) 132–144 139

1/ 3
and an increased fuel fraction always provides increased endurance where ψ(b, ρ ) = μ(b) νre f (ρ ) W re f and νre f (ρ ) is obtained from
and range, the weight of electrically-powered aircraft remains con- Eq. (30) for S = S re f .
stant and the beneficial effect of weight loss during cruise is not Finally, let σ = W b / E 0 = W b / (V0 C 0 ) indicate battery weight/
experienced, as qualitatively highlighted in the introduction. As a energy ratio (that is, the inverse of battery energy density). Hence,
consequence, increasing battery weight may not necessarily pro- aircraft total weight in Eq. (31) can be rewritten as
vide an increased endurance and/or range, if the energy cost of
lifting more weight during the whole cruise overcomes the benefit W = W p + ( W eo / W ) W + W b = W p + W γ + 1 + σ V0 C 0 (35)
of the increased battery-pack capacity.
In what follows, aircraft sizing will be performed assuming takeoff
In this section, the optimal value of the battery pack that max-
weight W as the independent variable and determining battery
imizes either endurance or range is determined, following an ap-
capacity from Eq. (35),
proach similar to that described in [15], where the optimal battery
pack sizing was obtained for a multirotor configuration using the
W − Wp − W γ +1
classical Peukert discharge model. In the present case, the novel C0 = (36)
battery discharge model is applied to the case of fixed-wing elec- σ V0
tric aircraft, in order to derive useful guidelines for sizing the
4.2. Battery sizing for endurance optimization
battery pack during the conceptual design phase for this class of
vehicles.
First of all, it is necessary to rearrange the expression of the Consider Eq. (16) and the expressions of battery power and
electrical power, P b , as a function of aircraft weight. From Eqs. (11) nominal capacity respectively given by Eqs. (34) and (36). Assume,
and (13) it is P b = D V /ηtot + P s . Taking into account the definition without loss of generality, that the nominal capacity C 0 is fully dis-
of aerodynamic efficiency, E, and the parametrization introduced charged during the cruise and that the√airspeed ratio b is equal to
the optimal value in Eq. (19), bbe = 1/ 3. Hence
4
in Eq. (18), the power requested from the battery can be expressed
as an explicit function of total aircraft weight, namely  β
  W − Wp − W γ +1
7/6
WV t = δ ψbe W + Ps (37)
Pb = + P s = ν (ρ ) μ(b) W 3/2 + P s (28) σ V0
E ηtot
where ψbe = ψ(bbe , ρ ). The battery configuration for maximum
where
endurance is obtained by taking the derivative of t with respect
1 + b4 to W and imposing dt /dW = 0. This leads to the equation
μ(b) = (29)  
2b E max ηtot f t ( W ) = 7 ψbe W 1/6 W − W p − W γ +1
is a function of the selected airspeed condition and   
 − 6β (γ + 1) W γ − 1 ψbe W 7/6 + P s = 0 (38)
2
ν (ρ ) = (30) To the best of the authors’ knowledge, Eq. (38) cannot be solved
ρ SC L Emax
analytically in the most general case. Hence, a numerical method
depends on cruise altitude. is required (e.g. Newton’s iterative scheme [26]) in order to ob-
Aircraft weight, which explicitly appears in Eq. (28), can be con- tain a solution for the optimal weight W be and the corresponding
veniently decomposed into C 0be . An analytical solution to Eq. (38) can be obtained when no
payload is required to be carried on board, such that W p = 0 N
W = W eo + W p + W b (31) and P p = 0 W. Assuming the power required for aircraft avionics
is negligible ( P a ≈ 0 W), one gets
where W eo is the operative empty weight (that includes structure,
propulsion system, and avionics weight), W p is payload weight,  1/γ
( 6β + 7 ) /
and W b is battery weight. At a conceptual design stage, it is pos- W be = (39)
6β (γ + 1) + 7
sible to express the operative empty weight fraction as a function
of takeoff weight, in the form while the corresponding battery capacity is given by Eq. (36) for
W = W be .
( W eo / W ) = Wγ (32)

where the regression coefficients and γ are obtained from statis- 4.3. Battery sizing for range optimization
tics based on available data for the considered class of aircraft [23].
In the present application, data from [24] are used for the de- The optimal aircraft configuration which maximizes range for
termination of values for and γ , valid for electrically-powered a given payload achieves the best performance by flying at the
remotely piloted vehicles. best range airspeed identified by Eq. (27). From the definitions in
As discussed in [25], it is possible to assume that wing surface Eqs. (15), (24), and (25), it is evident that the best range airspeed
S scales approximately as W 2/3 . Thus, assuming S re f and W re f ratio bbr depends on aircraft weight W and is influenced by bat-
as wing surface and takeoff weight of a reference aircraft, respec- tery parameters and power required by the payload. For the sake
tively, one gets of clarity, the functional argument of bbr will be omitted in what
follows.
S = S re f ( W / W re f )2/3 (33) Aircraft range is obtained by multiplying endurance in Eq. (37)
times the airspeed V br = bbr V E max , where V E max can be written
In order to allow for quantitative comparison, the aircraft model as
considered in [9] will be taken as a reference for the study in
−1
the numerical example proposed in the next section. Equation (28) V E max = ν (ρ ) W 1/2 = ψbr μbr W 1/6 (40)
thus achieves the form
according to Eqs. (30) and (35), provided ψbr = ψ(bbr , ρ ), μbr =
P b = ψ(b, ρ ) W 7/6 + P s (34) μ(bbr ), and bbr = bbr ( W ). Hence, since x = t V , it is
140 G. Avanzini et al. / Aerospace Science and Technology 59 (2016) 132–144

  
−1 (8), the battery-model coefficients are δ = 13.28 and  = −1.036.
x = bbr δ ψbr μbr W 1/6 ψbr W 7/6 + P s
 β Provided S = S re f = 0.32 m2 and W = W re f = 9.34 N, one has Ā =
W − W p − W γ +1 5.760 · 10−3 kg m−1 and B̄ = 1.181 · 102 kg m3 s−4 from Eq. (15).
× (41) Let P s = 5 W be the power required for avionics and payload. Pro-
σ V0
vided V E max = B̄ / Ā ≈ 12 m s−1 at the maximum aerodynamic
4

The size of the battery pack that maximizes range performance efficiency E max = 11.32, the total power required to the battery for
should be obtained by taking the derivative of x with respect to cruise flight is plotted in Fig. 8 as a function of the airspeed ratio,
W and solving the equation dx/dW = 0. Unfortunately, the com- b [see Eq. (14)].
plexity of the expression of the derivative precludes, to the best Flight endurance expressed in minutes as given by Eq. (16) is
of the authors’ knowledge, the possibility of an analytical solution. reported in Fig. 8 for different values of discharged capacity. Note
As a consequence, the function that relates aircraft total weight to that the best endurance
√ condition is exactly located by Eq. (19),
where b = bbe = 1/ 3 and P b = P bmin = 22.32 W (see Table 4).
4
the expected range performance in Eq. (41) is plotted and the best
range configuration is identified either graphically on the plot or An experiment is also performed where a FullPower® battery pack
numerically, by means of a simple root-finding algorithm, such as with the above characteristics is discharged by means of the elec-
the parabolic search or the simplex method [27,28]. trical load to the 80% of the nominal capacity C 0 at the constant
power P b = P bmin = 22.32 W. In such a case, the discharge time
5. Numerical example is equal to t exp = 54.6 min (see the circle marker in Fig. 8). In
order to validate the proposed battery model, the same values
5.1. Performance optimization of discharged capacity and power are respectively substituted in
Eq. (16), to obtain tmod = 55.1 min, with an estimation error about
In what follows, the battery model derived in Section 2 and the 0.86%.
approach presented in Sections 3 and 4 for endurance and range As a final consideration, the validity of the proposed battery
performance optimization and battery sizing of an electrically- model is compared to a classical Peukert-based approach, where
driven propeller aircraft are validated by means of numerical sim- the assumption of constant current during battery discharge is
ulations and an experimental campaign. The small-size electrical considered, as indicated in Eq. (9) from [9]. In the present frame-
aircraft used as an example in [9], taken from [29], is considered work, an equivalent constant voltage is considered as the mean
(see Table 3), with a Li-Po battery pack characterized by N = 3, value between the open-loop fully  charged battery voltage and the
V0 = 11.1 V, and C 0 = 2.2 Ah (battery type not tested in Sec- nominal voltage, namely Ve = V f + V0 /2 = 11.85 V, provided
tion 2 for model identification). Taking into account Eqs. (7) and V f = 12.6 V. It follows that the equivalent constant current relative
to a discharge at P b = P bmin = 22.32 W is I pk = P bmin /Ve = 1.88 A.
Table 3 Let R t be the battery hour rating, that is the discharge time over
Reference aircraft and battery data. which the nominal capacity was determined by the manufacturer
Parameter Symbol Value Units (typically 1 h for small rechargeable battery packs). Provided n =
Aircraft 1.107 is the Peukert coefficient estimated for the considered bat-
Drag polar coefficients CD0 0.015 tery, predicted endurance is t pk = R t1−n (0.8 · Ve C 0 / P bmin )n · 60 =
k 0.13 55.7 min (see the square marker in Fig. 8), with an estimation
Total efficiency ηtot 0.5
error about 2%. As a final consideration, note that both Peukert ap-
Air density ρ 1.2 kg m−3
Wing area S re f 0.32 m2
proach and the proposed method provide the same best endurance
Weight W re f 9.34 N airspeed ratio.
Battery pack
The procedure outlined above is also followed for range char-
Peukert coefficient n 1.107 acterization. Flight range as given by Eq. (20) is reported in Fig. 9
No. of cells N 3 for different values of discharged capacity. The best range condi-
Nominal voltage V0 11.1 V tion is evaluated by Eqs. (24)–(27), providing V br = 12.6 m s−1
Nominal capacity C0 2.2 Ah
and b = bbr = 1.051 at the constant power value P b = 25.84 W.
System power Ps 5 W
The considered battery pack with N = 3 cells is discharged by

Fig. 8. Endurance (—) and power (- - -) as a function of airspeed ratio (N = 3).


G. Avanzini et al. / Aerospace Science and Technology 59 (2016) 132–144 141

Table 4
Experimental validation of results for optimal performance.

Discharge model Endurance: tbe [min] Range: xbr [km]


bbe P b (bbe ) [W] Theory (exp.) bbr P b (bbr ) [W] Theory (exp.)

4
Proposed model 1/ 3 22.32 55.1 (54.6) 1.051 25.84 35.69 (36.08)

4
Peukert model 1/ 3 22.32 55.7 (54.6) 1.032 25.41 35.72 (–)

Fig. 9. Range (—) and power (- - -) as a function of airspeed ratio (N = 3).

means of the electrical load to the 80% of nominal capacity C 0 Fig. 10.a shows the variation of maximum flight endurance as
at the constant power P b = 25.84 W. In such a case, the dis- determined from Eq. (37) as a function of aircraft weight for two
charge time results to be t exp = 47.7 min, corresponding to a different payload configurations and three battery packs character-
range xexp = 36.08 km (see the circle marker in Fig. 9). In order ized by the same technology and N = 2, 3, and 4 cells connected in
to validate the proposed battery model, the same values of dis- series, respectively. The function f t ( W ) in Eq. (38), which is pro-
charged capacity and power are substituted in Eq. (20), to obtain portional to the derivative of the endurance dt /dW in Eq. (37), is
tmod = 47.3 min and xmod = 35.69 km, with an estimation error plotted as a function of aircraft weight for N = 3 and both pay-
about 1.08%. loads in Fig. 10.b, highlighting that best endurance configurations
With regard to the range equation, a comparison is also pro- are obtained when Eq. (38) is satisfied. Best endurance configura-
vided between the proposed model and a Peukert-based approach tions for all the payloads and cell number are listed in Table 5,
[see Eq. (5) from [11]]. Taking into account the contribution of
together with battery parameters obtained from Eqs. (7) and (8).
P s = 5 W (not considered in [11]), it is straightforward to prove
( pk) The presence of a maximum in all the curves clearly points out
that the best range condition results into b = bbr = 1.032, where that, if endurance is pursued as the most relevant goal in the de-
P b = 25.41 W, the equivalent current is I pk = P b /Ve = 2.14 A, and
sign process, it is useless to increase the size of the battery pack
x pk = R t1−n (0.8 · Ve C 0 /25.41)n · bbr V E max · 3.6 = 35.72 km (see
( pk)
beyond a certain limit, provided that the corresponding growth
the square marker in Fig. 9). The estimation error between x pk and in aircraft weight affects necessary power for the whole cruise
xexp is about 1%. and, beyond the maximum, this growth is not compensated by the
increase in available energy (thus decreasing aircraft endurance).
5.2. Battery sizing
Weight and power required by the payload significantly affect the
weight of the optimal-size aircraft by approximately 30% for the
In what follows, the battery-sizing problem analyzed in Sec-
considered cameras, whereas the optimal endurance is less influ-
tion 4 is investigated by scaling the fixed-wing platform considered
enced by changes in payload characteristics, with an approximate
above, mounting a Li-Po battery with a weight-energy ratio σ =
0.0763 N (Wh)−1 . The regression coefficients introduced in Eq. (32) endurance increase of 6% for the lighter camera. Conversely, the
are obtained from the data reported in the UAS database in Ap- number of cells N marginally affects the optimal weight, where
pendix C of [24]. The resulting coefficients are given by = 0.6998 variations as small as −1% and +2.5% are expected when chang-
and γ = −0.0890. It is assumed that P a = 1 W and a 80% usage ing the battery from a 3 cell to a 2 or a 4 cell pack, respectively.
of nominal capacity C 0 is considered. Provided C L Emax = 0.34, from In this case, the corresponding variation of maximum endurance is
Eq. (30) it is νre f (ρ ) = 3.92 kg−1/2 m1/2 . The non-dimensional pa- more significant, but still limited to −8% when changing the bat-
rameter defined by Eq. (29) is μ(bbe ) = 0.16 and the coefficient tery pack from N = 3 to 2, and +5% when changing N from 3 to 4.
in Eq. (34) is ψ(bbe , ρ ) = 1.28. The payload is represented by an In this latter respect, one should also consider that the number of
infrared camera designed for miniature unmanned aerial vehicles, cells needs to be increased with aircraft weight, provided that a
here named Camera 1 (W p = 0.186 kgf, P p = 2.5 W). A lighter heavier aircraft requires a more powerful engine, usually working
and smaller version of the same device, here named Camera 2 at a higher voltage. This means that, even if the results obtained
(W p = 0.114 kgf, P p = 1.5 W), is also considered in order to pro- from Eq. (37) for sizing the aircraft are plotted over a wide range
vide further insight into the sizing process under different payload of aircraft weight, a number of cells consistent with engine power
requirements [30]. needs to be selected in practical applications of the sizing process.
142 G. Avanzini et al. / Aerospace Science and Technology 59 (2016) 132–144


Fig. 10. Endurance (a) and f t ( W ) function (b) for different battery/payload configurations (b = bbe = 1/ 3).
4

Table 5 √
Endurance for different battery/payload configurations (b = bbe = 1/ 3).
4

Battery parameters Camera 1 Camera 2


W be [kgf] C 0be [Ah] tbe [min] W be [kgf] C 0be [Ah] tbe [min]
N =2
3.932 30.53 98.1 2.731 20.68 104.3
(δ = 8.336,  = −1.038)
N =3
3.970 20.59 106.5 2.759 13.96 113.1
(δ = 13.28,  = −1.036)
N =4
4.068 15.90 112.0 2.829 10.79 118.8
(δ = 18.09,  = −1.032)

It is interesting to investigate, for a given battery configura- suboptimal values obtained in Cases 1 and 2. Also, a higher
tion, how technological innovation in terms of payload weight value of endurance is obtained, as reported in the second line
and power absorption drives the sizing phase of new vehicles of Table 5, related to Camera 2.
or influences the performance of existing platforms. Consider the
aircraft configuration reported in Table 5 where N = 3 and Cam- In Fig. 11 flight range as given by Eq. (41) is reported as a func-
era 1 is considered. In such a case, maximum endurance is tmax = tion of the aircraft weight for the same battery/payload configu-
106.5 min and the aircraft is characterized by W = W p + W eo + rations introduced above. The best range configuration parameters
W b = 0.186 + 2.006 + 1.779 = 3.970 kgf and C 0 = 20.59 Ah. As- are summarized in Table 6. When range becomes the performance
sume that Camera 1 is replaced with Camera 2. Three scenarios objective of the sizing process, an optimum is obtained for large
can be envisaged: values of the total weight. Note that for all the considered con-
figurations (Camera 1 and 2, N = 2, 3, and 4), the maximum is
1. The new payload is mounted on the existing platform, such “flat”, meaning that very large variations of aircraft weight are nec-
that W = W p + W eo + W b = 0.114 + 2.006 + 1.779 = 3.970 kgf, essary for marginal gains in terms of expected range. From the
and the decrease in aircraft weight by  W = 0.114 − 0.186 = practical standpoint, this growth in aircraft weight for the “op-
−0.072 kgf allows increased endurance, namely t = 110.6 min. timal” design is clearly not justified, when one considers that a
2. The aircraft weight is maintained and the decrease in pay- heavier aircraft is more expensive and requires a more powerful
load weight is balanced by an equivalent increase in battery engine.
weight and capacity respectively given by | W | = 0.072 kgf In the attempt of identifying a good compromise for the op-
and C 0 = 0.83 Ah. In such a case, it results W = W p + W eo + timal aircraft weight, Fig. 12 provides the Pareto front of the op-
W b = 0.114 + 2.006 + 1.851 = 3.970 kgf, C 0 = 20.59 + 0.83 = timal designs in the tbe vs xbr plane, where both endurance and
21.43 Ah, and the endurance is 111.8 min. range performance indexes are scaled with respect to their opti-
3. A new aircraft is sized according to the weight and power ab- mal values, as a function of aircraft total weight (reported on the
sorption of Camera 2, leading to the configuration reported in dots distributed along the curve). The case considered is N = 3
Table 5, where N = 3 and Camera 2 is considered. It is noted for Camera 2. The plot is limited to the non-dominated solutions,
that, in this third case, the values of both aircraft weight and obtained for weights ranging between W be = 2.759 kgf for the
required battery capacity are smaller than the corresponding maximum endurance design, and W br = 25.437 kgf for the de-
G. Avanzini et al. / Aerospace Science and Technology 59 (2016) 132–144 143

Fig. 11. Range for different battery/payload configurations (b = bbr ( W )).

Table 6
Range for different battery/payload configurations (b = bbr ( W )).

Battery parameters Camera 1 Camera 2


W br [kgf] C 0br [Ah] xbr [km] W br [kgf] C 0br [Ah] xbr [km]
N =2
29.852 296.37 87.48 23.323 228.29 87.98
(δ = 8.336,  = −1.038)
N =3
32.326 215.25 95.32 25.437 167.09 95.82
(δ = 13.28,  = −1.036)
N =4
40.139 203.63 101.41 32.214 161.46 101.83
(δ = 18.09,  = −1.032)

Fig. 12. Pareto chart analysis (N = 3, Camera 2).

sign which maximizes range. Given the scale of the axes, each 6. Conclusions
point on the plot represents the fraction of the optimal perfor-
mance, where an ideal best would be represented by the point A model for predicting battery discharge time during a constant-
(1, 1), where both maximum endurance and range are simultane- power discharge process was developed and discussed. The model
ously achieved. As a consequence, the best design (indicated by is based on a campaign of experiments performed with the aim of
the subscript bd) is determined numerically by selecting the point describing the behavior of lithium-polymer batteries under differ-
on the plot that minimizes the distance from the ideal point (1, 1). ent operating conditions and battery pack configurations, including
For N = 3 and Camera 2 the weight of the best design is equal to power request and nominal voltage of the battery.
W bd = 5.397 kgf, with a battery pack weight equal to 2.630 kgf, This modeling approach allowed to estimate cruise performance
a reduction of endurance and range performance equal to only 3% of an electrically-driven propeller aircraft. The best endurance and
and 4%, respectively, with respect to the maxima in the single ob- best range conditions were derived in terms of cruise airspeed and
jective optimization problems, and a weight reduction as large as compared to the results available in literature, based on constant-
a factor approximately equal to 5 with respect to the design for current discharge models. A sizing process is also proposed, which
maximum range only. allows one to size an aircraft aiming at maximum endurance or
144 G. Avanzini et al. / Aerospace Science and Technology 59 (2016) 132–144

maximum range, thus providing an insight into the battery weight [12] C.M. Shepherd, Design of primary and secondary cells. An equation describing
estimation in the framework of aircraft preliminary sizing. battery discharge, J. Electrochem. Soc. 112 (7) (1965) 657–664, http://dx.doi.
org/10.1149/1.2423659.
The methodology was applied to a small-size electrically driven
[13] O. Tremblay, L.A. Dessaint, Experimental validation of a battery dynamic model
unmanned aircraft. The results show that the sizing procedure for EV applications, World Electr. Veh. J. (ISSN 2032-6653) 3 (2009) 1–10.
provides reasonable design in terms of battery-pack weight with [14] M.E. Fuller, A battery model for constant-power discharge including rate ef-
respect to total aircraft weight only if maximum endurance is the fects, Energy Convers. Manag. 88 (2014) 199–205, http://dx.doi.org/10.1016/j.
design goal. Conversely, when maximum range is pursued, the best enconman.2014.08.015.
[15] M. Gatti, F. Giulietti, M. Turci, Maximum endurance for battery-powered rotary-
design for the same payload is characterized by a higher weight,
wing aircraft, Aerosp. Sci. Technol. 45 (2015) 174–179, http://dx.doi.org/10.
with a modest range increment. A compromise between the two 1016/j.ast.2015.05.009.
optimal designs was thus identified on the plane of the two com- [16] T. Chang, H. Yu, Improving electric powered UAVs’ endurance by incorporating
peting objective functions. battery dumping concept, Proc. Eng. 99 (2015) 168–179, http://dx.doi.org/10.
1016/j.proeng.2014.12.522.
[17] M.D. Moore, B. Fredericks, Misconceptions of electric aircraft and their emerg-
Conflict of interest statement
ing aviation markets, in: Proc. of the 52nd Aerospace Sciences Meeting, 13–17
Jan. 2014, 2014, pp. 1–17, AIAA SciTech.
There is no conflict of interest. [18] P.A. Jackson, Jane’s All the World’s Aircraft: Development & Production:
2014–15, IHS Global, 2015, p. 335.
References [19] H.G. Schweiger, O. Obeidi, O. Komesker, A. Raschke, M. Schiemann, C. Zehner,
M. Gehnen, M. Keller, P. Birke, Comparison of several methods for determining
the internal resistance of lithium ion cells, Sensors 10 (6) (2010) 5604–5625,
[1] S. Harwey, A. Lucieer, Assessing the accuracy of georeferenced point clouds
http://dx.doi.org/10.3390/s100605604.
produced via multi-view stereopsis from Unmanned Aerial Vehicle (UAV)
imagery, J. Remote Sens. 4 (6) (2012) 1573–1599, http://dx.doi.org/10.3390/ [20] D. Linden, T.B. Reddy, Handbook of Batteries, 3rd edition, McGraw–Hill Hand-
rs4061573. books, New York, 2002, pp. 13–16, 64–66.
[2] D.W. Casbeer, R.W. Beard, T.W. McLain, Sai-Ming Li, R.K. Mehra, Forest fire [21] S. Dhameja, Electric Vehicle Battery System, 1st edition, Newnes, Oxford, Oct.
monitoring with multiple small UAVs, in: Proc. of the 2005 IEEE American Con- 2001, pp. 44–46.
trol Conference, 8–10 June 2005, in: IEEE Publ., vol. 5, 2005, pp. 3530–3535. [22] E.L. Rees, Graphical discussion of the roots of a quartic equation, The American
[3] P. Rudol, P. Doherty, Human body detection and geolocalization for UAV search Mathematical Monthly 29 (2) (1922) 51–55, http://dx.doi.org/10.2307/2972804.
and rescue missions using color and thermal imagery, in: Proc. of the 2008 [23] D.P. Raymer, Aircraft Design: A Conceptual Approach, 2nd edition, AIAA Educa-
IEEE Aerospace Conference, 1–8 March 2008, in: IEEE Publ., 2008, pp. 1–8. tion Series, AIAA, Washington, DC, 1992, pp. 11–14.
[4] J.D. Anderson, Aircraft Performance and Design, 1st edition, McGraw–Hill [24] H.B.C. Veerman, Preliminary Multi-mission UAS Design, M.Sc. Thesis, Delft
Higher Education, New York, ISBN 978-0070019713, 1999, pp. 293–316. University of Technology, Aug. 2012, pp. 67–68, 221–223, available online at
[5] E.R. Retana, H. Rodrigue-Cortes, Basic small fixed wing aircraft sizing optimiz- http://repository.tudelft.nl.
ing endurance, in: Proc. of the 4th International Conference on Electrical and [25] K.P. Valavanis, Advances in unmanned aerial vehicles, in: Intelligent Systems,
Electronics Engineering, 5–7 Sept. 2007, in: IEEE Publ., 2007, pp. 322–325. Control and Automation: Science and Engineering, Springer, 2007, pp. 387–388
[6] D.A. Lawrence, K. Mohseni, Efficiency analysis for long-duration electric MAVs, (Book 33).
in: Proc. of the 2005 Infotech@Aerospace, 26–29 Sept. 2005, 2005, pp. 1–7, [26] K.E. Atkinson, An Introduction to Numerical Analysis, 2nd edition, John Wiley
AIAA 2005-7090. & Sons, Hoboken, NJ, Apr. 1989, pp. 58–65.
[7] W. Peukert, Über die Abhängigkeit der Kapazität von der Entladestromstärke [27] R.P. Brent, Algorithms for Minimization Without Derivatives, 1st edition,
bei Bleiakkumulatoren, Elektrotech. Z. 20 (1897) 20–21. Prentice–Hall, Upper Saddle River, NJ, Mar. 2013, Ch. 5.
[8] D. Doerffel, S.A. Sharkh, A critical review of using the Peukert equation [28] W.H. Press, S.A. Teukolsky, W.T. Vetterling, B.P. Flannery, Numerical Recipes:
for determining the remaining capacity of lead-acid and lithium-ion bat- The Art of Scientific Computing, 3rd edition, Cambridge University Press, Cam-
teries, J. Power Sources 155 (2) (2006) 395–400, http://dx.doi.org/10.1016/j. bridge, UK, Sept. 2007, Ch. 10.
jpowsour.2005.04.030. [29] J.N. Ostler, Flight Testing Small Electric Powered Unmanned Aerial Vehicles,
[9] L.W. Traub, Range and endurance estimates for battery-powered aircraft, J. M.S. Thesis, Brigham Young Univ., Salt Lake City, UT, Apr. 2006, pp. 1–38.
Aircr. 48 (2) (2011) 703–707, http://dx.doi.org/10.2514/1.C031027. [30] J. Kostrzewa, W.H. Meyer, S. Laband, W.A. Terre, P. Petrovich, K. Swanson,
[10] L.W. Traub, Validation of endurance estimates for battery powered UAVs, Aero- C. Sundra, W. Sener, J. Wilmott, Infrared microsensor payload for miniature
naut. J. 117 (1197) (2013) 1155–1166. unmanned aerial vehicles, in: Proc. of the SPIE 5090, 18 Sept. 2003, in:
[11] G. Avanzini, F. Giulietti, Maximum range for battery-powered aircraft, J. Aircr. Unattended Ground Sensor Technologies and Applications V, vol. 5090, 2003,
50 (1) (2013) 304–307, http://dx.doi.org/10.2514/1.C031748. pp. 1–10.

You might also like