Design of the Battery Management System of
LiFePO
4
Batteries for Electric Off-Road Vehicles
F. Baronti, G. Fantechi, R. Roncella, R. Saletti
Dip. Ingegneria dell’Informazione, University of Pisa
Via Caruso 16, 56122 Pisa, Italy
G. Pede, F. Vellucci
ENEA - Centro Ricerche Casaccia
Via Anguillarese 301, 00123 S. Maria di Galeria (RM), Italy
Abstract—This paper describes the design of a modular battery
management system for electric off-road vehicles, where lithium-
ion batteries are expected to be widely used. A massive
electrification of off-road vehicles can be enabled by the
availability of a standard battery module, provided with an
effective management unit. The design and some preliminary
experimental results of the module management unit are
discussed in this paper. The unit contains a high current active
equalizer that enables the dynamic charge equalization among
cells and maximizes the usable capacity of the battery.
Keywords—Battery Management System, Active balancing,
Electric Off-Road Vehicles, Lithium-ion battery technology
I. INTRODUCTION
Off-road vehicles used in many commercial and industrial
activities are a large portion of means of transportation. These
vehicles include mobile work machines for various
applications, such as building sites, earth moving, street
cleaning, as well as agriculture, horticulture, greenhouse, and
gardening. An internal combustion engine commonly powers
off-road vehicles. The replacement or the combination of this
engine with an electric motor might be a remarkable step
toward energy sustainability by reducing CO
2
emissions and by
improving energy utilization efficiency [1], [2].
A recent study on the Italian market for mobile work
machines gives a good idea of the relevance of these kinds of
vehicles and the impact of their electrification [2]. Figure 1
shows the forecasted sales of electric off-road vehicles,
grouped by function and expressed in terms of onboard battery
energy. These numbers have been obtained by firstly
estimating the sales volumes of the various kinds of mobile
work machines in 2020. Then, the appropriate sizing of the
battery, in terms of stored energy and power, has been
evaluated for each category of work machine. Finally, the
penetration of the electric models has been estimated as 10 %
of the market. To gather the relevance of the potential battery
market for electric mobile work machines, let us express it in
terms of equivalent electric cars. Assuming 25 kWh as the
energy capacity of an electric car battery, the estimated market
for electric mobile work machines in Italy by 2020 is
equivalent to 20 k electric cars. Considering 2 M cars as the
overall Italian car market in 2020 and that optimistically 4 % of
the new cars is electric, we end up with the surprising
conclusion that the battery market for electric mobile work
machines will be 25 % of the batteries used in electric cars.
The above sketched scenario is very promising for the
electrification of off-road vehicles. The success of this
transformation is deeply connected to the availability of an
energy storage system that addresses the specific constraints of
each off-road vehicle category, which has its own requirements
in terms of battery voltage and capacity. It is thus important to
identify a standard battery module, which can be used to build
up the battery for each type of machine by connecting in series
or parallel the required number of standard modules. This
standardization effort is very significant, as it might help in
reducing the overall battery cost, which is the major limiting
factor for the electric transformation of these kinds of vehicles.
The standard battery module consists of 4 series-connected
lithium-ion cells (LiFePO
4
units), according to the study in [2].
Here, the focus is on the design of the relevant battery
management system (BMS). The BMS is a fundamental
component for an effective use of lithium-ion batteries.
Lithium-ion battery technology provides several advantages in
terms of higher energy and power densities, higher
charge/discharge efficiency, and longer lifetime as compared to
the more traditional valve-regulated lead-acid or nickel-metal-
hydride technologies. However, lithium-ion cells are very
fragile and sensitive to overcharge, deep discharge, and over
temperature. These conditions might shorten the battery
lifetime (accelerating capacity fading and internal resistance
degradation mechanisms due to battery aging), but even cause
destructive damages to the battery itself. Thus, the major
function of a BMS is to protect the battery against hazardous
Figure 1. Estimated sales volumes for electric mobile work machines,
expressed in terms of capacity of the onboard battery [2].
situations by monitoring the voltage and the temperature of the
cells that build up the battery. Extending battery life is
dramatically important because the battery cost represents a
large percentage of the overall cost of a typical electric mobile
work machine.
Furthermore, batteries consisting of series-connected cells
tend to become unbalanced over time (a condition in which
some cells store more energy than others). This is due to the
variability of the cell parameters (mainly the self-discharge
rate) and of the operating conditions (e.g. temperature gradient
in the battery). The net consequence is a reduction of the usable
battery capacity. In fact, when the least charged cell reaches the
discharge cut-off voltage, the battery current has to be stopped
to protect that cell, while there is still energy in the other cells.
Unfortunately, unbalancing in a lithium-ion battery cannot be
self-recovered as it can in more traditional technologies, which
can tolerate a controlled overcharge of the more charged cells.
On the contrary, recharging of a lithium-ion battery has to be
stopped when the most charged cell reaches the upper voltage
limit, causing the other cells not to be fully recharged. Thus, an
appropriate circuit is needed to restore the balanced condition.
The simplest technique consists of connecting a shunt resistor
in parallel to the more charged cells (passive balancing). This
way, the balanced condition is reached by dissipating the extra
energy stored in the more charged cells. Active techniques,
where charge equalization is obtained by transferring energy
from more charged cells to lower charged ones [3], have been
proposed to avoid wasting energy. Research focuses on
improving efficiency, reducing implementation complexity and
balancing time [3]–[7].
II. STANDARD BATTERY MODULE
The standard battery module consists of 4 series-connected
LiFePO
4
cells, as described in [2]. The module nominal voltage
is 12.8 V, whereas the discharge and charge cut-off voltages
are 10 V and 15.4 V, respectively. This allows the standard
module to be used in place of a conventional lead acid battery,
commonly found in any vehicle for starting, lighting, and
ignition [8]. A huge market with potential great benefits in
terms of cost reduction is expected. The proposed standard
module can be very effective in meeting the different battery
voltage specifications found in the electrification of the various
types of mobile work machines. For instance, a large portion of
these vehicles is powered by a 48 V battery, which can be
implemented by series-connecting 4 standard modules. The
standard module can host three different cell capacities: 30 Ah,
60 Ah, and 100 Ah, in order to address the different
requirements for the stored energy and battery runtime.
It is worth noting that the standard module contains a
relatively small number of cells. Thus, the cost overhead
related to module assembling and BMS, which do not scale
with the number of cells, is more significant than in a coarser-
grained partitioning of the battery pack. However, these costs
are by far counterbalanced by the availability of a standard
module fully compatible with a lead acid battery with the
benefits of the lithium-ion battery technology. LiFePO
4
cells
are more robust and cheaper than other lithium-ion cells, at the
expense of a lower energy density. On the other hand, electric
off-road vehicles have less stringent constraints on the battery
weight and volume compared to other electric vehicles.
Therefore, LiFePO
4
battery technology has been selected for
implementing the standard module. Figure 2 shows the
relationship between open circuit voltage (OCV) and state of
charge (SOC) measured on a 60 Ah LiFePO
4
(from Hipower),
showing the flatness of the characteristics and the hysteresis
effect between charge and discharge.
III. BMS DESIGN
A. BMS architecture
The effective partitioning and implementation of the BMS
functions is crucial when a large number of series-connected
cells are managed. We proposed to solve the problem by
considering the battery as a hierarchical platform consisting of
three layers: the elementary cell, the module (i.e. a subset of
adjacent series-connected cells, usually assembled in a
dedicated case) and the pack (a connection of modules) [9].
This perspective leads to a clear, general and easy to implement
partitioning of the BMS functions [10]. In fact, fundamental
monitoring tasks (cell voltage and temperature measurement),
as well as passive balancing, lie on the lowest layer of the
platform, i.e. the cell. Charge transfer between cells (to achieve
active equalization) and thermal management belong to the
intermediate layer, namely the module. Battery protection (by a
main switch or contactor) and more advanced functions, such
as state-of-charge (SOC) and state-of-health (SOH) estimation,
are mapped in the uppermost layer of the platform. SOC is
usually evaluated as the amount of charge stored in a cell as a
percentage of the actual capacity value. The latter might differ
from the nominal value due to variations in the manufacturing
process and cell ageing (capacity fading effect). SOH expresses
the ratio of the capacity value to the nominal one. The pack
layer is also provided with interfaces to communicate with
external systems, for example the vehicle management unit in
an electric vehicle application or a personal computer for data
acquisition and configuration of the BMS.
Given the above logic partitioning, which is independent of
the battery size and organization, the fundamental issue is how
to effectively map the logic layers of the platform into their
hardware counterparts. This step instead depends on the actual
composition of the battery and on the targeted application, as
the latter determines the requirements in terms of safety levels
and advanced function implementation. Although there are
attractive implementations of the logic cell layer into a
Figure 2. Open circuit voltage (OCV) versus State-of-charge (SOC) measured
on a 60 Ah LiFePO
4
cell (manufactured by Hipower).
dedicated electronic board applied to each single cell of the
battery [11], [12], these solutions are less effective when the
module consists of a small number of cells, as in our standard
module, which contains 4 cells. In addition, these solutions
may hardly be compatible with the module assembling
constraints that are determined by the application. Thus, the
cell layer merges with the module layer in our BMS. This
design choice is quite common and is confirmed by the
availability of off-the-shelf components designed to measure
the voltage of up to 12 cells, the temperature in one or two
spots and to carry out passive balancing. Figure 3 shows the
devised BMS architecture in the case of a battery consisting of
4 standard modules. The MMU (Module Management Unit)
implements the module logic layer functions, such as active
balancing and thermal management, but also the cell logic
layer functions, such as monitoring the voltage and temperature
of the 4 cells of the module.
The pack logic layer carries out functions such as SOC and
SOH estimation and data logging, which can be implemented
by software routines. Thus, there are various options for
mapping this logic layer into hardware. A possible choice is to
implement the high level BMS software functions as further
tasks of the firmware running on the main control unit of the
application, such as the vehicle management system. Although
this solution is attractive as it does not require any additional
piece of hardware, it is strongly dependent on the targeted
application, particularly on the free computational resources
available on the processor of the main control unit. As a
consequence, this approach is not appropriate to achieve our
goal of assembling the battery for a wide range of applications
by using standard battery modules. On the other hand, the use
of dedicated hardware (e.g. a centralized pack management
unit, PMU) to map the logic pack layer, as proposed in many
publications (e.g. [6], [13]–[15]), would be hardly affordable
for many mobile work machines, where the battery requires a
small number of standard modules.
We instead propose to distribute the pack logic layer
functions to the MMUs as much as possible. For instance,
advanced SOC and SOH algorithms based on a cell model
([16], [17]) can be executed at the module rather than at the
pack level. Distributing high level functions on different
processors instead of a single one (embedded in the PMU) also
reduces the communication requirements between MMU and
PMU and increases the system reliability. The remaining
centralized functions can be mapped onto a particular MMU,
which takes the role of master, whereas the other MMUs act as
slaves. It is worth noting that the centralized functions, such as
supervising and coordinating the behavior of the slave MMUs,
measuring the battery current and controlling the main battery
switch, are a very small portion of the overall functions of a
MMU. The corresponding overhead is minimal so that a
standard module can assume the role of master by only
changing some firmware configuration parameters that enable
the relevant centralized functions. In that case, the additional
peripherals (for instance the current sensor (I) and the pack
protection switch (PPS)) will be connected to specific I/O pins
on the MMU acting as master, as shown in Figure 3. As an
example that proves the power of this flexible design approach,
let us mention that a battery for starting, lighting, and igniting
an ordinary car can be obtained by just one standard module
configured as a master.
Finally, it is interesting to note how the battery protection
mechanism is distributed over all the MMUs. Each MMU
asserts an internal logical variable (Safe) when all the voltages
and temperatures of the 4 cells in the module are in their safe
operating ranges. This variable is ANDed with the input enable
signal produced by the preceding slave module in the chain to
generate the output enable signal, as indicated in Figure 3. The
output of the chain’s last unit (the master MMU) is finally
connected to the PPS, which breaks the battery current when an
unsafe condition occurs in whichever module. The master
MMU can also supervise the behavior of the slave MMUs and
transmit the battery current value to them (if required by the
SOC and SOH estimation algorithms implemented on each
MMU) by means of an isolated CAN (Controller Area
Network) bus. As most of the processing is performed at the
module level, the communication requirements on this bus are
low. Thus, the isolated CAN bus can also be used to connect
Figure 3. Architecture of the BMS sketched for a battery comprising of 4
standard modules. MMU is the module management unit and PPS is the pack
protection switch.
the intelligent battery to other electronic units or to a PC for
BMS configuration, diagnostics or data logging.
B. Module Management Unit design
As described above, the MMU is the hardware platform
that implements the three logic layers in which the BMS
functionalities are partitioned. In addition, the MMU hardware
is always the same independently of the module role in the
battery (master or slave), which will be defined by firmware
configuration. Figure 4 shows the schematic block diagram of
the MMU. The logic cell layer is implemented by a single chip
(i.e. LTC6803-3 from Linear Technology), which measures the
voltage (with 10 mV accuracy) and temperature of each
module cell. As the LTC6803-3 provides only two channels for
temperature measurements, the 4 NTC (Negative Temperature
Coefficient) resistors are connected by means of two 2-way
analog multiplexers controlled by the microcontroller. The
NTCs are glued to each cell case in the spot where the highest
temperature was measured by an infrared camera during high
current tests. The Monitor chip is connected to the
microcontroller via an SPI (Serial Peripheral Interface).
The microcontroller (i.e. LPC1754 from NXP, based on an
ARM Cortex M3 processor with a 32 b data bus) is indeed the
core of the MMU. It acquires the voltage and temperature of
the cells and (if the MMU is configured as master) the battery
current through a Hall Effect transducer (DHAB family from
LEM) with 1 % typical accuracy. This sensor family provides
two ratio-metric output channels with different measurement
ranges. The sensor model to be placed in the master module is
selected according to the capacity of the module cells (i.e.
30 Ah, 60 Ah, and 100 Ah) and the application specifications.
The outputs of the sensor are read by the ADC embedded in the
microcontroller. The sampling rate of the cell voltages and the
battery current is 10 Hz, whereas cell temperatures are acquired
at the lower rate of 1 Hz. The battery current value is
transmitted by the MMU master on the isolated CAN bus. The
microcontroller compares the voltage and temperature of the
module cells, as well as the battery current, to configurable
thresholds in order to carry out two of the major MMU tasks,
i.e. battery protection and thermal management. In fact, as soon
as one cell exceeds its safe operating range, the MMU enters
into an error state in which the local enable of the PPS is
disabled. This signal is ANDed with the Enable_in signal to
generate the MMU output Enable_out. The Fan Driver block
activates the module fan when one of the cell temperatures
exceed a given threshold.
The MMU is powered by the module cells so that its power
consumption is critical, as it might dramatically worsen the
self-discharge rate of the battery. Thus, the microcontroller
disables the power supply of certain blocks, when they are not
needed, and enter low-power operating modes to implement
the necessary power saving policy. Finally, the active equalizer
is an innovative part of this design as it is capable of equalizing
the energy stored in the 4 cells of the module, as well as to
cooperate with the active equalizer of the two adjacent MMUs
(through the Bal_bus_in and Bal_bus_out power buses) to
balance the charge among modules also. This circuit will be
described more in detail in the following sub-section.
C. Active balancing
The ultimate goal of an active balancing technique is to
maintain an even distribution of charge among series-
connected cells of a battery over time, with ideally no energy
losses for the equalization process. This makes it possible to
exploit all the energy stored in the battery (i.e. to maximize the
usable capacity of the battery). In contrast with passive
balancing, a non dissipative element, such as a capacitor or an
inductor, is used as tank to transfer energy between the cells of
the battery in every active balancing circuit. The circuits
presented in the literature so far can achieve very high energy
transfer, but usually provide very low balancing currents
(around a few hundreds of milliamperes), limiting the amount
of charge that can be transferred in a given time. Increasing the
balancing current requires the use of bulky passive components
(capacitors, inductors or transformers), which is not affordable
in many of the proposed techniques, and also negatively affects
the efficiency, as conduction losses increase.
In addition, these circuits aim at counterbalancing
differences in the cell self-discharge rates, which produce an
appreciable imbalance in the SOC of the battery cells in the
range of months (or even weeks). Thus, low balancing currents
can be used to periodically equalize the SOC of the cells. This
phase is usually carried out at the end of a battery recharge to
bring all the cells at 100 % SOC. However, this does not
necessarily mean that every cell stores the same amount of
charge or energy. This is due to mismatches in the capacity
values caused by variations in the manufacturing process or in
the operating conditions (particularly the temperature) that
significantly affects the speed of the capacity fading effect.
Then, even if all the cells are initially fully charged (100 %
SOC), the cell with the lowest capacity will reach the discharge
cut-off voltage before the others, causing the interruption of the
battery current while there is still energy in the battery. The net
consequence is that the usable capacity of the battery is
determined by the cell with the minimum capacity value.
Figure 4. Block diagram of the module management unit (MMU).
As stated above, active balancing should allow the battery
to deliver all the stored energy to the load. This way, the usable
capacity increases from the minimum value to the average one.
Such a goal can be achieved by dynamic charge equalization
that allows all the cells to reach the discharge cut-off voltage at
the same time. This means that the equalizer is used to transfer
charge from the cells with higher capacities to those with lower
capacity values during the battery discharging. Capacity values
can be calculated by accurate estimation algorithms [18], [19].
This attractive approach requires an equalizer circuit capable of
delivering high balancing currents. Equation (1) shows the
balancing current of the i-th cell (i.e. the charge current that
flows into the cell when the terminal battery current is zero) I
bali
needed to obtain a usable capacity equal to C
u
. I
pack
is the
average discharge battery current and C
i
is i-th cell capacity. If
the equalizer circuit has a unitary efficiency, the usable
capacity can reach the average capacity value.
(1)
One of the main goal of this work is to design an active
equalizer circuit that achieves the benefits of dynamic charge
equalization. This means that the circuit should deliver a
current high enough to perform the required charge
redistribution during the battery runtime. Usually, a mobile
work machine does not continuously operate between a full
charge and a full discharge. Dynamic charge equalization can
be applied during the pauses, where recharging might not be
available.
Let us now evaluate the performance of dynamic charge
equalization when it is applied to the medium-size standard
module (i.e. the one with 60 Ah cells), if a 1.5 A balancing
current were available. For this purpose, we consider a capacity
distribution of the module cells, where one cell capacity is
10 % lower than the nominal value, whereas the other 3 cells
have a capacity higher than the nominal one. The average
capacity is assumed to be equal to the nominal one. The usable
capacity of the module is reduced to 90 % of the nominal value
if dynamic charge equalization is not applied. Let us now apply
dynamic charge equalization with 1.5 A balancing current.
According to (1), all the energy in the module is utilized if the
average discharge current I
pack
is less than 15 A. This value is a
quarter of the battery C-rate (a quarter of the nominal cell
capacity in ampere), a reasonable value for a mobile work
machine. Indeed, it means that the battery has been sized to
operate 2 h with a 50 % duty cycle.
Figure 5 shows a schematic view of the active equalizer.
The key element to achieve high balancing currents is the very
low on-resistance switch matrix, which makes it possible to
selectively inject or draw current from each cell of the battery.
The isolated DC/DC converter can be designed to transfer
energy from the series of the 4 cells to the selected cell, from
the selected cell to the series of the 4 cells or in both directions
(bidirectional converter). This approach overcomes the
limitations in terms of maximum current, volume occupation,
and efficiency, which affect the balancing circuit based on
transformers with multiple windings or flying capacitors [4].
The design of the switch matrix is derived from that described
in [7]. In more detail, each cell terminals can be connected to
the inner balancing bus through a solid-state DPST (dual-pole
single-throw) switch (SW
1
-SW
4
). The additional switch SW
5
allows the module balancing to be connected to the balancing
bus of the adjacent module to form a closed loop (Bal_bus_out
of the last module is connected to Bal_bus_in of the first
module). In this way, energy can be transferred among cells
located in different modules (inter-module active balancing).
Each DPST switch consists of a couple of two series-
connected MOSFETs (to form a bidirectional switch). The
selected MOSFETs (SI7478DP from Vishay) are characterized
by a very low on-resistance (7.5 m!) and a breakdown voltage
of 60 V. The latter value sets the maximum number of standard
modules that can share the same balancing bus (which is 4, for
the selected MOSFETs). The two couples of parallel-connected
gates are driven by the two outputs of the isolated gate driver
HT0440 (from Supertex). The two inputs of the HT0440 are
driven by the same signal. The five input signals of the switch
matrix must be generated in such a way that only and only one
DPST switch is on at one time. The violation of this rule leads
to catastrophic consequences because it will short-circuit a
segment of the battery stack. Thus, the binary-coded selection
command of the switch matrix produced by the microcontroller
enters the safe logic driver (the design of which is derived from
that presented in [7]), which decodes the selection command,
and also assures an appropriate deadtime between two different
active configurations of the switch matrix.
An off-the-shelf isolated DC/DC converter simplifies the
design of the active equalizer. The selected component
(EC6A01 from Cincon) provides 7.5 W power on a nominal
5 V output, 9-18 V input range, and 400 V galvanic isolation.
As shown in Figure 5, the converter input is the module voltage
and the output is connected to the module balancing bus. An
Hall effect sensor is employed to measure the output current of
the converter. As the active equalizer is a crucial feature of our
design, we carried out a preliminary experiment to measure the
Figure 5. Schematic block diagram of the active charge equalizer.
![Figure 1. Estimated sales volumes for electric mobile work machines, expressed in terms of capacity of the onboard battery [2].](/figures/figure-1-estimated-sales-volumes-for-electric-mobile-work-19ute2tw.webp)
