.
`(cid:14)
`Journal of Power Sources 90 2000 156–162
`
`www.elsevier.comrlocaterjpowsour
`
`Use of lithium-ion batteries in electric vehicles
`B. Kennedy ), D. Patterson, S. Camilleri
`Northern Territory Centre for Energy Research, Northern Territory Uni˝ersity, Darwin, NT 0909, Australia
`
`Received 22 December 1999; received in revised form 1 February 2000; accepted 16 February 2000
`
`Abstract
`
`.
`(cid:14)
`An account is given of the lithium-ion Li-ion battery pack used in the Northern Territory University’s solar car, Fuji Xerox Desert
`(cid:14)
`.
`Rose, which competed in the 1999 World Solar Challenge WSC . The reasons for the choice of Li-ion batteries over silver–zinc batteries
`are outlined, and the construction techniques used, the management of the batteries, and the battery protection boards are described. Data
`from both pre-race trialling and race telemetry, and an analysis of both the coulombic and the energy efficiencies of the battery are
`presented. It is concluded that Li-ion batteries show a real advantage over other commercially available batteries for traction applications
`of this kind. q 2000 Elsevier Science S.A. All rights reserved.
`
`Keywords: Lithium-ion battery; Electric vehicle; Rechargeable battery
`
`1. Introduction
`
`.
`(cid:14)
`Lithium-ion Li-ion batteries are an attractive proposi-
`tion for use in high-performance electric vehicles. In com-
`parison with other rechargeable batteries, Li-ion provides
`very high specific energy and a large number of charge–
`discharge cycles. The cost is also reasonable. Thus, Li-ion
`batteries are the preferred choice over other technologies
`such as silver–zinc and nickel–metal-hydride. Presently,
`however, Li-ion batteries are only commercially available
`in small sizes. Accordingly, large numbers of cells have to
`be assembled in seriesrparallel configurations to achieve
`the desired battery sizes. This, combined with safety is-
`sues, presents the challenge of making highly efficient,
`highly reliable, battery packs for use in electric vehicles.
`
`2. Background on World Solar Challenge
`
`.
`(cid:14)
`The World Solar Challenge WSC is the premiere
`(cid:14)
`.
`event for solar-powered vehicles so-called, ‘solar cars’ .
`Starting from Darwin in Northern Australia, it crosses the
`continent and finishes in Adelaide, a distance of 3000 km.
`
`)
`
`Corresponding author.
`.
`(cid:14)
`E-mail address: byron.kennedy@ntu.edu.au B. Kennedy .
`
`The WSC has been run every 3 years since 1987. The
`(cid:14)
`.
`winners have been: General Motors, USA 1987 ;
`the
`(cid:14)
`.
`Engineering College of Biel, Switzerland 1990 ; Honda,
`(cid:14)
`.
`Japan 1993, 1996 ; the Aurora Vehicle Association, Aus-
`(cid:14)
`.
`tralia 1999 . The Northern Territory University is one of
`only four teams to compete in all five of the WSC events
`held to date, and the only team to complete successfully all
`(cid:14)
`.
`races. In 1999,
`the Fuji Xerox Desert Rose Fig. 1
`finished fourth overall and won its class, ‘Silicon Solar
`Cells, Exotic Batteries’, at an average speed of 71.00
`kmrh, just 68 min behind the winner after 5 days of
`racing.
`
`3. WSC rules and battery regulations
`
`w x
`The rules for building a solar car are simple 1 and can
`be summarized as follows.
`
`fl The box rule: the car must fit in a box which is 6 m
`long, 2 m wide, and 1.6 m high. The solar cells must
`have a projected horizontal area of no more than 8 m2.
`fl The clock rule: cars may race only between 0800 and
`1700 h; stationary charging of the batteries from the sun
`is allowed before and after these times, respectively.
`
`0378-7753r00r$ - see front matter q 2000 Elsevier Science S.A. All rights reserved.
`(cid:14)
`.
`PII: S 0 3 7 8 - 7 7 5 3 0 0 0 0 4 0 2 - X
`
`1
`
`APPLE 1006
`
`

`

`B. Kennedy et al.r Journal of Power Sources 90 2000 156–162
`)
`(
`
`157
`
`of the vehicle, the tyre rolling resistance and the weight of
`the vehicle. In parallel,
`the energy efficiencies of the
`vehicle components, i.e., solar array, motor and battery,
`must all be maximized.
`A common strategy for a solar car is to race throughout
`the day at a constant speed and to end the day with a
`precalculated amount of energy stored in the batteries.
`Typically, the battery capacity will be at 40% state-of-
`(cid:14)
`.
`charge SoC at the end of the first day, and at zero SoC at
`the end of the fifth day. As mentioned above, after racing
`has finished, time is available in the afternoon and next
`morning to recharge the batteries from the sun. The battery
`profile for the Fuji Xerox Desert Rose solar car on day 5 is
`shown in Fig. 2. The data show the following.
`
`fl The SoC was taken from nearly 60 A h at the start of
`the day to less than 5 A h at the finish. This was based
`on the expectation of good weather for the charging
`periods.
`fl The battery voltage started to fall away at 10 A h.
`Consequently, the vehicle speed was reduced to con-
`serve energy.
`fl Strong head winds were encountered late in the after-
`noon. This accounts for the slower speeds, yet similar
`power consumption.
`
`5. Li-ion battery configuration
`
`When determining what sort of battery to use, a number
`of factors must be considered. These include minimization
`of weight and maximization of both battery efficiency and,
`of course, cost. This is the reason why lead–acid batteries
`are still extremely popular with solar car teams who
`operate on low budgets.
`
`Fig. 1. Fuji Xerox Desert Rose solar car.
`
`Other rules also govern the exact dimensions of the
`solar array,
`the procedure at media stops, and general
`safety criteria for the vehicles. Of particular interest are the
`regulations for battery size. The cells or battery modules
`must be rechargeable by the vehicles in which they are
`fitted. The total energy must be less than a nominal 5 kW
`(cid:14)
`.
`h 20-h rate . This is determined on the basis of weight,
`(cid:14)
`namely, not more than 40 kg for silver–zinc on the basis
`that the specific energy of this system is typically 125 W
`hrkg, so that 40 kg equates to 5 kW h , 40 kg for
`.
`(cid:14)
`.
`‘advanced’ battery systems e.g., lithium-ion , 75 kg for
`nickel–zinc, 71 kg for nickel–metal-hydride, 100 kg for
`nickel–cadmium and nickel–iron, and 125 kg for lead–
`acid. The limit applies only to the weight of the individual
`cells or batteries, that is, the weight of peripheral equip-
`(cid:14)
`.
`ment connectors, battery box, etc.
`is not counted.
`
`4. Solar car strategy
`
`is essential to
`it
`In order to build a fast solar car,
`minimize the aerodynamic drag coefficient, the frontal area
`
`Fig. 2. Telemetry data from Fuji Xerox Desert Rose on fifth day of 1999 WSC.
`
`2
`
`

`

`158
`
`B. Kennedy et al.r Journal of Power Sources 90 2000 156–162
`)
`(
`
`The following configuration of cells was therefore de-
`vised for the vehicle:
`
`.
`(cid:14)
`fl 60 cells in parallel to form one module Fig. 3
`fl 15 modules in series, i.e., 900 cells in total, to form the
`(cid:14)
`.
`battery pack Fig. 4
`fl total battery weight: 39.6 kg, or 47 kg when including
`busbars, protections and battery box
`fl bus voltage: 55.5 V; maximum bus: 64.8 V
`fl total nominal capacity: 102 A h at 55.5 V, i.e., 5.6 kW
`h
`
`6. Comparison of Li-ion and silver–zinc battery config-
`urations
`
`A comparison of the two different battery types used in
`(cid:14)
`the Fuji Xerox Desert Rose silver–zinc in 1996, Li-ion in
`.
`1999 is given in Table 1. The data was obtained both
`from the solar car telemetry and from pre-race testing. The
`weight includes that of the box and all interconnects.
`The coulombic efficiency was calculated by measuring
`the ampere hours put into the battery and ampere hours
`taken out of the battery over a full charge–discharge cycle.
`The ‘Labview’ program was used to monitor Ah-in and
`Ah-out.
`The energy efficiency, the ratio of the joules-out to
`joules-in over a complete cycle, was calculated in several
`ways. First, the battery pack was put on charge at 20 A
`and the voltage was measured. Immediately following a
`discharge current of 20 A was established, the voltage was
`remeasured. The ratio of these two voltages gives an
`indication of the energy efficiency, with further tests post-
`race required for more accurate results.
`Two different tests were conducted post race, which
`were used to calculate energy efficiency. First, the voltage
`at a particular SoC was recorded during both charge and
`discharge. These voltages were recorded and were used to
`calculate the efficiency. This result, once again, only gave
`an indication of energy efficiency with further tests still
`required.
`A more accurate way to calculate energy efficiency is to
`measure the joules-in and joules-out of the battery between
`a predefined SoC window. This was conducted between 10
`
`Table 1
`Comparison of lithium-ion and silver–zinc cells
`
`Li-ion
`
`Silver–zinc
`
`.
`. (cid:14)
`(cid:14)
`Bus voltage average measured V
`(cid:14)
`. (cid:14)
`.
`Capacity measured A h
`(cid:14)
`.
`Capacity W h
`.
`(cid:14)
`Coulombic efficiency %
`.
`(cid:14)
`Energy, or charge–discharge, efficiency %
`(cid:14)
`.
`Total weight kg
`
`57
`106
`6042
`100
`95
`47
`
`48
`126
`6048
`100
`81
`40
`
`.
`(cid:14)
`Fig. 3. Li-ion module 60 cells .
`
`The cell chosen by the Fuji Xerox Desert Rose team
`was the NEC ICR18650E Li-ion design made by NEC
`w x
`Moli Energy. The basic specifications of this cell are 2 :
`(cid:14)
`.
`voltage, 3.7 V; capacity, 1.7 A h typical ; weight, 43 g
`(cid:14)
`.
`44 g maximum ; life, 500 cycles. The cell was selected on
`the basis of the following criteria.
`fl High specific energy, viz., 146 W hrkg; in previous
`events, silver–zinc cells were used with a specific
`energy of ;125 W hrkg.
`fl Good life; the cells will last the life of the solar car,
`whereas silver–zinc batteries last only a single race.
`fl Lower cost; a comparable pack of silver–zinc cells
`costs an additional US$10 000.
`fl Greater reliability and handleability; silver–zinc cells
`used in previous events proved unreliable when cycled
`to deep SoCs. Problems encountered included cracking
`of cells, which resulted in electrolyte leakage.
`fl Adaptability to existing electronics designed for a 48-V
`nominal bus; this is easily achievable with Li-ion batter-
`ies.
`
`(cid:14)
`Fig. 4. Ten Li-ion modules viewed from above bypass diodes shown on
`.
`top of cells .
`
`3
`
`

`

`B. Kennedy et al.r Journal of Power Sources 90 2000 156–162
`)
`(
`
`159
`
`Fig. 5. Capacity and voltage vs time.
`
`and 80 A h for both charge and discharge and corresponds
`to the linear charge–discharge region. The curves, shown
`in Figs. 5 and 6, outline the results used to calculate
`efficiency via counting the joules-in and joules-out of the
`battery. All three methods showed the same efficiency,
`viz., ;95%. This value also includes the losses due to the
`(cid:14)
`.
`MOSFETs used on the protection boards see later , which
`are calculated at 0.7%. Similarly, joules-in and joules-out
`were computed over a complete cycle of a silver–zinc
`battery and an efficiency of 81% was calculated. The data
`were obtained from a previous WSC race.
`A comparison of the available energy from the batter-
`ies, or the actual energy used to propel the vehicle, can
`now be made between Li-ion and silver–zinc batteries
`during race conditions. This is achieved by calculating the
`total energy from the solar array and multiplying this by
`the battery efficiency.
`Under perfect meteorological conditions, about 2.2 kW
`h could be stored in the batteries from the solar array of
`the Fuji Xerox Desert Rose during the combined evening
`and morning charges. If it is assumed that, during racing,
`all the energy from the sun is used by the vehicle, then the
`total extra energy available from the morning and after-
`noon charging periods, as well as before the start of the
`race, can be calculated.
`The available energy at the beginning of the race is
`given in Table 2, together with the calculated available
`energy from the morning and afternoon charging periods.
`
`Table 2
`Total energy available for the race
`
`Li-ion Silver–zinc
`
`.
`. (cid:14)
`(cid:14)
`6042 6048
`Initial capacity measured W h
`Morningrafternoon energy available W h — 4 days 8800 8800
`.
`(cid:14)
`(cid:14)
`.
`Energy, or charge–discharge, efficiency %
`96
`81
`.
`(cid:14)
`Actual energy stored in batteries during race W h
`8448 7128
`(cid:14)
`.
`Total race capacity W h
`14490 13176
`
`This shows the real advantage of Li-ion battery packs over
`a silver–zinc counterpart. Even though the Li-ion battery
`pack weighs slightly more because of the multiplicity of
`interconnects and protection circuits, the increase in effi-
`ciency provides a much greater available energy. This
`equates to a speed increase of approximately 1 kmrh over
`the whole race for an assumed average speed of 75 kmrh.
`
`7. Construction of Li-ion battery pack
`
`The real disadvantage of present Li-ion batteries for use
`in electric vehicles is their small physical cell size and,
`hence, small stored capacity. As already mentioned, 900
`cells were required for the battery pack of the Fuji Xerox
`Desert Rose solar car.
`Commercial units of 10 cells in parallel s1 string ,
`.
`(cid:14)
`which were connected via spot-welded tabbing material
`with individual protection boards for each cell, were pur-
`chased.1 Six of these strings were then soldered together
`and joined via nickel-plated, 1-mm thick, copper busbars
`to make one module. These busbars were tapered accord-
`ing to the magnitude of the current flowing through the
`copper. Nominally, the Fuji Xerox Desert Rose consumes
`1 kW of power throughout the day when running at 78
`kmrh. This corresponds to a nominal current of 18 A from
`the batteries. Thus, copper busbars and all cabling within
`the vehicle were sized according to a current density of
`1.25 Armm2 or, using the aforementioned current density,
`a cross-sectional area of 15 mm2. This low current density
`w x
`results from a power-to-weight
`trade-off 3 which,
`in
`brief, is a calculation of the power required to carry each
`kg within the vehicle.
`Thermal considerations were also a concern in design-
`ing the battery pack. A gap of 1.5 mm was left between
`each string of 10 cells to allow sufficient air flow, and
`adequate ventilation holes were provided in the bottom of
`the battery box. It was subsequently found that, even under
`the severest load, the batteries did not heat up to any
`significant extent.
`
`Fig. 6. Li-ion discharge characteristics at 20 A.
`
`1 Contact Johnny Grzan at angmo72@hotmail.com
`
`4
`
`

`

`160
`
`B. Kennedy et al.r Journal of Power Sources 90 2000 156–162
`)
`(
`
`8. Charge–discharge characteristics
`
`Before the race, each 3.7-V battery module was charged
`separately. A constant-current throughout the linear region
`followed by constant voltage was found to obtain the
`greatest stored energy. The linear region is shown in Fig. 5
`and is that region where the voltage is approximately
`linear for a constant-current charge rate.
`The method used to charge the batteries is shown in
`Fig. 7. A Cr3 rate is used to charge the batteries until 4.2
`Vrcell is reached. This is followed by progressively de-
`creasing the current in order to keep a relatively constant
`voltage at a maximum of 4.32 Vrcell. This voltage is
`stabilized overnight to an average of 4.29 Vrcell. The
`lower value is due to the fact
`that, under charge,
`the
`voltage is slightly higher because of the battery’s internal
`resistance.
`The variation of voltage with SoC, using the charging
`technique shown in Fig. 7, is presented in Fig. 5. The
`battery module could be charged to a peak SoC of 106 A
`h, or 1.76 A hrcell.
`Of equal importance, particularly to solar car teams, is
`the discharge performance of the batteries. A constant load
`of approximately 20 A from the module is shown in Fig. 6.
`This equates to driving the solar car at just above 78 kmrh
`without sun. Of most importance to solar car teams is the
`‘knee’ point of the discharge curve, i.e., the point at which
`the voltage starts to fall rapidly. The car will still travel
`after this point, but the speed will be reduced and hill
`climbing will prove very difficult.
`
`9. Protection of Li-ion batteries
`
`Safety concerns regarding Li-ion batteries have led to
`(cid:14)
`.
`the development of specialized integrated circuits ICs to
`protect against damaging the battery and, hence, causing
`battery failure. Such an IC, fitted with two small MOS-
`FETs, sense resistors, capacitors, fuses, etc. and a small
`printed-circuit board, was used in the Fuji Xerox Desert
`Rose. A schematic of the protection circuit is presented in
`w x
`Fig. 8 4 . In normal operation, the MOSFET switches,
`FET1 and FET2, are both closed. The power consumed by
`
`Fig. 8. Schematic of Li-ion protection circuit.
`
`all ICs and all MOSFETs for the complete pack equates to
`(cid:14)
`.
`6.6 W under 1 kW discharge .
`The printed-circuit boards are designed to protect small
`numbers of cells. Problems arise both in parallelling large
`numbers of cells and connecting these parallel strings in
`series to form a high-voltage pack. In particular, this type
`of protection is not designed to interrupt high currents as
`the semiconductor switches have a low current rating.
`A problem could result when parallel cells are discon-
`nected one-by-one from the module. Under a high load, up
`to 100 A will be discharged from each module of 60 cells
`in parallel, i.e., each cell contributes 1.67 A. If a cell is
`disconnected, the remaining cells have to provide extra
`current. At a low SoC, this could result in more cells being
`disconnected so that, eventually, one cell tries to provide
`all the 100 A. This would result in the destruction of the
`MOSFET and the associated printed-circuit board.
`The protection board offers three different forms of
`(cid:14) .
`protection namely:
`i overvoltage with a hysteresis band;
`(cid:14) .
`(cid:14)
`.
`ii undervoltage with a hysteresis band;
`iii an overcur-
`rent trip. The hysteresis band for overvoltage and under-
`w x
`voltage is shown in Fig. 9 4 .
`
`Fig. 7. Charging regime per battery module.
`
`Fig. 9. Mode of operation of protection board.
`
`5
`
`

`

`B. Kennedy et al.r Journal of Power Sources 90 2000 156–162
`)
`(
`
`161
`
`and V
`The hysteresis band is the region between V
`CD
`CU
`during charge, and between V
`and V
`during dis-
`DD
`DU
`charge. Considering the charging hysteresis band, the volt-
`age V
`is the maximum voltage to which the cell can be
`CU
`charged before disconnection from the circuit. This cell is
`only reconnected once V minus the hysteresis band, or
`CU
`is reached. A similar analogy applies to the discharge
`V
`CD
`hysteresis band.
`(cid:14) .
`Protection mode i prevents overcharging of the cell. If
`the cell voltage exceeds V , the charge switch FET2 opens
`cu
`and disables charging to the cell. The switch will close
`once the voltage falls below V . If this mode of protection
`cd
`is activated in a cell and a normal load is applied, the cell
`will start to discharge when the module bus voltage reaches
`3.55 to 3.65 V. This is equal to the cell voltage minus the
`(cid:14)
`.
`drop ;0.7 V across the body diode of the charge switch
`FET2. This load will ultimately bring the cell voltage
`below V and the switch will again close to equalize the
`cd
`cell and bus voltages. The problem this presents is that if,
`in a pack of 60 cells in parallel, overcharging occurs and
`the cells turn off, the various protection MOSFETs will
`turn back on at different times due to production differ-
`ences. At typical loads, this could result in a small number
`of the cells providing power until they reach 20% SoC.
`The remaining cells of the module will then turn back on
`and produce an inefficient discharge regime.
`(cid:14)
`The following methods were used both in the labora-
`.
`tory and in the vehicle to reset the mode of overvoltage
`shutdown.
`.
`(cid:14)
`fl A small resistor 1.5 V was connected momentarily
`across each individual cell. This reduced the cell voltage to
`below V
`such that the protection board IC reset the
`CD
`charge switch FET2 to on. This cannot be done within the
`vehicle, as access to each individual cell is not available.
`fl Within the solar car, resetting can be achieved by
`putting a large capacitor across the battery pack, i.e., a
`discharged motor controller capacitor. A discharged capac-
`itor is effectively a short circuit for a brief period of time.
`Therefore, as charging begins, this capacitor will cause a
`high inrush current from the battery pack that results in the
`cell voltage falling below V
`briefly, due to the internal
`CD
`resistance of the cell. This, in turn, results in resetting of
`the overvoltage mode of each of the protection boards.
`The second method is convenient for resetting the over-
`voltage mode, as it will preferentially shutdown those
`modules in which not all cells are connected. For a given
`high current load, in a single module, as one cell trips so
`the current required from the other cells increases. This
`results in a ‘domino’ effect, which means that the protec-
`tion on the last cell of the module will interrupt a very
`high current. A side effect is that the cell voltage can also
`drop below V . This results in the undervoltage lock-out
`DD
`mode, which can be reset as described below.
`(cid:14) .
`Protection mode ii prevents excessive discharge from
`(cid:14)
`.
`i.e., low battery voltage . If the cell voltage drops
`the cell
`below V , the discharge switch FET1 will turn off and
`DD
`
`Fig. 10. Schematic of bypass diode.
`
`prevent further discharge. Two conditions must be fulfilled
`to reset this mode of operation. First, the cell voltage must
`be above V
`and, second the module bus voltage, con-
`DU
`(cid:14)
`.
`nected to a charger or solar array , must be greater than
`2.0 V and supplying a small current.
`(cid:14)
`.
`Protection mode iii prevents excessive discharge cur-
`(cid:14)
`.
`rent overcurrent from the cell. Once again, the discharge
`switch FET1 turns off, but this time it does not latch. To
`reset, the load resistance must be increased above 50 M V.
`This mode was activated several times within the solar car
`as a pre-charge circuit for the motor controller did not
`initially exist. Interestingly, initial tests did not find this to
`be a problem as the solar array was continually plugged in
`(cid:14)
`.
`i.e., constant reset . This mode was only found during
`trialing without the solar array.
`The Li-ion battery also has an internal seal as an
`ultimate protection device. The seal will blow if the inter-
`nal pressure goes too high.
`Implemented into the battery pack were bypass diodes,
`as shown in Fig. 10. These diodes allow current to con-
`tinue to flow if all of one module’s protection boards
`operate and the complete module goes, open-circuit.
`Therefore, if one module is on open-circuit, the vehicle
`can still be driven until
`time is available to reset
`the
`protection mode.
`The bypass diodes also protect the ICs of the protection
`boards. In the case of the protection board shutting down
`one module, the battery pack effectively reverse biases
`itself across the protection circuit, i.e., y55 V appears
`across the IC. Testing this in the laboratory resulted in the
`destruction of the IC. Thus, implementation of the bypass
`diodes protects the IC against the module reverse biasing.
`
`10. Testing of protection boards
`
`A procedure was implemented to test every protection
`board on every cell individually. First under no load, the
`
`6
`
`

`

`162
`
`B. Kennedy et al.r Journal of Power Sources 90 2000 156–162
`)
`(
`
`voltage of each cell was compared with the module bus
`voltage. These voltages should be equal, and for those that
`were found not to be, the protection boards were replaced.
`Second, a Cr5 discharge test was performed and the
`voltage measured. This was to confirm that the charge
`switch FET2 was on and conduction was not through the
`body diode of the switch. A 30-mV drop should be seen
`from the module bus voltage to the cell voltage. A Cr5
`charge test was then implemented to determine that switch
`FET 1 was turned on and not conducting through its body
`(cid:14)
`.
`diode. Lastly, a full discharge until cut-off and a full
`(cid:14)
`.
`charge until cut-off of each module were conducted and
`the voltage monitored after each cut-off. This series of
`tests will reveal any faulty protection boards. Between 15
`and 20 of the boards were found faulty out of the 900
`received. In addition, two suspect Li-ion cells were identi-
`fied.
`
`11. Conclusions
`
`Li-ion batteries for use in electric vehicles are a very
`promising technology to provide longer range, a high
`lifetime, and a very high efficiency. The small size avail-
`able and safety concerns demand that time and care must
`be exercised in order to make a reliable and safe battery
`pack. If this is done, the batteries should perform reliably
`and pose no safety threat. For solar-car racing,
`these
`
`batteries will become the batteries of choice for medium-
`to-large budget teams.
`
`Acknowledgements
`
`The Fuji Xerox Desert Rose Solar Car Team would like
`to thank all their sponsors, namely: Fuji Xerox — The
`Document Company, Northern Territory University,
`Holden, Integrated Technical Services, CEANET Techni-
`cal Computing Solutions, The Math Works, Gulf Trans-
`port, Michelin Tyres, Perkins Shipping, N.T. Power and
`Water Authority and the NTU Sport Association. The
`authors are also grateful to Dr. D.A.J. Rand for his support
`and ongoing discussions.
`
`References
`w x1 D.M. Roche, A.E.T. Schinckel, J.W.V. Storey, C.P. Humphris, M.R.
`Guelden, in: Speed of Light — The 1996 World Solar Challenge,
`Photovoltaics Special Research Centre, University of New South
`Wales, Sydney, Australia, 1997, pp. 7–9.
`w x2 NEC Data Sheet — Lithium-Ion Rechargeable Battery —
`ICR18650E, NEC Moli Energy.
`w x3 Patterson, D.J., An efficiency optimized controller for a brushless d.c.
`machine, and loss measurement using a simple calorimetric tech-
`nique, Record of the 26th Annual IEEE Power Electronics Specialist
`Conference, PESC’95, Atlanta, pp. 22–27.
`w x4 Seiko Battery Protection IC Data Sheet — S-8491 Series, Seiko
`Instruments.
`
`7
`
`