You are on page 1of 7

AUTOMATIC BATTERY HEALTH MONITORING

USING MACHINE LEARNING FOR E-VEHICLES


Sarupriya S1, Abitha N2, Aishwarya Lakshmi R3, Merscia Serylin Jefy A4, Shrinidhi N A 5
Assistant Professor1, UG Students 2 ,3, 4, 5
Electronics and Communication Engineering1, 2, 3, 4, 5
Velammal Engineering College 1, 2, 3, 4, 5

Abstract- Batteries are combinations of solutions of extensive usage in low power portable
electrochemical cells that generate electricity to electronic devices, power sources for electric
power electrical devices. Batteries are vehicles and for future grid-scale energy storage.
continuously converting chemical energy to High energy density, high power density, safety,
electrical energy, and require appropriate large number of life-cycles are the main attributes
maintenance to provide maximum efficiency. of Li-ion, which make it superior to other batteries
Management systems having specialized such as lead-acid, Sodium Sulfur (NaS) and Nickel
monitoring features; such as charge controlling Metal hydride (NiMH). The electrical
mechanisms and temperature regulation are characteristics of Li- ion batteries, i.e., energy-
used to prevent health, safety, and property density, power-density, and C-rate, have upgraded
hazards that complement the use of batteries. significantly, albeit the problem remains with the
These systems utilize measures of merit to non-linear and nonuniform charging capacity and
regulate the battery performance . In this work, aging retention with a large number of cells in
we propose a continual learning method of battery packs .To overcome this inherent pitfall of
neural network to monitor the aforementioned Li-ion batteries, a battery management system is a
parameters using a data-driven approach. We prerequisite to protect the overall system and
use a machine learning algorithm that extracts monitor the optimal usage of energy Such battery
significant features from the discharge curves to management systems need to ensure and monitor
estimate these parameters. Extensive simulations the state of functions of batteries including the state
have been carried out to evaluate the of health (SOH) and state of charge (SOC), which
performance of the proposed method under in return, optimize the battery performance and life.
different currents and temperatures. As an example, in electrical vehicles, the SOH is
analogous to an odometer and SOC mimics the fuel
1. INTRODUCTION gauge used in fueled vehicles. If SOH and SOC are
estimated accurately, they can prevent
Recent decades, global climate has degraded, and overcharging, overheating and extend battery life,
greenhouse effect has become prominent due to the making them highly important parameters.
rampant increase in earth’s temperature by carbon- However, they cannot be measured directly by any
emitting fossil fueled vehicles and conventional electrical equipment, as a substitute, they are
power generation resources. This has globally estimated by variables of batteries like input load
driven and endorsed the research and development current (1A and 2A) and terminal voltage at
of energy storage systems. In the evolution of output.The growing understanding of global warm
several energy storage devices, Lithium-ion (Li-ion) around the world, the demand for clean fuel/energy
has emerged as one of the best and widely is on the rise and as a result there is a continuous
implemented shift towards the electric vehicles and hybrid
electric
vehicles. Battery performance is influenced by Noman Khan et al proposes MCCPs based battery
factors such as depth of discharge (DOD), management system (BMS) to estimate batteries
temperature and charging algorithm. This paper health/capacity through the deep learning (DL)
attempts to provide a measurement of voltage and concept where the patterns in these CPs are changed
current level of the battery using internet of things. as the battery ages with time and cycles. Thus, we
Lead-acid batteries are very effective at powering deeply investigate both machine learning (ML) and
many different applications. They are easy to DL based methods to provide a concrete
obtain, relatively inexpensive, and provide a lot of comparative analysis of our method. The adaptive
power to whatever they are hooked up to. boosting (AB) and support vector regression (SVR)
Regrettably, if there is nil monitoring the charge, are widely compared with long short-term memory
the battery will eventually run out of power. In (LSTM), multi-layer perceptron (MLP), bi-
addition to determine the charge of the battery, the directional LSTM (BiLSTM), and convolutional
current voltage of the battery is needed. By neural network (CNN) to attain the appropriate
depending on the output voltage of the battery, the approach for battery capacity and state of health
approximate charge of the battery can be estimated. (SOH) estimation.
Hu et al concerned with machine learning enabled
2. RELATED WORK battery State-of-Health (SOH) indication and
Few studies have shown the possibilities of an prognosis. The sample entropy of short voltage
automatic battery health monitoring using different sequence is used as an effective signature of
ideologies. capacity loss. Advanced sparse Bayesian predictive
modeling (SBPM) methodology is employed to
Yu, Jianbo et al developed an adaptive Gaussian
capture the underlying correspondence between the
mixture model (AGMM) for online learning
capacity loss and sample entropy.
dynamic changes of battery health in its full life. A
Bayesian-inference-based probability indication is Xing, et al discussed the latest known methods of
developed for detection of new health states that are state estimation and life prediction based on battery
online modeled by deleting and adding of health monitoring in this paper. Through comparing
components in AGMM. their characteristics respectively, a prognostics-
based fusion technique is proposed that combines
Banaei, et al r reports on a new method for
physics-of-failure (PoF) with data-driven
estimating the SOH of a Li-Ion battery using its
technology.
impulse response. Having the impulse response of a
healthy battery, the proposed method predicts the Li, Hong et al developed a multistep-ahead
terminal voltage of a battery using the terminal prediction model based on the mean entropy and
current measurement. relevance vector machine (RVM), and applied to
state of health (SOH) and remaining life prediction
Lee, et al implemented a system to diagnosis the
of the battery. A wavelet denoising approach is
fault of a battery cell through the Deep Neural
introduced into the RVM model to reduce the
Network state classifier . In this method , DNN state
uncertainty and to determine trend information. The
classifier utilized the discharge voltage data that
mean entropy based method is then used to select
was obtained by operating the lithium battery cell at
the optimal embedding dimension for correct time
high temperature .
series reconstruction.
Yu et al developed a battery health prognostics
Tang, et al , presented a health monitoring method
system based on Bayesian-inference probabilistic
for the application of battery management system
(BIP) indication and state-space model (SSM) that
(BMS) to solve this issue. First of all, the necessity
integrates logistic regression (LR) and particle
of multiple indicators (MIs) is demonstrated
filtering (PF). In this system, generative
through the analysis of estimation dispersion of
topographic mapping is constructed to model
single indicator (SI), and three health indicators are
distribution of multisensor data from healthy battery
extracted from actual operating conditions of EVs.
under an assumption that predictable fault patterns
M.Senthilkumar; et al proposed a battery
are not available.
management system can be integrated with the
monitoring structure which is capable of both
managing, monitoring and logging the data to an of GSM module. For coding part of system of
online database. This system monitors the battery system we have used arduino IDE and embedded c.
parameters like voltage, current, temperature, power first of all components and check if input voltage
and state of charge. These parameters are then sent and current are there are in defined limits then
and stored in a database via internet which is then check if battery is 100%charged or not charged
shown to the user by means of an android app. fully then it will enable battery charging circuit also
Chetan et al proposed a vibration based health when battery goes below 10% user will be notified.
monitoring. A regular data of the vibration for A buzzer based alerting system is also designed to
batteries used in vehicles has proved to be helpful in indicate the raise and fall in the values. This system
realizing the need of repair or replacement. To will analyse the exact state of the battery and alert
collect this data, a set of vibration sensors are used the user about it. Another part of this system has a
along with a set-up placed around a battery, a GSM Module to ensure that the batteries status is
microcontroller for data acquisition from the notified to the users properly. It is integrated with
sensors. the controller which collects the data from the
sensors. By using these values the controller sends
Miftahullatif et al developed a diagnostic the messages about the status of the battery to the
architecture of lithium-ion batteries that finds the user alongside the result displayed in the LCD
variation of State of Health (SOH) by actively Display and the buzzer.
sensing the voltage transient response at rest period
after discharge, that is where each battery cell
returns to an equilibrium state. We found that this
transient response represents SOH of the battery
cells. Since at this rest period after discharge the
cell balance controller is not activated, by sensing
the voltage variation we can also predict the SOH
variation inside the Battery Storage System (BSS).

3. PROPOSED SYSTEM

First system comprises a battery health monitoring


system which comprises Temperature sensor,
current transformer, a LCD Display, buzzer and a Fig. Proposed Block Diagram
GSM Module. Once the vehicle is started the
system gets started and the temperature, current and WORKING
voltage values are being continuously monitored.
It is mainly consists of three major parts collecting
This provides timely readings of the basic
data from all sensor sending the data and processing
parameters via a liquid crystal display connected to
it with microcontroller sending user feedback. For
it. The system mainly consists of 3 main parts:
monitoring input current and voltage we are using
collecting data from all sensors, sending that data
voltage sensor and current sensor. These sensor will
and processing it with a microcontroller sending
be monitoring incoming voltage and current to UPS
user feedback for monitoring input current and
battery and send the data to microcontroller. After
voltage we are using voltage sensor and current
collecting all these data user will be alerted
sensor ,these sensors will be monitoring incoming
whenever any predefined parameters will change by
voltage and current to UPS battery and sending that
using GSM technology. SMS will be sent user on
data to microcontroller arduino UNO will monitor
his/her phone with help of GSM module.With this
battery parameters like remaining percentage,
the user will be able to know the current status of
battery etc.this data will also be sent to
the battery. In case of sudden rise in temperature or
microcontroller. After collecting all these data user
the temperature levels going beyond the threshold
will be alerted whenever any predefined parameter
value say 40 degree Celsius the code is set in a way
will change input voltage, current and battery
such that the motor will rotate to indicate the
percentage etc.by using GSM technology. SMS will
change in the value.
be sent user on his/her phone with the help
4. COMPONENTS Piezoelectric Buzzer

DHT11 Sensor

Fig. Piezoelectric Buzzer


Fig.DHT11 Sensor The Piezoelectric Buzzer is used to indicate the
result in the form of sound or alarm. They are
This DHT11 Temperature and Humidity Sensor electrically run sounding which are basically run on
includes an aligned advanced flag output with the DC Voltage. The core of this buzzer is made up of a
temperature and mugginess sensor ability. It is piezo electric element. Due to this element the core
incorporated with an elite 8-bit microcontroller. Its of the buzzer constantly expands and shrinks when
innovation guarantees the high dependability and Dc Voltage passes through it. This continuous
magnificent long haul steadiness. This sensor is action produces vibrations which is responsible for
mainly responsible for the measuring of the the production of the sound. The use of Piezo
temperature of the batteries in real time. It has great electric buzzer in this project is to indicate and alert
quality, quick reaction, hostile to impedance the user about the changes in voltage, current and
capacity and high performance. temperature.

Liquid Crystal Display Regulator

Fig: Regulator

Fig. Liquid Crystal Display The simplest regulator is a large capacitor in


parallel with the load. The capacitor converts the
LCD’s are used to display the outputs and the pulsating DC voltage of a rectifier into a smooth Dc
desired results that are obtained from the sensors by load voltage. Two important parameters of a
the means of controllers. The LCD are extensively capacitor regulator are its working voltage and its
used in many telecommunications and capacitance. The working voltage must be at least
entertainment electronics. The LCD have even equal to no-load output voltage of power supply.
started replacing the cathode ray tubes (CRTs) used The capacitance determines the amount of ripple
for the display of text and graphics, and also in that appears on the Dc output when current is drawn
small TV applications. The main use of LCD from the circuit. The amount of ripple decreases
Display in our project is to show the realtime values with increase in capacitance.MC 7800 and 7900
of current, temperature, voltage and also to indicate Regulators standard application circuit. The Mc
the state of charge. 7800 is a 3 terminal, positive, fixed voltage
integrated circuit regulator
These regulators employ internal current limiting,
thermal shutdown and safe area compensation. The
Mc 7900 is a 3 terminal, negative, fixed voltage
integrated circuit regulator. Mc 7800 and Mc 7900
series requires no external components.

GSM Module

Fig. Hardware part consisting of sensors to monitor


health parameters of the battery and display in the Liquid
Fig. GSM module Crystal display
GSM modem or GSM module is a device that uses
GSM mobile telephone technology to provide a The hardware part is designed using sensors to
wireless data link to a network. GSM modems are measure the health of the batteries. These sensors
used in mobile telephones and other equipment that include DHT11 Temperature Sensor, regulators to
communicates with mobile telephone networks. measure the current and voltage of the battery. Live
They use SIMs to identify their device to the monitoring is done with the help of these sensors
network. The module offers GPRS/GSM and the readings are displayed by using a LCD
technology for communication with the uses of a Display. Here once the vehicle is started the
mobile sim. It uses a 900 and 1800MHz frequency temperature, current, voltage and the state of charge
band and allows users to receive/send mobile calls parameter readings are shown in the Liquid Crystal
and SMS. The keypad and display interface allows display.
the developers to make the customized application
with it. Furthermore, it also has modes, command
mode and data mode. In every country the
GPRS/GSM and different protocols/frequencies
operate. Command mode helps the developers to
change the default setting according to their
requirements.

The main use of GSM Module in this project is


to send Messages to the user about the changes in
the temperature if the battery is not in the optimized
temperature condition.

Fig. Hardware Part


5. RESULTS AND DISCUSSION:
Any abnormality in the readings will
trigger an alert message to the user’s phone. If there
is a rise in temperature, it will send a message as
“Temp High” to the users mobile. this will Also be
Displayed on the LCD Screen. The readings are
also recorded in the form of excel sheets. This
provides a
clear view of increase/decrease in battery’s Error percentage is calculated and the system is
condition. again trained to reduce the errors.. The minute
variations in readings are displayed in a graphical
manner. Optimum performance is achieved.

6. CONCLUSION

Battery health degradation detection and


monitoring are crucial to realize equipment’s near-
zero downtime and maximum productivity. A big
challenge is how to construct an effective
monitoring system that can consistently exemplify
the degradation propagation of battery health over
its whole life. This work proposes a continual
learning scheme for monitoring of state-of-health
(SOH) of batteries. An ANN is developed for online
learning dynamic changes of battery health in its
full life. This paper provides guidance for the
development of battery SOH monitoring models
based on adaptive learning schemes without too
much human intervention.

Fig. Message sent when High Temperature is measured The future work proposes a deep learning-
based optimal battery management scheme for
The changes that occur in the battery are also frequency regulation (FR) by integrating model
displayed in the LCd Display screen. In case of predictive control (MPC), supervised learning (SL),
voltage fluctuation the display will show a message reinforcement learning (RL), and high-fidelity
as “Low Voltage”. If the battery gets overpowered battery models. By taking advantage of deep neural
the system shows a message as the “ Overload”. networks (DNNs), the derived DNN-approximated
policy is computationally efficient in online
implementation.

REFERENCES
1. Preethika.J,Priya.P,Ramya.U.P,Surendhar.K
,Vijayakumari.G(2016)- “Anti-Kidnapping
System for Women with Security Alert”,
International Journal of Advanced Research
in Biology Engineering Science and
Technology (IJARBEST), Volume 2,
Special Issue10.

2. Hema
Fig. Message is displayed when change in battery parameters Shalini.M,Liya.B.S,Monisha.V,Nandhini
are measured is measured devi.R(2018 January)- Prathyusha
Engineering College,Chennai “Smart Device
Readings were taken with the system designed and to alert Bystanders of the abducted
compared with the readings taken from the sensors. child”,IJCRT,Volume 6,ISSN:2320-2882.
10. AlbersheimS,HolstiL,Mahon P (2015) ‘The
3. Shivashankar B.Nair,Sonia,Suraj Kumar fathers support scale:neonatal intensive care
Pandey,TusharSemwal,(2020) “Smart Patch: unit (FSS: NICU): development and initial
An IoT Based Anti Child-Trafficking content validation’,Journal of Neonatal
Solution” The 2020 IEEE International Nursing,, pp. 63–71, Volume 21, Issue:2.
Conference on Internet of Things and
Intelligence System (IoTaIS),Volume 2. 11. Chaseling G.K, DabovalT,Molgat-Seon
Y(2016) ‘Body temperature mapping in
4. Sachin Damke,YogitaK.Dubey (August critically ill newborn infants nursed under
2019) “Baby Monitoring System using radiant warmers during intensive
Image Processing and IoT” , International care’,Journal of Perinatology, Volume 36.
Journal of Engineering and
Advanced Technology(IJEAT),Volume 12. Basha.C.Z, Durga.S, Itnal.S,
8,ISSN:2249- 8958. Saxena.C,Soujanya.K (2021) ‘Advanced and
effective baby care monitoring Smart Cradle
5. Deepika.L,Indu.R,JasminBanu.R,Shakuntha System using Internet of things’,2nd
la.M (June 2018) “Neonatal Healthcare International Conference on Smart
Monitoring in Incubator Using IoT”, Electronics and
International Journal of Communication(ICOSEC),pp. 35-42.
Electrical,Electronics and Data
Communication, Volume 6,Issue 13. AkbiyikM.E,AygunE,Coban C
6,ISSN(p):2320-2084. S,GurgunogluD,IderD,Imamoglu H Z
(2020) “ThermoCam: Smart Baby
6. Anwesha Mukherjee,Arkaprabha Sau, Monitoring Assistant”,2020 IEEE 44th
DebashisDe.Ishita Bhakta (August 2016) Annual Computers,Software,and
“Design of Smart Neonatal Health Applications Conference
Monitoring System using (COMPSAC),pp.636-643.
SMCC”,HealthCare Technology, pp.1-
7,Volume 4,Issue 1. 14. Ahmad M.I,Ali.H,Malik.A (2019) “Li-Fi
Based Health Monitoring System for
7. Chiranjeevi Mondikathi,SilvyAchankunju Infants,”2nd International Conference on
(2015) “Voice based Security System Using Communication,Computing and Digital
Matlab& Embedded System”,International Systems (C-CODE),pp.69-72.
Journal of Scientific
Research,Volume:4,Issue:5,ISSN No 2277- 15. Raina G,Semwal.S,Tiwari R (2022) “An
817. Innovative Approach to Neonatal Intensive
Unit Care System for New Born babies”,
8. GuazziA,JorgeJ,Villarroel M (2014) 2022 International Conference on Advances
‘Continuous non-contact vital sign in Computing,Communication and Materials
monitoring in neonatal intensive care (ICACCM),pp.1-6.
unit’,Healthcare Technology Letters,pp. 87–
91, Volume 1, Issue 3. 16. AnnadO,CheggouR,KhoumeriE.H,Mohand
S.H (2020) “An Intelligent Baby Monitoring
9. Chen W,Coops R, Nguyen S.T (2009): System based on Raspberry Pi,IoT sensors
‘Wireless transmission design for health and Convolutional Neural network”, 2020
monitoring at neonatal intensive care IEEE 21st International Conference on
units’.Second International Symposium on Information Reuse and Integration for Data
Applied Sciences in Biomedical and Science (IRI),pp.365-371.
Communication Technologies, pp. 1–
6,Volume 4.

You might also like