Notes On BMS Project
Notes On BMS Project
Yoshikazu FUJITA
Yasuyuki HIROSE
Yusuke KATO
Takahiro WATANABE
Abstract
Due to their high efficiency and high energy density, lithium-ion batteries have been adopted for mobile
electronic devices and electric vehicles. They have been increasingly used further for various applications,
such as small mobility vehicles (electric motorcycles, golf carts, etc.), stationary batteries for HEMS (Home
Energy Management System), trucks/buses and industrial machinery. However, they have risks of fire
hazard and electric shock if being used incorrectly. In order to use the highly efficient lithium-ion batteries
safely and effectively, a battery management system (BMS) is needed. Among the BMS, technologies of the
battery capacity estimation and the malfunction detection are important.
FUJITSU TEN has developed a universal BMS PF (platform) that can be used for a variety of applica-
tions. This paper elaborates the development concept, the safety design technology and the highly-accurate
battery capacity estimation technology of the universal BMS PF.
1 Introduction
Among secondary batteries (chargeable bat-
teries), lithium-ion batteries, with higher energy
density, have been reduced in size and cost with
more capacity due to higher environmental
awareness and technological innovation. The
market of the batteries is expected to grow about
20% annually from 2015 to 2025. Having the char-
acteristics of no memory effect and small self-
discharge as well as being compact and light-
weight, they have been used as a power supply
for mobile devices, such as cell phones and laptop
computers, since 1990s. Their use has been Fig. 1 Block Diagram of Electric Vehicle
expanding since 2009 when they were mounted
on electric vehicles and hybrid vehicles. The BMS estimates the state of charge (SOC)
In the future, the lithium-ion batteries will be of the lithium-ion battery. The EV ECU controls
increasingly employed in more various fields, for a power supply amount (discharged amount from
example, the power supply) and a regeneration amount
・stationary batteries for HEMS (Home Energy (charging amount to the power supply) to/from
Management System) for storing electricity the motor based on the estimated value and bat-
generated at lower night rate or electricity tery temperature information. Moreover, the
generated by solar power charging control ECU controls a charging
・power supplies for industrial machinery, such amount. Those collaborative controls can prevent
as construction machines and devices overcharge of the lithium-ion battery so that the
・small mobility vehicles, such as electric motor- battery can be charged and discharged safely.
cycles and electric carts, and This time, FUJITSU TEN has developed a
・commercial vehicles, such as large buses and universal BMS platform (PF) with the aim that it
trucks. will be used in wider fields having a possibility of
However, those lithium-ion batteries have a using lithium-ion batteries, besides the automo-
smoke/fire hazard and electric shock if they are bile industry for which we have developed BMS
handled incorrectly. Therefore, a battery man- before. This paper elaborates the development
agement system (BMS) is important to secure concept of the BMS, a safety design technology
safety. and a highly-accurate SOC estimation technology
The BMS plays the roles of: used for it.
・detecting malfunctions, such as overcharge,
excessive rise of temperature and electric leak;
2 Development Concept
and
・estimating state of charge at temperatures and 2.1 Change in Development Process
in the charging/discharging environment Our conventional development process has
Thanks to those functions of the BMS, the taken a few years to develop a vehicle-mounted
lithium-ion batteries having large energy density product for each vehicle in line with the develop-
can be used safely. ment schedule of the vehicle, which means that
Fig. 1 illustrates a block diagram of the func- we have tailored products based on customer
tions of an electric vehicle. requirements (Fig. 2).
software and relay sequence varying depending battery cell capacities by eliminating differences
on customer specifications can be easily changed in capacity among the battery cells, as shown in
only by replacing software parts. Fig. 8. If there are differences in voltages among
When the number of the cells is changed, a battery cells in a cell pack, those cells cannot be
position of a service plug (SP) in a battery cell charged or discharged to their limits. Thus, their
pack also needs to be changed (refer to Fig. 5). practical capacities decrease. The cell balancing
The service plug is an insulation plug for safety prevents this. In the case of a BMS having cells
to protect service person from electric shock dur- connected (packed) in series, this function is nec-
ing maintenance work of the vehicle. When the essary to maximize performance of those cells.
service plug is detached, the current pathway is
cut off so that safety can be ensured.
Normally, the SP is provided near the center
of the battery cells connected in series. For the
BMS circuit board, a pulse transformer should be
provided for insulation of the circuit in accor-
dance with the position of the SP. Moreover, com-
munication from a microcomputer on the board
to each cell monitoring IC is established via the
lowest IC (the IC1 in case of Fig. 5). Therefore,
when the circuit board is adjusted to a cell pack
having a small number of cells, ICs are removed
in a decreasing order of hierarchy. In order to
materialize a board design flexibly adjustable to
a cell pack having the SP as well as fewer cells,
the circuit board can be designed so as to provide
the pulse transformer between any two ICs.
However, that will make the board larger and the
cost will exceed the target.
Fig. 8 Cell Balancing
Meanwhile, in the structure of the universal
BMS, a location of the IC for the communication The "passive cell balancing" is employed for
can be changed between the low voltage side and our universal BMS PF. In the passive cell balanc-
the high voltage side of the board, as shown in ing, a discharging switch (SW) forcibly dis-
Fig. 5, so that the position of the pulse trans- charges battery cells having higher voltages. In a
former for the circuit insulation can be fixed at conventional structure, as shown in Fig. 9, one
one position. Accordingly, the structure is adapt- same pathway is used both for voltage monitor-
able to the cell pack having fewer cells and a SP, ing and cell balancing current so that it is difficult
as shown in Fig. 7, because the IC can be to monitor an accurate voltage during the cell
changed. Thus, the structure helps to realize a balancing because a voltage drops due to resis-
smaller-sized circuit board and a reduction in the tance and current of the pathway. Therefore,
board cost. there is a problem of low accuracy of battery cell
voltages measured during charging/discharging
2.4 Satisfaction of Required Functions and travelling of the vehicle.
(Cell Balancing)
One of important functions of the BMS is cell
balancing. It is a function of optimally using the
developed the universal BMS PF based on the redundant design, the main function, the 1st SM
design specifications that comply with the level and the 2nd SM (Fig. 13).
D of the ASIL.
This section explains the safety design dis-
tinctive to the universal BMS PF.
routes. Thus, dangerous malfunctions that may and the "memory protection unit (MPU)" function
lead to overcharge of the cells can be detected at that separates memory regions used for the MF
99% or higher, the malfunction detection rate and the 1st SM from each other. In the case
required for the ASIL D. where the calculation results are determined not
Both of the cell monitoring IC and the micro- to be the same by the dual core lock step func-
computer are configured to include the MF and tion, there is a possibility that the microcomputer
the 1st SM in the same devices. Therefore, mea- is in a state in which it cannot turn off the relay.
sures are taken against malfunctions common to Therefore, it is designed to send a notification of
those safety functions. Some of those measures occurrence of the malfunction to the power sup-
included in the design will be explained below. ply IC. Once receiving the notification, the power
(1) Measurements against malfunction common supply IC can turn off the relay (Fig. 16).
to cell monitoring ICs
Fig. 15 illustrates the internal structure of the
cell monitoring IC. We adopted an IC including
the voltage measuring route and the voltage
measurement monitoring route. Each of those
routes has a power supply, a MUX, an ADC, a
data processor and a clock, and they are structur-
ally separated from each other. In addition, the
data communication line connected to the micro-
computer is shared by the two routes so that
problem detection is performed for the communi- Fig. 16 Turning-off of the Relay by Power Supply
IC in Case of Microcomputer’s Malfunction
cation.
3.2.3 Design of Second Safety Mechanism
In order to detect a malfunction of the 1st SM,
various diagnoses are performed as the 2nd SM.
The diagnoses performed by the circuit blocks
will be explained below.
(1) Diagnosis of cell monitoring IC (2nd SM)
It compares the cell voltage values individu-
ally measured by the voltage measuring route
and by the voltage measurement monitoring
route, and then diagnoses whether the measured
values are correct.
(2) Diagnosis of microcomputer and power supply
IC (2nd SM)
Fig. 15 Internal Structure of Cell Monitoring IC The diagnosis called the built-in-self-test
(BIST) is performed. The BIST is a diagnosis
(2) Measurement against malfunction common
method including test pattern generation
to microcomputers
through test result determination performed
We adopted a microcomputer including: the
within the chip itself. It can diagnose logics,
"dual core lock step" function that causes the pro-
memories and analogue circuits.
cessing core and the monitoring core to perform
(3) Diagnosis of relay control circuit and relay (2nd
the same calculation and that confirms whether
SM)
results from both the calculations are the same;
It diagnoses the relay control circuit and the
relays to find a malfunction, by turning on and off redundancy. Thus, this design can perform a
the relays based on a predetermined pattern. self-diagnosis by comparing the input logic from
the circuits (Fig. 18).
3.3 Safety Design against Electric Shock
Circuits for measuring an insulation resis-
tance value and a voltage value at the charging
inlet are provided to the universal BMS PF to
protect a user from electric shock. The circuits
will be explained below.
(1) Insulation resistance detection circuit
A ground of the vehicle is connected to a
ground of the battery cell by coupling connec-
tion. An insulation resistance value is converted
to a voltage value by applying a pulse voltage
and then the converted voltage value is mea-
sured. A capacitor is used for the coupling con-
nection. A short circuit of the capacitor directly
leads to a decrease in the insulation resistance Fig. 18 Circuit for Voltage Measurement at Charging Inlet
value. Thus, three capacitors are connected in
series as safety design. Moreover, this circuit has
4 Development of Battery Status
Estimation Algorithm
a self-diagnosis function of inputting test signals
to confirm whether the circuit is normal (Fig. 17). 4.1 Necessity for Accurate Estimation of
Battery Status
Lithium-ion batteries have a region that is
prohibited from being used because using the
region carries a fire hazard and other risks.
Therefore, a margin is set to the lithium-ion bat-
teries to prevent the region from being used (Fig.
19).
A size of the margin depends on accuracy in
estimating the state of charge(SOC)of the bat-
tery. Therefore, an accurate estimation for the
SOC makes the margin smaller and makes the
usable capacity of the cell larger, which leads to
better fuel efficiency of the vehicle and longer
Fig. 17 Insulation Resistance Detection Circuit
cruising distance. Therefore, accurate SOC esti-
(2) Voltage measuring circuit at charging inlet mation is very important to maximize the perfor-
When there is a possibility that the user mance of the cell.
touches the charging inlet, the voltage measuring
circuit measures whether voltage comes out of
the charging inlet. The circuit is configured with
an input circuit connected to the positive side of
the charging inlet and an input circuit connected
to the negative side of the charging inlet. Two
sets of those two input circuits are provided for
Generally, an estimation error of the SOC is The battery equivalent circuit model, having
regarded as 5% to 10%. We worked with resistances and capacitors as structure elements,
FUJITSU Limited to develop an SOC estimation is used to estimate the state of the battery (Fig.
algorithm with an estimation error target of 3% 20). The circuit model is derived based on inputs
or less. (current and temperature) and an output
(voltage) to/from the battery.
4.2 Battery Status Estimation Algorithm The mathematical expression of the model is
For estimating the SOC, it is important to shown below.
measure an open-circuit voltage (OCV) of the bat-
tery because the OCV is closely related to the
SOC. However, since the OCV means a terminal
voltage measured when the battery is completely
disconnected from a load and is in a stable state,
it is difficult to measure the OCV while current is V1, R1 and C1 represent the voltage, the resis-
flowing. Therefore, in order to overcome this dif- tance and the capacity of the electrode surface,
ficulty, we used the characteristics that a voltage respectively, and V2, R2 and C2 represent the volt-
at the battery terminal is equal to a sum of the age, the resistance and the capacity of the active
OCV and an overvoltage caused by an internal material, respectively. R0 represents the electro-
resistance of the battery, while the current is lyte resistance, and I represents the input cur-
flowing. rent. SCA represents the full capacity. These
In our algorithm, once the internal resistance parameters need to be defined whenever a type
of the battery is calculated based on a battery of the battery is changed. R1, R2, C1 and C2 are
equivalent circuit model, a terminal voltage is expressed by functions using R0 and the input
consecutively estimated. After the estimated ter- current so that current dependency of the inter-
minal voltage is compared with an actually- nal resistance can be considered.
measured voltage, the value is corrected by The OCV of the battery is expressed in a
using Kalman filtering, a statistic method, so as to mathematical equation based on curve fitting
reduce the estimation error, and then the SOC is technique, using measured values of SOC and
estimated. OCV, and is adjusted based on the temperature
and the reduced capacity of the battery. Then,
the dependency is reflected. Moreover, an OCV
for a charging side (VOCV_CC) is defined differ-
ently from an OCV for a discharging side
(VOCV_DC), and they are switched based on a estimated based on the flow from ① to ③.
direction of the current input to the battery. By ①Calculate a rise in temperature (heat value)
using this method, the SOC dependency of the from the internal resistance and the input cur-
OCV can be accurately expressed as a function rent.
so that the estimation error of the SOC can be ②Calculate change in temperature (increase and
reduced. decrease of heat) from a difference between the
estimated internal temperature and the surface
4.2.2 Structure of Battery Capacity Estimation temperature.
Algorithm ③Estimate the internal temperature from ①
and ②.
Temperature change = ① + ②
Estimated internal temperature of battery=pre-
viously estimated internal temperature of bat-
tery + temperature change.
(4) SOC estimation
The estimated battery terminal voltage Vp is
calculated from the equation of the battery
equivalent circuit model.
Fig. 21 Configuration of Battery Capacity The estimated value is compared with the
Estimation Algorithm
actually-measured terminal voltage V. Then, Kal-
In addition to the SOC, a reduced amount of man gain G, a weight coefficient for the estima-
the capacity of the battery also can be estimated tion error, is calculated from the Kalman filtering,
by our battery estimation algorithm configured and then the estimated value is corrected. This
as shown in Fig. 21. The next section will explain process is repeated at every measurement of a
an outline of each step of the process. value input to the battery.
(1) Calculation of internal resistance
The internal resistance R0 in the battery
equivalent circuit model is calculated based on
the equation below using a cell voltage and an
input current at the time. (5) Reduced capacity estimation
R0 = ΔV(cell) / ΔI(input) The reduced capacity is defined by a sum of
Moreover, R0 is calculated based on the natural degradation that has a greater influence at a
actually-measured voltage and the input current. high temperature and current degradation that has
Therefore, this method allows the internal resis- a greater influence during charging/discharging
tance to be derived, considering the temperature at a low temperature.
and the reduced capacity of the battery. The natural degradation is a function of tem-
(2) Calculation of charge amount by charging perature, SOC and time, and the current degra-
and discharging dation is a function of temperature and moved
An amount of charge changed at each time of charge amount.
charging and discharging is calculated based on
the input current. 4.3 Estimation Accuracy Evaluation
(3) Estimation of internal temperature of battery We evaluated accuracy of the SOC estimated
The internal temperature of the battery is by our battery estimation algorithm based on a
Fig. 23 Transition of Estimated Value and True Value of Reduced Capacity (Degradation to Capacity)
Profiles of Writers
Takahiro WATANABE
AE Engineering Group
Control System Engineering
Department