DESC:TECHNICAL FIELD
[0001] The present disclosure relates generally to the field of battery management systems components capable of traversing over uneven surfaces. In particular, the present disclosure relates to a simple, compact, efficient, cost-effective, and improved battery management system and a method to predict happening of a fire hazard within a battery pack by accurately determining the State of Charge (SoC) and State of Health (SoH) of the battery pack in an electric vehicle (EV).

BACKGROUND
[0002] The field of advanced mobility solutions encompasses mechanisms and devices designed to facilitate movement across various terrains, including staircases. These solutions often involve intricate mechanical systems that enable devices such as wheelchairs, robots, and other mobility aids to navigate steps and uneven surfaces. Applications of these technologies range from personal mobility aids to autonomous delivery systems and search and rescue robots.
[0003] Electric vehicles (EVs) represent a significant shift toward sustainable transportation, relying heavily on battery systems for energy storage and delivery. However, the performance, safety, and longevity of these batteries remain substantial challenges. Conventional Battery Management Systems (BMS) primarily focus on monitoring basic parameters such as voltage and current, which limits their ability to provide accurate predictions of battery health and safety risks. This lack of comprehensive monitoring and predictive capabilities often results in less-than-optimal performance, reduced battery lifespan, and an inability to proactively address potential hazards such as thermal runaway or fire incidents. Furthermore, existing systems typically rely on singular machine learning (ML) models, which may fail to capture the complex temporal and nonlinear relationships associated with battery behavior, leading to inaccuracies in predicting the State of Charge (SoC) and State of Health (SoH).
[0004] There is, thus, a need for mitigating the above-stated challenges by providing an improved battery management system and method implemented by the said system integrating advanced sensors, hybrid learning modules implemented by a processor to predict happening of a fire hazard within a battery pack by accurately determining the State of Charge (SoC) and State of Health (SoH) of the battery pack in an electric vehicle (EV).

OBJECTIVE OF THE PRESENT DISCLOSURE
[0005] A general objective of the present disclosure is to overcome the problems associated with existing conventional battery management systems, by providing a simple, compact, efficient, and cost-effective, and improved battery management system to predict happening of a fire hazard within a battery pack of an electric vehicle.
[0006] An objective of the present disclosure is to use a combination of two or more learning modules implemented by a processor to predict occurrence of the fire hazard.
[0007] Another objective of the present disclosure is to accurately determine the State of Charge (SoC) and State of Health (SoH) of cells within the battery pack.
[0008] Yet another objective of the present disclosure is to provide a method implemented by the battery management system to predict happening of a fire hazard within a battery pack, and accurately determine the State of Charge (SoC) and State of Health (SoH) of cells within the battery pack of an electric vehicle.

SUMMARY
[0009] Aspects of the present disclosure pertain to the field of battery management systems for battery packs. In particular, the present disclosure relates to a simple, compact, cost-effective, and improved battery management system and a method to predict happening of a fire hazard within a battery pack by accurately determining the State of Charge (SoC) and State of Health (SoH) of the battery pack in an electric vehicle (EV).
[0010] According to an aspect, the proposed battery management system for monitoring performance of a battery pack of an electric vehicle (EV) includes sensors configured inside a housing of the battery pack to sense predefined parameters of cells within the battery pack. The system comprises a processor implementing a combination of first and second learning modules is configured to analyse sensed data received from the sensors using a first learning module to capture temporal patterns by identifying sequential dependencies, trends in the sensed data over time, and using the second learning module to capture nonlinear relationships by modeling complex interactions between input features, output variables.
[0011] The processor determines a state of charge (SoC) and a state of health (SoH) of the cells within the battery pack upon capturing temporal patterns and nonlinear relationships. The processor predicts a risk of fire hazard upon determining the SoC and the SoH and generates alerts for a user regarding the sensed parameters, the determined SoC, SoH, and the predicted risk of fire hazard using a mobile device associated with a user of the system. The system accurately predicts the occurrence of fire hazard and reduces errors in determining the SoC and the SoH of the cells within the battery pack.
[0012] In an embodiment, the processor may be a Raspberry Pi 4B.
[0013] In an embodiment, the first learning module may be a Long Short-Term Memory (LSTM) module for capturing temporal patterns. The second learning module may be a Support Vector Regression (SVR) module for capturing nonlinear relationships.
[0014] In an embodiment, the predefined parameters may be selected from a group comprising temperature, voltage, current, flame presence, and gas concentration.
[0015] In an embodiment, the system may include a remote server. The remote server is in communication with the processor to store the sensed data, determined SoC, SoH, and the predicted risk of fire hazard.
[0016] In an embodiment, the remote server may be in communication with the processor and a mobile device associated with the user. The mobile device may enable the user to access the stored data within the remote service in real-time.
[0017] In an embodiment, the sensors may include a temperature sensor to measure the temperature of the battery pack, a voltage sensor to measure the voltage of the battery pack, a current sensor to measure the current flowing through the battery pack, a flame sensor to detect the presence of flames; and a gas sensor to detect the concentration of gases such as carbon monoxide, methane, sulfur dioxide, and hydrogen sulfide.
[0018] In an embodiment, the processor may train the first and second learning modules using historical data on performance of the battery pack to improve the accuracy of SoC and SoH predictions. The processor may detect anomalies in the sensed data and generate alerts for potential faults in the battery pack. The processor may calculate performance metrics, including R2 score, root mean square error (RMSE), and mean absolute error (MAE), to evaluate the accuracy of the SoC and SoH predictions.
[0019] According to an aspect, a method implemented by a system for monitoring the health of a battery pack in an electric vehicle (EV) begins with a step of sensing predefined parameters like temperature, voltage, current, flame presence, and gas concentration of cells within the battery pack using sensors.
[0020] The method includes a step of analyzing the sensed data received from the sensors such that a first learning module that correspond to a Long Short-Term Memory (LSTM) module implemented by the processor captures temporal patterns in the sensed data by identifying sequential dependencies and the second learning module that correspond to a Support Vector Regression (SVR) module captures nonlinear relationships in the sensed data by modeling complex interactions between input features, output variables.
[0021] The method includes another step of determining a state of charge (SoC) and a state of health (SoH) of the battery pack using the processor and predicting a risk of fire hazard using the processor. The method includes a step of storing the sensed parameters, the determined Soc, SoH, and the predicted risk of fire hazard using a remote server in communication with the processor. The method ends with a final step of generating alerts for the user regarding the determined SoC, SoH, and the predicted risk of fire hazard using a mobile application in a mobile device associated with the user.
[0022] Various uneven surface / obstacles, features, aspects, and advantages of the subject matter will become more apparent from the following detailed description of preferred embodiments, along with the accompanying drawing figures in which like numerals represent like components.

BRIEF DESCRIPTION OF DRAWINGS
[0023] The accompanying drawings are included to provide a further understanding of the present disclosure and are incorporated in and constitute a part of this specification. The drawings illustrate exemplary embodiments of the present disclosure and, together with the description, serve to explain the principles of the present disclosure. The diagrams are for illustration only, which thus is not a limitation of the present disclosure.
[0024] FIG. 1 illustrates a block diagram of the proposed battery management system for accurately determining the State of Charge (SoC) and State of Health (SoH) of a battery pack in an electric vehicle (EV), in accordance with embodiments of the present disclosure.
[0025] FIG. 2 illustrates a flow diagram describing a method implemented by the proposed battery management system for accurately determining the State of Charge (SoC) and State of Health (SoH) of a battery pack in an electric vehicle (EV), in accordance with embodiments of the present disclosure.
[0026] FIG. 3 illustrates a flow diagram describing the monitoring performance of a battery pack, in accordance with embodiments of the present disclosure.
[0027] FIGs. 4 and 5 illustrate a graphical representation showing the dynamic nature of the voltage of the battery where the voltage fluctuates between 1.02V to 1.05V reflecting on the variations for understanding the performance and stability of the battery pack, in accordance with embodiments of the present disclosure.
[0028] FIG. 6 illustrates a graphical representation showing the voltage, current and temperature of the cells stored and shown on the dashboard every second,in accordance with embodiments of the present disclosure.
[0029] FIG. 7 illustrates a graphical representation showing flame status, in accordance with embodiments of the present disclosure.
[0030] FIGs. 8 and 9 illustrates a graphical representation showing gas concentration, in accordance with embodiments of the present disclosure.
[0031] FIG. 10 illustrates alerts generated on a mobile device, in accordance with embodiments of the present disclosure.

DETAILED DESCRIPTION
[0032] For the purpose of promoting an understanding of the principles of the present disclosure, reference will now be made to the various embodiments and specific language will be used to describe the same. It will nevertheless be understood that no limitation of the scope of the present disclosure is thereby intended, such alterations and further modifications in the illustrated system, and such further applications of the principles of the present disclosure as illustrated therein being contemplated as would normally occur to one skilled in the art to which the present disclosure relates.
[0033] It will be understood by those skilled in the art that the foregoing general description and the following detailed description are explanatory of the present disclosure and are not intended to be restrictive thereof.
[0034] Whether or not a certain feature or element was limited to being used only once, it may still be referred to as “one or more features” or “one or more elements” or “at least one feature” or “at least one element.” Furthermore, the use of the terms “one or more” or “at least one” feature or element do not preclude there being none of that feature or element, unless otherwise specified by limiting language including, but not limited to, “there needs to be one or more…” or “one or more elements is required.
[0035] Reference is made herein to some “embodiments.” It should be understood that an embodiment is an example of a possible implementation of any features and/or elements of the present disclosure. Some embodiments have been described for the purpose of explaining one or more of the potential ways in which the specific features and/or elements of the proposed disclosure fulfil the requirements of uniqueness, utility, and non-obviousness.
[0036] Use of the phrases and/or terms including, but not limited to, “a first embodiment,” “a further embodiment,” “an alternate embodiment,” “one embodiment,” “an embodiment,” “multiple embodiments,” “some embodiments,” “other embodiments,” “further embodiment”, “furthermore embodiment”, “additional embodiment” or other variants thereof do not necessarily refer to the same embodiments. Unless otherwise specified, one or more particular features and/or elements described in connection with one or more embodiments may be found in one embodiment, or may be found in more than one embodiment, or may be found in all embodiments, or may be found in no embodiments. Although one or more features and/or elements may be described herein in the context of only a single embodiment, or in the context of more than one embodiment, or in the context of all embodiments, the features and/or elements may instead be provided separately or in any appropriate combination or not at all. Conversely, any features and/or elements described in the context of separate embodiments may alternatively be realized as existing together in the context of a single embodiment.
[0037] Any particular and all details set forth herein are used in the context of some embodiments and therefore should not necessarily be taken as limiting factors to the proposed disclosure. The terms “comprises”, “comprising”, or any other variations thereof, are intended to cover a non-exclusive inclusion, such that a process or method that comprises a list of steps does not include only those steps but may include other steps not expressly listed or inherent to such process or method. Similarly, one or more devices or sub-systems or elements or structures or components proceeded by “comprises... a” does not, without more constraints, preclude the existence of other devices or other sub-systems or other elements or other structures or other components or additional devices or additional sub-systems or additional elements or additional structures or additional components.
[0038] Embodiments explained herein relate to a simple, compact, efficient, cost-effective, and improved battery management system and a method to predict happening of a fire hazard within a battery pack by accurately determining the State of Charge (SoC) and State of Health (SoH) of the battery pack in an electric vehicle (EV).
[0039] According to an aspect, the proposed system and a method for monitoring performance of a battery pack of an electric vehicle (EV) includes sensors and a processor configured inside the battery pack. The processor includes a first learning module and a second learning module configured in the battery pack to analyse the predefined parameters of cells within the battery pack sensed by the sensors using a combination of the first learning module that captures temporal patterns, and the second learning module that captures nonlinear relationships. The processor determines a state of charge (SoC) and a state of health (SoH) of the cells and predicts a risk of fire hazard. The processor generates alerts for a user regarding the sensed parameters, the determined SoC, SoH, and the predicted risk of fire hazard using a mobile device.
[0040] Referring to FIGs.1 to 10, a battery management system (hereinafter referred to as ‘battery management system 100’ or simply ‘system 100’) for monitoring health of a battery pack 200 of an electric vehicle (EV) is described. The system 100 includes a plurality of sensors 102 and a processor 104 assembled within a housing 202 of the battery pack 200 to accurately predict occurrence of fire hazard and reduce errors in determining state of charge (SoC) and state of health (SoH) of the cells 204 within the battery pack 200 within the battery pack 200.
[0041] The plurality of sensors 102 sense predefined parameters of cells 204 within the battery pack 200. The predefined parameters may be selected from a group comprising temperature, voltage, current, flame presence, and gas concentration.
[0042] The plurality of sensors 102 may include a temperature sensor 102A that is a BMP280 Barometer to measure the temperature of the battery pack, a voltage sensor 102B that is an analog voltage sensor to measure the voltage of the battery pack; a current sensor 102C that is a INA219 Current sensor to measure the current flowing through the battery pack, a flame sensor 102D that is a Digital Flame sensor to detect the presence of flames; and a gas sensor 102E that is a MQ2 Gas sensor to detect the concentration of gases such as carbon monoxide, methane, sulfur dioxide, and hydrogen sulfide.
[0043] The processor 104 is a Raspberry Pi 4B in communication with the plurality of sensors 102 analyses the sensed data received from the sensors 102 using a combination of the first learning module 104A and the second learning module 104B implemented by the processor 104 to determine a state of charge (SoC) and a state of health (SoH) of the cells 204 within the battery pack 200.
[0044] The first learning module 104A is a Long Short-Term Memory (LSTM) module to capture temporal patterns by identifying sequential dependencies, and trends in the sensed data over time. The second learning module 104B to capture nonlinear relationships by modeling complex interactions between input features, and output variables. The processor 104 predicts a risk of fire hazard upon determining the SoC and the SoH and generates alerts for a user regarding the sensed parameters, the determined SoC, SoH, and the predicted risk of fire hazard using a mobile device 106 associated with a user of the system 100.
[0045] Additionally, the processor 104 may further be configured to train the first and second learning modules 104A,104B using historical data on performance of the battery pack 200 to improve the accuracy of SoC and SoH predictions. The processor 104 may detect anomalies in the sensed data and generate alerts for potential faults in the battery pack 200. The processor 104 may calculate performance metrics, including R2 score, root mean square error (RMSE), and mean absolute error (MAE), to evaluate the accuracy of the SoC and SoH predictions.
[0046] Further, the system 100 may include a remote server 108 in communication with the processor 104 to store the sensed data, determined SoC, SoH, and the predicted risk of fire hazard. The remote server 108 in communication with the processor 104 and the mobile device 106 associated with the user may enable the user to access the stored data transmit in real-time.
[0047] In an embodiment, the sensed data is processed in a Raspberry Pi 4B which then stores the data in the cloud for analysis and visualization. The visualization is done using Thingspeak IoT Cloud Platform. The system is configured to send alerts to mobile user whenever required. The data processing is done using a hybrid model combining LSTM and SVR neural network models for SoC and SoH prediction, where LSTM is a Recurrent Neural Network which has an application in sequence prediction and SVR is effective for Regression scenarios. LSTM works by employing three steps which involves utilizing the historical data, training the model by identifying the patterns and temporal dependencies and final step to predict the SoH or SoC of the battery system. On the other hand, SVR incorporates the capturing of complex nonlinear input features and output variables relationships. A combination of the LSTM and SVR models avails the capturing of temporal patterns by using LSTM and capturing complex nonlinear relationships by using SVR leading to reduction in errors, losses, and storage consumption whereas increasing the performance metrics such as R2 score resulting in more accurate predictions of SoH and SoC of the battery.
[0048] A system to monitor the health of an electric vehicle (EV) battery uses a combination of models to predict its SoC and SoH of the battery with an EV Fire Detection system can give EV owners a sense of security. By identifying potential problems early on, it ensures that the battery performs well and lasts longer, ultimately saving them money on maintenance and replacement expenses and prevent mishaps in the future.
[0049] Referring to FIG. 2, a method (hereinafter referred as ‘method 200’) implemented by a system for monitoring performance of a battery pack 200 in an electric vehicle (EV) is described. The method 200 includes a step 202 of sensing predefined parameters like temperature, voltage, current, flame presence, and gas concentration of cells within the battery pack 200 using a plurality of sensors. The method 200 includes a step 204 of analyzing the sensed data received from the plurality of sensors such that a first learning module that correspond to a Long Short-Term Memory (LSTM) module implemented by a processor captures temporal patterns in the sensed data by identifying sequential dependencies, trends in the sensed data over time, and a second learning module that correspond to a Support Vector Regression (SVR) module implemented by the processor captures nonlinear relationships in the sensed data by modeling complex interactions between input features, output variables.
[0050] The method 200 includes a step 206 of determining a state of charge (SoC) and a state of health (SoH) of the battery pack using the processor and another step 208 of predicting a risk of fire hazard upon determining the SoC, and the SoH using the processor.
[0051] The method 200 includes a step 210 of storing the sensed parameters, the determined Soc, SoH, and the predicted risk of fire hazard and another step 212 of generating alerts for the user regarding the determined SoC, SoH, and the predicted risk of fire hazard using a mobile application in a mobile device associated with the user.
[0052] Referring to FIG.3, illustrates a flowchart for monitoring the EV battery. An EV Battery Detection System is developed to collect data for fire detection FIG. 2 illustrates a flowchart for monitoring the EV battery. An EV Battery Detection System is developed to collect data for fire detection and battery parameters. FIG. 2 illustrates a flowchart for monitoring the EV battery. An EV Battery Detection and battery parameters. It collects tery in EV. All temperature, gas concentration, flame presence, voltage and current of the bat the EV data is sent to the Raspberry Pi 4B which process the EV data to send alert to mobile and to the cloud for data visualization and storage of data for longer period. This also includes prediction the SoC and SoH of the battery for pr vehicle using ML hybrid models. The ML model which is LSTM temperature, gas concentration, flame presence, voltage and current of the battery in EV. All the EV data is sent to the Raspberry Pi 4B which process the EV data to send alert to mobile and to the cloud for data visualization and storage of data for longer period. This also includes prediction the SoC and SoH of the battery for predictive and preventive maintenance of the vehicle using ML hybrid models. The ML model which is LSTM-SVR is trained and tested in the Raspberry Pi 4B using the EV battery parameters for prediction of SoC and SoH which is further sent to the smartphone and to the cloud storage which is also used for data the EV data is sent to the Raspberry Pi 4B which process the EV data to send alert to mobile and to the cloud for data visualization and storage of data for longer period. This also includes edictive and preventive maintenance of the SVR is trained and tested in the Raspberry Pi 4B using the EV battery parameters for prediction of SoC and SoH which is to the cloud storage which is also used for data the Raspberry Pi 4B using the EV battery parameters for prediction of SoC and SoH which is further sent to the smartphone and visualization as well.
[0053] HARDWARE COMPONENTS USED:
EV Battery Detection System
Flame Sensor (KY-026)– This sensor is used to detect the presence of fire.
Gas Sensor (MH MQ2) – This sensor is used for detecting various gases.
Temperature Sensor (BMP280) – This sensor is used to measure the temperature of the battery.
Analog Voltage Sensor - This sensor is used to measure the voltage values of the battery.
ADC 1115 – It is used to provide analog values of the battery’s voltage.
Digital Power Monitor (INA219) this sensor is used to measure the current This sensor is used to measure the current of the Raspberry pi 4B (8 GB RAM) for SoH and SoC prediction are done here. Processing of sensors data and ML Hybrid algorithms Raspberry pi 4B (8 GB RAM) – Processing of sensors data and ML Hybrid algorithms d SoC prediction are done here.
[0054] SOFTWARE COMPONENTS:
To enable sensors interact with the raspberry pi, this software is utilized. To enable sensors interact with the raspberry pi, this software is utilized. Through this way, it becomes possible for the sensors to work, sense any signals, and alert the Thonny – To enable sensors interact with the raspberry pi, this software is utilized. Through this way, it becomes possible for the sensors to work, sense any signals, and alert the mobile user when necessary, using the API key. Through this way, it becomes possible for the sensors to work, sense any signals, and alert the ssary, using the API key.
Sharing the required data or alert over a phone could be safe and secure Sharing the required data or alert over a phone could be safe and secure
Pushbullet – Sharing the required data or alert over a phone could be safe and secure from the push bullet software.
Python IDLE - This is used for ML Hybrid algorithms to predict the SoH and SoC. This is used for ML Hybrid algorithms to predict the SoH and SoC.
[0055] Referring to FIG.4. shows the dynamic nature of the voltage of the battery where the voltage fluctuates between 1.02V to 1.05V reflecting on the variations which could be helpful for understanding the performance and stability of the battery system.
[0056] The FIG.5 shows the dynamic nature of the current of the battery where the current fluctuates between 0A to 0.2A due to the shunt resistor of the current sensor that is 0.1.
[0057] The FIG. 6 show the voltage, current and temperature of the battery stored and shown on the dashboard every second.
[0058] FIG. 7 shows the flame status and FIG. 8 shows the gas concentration. FIG. 9 depicts the range of gas concentration to a numerical value. 400-500 ppm gas concentration depicts ‘0’ in the Thingspeak dashboard, which further depicts that type of gas could be carbon dioxide (CO2). If it is less than the range it indicates the presence of harmful gases such as CO(Carbon Monoxide) which could be commonly detected or methane(CH4). Even SO2 or H2S presence can also be detected below 350 ppm. This helps in monitoring the battery status and provides insights for preventing fire mishaps.
[0059] To provide alerts with respect to the gas and flame detection of the battery, an api key was generated and given to the code which would send the necessary commands to the mobile through the pushbullet application as shown in FIG.10.
[0060] In this study, the focus lies on accurate estimation and prediction of SoC (State of Charge) and State of Health (SoH) of the battery for an improved performance of the EV battery health. An integration of Machine Learning (ML) techniques and Batter Management System brings a spectrum of redefined optimization, monitoring, and controlling strategies with accuracy and precision. ML leverages data analytics and intelligent algorithms to provide formerly unseen opportunities for enhanced efficiency, sustainability, and user experience in built environments.
[0061] There are various ML methods based on variable parameters used in BMS for SoC and SoH prediction and few of those methods are Support Vector Regression (SVR), Long-short Term Memory (LSTM), Autoregressive Integrated Moving Average (ARIMA), Attention, etc. All ML techniques will have their respective leading behavior for selectivity towards the most suitable method for prediction. Moreover, it will also possess the drawbacks points where scope of improvement is present. However, combining more than one method will leverage the strengths of each algorithm to offset the limitations of the others. Therefore, Hybrid models where two algorithms are used together for SoC and SoH prediction to counter the weaknesses leading to better desired results. The hybrid model incorporates the performance optimization; By using two algorithms together it counters the issue of losing data points, resulting in the reduction of the chances of failing in prediction of SoC and SoH value. The losses are reduced with minor differences in training and testing time. Moreover, when the data goes under the training process it gives more accurate and precise estimation and prediction. The Hybrid Model – LSTM-SVR with their results in Python IDLE and RaspberryPi 4B are discussed below:
[0062] LSTM-SVRis the blend of Support Vector Regression (SVR) processes and successive Long Short-Term Memory (LSTM) NN, which allows it to make regression predictions and learn orders at the same time. The hybrid model for time series forecasting situations where it is all about taking long-term dependences and nonlinear relations will work best. The below subsections provide detailed results of SoH and SoC predictions using this ML algorithm in Python IDLE and RaspberryPi 4B.
[0063] Prediction of State of Health (SoH)
[0064] In Python IDLE

Epoch s RMSE*10-3 MAE*10-3 R2 Score Explained Variance*10-3 MSLE*10-3 RMSLE*10-3
40 41.7560 16.9830 99.99884 99.99887 0.9040 30.0671
60 39.1151 14.8152 999.9898 999.9899 0.8455 29.0778
80 57.4730 19.1150 999.9781 999.9784 1.6887 41.0936
100 44.0340 18.3607 999.9871 999.9873 1.0142 31.8464
120 36.23651 20.72522 999.9913 999.99280 0.54491 23.34301
Table.1. shows the performance metrics such as R2 score using Python IDLE, which represents the efficiency of the SOH predicted comes out to be 0.99 whereas the errors generated to be 0.036, 0.020.

Epochs RMSE *10-3 MAE*10-3 R2
Score Explained Variance *10-3 MSLE**10-3 RMSLE*10-3

40 72.6091 39.3686 999.9650 999.9979 2.0565 45.3484
60 75.8280 42.8333 999.9619 999.9690 2.1537 46.4084
80 76.4011 29.6330 999.9613 999.9635 2.5279 50.2781
100 72.0652 32.2187 999.9655 999.9687 2.2437 47.3678
Table 2 shows the performance metrics such as R2 score using Raspberry Pi which represents the efficiency of the SOH predicted comes out to be 0.99 whereas the errors generated are 0.072, and 0.032.

Epochs RMSE*10-3 MAE*10-3 R2
Score Explained Variance*10-3 MSLE**10-3 RMSLE**10-3
40 648.5616 789.5462 978.4562 988.632 54.6289 245.7812
60 384.8762 314.6299 997.8456 982.155 5.4786 388.7921
80 345.6799 254.8567 934.9638 958.7966 4.6543 364.8763
100 248.7938 265.7893 967.6627 978.3786 5.6393 694.8991
120 217.8952 365.4866 997.8933 997.7951 3.5254 438.1284
Table 3 shows the R2 score which represents the efficiency of the SOC predicted using Python IDLE, which comes out to be 0.9978933 whereas the errors generated are 0.02178592, and 0.3654866.

[0065] Among the Hybrid Models for SoC and SoH prediction, Hybrid model of LSTM and SVR used finds to be the best for SoC and SoH prediction where LSTM is a Recurrent Neural Network which has an application in sequence prediction and SVR is effective for Regression scenarios. LSTM works by employing three steps which involve utilizing the historical data, training the model by identifying the patterns and temporal dependencies and final step to predict the SoH or SoC of the battery system. On the other hand, SVR incorporates the capturing of complex nonlinear input features and output variables relationships. Combination of these two avails the capturing of temporal patterns by using LSTM and capturing complex nonlinear relationships by using SVR leading to reduction in errors, losses, and storage consumption whereas increasing the performance metrics such as R2 score resulting in more accurate predictions of SoH and SoC of the battery.
[0066] Thus, among the Hybrid Machine Learning (ML) methods such as LSTM- ARIMA and LSTM-Attention used for prediction of SoC and SoH, LSTM-SVR ML Hybrid method proved to be the best based on performance metrics such as RMSE, MAE, R2 Score. The Results achieved from the hardware setup for battery detection give an insight on condition of the battery and if any necessary actions can be taken. Designing a system to monitor the health of an electric vehicle (EV) battery using a combination of models to predict its SoC and SoH of the battery with an EV Fire Detection system can give EV owners a sense of security. By identifying potential problems early on, it ensures that the battery performs well and lasts longer, ultimately saving them money on maintenance and replacement expenses and prevent mishaps in the future.
[0067] While the foregoing describes various embodiments of the invention, other and further embodiments of the invention may be devised without departing from the basic scope thereof. The scope of the invention is determined by the claims that follow. The invention is not limited to the described embodiments, versions, or examples, which are included to enable a person having ordinary skill in the art to make and use the invention when combined with information and knowledge available to the person having ordinary skill in the art.

ADVANTAGES OF THE INVENTION
[0068] The present disclosure overcomes the problems associated with existing rotating components capable of traversing over uneven surfaces, by providing a simple, compact, efficient, and cost-effective wheel to traverse over an uneven surface / obstacle.
[0069] The present disclosure is to provide an improved battery management system to predict happening of a fire hazard within a battery pack of an electric vehicle.
[0070] The present disclosure uses a combination of two or more learning modules implemented by a processor to predict occurrence of the fire hazard.
[0071] The present disclosure accurately determines the State of Charge (SoC) and State of Health (SoH) of cells within the battery pack.
[0072] The present disclosure provides a method implemented by the battery management system to predict happening of a fire hazard within a battery pack, and accurately determine the State of Charge (SoC) and State of Health (SoH) of cells within the battery pack of an electric vehicle.
,CLAIMS:1. A system for monitoring health of a battery pack of an electric vehicle (EV), the system (100) comprising:
a plurality of sensors (102) configured inside a housing (202) of the battery pack (200) to sense predefined parameters of cells (204) within the battery pack (200);
a processor (104) comprising a first learning module (104A) and a second learning module (104B), in communication with the plurality of sensors (102), and configured in the battery pack (200) to:
analyse sensed data received from the one or more sensors (102) using a combination of the first learning module (104A) that captures temporal patterns by identifying sequential dependencies, and trends in the sensed data over time, and the second learning module (104B) that captures nonlinear relationships by modeling complex interactions between input features, output variables;
determine a state of charge (SoC) and a state of health (SoH) of the cells (204) within the battery pack (200) upon capturing temporal patterns and nonlinear relationships;
predict a risk of fire hazard upon determining the SoC and the SoH;
generate alerts for a user regarding the sensed parameters, the determined SoC, SoH, and the predicted risk of fire hazard using a mobile device (106) associated with a user of the system (100);
wherein the system (100) accurately predicts the occurrence of fire hazard and reduces errors in determining the SoC and the SoH of the cells (204) within the battery pack (200).
2. The system (100) as claimed in claim 1, wherein the processor (104) is a Raspberry Pi 4B.
3. The system (100) as claimed in claim 1, wherein the first learning module (104A) is a Long Short-Term Memory (LSTM) module for capturing temporal patterns and the second learning module (104B) is a Support Vector Regression (SVR) module for capturing nonlinear relationships.
4. The system (100) as claimed in claim 1, wherein the predefined parameters are selected from a group comprising temperature, voltage, current, flame presence, and gas concentration.
5. The system (100) as claimed in claim 1, wherein the system (100) comprises a remote server (108) in communication with the processor (104) to store the sensed data, determined SoC, SoH, and the predicted risk of fire hazard.
6. The system (100) as claimed in claim 5, wherein the remote server (108) in communication with the processor (104) and a mobile device (106) associated with the user, enables the user to access the stored data transmit in real-time.
7. The system (100) as claimed in claim 1, wherein the plurality of sensors (102) comprises a temperature sensor (102A) to measure the temperature of the battery pack, a voltage sensor (102B) to measure the voltage of the battery pack; a current sensor (102C) to measure the current flowing through the battery pack, a flame sensor (102D) to detect the presence of flames; and a gas sensor (102E) to detect the concentration of gases such as carbon monoxide, methane, sulfur dioxide, and hydrogen sulfide.
8. The system (100) as claimed in claim 1, wherein the processor (104) is further configured to:
train the first and second learning modules (104A,104B) using historical data on performance of the battery pack (200) to improve the accuracy of SoC and SoH predictions;
detect anomalies in the sensed data and generate alerts for potential faults in the battery pack (200); and
calculate performance metrics, including R2 score, root mean square error (RMSE), and mean absolute error (MAE), to evaluate the accuracy of the SoC and SoH predictions.
9. A method implemented by a system for monitoring the health of a battery pack in an electric vehicle (EV), the method (200) comprising:
sensing (202), using a plurality of sensors, predefined parameters like temperature, voltage, current, flame presence, and gas concentration of cells within the battery pack;
analyzing (204), using a combination of a first and second learning modules implemented by a processor, the sensed data received from the plurality of sensors such that the first learning module that correspond to a Long Short-Term Memory (LSTM) module captures temporal patterns in the sensed data by identifying sequential dependencies, trends in the sensed data over time, and the second learning module that correspond to a Support Vector Regression (SVR) module captures nonlinear relationships in the sensed data by modeling complex interactions between input features, output variables;
determining (206), using the processor, a state of charge (SoC) and a state of health (SoH) of the battery pack upon capturing temporal patterns and nonlinear relationships;
predicting (208), using the processor, a risk of fire hazard upon determining the SoC, and the SoH;
storing (210), using a remote server in communication with the processor, the sensed parameters, the determined Soc, SoH, and the predicted risk of fire hazard; and
generating alerts (212), using a mobile application in a mobile device associated with the user, for the user regarding the determined SoC, SoH, and the predicted risk of fire hazard.