DESC:SYSTEM AND METHOD FOR CREATING USAGE PROFILE FOR BATTERY PACK
CROSS REFERENCE TO RELATED APPLICATIONS
The present application claims priority from Indian Provisional Patent Application No. 202421070094 filed on 17/09/2024, the entirety of which is incorporated herein by a reference.
TECHNICAL FIELD
Generally, the present disclosure relates to a battery pack. Particularly, the present disclosure relates to a system and method for creating a usage profile for a battery pack.
BACKGROUND
Electric vehicles (EVs) operate using an electric powertrain powered by an onboard rechargeable battery pack that serves as the primary energy storage unit. The battery pack supplies electrical energy to propulsion blocks and auxiliary subsystems, making the EV's operational efficiency and reliability critical to overall vehicle performance. The state of health, charge discharge cycles, and operational load patterns of the battery pack directly influence range, durability, and safety. Further, monitoring and managing the operational profile of the battery pack requires continuous acquisition and processing of parameters from multiple vehicle subsystems to ensure optimal performance under varying driving conditions.
Conventionally, the battery management systems and vehicle monitoring platforms employ real-time data acquisition and basic signal processing techniques to assess the battery status and predict performance trends. The above-mentioned methodology utilizes quantization to reduce data size, filtering to remove noise, and statistical analysis to identify operational patterns. Further, in certain advanced systems, frequency-domain analysis and limited adaptive algorithms are applied to enhance diagnostic accuracy. The aforementioned solutions are functional and rely on fixed configuration parameters, offline recalibration, and simplified feature extraction mechanisms that do not completely capture the complexity of dynamic driving environments and evolving battery characteristics.
However, there are certain problems associated with the existing or above-mentioned mechanism for generating an operational profile for a battery pack of an electric vehicle. Specifically, the above-mentioned approaches face challenges in adapting to rapidly changing operational conditions, maintaining data precision with reduced computational resources, and performing continuous self-optimization without manual intervention. Further, the static quantization and filtering parameters result in loss of critical signal data, and fixed-step retraining mechanisms lead to slow adaptation or instability in predictive models. Additionally, limited integration of frequency-domain features in retraining workflows reduces the ability to detect subtle degradation trends and real-time operational anomalies.
Therefore, there exists a need for a secure, interoperable, and automated alternative for tracking the swappable battery of an electric vehicle.
SUMMARY
An object of the present disclosure is to provide a system for generating an operational profile for a battery pack of an electric vehicle.
Another object of the present disclosure is to provide a method for generating an operational profile for a battery pack of an electric vehicle.
Yet another object of the present disclosure is to provide a system and method for ensuring continuous and independent tracking of a swappable battery of an electric vehicle during both coupled and detached states.
In accordance with a first aspect of the present disclosure, there is provided a system for generating an operational profile for a battery pack of an electric vehicle, the system comprising:
- a sensing module operatively coupled to a plurality of vehicle subsystems and the battery pack, and configured to acquire real-time vehicle parameters from the plurality of vehicle subsystems and the battery pack;
- a memory module operatively coupled to the sensing module and configured to store real-time vehicle parameters acquired from the sensing module and historical vehicle parameters collected from past operational cycles of the vehicle;
- a processing unit operatively coupled to the memory module and configured to execute at least one data processing routine, the data processing routine comprising:
- quantizing real-time vehicle parameters received from the memory module;
- applying a temporal consistency filtering to the quantized real-time vehicle parameters; and
- extracting a plurality of frequency-domain attributes from the filtered real-time vehicle parameters,
wherein the processing unit iteratively retrains the data processing routines within the memory module based on the plurality of frequency-domain attributes extracted from the real-time vehicle parameters and the historical vehicle parameters.
The system for generating an operational profile for a battery pack of an electric vehicle, as described in the present disclosure, is advantageous in terms of producing highly accurate operational profiles for electric vehicle battery packs. Further, the approach reduces computational and memory overhead by employing reduced-bit fixed-point representations and optimized quantization parameters while preserving critical signal integrity. Furthermore, the adaptive learning through iterative retraining with dynamically adjusted step sizes ensures continuous system optimization and alignment with evolving vehicle conditions. Moreover, the integration of temporal and spectral features enhances diagnostic accuracy, predictive capability, and robustness under varying operational states. Subsequently, the overall design supports scalability across different vehicle platforms, delivering long-term reliability and performance efficiency.
In accordance with another aspect of the present disclosure, there is provided a method for generating an operational profile for a battery pack of an electric vehicle, the method comprising:
- acquiring real-time vehicle parameters from a plurality of vehicle subsystems, via a sensing module;
- executing at least one data processing routine, via a processing unit;
- applying adaptive quantization by selecting the quantization parameters based on real-time operational state variables of the vehicle, via the processing unit;
- applying temporal consistency filtering by employing a sliding-window temporal averaging, via the processing unit; and
- extracting a plurality of frequency-domain attributes from the filtered real-time vehicle parameters, via the processing unit.

Additional aspects, advantages, features, and objects of the present disclosure would be made apparent from the drawings and the detailed description of the illustrative embodiments constructed in conjunction with the appended claims that follow.
It will be appreciated that features of the present disclosure are susceptible to being combined in various combinations without departing from the scope of the present disclosure as defined by the appended claims.
BRIEF DESCRIPTION OF DRAWINGS
The summary above, as well as the following detailed description of illustrative embodiments, is better understood when read in conjunction with the appended drawings. For the purpose of illustrating the present disclosure, exemplary constructions of the disclosure are shown in the drawings. However, the present disclosure is not limited to specific methods and instrumentalities disclosed herein. Moreover, those in the art will understand that the drawings are not to scale. Wherever possible, like elements have been indicated by identical numbers.
Embodiments of the present disclosure will now be described, by way of example only, with reference to the following diagrams wherein:
Figure 1 illustrates a block diagram of a system for generating an operational profile for a battery pack of an electric vehicle, in accordance with an embodiment of the present disclosure.
Figure 2 illustrates a flow chart of a method for generating an operational profile for a battery pack of an electric vehicle, in accordance with another embodiment of the present disclosure.
In the accompanying drawings, an underlined number is employed to represent an item over which the underlined number is positioned or an item to which the underlined number is adjacent. A non-underlined number relates to an item identified by a line linking the non-underlined number to the item. When a number is non-underlined and accompanied by an associated arrow, the non-underlined number is used to identify a general item at which the arrow is pointing.
DETAILED DESCRIPTION
The following detailed description illustrates embodiments of the present disclosure and ways in which they can be implemented. Although some modes of carrying out the present disclosure have been disclosed, those skilled in the art would recognize that other embodiments for carrying out or practicing the present disclosure are also possible.
As used herein, the term “operational profile” refers to a structured and quantifiable representation of a battery’s behavior under real-time driving and usage conditions. Specifically, the operational profile of the battery pack is generated by acquiring real-time vehicle parameters through a sensing module coupled with vehicle subsystems, followed by systematic data processing to derive frequency-domain attributes that characterize battery usage patterns. Further, the operational profile is distinctly identified by quantized and temporally filtered sensor data that reflect instantaneous and historical operational states, comprising load demand, temperature variations, charge/discharge cycles, and driving dynamics. Furthermore, the system continuously updates the operational profile by retraining internal data processing routines using the extracted frequency-domain attributes from both real-time and historical datasets. Additionally, the types of operational profiles include, but not limited to, high-frequency dynamic profiles reflecting aggressive driving or rapid charging, low-frequency steady-state profiles representing cruising or idling conditions, and transitional profiles capturing variations between the states. Furthermore, each profile type supports precise, real-time battery management decisions and enables predictive maintenance by correlating frequency-domain patterns with battery health and performance degradation.
As used herein, the terms “battery pack”, “battery module”, and “power pack” are used interchangeably and refer to an integrated assembly of electrochemical energy storage units designed to power the propulsion and auxiliary modules of an electric vehicle, with precise control and monitoring of operational parameters. Specifically, the battery pack comprises multiple interconnected cells configured to deliver the required voltage, current, and energy density to meet dynamic vehicle load demands under varying environmental and operational conditions. Further, each battery pack interfaces with a plurality of vehicle subsystems, and the behavior of the battery pack is influenced by numerous factors such as, but not limited to, State of Charge (SoC), State of Health (SoH), internal resistance, temperature distribution, and load cycles. Furthermore, the battery pack functions as a dynamic arrangement, and the performance of the battery pack is continuously evaluated through adaptive quantization, temporal consistency filtering, and frequency-domain feature extraction, enabling iterative retraining of data processing routines. Additionally, based on structural configuration and application, the battery packs include, but are not limited to, types such as cylindrical cell packs, prismatic cell packs, and pouch cell packs, each exhibiting distinct thermal, electrical, and mechanical characteristics that influence quantization parameters and frequency response behavior.
As used herein, the terms “electric vehicle”, “EV”, and “vehicle” are used interchangeably and refer to a transport mechanism powered entirely or primarily by electric energy stored in a rechargeable battery pack. Specifically, the electric vehicle integrates multiple vehicle subsystems, including but not limited to propulsion, energy storage, thermal management, and control electronics, that operate in coordination to manage energy efficiency, performance, and sustainability through real-time monitoring and computational intelligence. The electric vehicles are classified into several types based on drivetrain architecture, including Battery Electric Vehicles (BEVs) powered solely by electric motors and rechargeable batteries, Plug-in Hybrid Electric Vehicles (PHEVs) that combine electric propulsion with an auxiliary internal combustion engine, and Hybrid Electric Vehicles (HEVs) that utilize both electric drive and fuel-based systems but lack external charging capability.
As used herein, the term “sensing module” refers to a set of hardware components capable of detecting and measuring physical parameters generated by various subsystems of an electric vehicle and subsequently converting the measurements of the physical parameters into electrical signals for processing. Specifically, the sensing module comprises sensors to sense parameters such as, but not limited to, voltage, current, temperature, acceleration, the SoC, the SoH, speed, and torque. Further, generating an operational profile for the battery pack, the sensor is operatively coupled to the vehicle subsystems to acquire real-time vehicle parameters with high temporal resolution. Further, the aforementioned physical parameters include, but are not limited to, voltage, current, temperature, acceleration, the SoC, the SoH, speed, and torque, which directly influence the performance and aging behavior of the battery pack. Furthermore, each sensor operates as an interface between the physical environment and the data processing architecture, ensuring that dynamic and transient behaviors of the vehicle subsystems are continuously captured. Moreover, the sensor is configured to deliver data streams that are subject to adaptive quantization, temporal consistency filtering, and frequency-domain transformation to extract features relevant for training and updating data processing routines. Additionally, the types of sensors include, but are not limited to, thermocouples or resistance temperature detectors (RTDs) for temperature monitoring, Hall effect sensors or shunt resistors for current measurement, voltage dividers or isolated differential amplifiers for voltage acquisition, inertial measurement units (IMUs) for acceleration and orientation, and position or speed sensors for drivetrain and wheel subsystem feedback. Further, each sensor type is selected based on the accuracy, sensitivity, and dynamic range required for reliable operational profiling under real-world driving conditions.
As used herein, the term “vehicle subsystem” refers to a distinct functional unit within the vehicle used to perform specific tasks essential to vehicle operation and that contributes real-time operational data. Specifically, each vehicle subsystem is defined by physical components, control logic, and interaction interfaces of the particular vehicle subsystem, and is responsible for producing measurable parameters, thereby reflecting the instantaneous and cumulative performance of the vehicle subsystem. The vehicle subsystems include, but are not limited to, the powertrain subsystem, thermal management subsystem, braking system, charging unit, battery management system (BMS), and traction control system. The powertrain subsystem delivers torque and power data, and the thermal management subsystem provides coolant temperature, heat flux, and thermal load information. Further, the braking system outputs braking force, hydraulic pressure, and deceleration metrics, and the charging unit reports state-of-charge (SoC), charging current, and voltage levels during both AC and DC charging events. Furthermore, the BMS yields internal battery metrics such as, but not limited to, cell voltage balance, internal resistance, temperature gradients, and charge/discharge cycles. Moreover, the traction control subsystem monitors wheel slip, torque vectoring, and longitudinal/lateral acceleration. Additionally, each vehicle subsystem interfaces with dedicated sensors that generate time-series data streams, which are subsequently acquired and quantized by the processing unit. Consequently, the real-time vehicle parameters serve as the foundational input for the data processing routines, enabling temporal filtering and frequency-domain feature extraction. Further, the operational profile derived from the vehicle subsystems directly informs adaptive quantization and model retraining strategies, ensuring the system maintains accuracy under varying driving conditions and operational loads.
As used herein, the term “vehicle parameters” refers to quantifiable characteristics generated by subsystems of the electric vehicle that influence the operational behavior of the battery pack. Specifically, the vehicle parameters are real-time data acquired by the sensing module operatively coupled to the vehicle subsystems. Further, the vehicle parameters include, but are not limited to, current, voltage, temperature, state of charge, state of health, acceleration profiles, torque demands, regenerative energy levels, vehicle speed, and environmental conditions such as, but not limited to, ambient temperature and humidity. Furthermore, each parameter provides a specific dimension of the vehicle's operational state and contributes to constructing an accurate and temporally coherent representation of the battery load profile. Moreover, the system acquires the aforementioned parameters in real-time and stores parameters with historical vehicle parameters collected from previous operational cycles in a memory module. Additionally, the processing of the vehicle parameters involves adaptive quantization, using numerical values and thereby transforming the vehicle parameters into reduced-bit fixed-point representations using quantization intervals selected based on operational state variables. Consequently, the combination of real-time and historical vehicle parameters enables the system to dynamically generate a robust and adaptive operational profile for the battery pack, optimizing performance and predictive analytics under varying load conditions.
As used herein, the term “memory module” refers to a hardware component operatively coupled to the sensors and the processing unit, and serves as the dedicated medium for storing both real-time vehicle parameters acquired from the vehicle subsystems and the historical vehicle parameters collected from past operational cycles. Functionally, the memory module enables persistent and structured storage of data required for executing the data processing routines associated with generating the operational profile for the battery pack of the electric vehicle. Further, the memory module provides direct data access to the processing unit, facilitating efficient retrieval, update, and retention of quantized vehicle parameters, filtered sequences, and computed frequency-domain attributes. Specifically, the memory module stores and supports iterative retraining of the data processing routines by maintaining updated batches of feature representations and associated parameters. Furthermore, the memory module includes, but is not limited to, volatile memory such as Static Random-Access Memory (SRAM) and Dynamic Random-Access Memory (DRAM), used for rapid access to temporary data during active processing operations, and non-volatile memory such as flash memory or electrically erasable programmable read-only memory (EEPROM), used to retain historical datasets, trained model parameters, and configuration metadata across operational cycles. Moreover, the memory architecture supports synchronous data exchange with the processing unit to ensure deterministic timing and coherence in real-time computation and storage workflows.
As used herein, the term “historical vehicle parameters” refers to previously collected data points that represent the operational behavior of an electric vehicle over multiple past driving cycles. The parameters serve as a foundational dataset for analyzing and modeling vehicle performance trends over time. Specifically, the historical vehicle parameters are persistently stored in the memory module and are utilized by the processing unit during the iterative retraining of data processing routines. Further, the parameters encompass time-stamped records of various subsystems, including but not limited to battery state-of-charge levels, power demand profiles, regenerative braking activity, motor current and voltage characteristics, thermal state transitions, ambient and internal temperature variations, charging patterns, and vehicle speed profiles. Furthermore, each of the parameters contributes to a comprehensive operational history, enabling the system to extract statistically significant patterns and frequency-domain attributes that reflect long-term behavioral characteristics. Moreover, the integration of historical data with real-time parameters allows the system to refine the operational profiling logic, improve predictive accuracy, and adapt to evolving usage conditions through enhanced feature extraction and adaptive quantization mechanisms.
As used herein, the term “operational cycle” refers to a complete sequence of activities states experienced by an electric vehicle during real-world usage, encompassing driving, idling, charging, regenerative braking, and environmental interactions. Specifically, each operational cycle captures the dynamic behavior of vehicle subsystems over time, defined by quantifiable real-time parameters such as, but not limited to, current, voltage, temperature, speed, torque, and battery state-of-charge. Further, the operational cycles serve as temporal windows of historical and real-time data to acquire, quantize, filter, and transform into frequency-domain representations for adaptive processing. Furthermore, the operational cycle constitutes a basis for iterative learning and refinement of data processing routines, which involves the frequency-domain attributes to be extracted from filtered parameter sequences to inform pattern recognition and profile optimization. Moreover, the operational cycles are classified into distinct types based on usage context and subsystem activity patterns, including, but not limited to, drive cycles characterized by propulsion load variations, charge cycles defined by energy inflow patterns, and hybrid cycles involving transitions between charge and discharge states under mixed operational loads. Additionally, each type of operational cycle presents unique signal characteristics and variability in sensor inputs, necessitating adaptive quantization and filtering mechanisms to ensure fidelity and robustness in operational profile generation.
As used herein, the term “processing unit” refers to a dedicated computational element configured to perform data processing routines essential for generating an operational profile of a battery pack in an electric vehicle. Specifically, the processing unit is operatively coupled to both the memory module and sensors and is responsible for executing deterministic operations on real-time and historical vehicle parameters acquired from various vehicle subsystems. Further, the processing unit applies quantization to the real-time parameters by converting the parameters into reduced-bit fixed-point representations and determining quantization parameters that include quantization intervals. Furthermore, the adaptive quantization is performed based on real-time operational state variables, enabling dynamic adjustment of data resolution according to current driving or environmental conditions. Moreover, the temporal consistency filtering is implemented through sliding-window temporal averaging, which evaluates deviations from a predefined temporal threshold to ensure signal stability before further analysis. Additionally, after filtering, the processing unit extracts frequency-domain attributes such as spectral components or power density metrics, which are critical for characterizing the dynamic behavior of the battery system under varying operational conditions. Further, the attributes are used in iterative retraining of the data processing routines stored in the memory module, allowing for continual refinement of the operational profile based on both current and historical data. Furthermore, the types of processing units suitable for this invention include microcontrollers, Digital Signal Processors (DSPs), Field-Programmable Gate Arrays (FPGAs), and System-on-Chip (SoC) architectures, each selected based on computational complexity, power efficiency, and integration requirements within the vehicle’s electronic architecture.
As used herein, the term “data processing routine” refers to a computational sequence executed by the processing unit to transform raw real-time vehicle parameters into actionable data representations used to generate an operational profile for the battery pack of an electric vehicle. Specifically, the routine consists of a structured pipeline beginning with quantization of the acquired real-time vehicle parameters, which involves numerical values to be converted into the reduced-bit fixed-point representations to enable efficient storage and computation. Further, the quantization process is refined using adaptive quantization, which involves quantization parameters, including, but not limited to, bit resolution and quantization intervals, that are dynamically selected based on real-time operational state variables of the vehicle. Subsequently, a temporal consistency filtering is applied using sliding-window temporal averaging techniques that evaluate deviations of quantized parameters against a predefined temporal threshold to ensure stability and suppress transient anomalies. Moreover, post-filtering, the routine extracts a plurality of frequency-domain attributes from the temporally smoothed data, using spectral analysis methods such as, but not limited to, discrete Fourier transforms or similar frequency decomposition techniques, enabling the identification of periodic or oscillatory patterns within operational behavior. Additionally, the data processing routine includes, but is not limited to, iterative retraining of embedded models or algorithms stored in memory, wherein the computed frequency-domain attributes derived from both current and historical vehicle data are used as input training data. Consequently, retraining operations utilize batch-wise updates with dynamically adjusted step sizes to ensure convergence and adaptation over time. Further, the types of data processing routines encompassed by the invention include quantization routines, temporal filtering routines, spectral feature extraction routines, and model retraining routines, each contributing to the accurate, efficient, and adaptive profiling of battery usage characteristics under varying vehicle operating conditions.
As used herein, the term “quantizing” refers to the process of converting continuous real-time vehicle parameters into discrete numerical values using reduced-bit fixed-point representations. In general, the quantization involves mapping a large set of input values to a smaller set, enabling efficient data storage and processing in embedded systems such as electric vehicle battery management units. Specifically, the quantization mechanism involves determining quantization parameters that include both the fixed-point representation format and associated quantization intervals. Further, the parameters are selected adaptively based on real-time operational state variables of the vehicle to ensure high fidelity of representation under dynamic driving conditions. Furthermore, the quantizing transforms high-resolution sensor data into a form that minimizes computational overhead without compromising the temporal or spectral characteristics required for downstream processing. Moreover, the types of quantization implemented in the system include, but are not limited to, uniform quantization, which involves input values being divided into equally spaced intervals, and non-uniform or adaptive quantization, which involves interval widths varying according to the statistical properties or operational states of the incoming data. Additionally, the adaptive approach ensures that precision is preserved in regions of interest while reducing resolution in less critical ranges, enabling an optimal balance between accuracy and resource usage.
As used herein, the term “temporal consistency” refers to the property of maintaining stable and coherent parameter values over time during dynamic system operations. Specifically, the temporal consistency involves enforcing statistical and structural uniformity in real-time vehicle parameters acquired from multiple subsystems. Further, the temporal consistency is achieved by applying a sliding-window temporal averaging process, and each quantized parameter is compared against a predefined temporal threshold to determine deviation. Furthermore, as the deviations exceed the threshold, parameter smoothing is triggered to ensure the continuity of data trends across consecutive time steps. Moreover, the process filters transient anomalies or abrupt fluctuations that do not reflect actual operational behavior, thus preserving the underlying temporal structure of the data. Additionally, the temporal consistency is classified into short-term consistency, which directs rapid changes within a narrow time window, and long-term consistency, thereby evaluating trends across extended operational periods. Further, the above-mentioned types are implicitly addressed through the adaptive sliding-window mechanism, which balances responsiveness with noise suppression. Furthermore, the filtered data, exhibiting strong temporal consistency, serves as the basis for reliable frequency-domain attribute extraction and iterative retraining of data processing routines, thereby enhancing the accuracy and adaptability of the operational profile generated for the battery pack.
As used herein, the term “frequency-domain attributes” refers to quantifiable characteristics of vehicle parameter signals after transformation from the time domain to the frequency domain via frequency transform techniques such as, but not limited to, the Fast Fourier Transform (FFT) or Discrete Cosine Transform (DCT). Further, the attributes serve as critical indicators of the dynamic behavior and operational patterns of the electric vehicle subsystems, including, but not limited to, the battery pack, by capturing periodicities, harmonics, and spectral energy distributions within the real-time vehicle parameters. Specifically, the frequency-domain attributes are extracted from temporally filtered and quantized real-time parameters to enable robust analysis under varying operational conditions. Moreover, the types of frequency-domain attributes include, but are not limited to, spectral magnitude, phase spectrum, power spectral density, dominant frequency components, and bandwidth concentration, each offering insight into specific operational characteristics. The spectral magnitude provides the amplitude distribution across frequencies, and power spectral density quantifies signal power per frequency band, required for detecting oscillatory patterns and anomalies. Further, the phase spectrum captures timing relationships between signal components, enabling precise reconstruction or phase-based diagnostics. Furthermore, the iterative retraining of the data processing routines leverages the attributes as training vectors, allowing continuous adaptation of operational profiles to reflect real-world usage patterns and ensuring optimal energy management and fault prediction in the battery.
As used herein, the term “reduced-bit fixed-point representation” refers to a numerical encoding scheme with real-time vehicle parameters converted into a limited number of bits using a fixed scaling factor, enabling efficient storage and processing within resource-constrained embedded systems. Specifically, the processing unit performs the conversion to minimize computational overhead and preserve critical signal characteristics necessary for subsequent data processing routines. Further, the fixed-point representation maps real-valued signals into discrete levels determined by a predefined quantization interval and a fixed number of fractional and integer bits, without relying on floating-point arithmetic. Furthermore, the reduced-bit variants specifically limit the total bit width, such as, but not limited to, 8-bit, 12-bit, or 16-bit formats, thereby constraining precision and dynamic range but enhancing memory efficiency and execution speed within the processing unit. Moreover, the types of reduced-bit fixed-point representations include, but are not limited to, uniform quantization with symmetric or asymmetric intervals, logarithmic fixed-point representation for signals with high dynamic range, and adaptive fixed-point formats where the scaling factor is dynamically selected based on real-time operational state variables of the vehicle. Additionally, the aforementioned representations enable consistent handling of real-time vehicle parameters across temporal filtering and frequency-domain feature extraction processes, forming a foundation for adaptive quantization and iterative retraining of data processing routines.
As used herein, the term “quantization parameters” refers to a set of numerical values and associated configurations used to map continuous real-time vehicle parameters into a reduced-bit fixed-point representation, enabling efficient storage, processing, and transmission within the electric vehicle’s data management architecture. Specifically, the quantization parameters comprise, but are not limited to, the reduced-bit fixed-point representations and quantization intervals, both determined by the processing unit in response to operational state variables of the vehicle. Further, the reduced-bit fixed-point representations define the number of bits allocated to represent a quantized parameter value, ensuring a controlled trade-off between precision and computational efficiency. Furthermore, quantization intervals define the step sizes or resolution levels at which input values are discretized, directly influencing the fidelity of the quantized data. Moreover, the processing unit selects and applies quantization parameters adaptively, modifying both the fixed-point representations and quantization intervals in real time according to dynamic operating conditions such as, but not limited to, speed, load, temperature, or battery state-of-charge. Additionally, the types of quantization parameters utilized include but are not limited to static quantization parameters and adaptive quantization parameters. The static quantization parameters remain constant and are configured for baseline operations, thereby providing uniform discretization across a fixed input range. Furthermore, the adaptive quantization parameters are dynamically selected or computed based on real-time operational state variables, allowing variable resolution tailored to the sensitivity or variability of the monitored signals. Furthermore, using the quantization parameters, the system reduces memory overhead and enhances the efficiency of subsequent temporal filtering and frequency-domain feature extraction operations, ensuring accurate and scalable generation of operational profiles for the battery pack.
As used herein, the term “quantization interval” refers to discrete numerical ranges used to map continuous real-time vehicle parameters into reduced-bit fixed-point representations. Generally, a quantization interval is the spacing between adjacent quantization levels in a digitized representation of analog or high-resolution data. Specifically, the quantization intervals serve as the foundational metric for compressing real-time operational data collected from the various vehicle subsystems, ensuring efficient storage and processing within constrained computational resources. Further, each quantization parameter comprises, but is not limited to, the specific reduced-bit fixed-point representation along with the associated quantization interval, enabling systematic discretization of input parameters such as, but not limited to, temperature, voltage, current, and vehicle speed. Furthermore, the quantization intervals are determined adaptively based on real-time operational state variables, ensuring alignment with varying dynamics of vehicle behavior. Moreover, the primary types of quantization intervals utilized include, but are not limited to, uniform intervals and non-uniform intervals. The uniform quantization intervals maintain constant step sizes across the range of input values and are suited for parameters with stable variance. Further, the non-uniform quantization intervals apply variable step sizes, allocating finer resolution to parameter ranges with higher sensitivity or more frequent changes, thereby preserving critical information while reducing data dimensionality. Moreover, the adaptive selection of the quantization intervals ensures that the system maintains accuracy in signal representation, along with optimizing the input for subsequent temporal filtering and frequency-domain analysis, forming a critical step in the data processing routine that governs battery operational profiling.
As used herein, the term “adaptive quantization” refers to a dynamic process of selecting quantization parameters such as, but not limited to, quantization intervals and fixed-point representations based explicitly on real-time operational state variables of the electric vehicle. Specifically, the quantization involves mapping a continuous or high-resolution signal to a discrete and lower-resolution representation to reduce data size and computational complexity. Explicitly, the adaptive quantization ensures the real-time vehicle parameters acquired from multiple subsystems are encoded with precision levels reflecting the current driving or environmental conditions, such as, but not limited to, temperature, speed, acceleration, and load demands. Further, the processing unit continuously assesses the real-time operational state variables to determine the appropriate quantization scheme, thereby enabling the system to preserve critical signal characteristics while minimizing redundant or non-informative data. Furthermore, the types of adaptive quantization employed include, but not limited to, uniform quantization with variable step sizes. The quantization intervals are adjusted according to the signal's dynamic range, and non-uniform quantization using logarithmic or linear mappings tailored to signal distributions observed in different operational states. Moreover, by implementing the adaptive schemes, the system maintains accuracy in downstream frequency-domain attribute extraction while optimizing storage and computational efficiency for real-time and historical data processing.
As used herein, the term “operational state variables” refers to measurable parameters that characterize the dynamic behavior and functional condition of an electric vehicle during operation. Specifically, the operational state variables serve as deterministic inputs for adaptive quantization and temporal filtering processes. Further, the variables are derived in real-time from multiple vehicle subsystems through sensor data acquisition and represent the current state of critical components and environmental conditions influencing battery usage. Furthermore, the operational state variables include instantaneous metrics such as, but not limited to, vehicle speed, acceleration, torque demand, motor current, battery state-of-charge (SoC), battery temperature, ambient temperature, regenerative braking status, and power output levels. Moreover, the parameters directly affect the electrical and thermal load on the battery system and are used to determine quantization parameters and filtering thresholds in the data processing routines. Additionally, the types of operational state variables include electrical variables (for instance, voltage, current, power), mechanical variables (for instance, speed, load, gear position), thermal variables (for instance, battery and ambient temperatures), and usage context variables (for instance, driving cycle classification, terrain gradient, and stop-start frequency). The system uses the variables to adaptively select quantization intervals and step sizes for retraining, ensuring that the extracted frequency-domain attributes accurately reflect current operating conditions of the vehicle and battery pack.
As used herein, the term “temporal consistency filtering” refers to a data refinement procedure that ensures the stability and reliability of time-series signals by mitigating abrupt fluctuations that do not align with expected temporal patterns. Specifically, the temporal consistency filtering is implemented to process the quantized real-time vehicle parameters, enhancing the accuracy of operational profiling for electric vehicle battery packs. Further, the filtering is achieved by employing sliding-window temporal averaging, which evaluates sequences of quantized data points over a predefined time interval. Furthermore, the deviations from a temporal threshold are computed within the sliding window, and values exceeding the threshold are adjusted to maintain continuity in the signal. The aforementioned procedure directly improves the quality of input data for downstream frequency-domain analysis and adaptive retraining of data processing routines. Moreover, the types of temporal consistency filtering include fixed-window averaging, exponential moving average, and adaptive window smoothing, each selected based on the variability of operational state variables. Additionally, the fixed-window averaging applies equal weighting across a fixed number of prior samples, and the exponential moving average assigns decreasing weights to older values, providing responsiveness to recent changes. Further, the adaptive window smoothing dynamically adjusts the window size and averaging behavior based on local signal variance, ensuring optimal temporal coherence under varying operational conditions. Furthermore, the application of temporal consistency filtering guarantees that only temporally stable and contextually valid vehicle parameters contribute to the generation of frequency-domain attributes, thereby reinforcing the robustness and learning accuracy of the retraining loop within the processing unit.
As used herein, the term “sliding-window temporal averaging” refers to a deterministic signal processing technique that smooths variations in quantized real-time vehicle parameters by computing a running average over a fixed-size or adaptive-size temporal window. Specifically, the sliding-window temporal averaging is applied as part of temporal consistency filtering to enforce stability and continuity in the processed parameter stream. Further, the sliding-window temporal averaging technique operates by systematically evaluating successive subsets of quantized data points, each corresponding to a defined time window, and replacing the current value with the average of all values within the window. Furthermore, the primary objective of the sliding-window temporal averaging is to reduce high-frequency fluctuations and eliminate transient anomalies that reflect inconsistent operational trends of the vehicle. Moreover, the technique utilizes either fixed-window or adaptive-window configurations. In the fixed-window averaging, the number of time steps or samples within the window remains constant throughout operation, providing uniform smoothing regardless of variability in input parameters. Further, in the adaptive-window averaging, the window size dynamically adjusts based on the deviation of incoming quantized data from a predefined temporal threshold, allowing the filter to adapt the responsiveness to underlying operational dynamics. Furthermore, by implementing the filtering step, the system ensures that only temporally stable parameter sequences contribute to the extraction of frequency-domain attributes, thereby enhancing the reliability of subsequent data processing routines and the retraining of models stored in the memory module.
As used herein, the term “temporal threshold” refers to a predefined limit used to assess variations in time-series data, enabling consistent filtering and detection of relevant temporal patterns in operational parameters. Specifically, the temporal threshold defines the allowable deviation of quantized real-time vehicle parameters within a specified time window, facilitating the execution of temporal consistency filtering. Further, the temporal consistency filtering employs a sliding-window averaging mechanism, with the inclusion or exclusion of parameter values in the averaging process governed by the compliance with the temporal threshold. Furthermore, a value exceeding the threshold indicates an abrupt or transient fluctuation, prompting the system to either exclude or attenuate the influence, thereby maintaining temporal smoothness and reducing noise. Moreover, the temporal thresholds are classified based on the nature of deviations monitored, including absolute thresholds, with the deviations measured against fixed numerical limits; relative thresholds, with the deviations assessed as a percentage change from previous values; and statistical thresholds, with the limits derived from standard deviations or variance metrics computed from historical data. The aforementioned threshold types enable adaptive and precise filtering strategies aligned with real-time operational state variables, ensuring that only temporally coherent data contribute to downstream processing such as frequency-domain attribute extraction and iterative retraining of data processing routines.
As used herein, the term “input training data” refers to structured datasets consumed by the processing unit for the purpose of iterative model retraining within a vehicle-based data processing system. Specifically, the input training data definitely comprises frequency-domain attributes derived from real-time vehicle parameters and historical operational data. The aforementioned parameters are acquired via the sensors coupled to the various vehicle subsystems and are subjected to quantization and the temporal consistency filtering. Further, the resulting frequency-domain attributes serve as precise input vectors representing the dynamic and spectral characteristics of vehicle operation over time. Furthermore, the input training data includes reduced-bit fixed-point representations of filtered parameters, spectral components extracted through transformation routines such as, but not limited to, Fast Fourier Transform (FFT), and temporally averaged values computed using sliding-window mechanisms. Moreover, each training data instance directly corresponds to operational cycles of the vehicle and encapsulates both transient and steady-state behaviors relevant to battery performance. Additionally, the processing unit utilizes the inputs to update data processing routine parameters, employing batch-wise retraining with dynamically adjusted step sizes to refine the accuracy and responsiveness of the system.
As used herein, the term “data processing routine parameters” refers to a definitive set of computational coefficients, thresholds, and transformation rules that govern the execution of data processing routines used in generating the operational profile of the battery pack for an electric vehicle. Further, the parameters establish the operational behavior of quantization, temporal filtering, and frequency-domain analysis algorithms applied to real-time and historical vehicle parameters. Specifically, the data processing routine parameters include quantization parameters such as, but not limited to, fixed-point bit-widths and quantization intervals, thereby dictating the resolution and scaling applied during signal compression. Furthermore, additional parameters include sliding-window sizes, temporal deviation thresholds for consistency filtering, and spectral feature extraction settings such as frequency resolution and windowing functions used during frequency-domain transformation. Moreover, during iterative retraining, the data processing routine parameters are adaptively updated based on the statistical characteristics of computed frequency-domain attributes, enabling optimization of model performance with respect to evolving vehicle dynamics. Additionally, the types of data processing routine parameters include but are not limited to static configuration parameters for initial arrangement deployment, dynamic adaptation parameters adjusted during runtime based on operational state variables, and learning parameters such as step sizes or convergence criteria used during retraining cycles. The parameters enhance the system’s ability to accurately encode, filter, and extract meaningful patterns from vehicular data streams in real time.
In accordance with a first aspect of the present disclosure, there is provided a system for generating an operational profile for a battery pack of an electric vehicle, the system comprising:
- a sensing module operatively coupled to a plurality of vehicle subsystems and the battery pack, and configured to acquire real-time vehicle parameters from the plurality of vehicle subsystems and the battery pack;
- a memory module operatively coupled to the sensing module and configured to store real-time vehicle parameters acquired from the sensing module and historical vehicle parameters collected from past operational cycles of the vehicle; and
- a processing unit operatively coupled to the memory module and configured to execute at least one data processing routine, the data processing routine comprising:
- quantizing real-time vehicle parameters received from the memory module;
- applying a temporal consistency filtering to the quantized real-time vehicle parameters; and
- extracting a plurality of frequency-domain attributes from the filtered real-time vehicle parameters,
wherein the processing unit iteratively retrains the data processing routines within the memory module based on the plurality of frequency-domain attributes extracted from the real-time vehicle parameters and the historical vehicle parameters.
Referring to figure 1, in accordance with an embodiment, there is described a system 100 for generating an operational profile for a battery pack 102 of an electric vehicle 104 is described. The system 100 comprises a sensing module 106 operatively coupled to a plurality of vehicle subsystems 108 and configured to acquire real-time vehicle parameters from the plurality of vehicle subsystems 108. Further, the system 100 comprises a memory module 110 operatively coupled to the sensing module 106 and configured to store real-time vehicle parameters acquired from the sensing module 106 and historical vehicle parameters collected from past operational cycles of the vehicle 104. Furthermore, the system 100 comprises a processing unit 112 operatively coupled to the memory module 110 and configured to execute at least one data processing routine. Moreover, the data processing routine comprises quantizing real-time vehicle parameters received from the memory module 110. Additionally, the data processing routine comprises applying a temporal consistency filtering to the quantized real-time vehicle parameters. Subsequently, the data processing routine comprises extracting a plurality of frequency-domain attributes from the filtered real-time vehicle parameters. Consequently, the processing unit 112 is configured to iteratively retrain the data processing routines within the memory module 110 based on the plurality of frequency-domain attributes extracted from the real-time vehicle parameters and the historical vehicle parameters.
The system 100 for generating an operational profile for the battery pack 102 of the electric vehicle 104 initiates operation by acquiring real-time vehicle parameters through the sensing module 106 operatively coupled to the plurality of vehicle subsystems 108. Further, the acquired real-time vehicle parameters are stored in the memory module 110, along with historical vehicle parameters from past operational cycles. Furthermore, the processing unit 112 receives the stored vehicle parameters and executes at least one data processing routine to enable efficient and structured profiling. The data processing routine is initiated by the quantization of real-time vehicle parameters using a plurality of reduced-bit fixed-point representations, compressing signal data and simultaneously preserving critical information within defined quantization intervals. Moreover, the quantization parameters are selected based on operational state variables, allowing adaptive transformation of input data into memory-efficient representations suitable for high-speed processing. Further, following quantization, the processing unit 112 applies temporal consistency filtering using a sliding-window temporal averaging method. The aforementioned filtering process ensures short-term fluctuations or transient noise in the quantized parameters are smoothed according to a predefined temporal threshold, improving signal stability and reliability. Furthermore, the filtered parameters are subjected to frequency-domain transformation, as a plurality of frequency-domain attributes are extracted. Moreover, the frequency-domain attributes represent the spectral behavior of the vehicle parameters over time and provide a robust feature space for operational profile analysis. Additionally, the extraction of frequency-domain attributes enables the system 100 to capture periodic patterns, anomalies, and dynamic load signatures critical for understanding battery usage and performance trends. Consequently, the technical effect of the system 100 includes enhanced data compression, improved temporal signal stability, and detailed spectral characterization of vehicle operational behavior. Further, the use of quantized fixed-point representations reduces memory usage and computational overhead and maintains signal fidelity. Furthermore, the temporal filtering ensures consistency in parameter streams, minimizing the influence of outliers or short-term disturbances. Moreover, the frequency-domain attributes enable high-resolution pattern recognition, supporting predictive modeling and real-time diagnostics. Additionally, the advantages of the system 100 include, but are not limited to, efficient resource utilization, scalability across different vehicle platforms, and the ability to perform iterative retraining of data processing routines using current and historical frequency-domain data, thereby enabling adaptive learning and continuous optimization of the operational profile with minimal manual intervention.

In an embodiment, the processing unit 112 is configured to convert the real-time vehicle parameters into a plurality of reduced-bit fixed-point representations. Specifically, each of the real-time vehicle parameters, acquired from the sensing module 106 coupled to various vehicle subsystems 108, undergoes numerical transformation using a fixed number of bits to represent the magnitude of the vehicle parameter, within a defined quantization range. Further, the reduced-bit fixed-point format eliminates the need for floating-point operations, thereby lowering computational complexity and improving processing speed. Furthermore, the conversion process preserves the essential characteristics of the original signals and aligns the resolution and numerical limits with the operational needs of the system 100. The bit-width selection balances precision and memory constraints, enabling consistent signal representation across variable operational states. Moreover, the process of applying reduced-bit fixed-point representations involves predefined scaling and rounding techniques to map continuous and high-resolution digital values into discrete levels. Additionally, the processing unit 112 determines the appropriate scaling factors based on the expected operating range of each parameter and executes the conversion before any downstream operations, such as, but not limited to, filtering or frequency-domain analysis. The fixed-point data format supports efficient arithmetic operations within the processing unit 112, particularly in embedded environments such as hardware resources are constrained. Further, by performing quantization early in the data flow, the system 100 ensures uniformity in data representation, thereby reducing variance across sensor inputs and enabling consistent processing logic. Consequently, using the reduced-bit fixed-point representations includes, but is not limited to, minimized memory usage, faster arithmetic operations, and enhanced predictability in execution time. Further, the lack of floating-point variability supports deterministic computation, thereby aiding real-time systems in the electric vehicle 104. Furthermore, the advantages of the plurality of reduced-bit fixed-point representations are improved energy efficiency of the processing unit 112, reduced latency in data processing routines, and the ability to scale the system 100 across different vehicle configurations without extensive hardware modifications. Moreover, the aforementioned approach also supports robust integration with adaptive quantization and filtering stages by ensuring all intermediate data remains within a constrained numerical domain, facilitating high-speed, low-power computational workflows for battery profile generation.
In an embodiment, the processing unit 112 is configured to determine a plurality of quantization parameters, wherein each quantization parameter comprises the reduced-bit fixed-point representations and quantization intervals. Specifically, the quantization interval defines the resolution or step size used to map a continuous signal into discrete fixed-point levels. Further, the selection of quantization parameters aligns with the dynamic range and variance of each vehicle parameter, ensuring accurate encoding within the constraints of the reduced-bit precision. Furthermore, the processing unit 112 applies the quantization parameters prior to any signal analysis, enforcing a standardized data format that maintains signal integrity and optimizes storage and processing efficiency. Moreover, the process of determining quantization parameters involves statistical evaluation of the signal distributions received from the sensing module 106. Further, the processing unit 112 analyzes historical vehicle parameters stored in the memory module 110 to establish baseline characteristics for each signal type. Furthermore, based on the aforementioned analysis, the processing unit 112 selects bit-widths for fixed-point representation and computes optimal quantization intervals that minimize quantization errors. Moreover, the procedure accounts for signal volatility, sensor noise margins, and operational thresholds specific to vehicle subsystem 108. Additionally, once established, the quantization parameters are applied uniformly across incoming real-time parameters to ensure deterministic signal behavior during downstream operations, such as, but not limited to, temporal filtering and frequency-domain attribute extraction. Consequently, determining the quantization parameters with explicit control over fixed-point formats and intervals enables reduced quantization noise, improved signal consistency, and enhanced compatibility with resource-constrained processing environments. Additionally, the advantages of determining quantization parameters include, but are not limited to, improved data throughput due to lower memory bandwidth requirements, increased processing efficiency through fixed-width data operations, and reduced algorithmic complexity in the processing unit 112. Further, the system 100 supports scalable deployment by allowing quantization parameters to be tailored for different vehicle platforms or operational scenarios without modifying the core data processing routines. Furthermore, the structured quantization framework ensures reproducible data behavior and supports high-fidelity analysis within the frequency domain, directly contributing to the generation of an accurate and computationally efficient operational profile for the battery pack 102.
In an embodiment, the processing unit 112 is configured to apply adaptive quantization by selecting the quantization parameters based on real-time operational state variables of the vehicle. Specifically, each operational state variable, such as but not limited to vehicle speed, torque demand, ambient temperature, or battery current, is continuously monitored and used as an input condition for determining the most suitable quantization configuration. Further, the selection process involves mapping operational states to the predefined quantization profiles or computing quantization intervals and bit-widths in real time, based on current signal dynamics. Furthermore, the adaptive mechanism ensures the signals with high variability or critical operational importance are represented with finer resolution. In contrast, more stable or less critical signals are assigned to coarser quantization levels to conserve computational and memory resources. Moreover, the algorithm of implementing adaptive quantization involves real-time analysis of incoming vehicle parameters and contextual interpretation of vehicle operating conditions. Additionally, the processing unit 112 accesses operational state variables directly from sensor inputs or derived metrics and applies a decision algorithm to select optimal quantization parameters. Further, the algorithm adjusts the bit-width and the interval definitions for the fixed-point representation in response to shifts in vehicle dynamics, such as but not limited to transitions from idle to acceleration or from urban to highway driving. Further, the system 100 enforces immediate reconfiguration of quantization parameters, ensuring seamless data processing without interrupting the flow of signal transformation, filtering, or frequency-domain extraction. Consequently, applying the adaptive quantization enables improved data representation accuracy under varying operating conditions, enhanced flexibility in signal processing, and optimized resource utilization. Further, the advantages of adaptive quantization include reduction of quantization error during high-load or high-variability scenarios, preservation of critical signal features during dynamic transitions, and avoidance of unnecessary precision in low-impact states, resulting in lower memory and processing overhead. Furthermore, the adaptive approach ensures consistent performance of data processing routines across a wide range of operational environments and enables the system to maintain high fidelity in the generation of the battery pack's 102 operational profiles under both nominal and extreme conditions.
In an embodiment, the processing unit 112 is configured to apply the temporal consistency filtering by employing a sliding-window temporal averaging based on a deviation of the quantized parameters with a predefined temporal threshold. Specifically, each sliding window operates over a fixed or adaptively sized sequence of quantized data points, capturing short-term variations and maintaining temporal continuity. Further, the filtering mechanism calculates the average value within the window and evaluates deviations of each quantized parameter against the temporal threshold to determine whether the parameter represents a consistent trend or a transient fluctuation. Furthermore, the parameters exceeding the temporal threshold are either smoothed or suppressed to ensure that only temporally stable information is retained for further processing. Moreover, the process involves initializing a temporal buffer for each quantized parameter stream, and the sliding window is continuously updated with incoming data. Further, for each window shift, the processing unit 112 computes a local average and compares each parameter’s deviation from the aforementioned average against the predefined threshold. The predefined temporal threshold is selected based on expected noise levels, system dynamics, and signal volatility specific to each vehicle subsystem 108. Furthermore, the parameters falling within the threshold are accepted as consistent and passed forward in the processing chain, and the parameters outside the limit are filtered or weighted accordingly. The entire filtering operation executes in real time, ensuring immediate stabilization of fluctuating data before frequency-domain attribute extraction or retraining processes. Consequently, applying temporal consistency filtering includes, but is not limited to, suppression of transient noise, enhancement of signal reliability, and reinforcement of stable behavioral patterns in the vehicle parameter streams. Additionally, the advantages include, but are not limited to, improved accuracy in downstream frequency-domain analysis due to the elimination of short-term irregularities, increased robustness in operational profile generation, and enhanced model generalization during retraining. Further, the sliding-window approach ensures high responsiveness to genuine parameter shifts. Furthermore, the resulting filtered data stream serves as a stable and consistent input for spectral processing, enabling precise and computationally efficient tracking of battery performance under varying operational conditions.
In an exemplary embodiment, the sliding window comprises the last 10 temperature readings: [28.0, 28.3, 28.1, 28.2, 28.4, 28.2, 28.3, 28.1, 28.2] (in °C). In order to estimate the temporal average, a mean is calculated: Mean = (sum of readings) / 10 = 28.19°C. Further, upon receiving a new temperature reading, the next sensor value is 33.5°C. Herein, a deviation check is performed using the difference between the new reading and the mean, as Difference = |33.5 - 28.19| = 5.31°C. Simultaneously, the predefined threshold is determined, and for instance, the threshold is set at 2°C. A filtering decision is reached upon comparison of the threshold with the difference. Since 5.31°C is greater than 2°C, the new reading is treated as a deviation.
In an embodiment, the processing unit 112 is configured to compute the plurality of frequency-domain attributes based on the temporal consistency filtered vehicle parameters. Specifically, each filtered parameter stream undergoes a transformation from the time domain to the frequency domain using analytical techniques such as, but not limited to, Fast Fourier Transform or equivalent spectral decomposition methods. Further, the transformation extracts key frequency components, including dominant frequencies, spectral energy distributions, and harmonic patterns that characterize the operational behavior of the associated vehicle subsystems 108. Furthermore, the frequency-domain attributes reflect the dynamic characteristics of real-time vehicle parameters over time, providing a multidimensional representation of signal behavior relevant to battery pack 102 usage and stress conditions. Moreover, the algorithm initiates by segmenting the filtered parameter data into analysis windows synchronized with the operational cycle of the vehicle 104. Additionally, for each segment, the processing unit 112 applies the selected transformation method to derive spectral coefficients and statistical measures such as spectral centroid, bandwidth, and peak amplitude. Further, the frequency-domain attributes are computed with precision and mapped to feature vectors for use in data modeling, retraining, and profile generation. The computation process accounts for signal normalization and windowing to minimize spectral leakage and ensure accurate frequency resolution. Furthermore, the extracted attributes are stored in the memory module 110 as structured data, linked to both real-time inputs and corresponding historical profiles, forming the basis for condition-aware modeling of battery pack 102 performance. Consequently, computing frequency-domain attributes includes, but is not limited to, improved visibility into long-term behavioral trends, identification of hidden cyclic patterns, and differentiation of normal versus anomalous operational states. Further, the advantages of using frequency-domain attributes include, but are not limited to, enhanced feature richness for training data models, improved sensitivity to operational stress indicators, and refined accuracy in the generation of battery pack 102 operational profiles. Furthermore, the frequency-domain representation complements time-domain analysis by revealing signal properties that remain undetectable in raw or filtered temporal data. The aforementioned approach supports real-time diagnostics, predictive analytics, and adaptive learning by supplying the processing unit 112 with high-fidelity, spectrally-informed insights into the functional dynamics of the electric vehicle subsystems 108.
In an embodiment, the processing unit 112 is configured to employ computed frequency-domain attributes of the filtered vehicle parameter sequences as input training data for iterative retraining of the data processing routines. Specifically, each frequency-domain attribute set, derived from temporally consistent and quantized vehicle parameters, is formatted into a structured feature vector. The feature vectors serve as inputs to machine learning models embedded within the data processing routines, enabling continuous refinement of internal parameters. Further, the retraining process uses both current real-time frequency-domain attributes and historical data stored in the memory module 110 to enhance the accuracy and adaptability of the system 100 in generating operational profiles. Furthermore, the method of retraining begins by aggregating batches of frequency-domain attributes over defined operational intervals. The processing unit 112 initializes a training cycle where existing model parameters are adjusted through optimization algorithms such as stochastic gradient descent or equivalent techniques. Moreover, the training data consists of labeled or unlabeled frequency-domain patterns associated with known operational outcomes or previously recorded battery behaviors. Additionally, during each iteration, the model updates internal weights and thresholds based on observed discrepancies between predicted and actual performance indicators. Further, the retraining mechanism operates in a looped sequence, enabling continuous learning from incoming vehicle data without requiring manual intervention or offline recalibration. Consequently, employing frequency-domain attributes for iterative retraining includes, but is not limited to, improved model accuracy, enhanced adaptability to evolving vehicle conditions, and sustained alignment between real-world operations and predictive analytics. Further, the advantages include the ability to personalize data processing routines to individual vehicle usage patterns, increased robustness in profile generation under variable operating states, and long-term system resilience through ongoing self-optimization. Furthermore, the integration of spectral features into the learning process enables the system 100 to capture subtle shifts in operational behavior, detect degradation trends, and maintain high fidelity in profiling outputs. Moreover, the retraining process ensures that the operational profile remains relevant, precise, and reflective of actual usage across the vehicle’s lifecycle.
In an embodiment, the processing unit 112 iteratively updates data processing routine parameters via batches of computed frequency-domain attributes with dynamically adjusted step sizes. Specifically, each batch comprises a sequence of frequency-domain attribute vectors derived from filtered vehicle parameter streams over specified time windows. The processing unit 112 evaluates the batches using model parameters and applies optimization routines that adjust internal coefficients, thresholds, and transformation rules. Further, the step size for each parameter update is dynamically adjusted based on convergence behavior, variance in incoming data, and rate of model improvement, ensuring that the system 100 adapts effectively without overfitting or instability. Furthermore, the method for updating routine parameters involves initialization of an adaptive learning loop, where the processing unit 112 computes parameter gradients from error functions or deviation metrics. Moreover, for each batch, the system 100 assesses the influence of the frequency-domain attributes on model performance and modulates the learning step size based on predefined criteria such as gradient magnitude, historical performance metrics, or operational confidence intervals. Additionally, smaller step sizes are assigned when the model approaches convergence; alternatively, larger step sizes are used in early or uncertain phases to accelerate adaptation. The processing unit 112 applies the updated parameters to the data processing routines in real time, maintaining synchronization between the retraining process and ongoing vehicle data acquisition. Consequently, updating data processing routine parameters with dynamically adjusted step sizes includes increased training efficiency, stabilized convergence behavior, and higher model precision. Further, the advantages include faster adaptation to shifts in vehicle operational states, reduced computational load through controlled update magnitudes, and sustained performance across varying driving conditions. The use of frequency-domain attributes as structured inputs enhances the relevance of each update cycle, while dynamic step size modulation ensures optimal balance between learning speed and model stability. The aforementioned mechanism enables the system 100 to retain long-term adaptability while minimizing retraining overhead, directly contributing to the generation of a responsive and reliable operational profile for the battery pack 102.
In accordance with a second aspect, there is described a method for generating an operational profile for a battery pack of an electric vehicle, the method comprising:
- acquiring real-time vehicle parameters from a plurality of vehicle subsystems, via a sensing module;
- executing at least one data processing routine, via a processing unit;
- applying adaptive quantization by selecting the quantization parameters based on real-time operational state variables of the vehicle, via the processing unit;
- applying temporal consistency filtering by employing a sliding-window temporal averaging, via the processing unit; and
- extracting a plurality of frequency-domain attributes from the filtered real-time vehicle parameters, via the processing unit.
Referring to figure 2, in accordance with an embodiment, there is described a method 200 for generating an operational profile for a battery pack 102 of an electric vehicle 104. At step 202, the method 200 comprises acquiring real-time vehicle parameters from a plurality of vehicle subsystems 108 via a sensing module 106. At step 204, the method 200 comprises executing at least one data processing routine, via a processing unit 112. At step 206, the method 200 comprises applying adaptive quantization by selecting the quantization parameters based on real-time operational state variables of the vehicle 104, via the processing unit 112. At step 208, the method 200 comprises applying temporal consistency filtering by employing a sliding-window temporal averaging, via the processing unit 112. At step 210, the method 200 comprises extracting a plurality of frequency-domain attributes from the filtered real-time vehicle parameters via the processing unit 112.
In an embodiment, the method 200 comprises iteratively retraining the data processing routines within the memory module 110 based on the plurality of frequency-domain attributes, via the processing unit 112.
In an embodiment, the method 200 comprises generating an operational profile for a battery pack 102 of an electric vehicle 104. Furthermore, the method 200 comprises acquiring real-time vehicle parameters from a plurality of vehicle subsystems 108, via a sensing module. Furthermore, the method 200 comprises executing at least one data processing routine, via a processing unit 112. Furthermore, the method 200 comprises applying adaptive quantization by selecting the quantization parameters based on real-time operational state variables of the vehicle, via the processing unit 112. Furthermore, the method 200 comprises applying temporal consistency filtering by employing a sliding-window temporal averaging, via the processing unit 112. Furthermore, the method 200 comprises extracting a plurality of frequency-domain attributes from the filtered real-time vehicle parameters, via the processing unit 112. Furthermore, the method 200 comprises iteratively retraining the data processing routines within the memory module 110 based on the plurality of frequency-domain attributes.
The system for generating an operational profile for a battery pack of an electric vehicle, as described in the present disclosure, is advantageous in terms of producing highly accurate operational profiles for electric vehicle battery packs. Further, the approach reduces computational and memory overhead by employing reduced-bit fixed-point representations and optimized quantization parameters while preserving critical signal integrity.
It would be appreciated that all the explanations and embodiments of the system 100 also apply mutatis-mutandis to the method 200.
In the description of the present invention, it is also to be noted that, unless otherwise explicitly specified or limited, the terms “disposed,” “mounted,” and “connected” are to be construed broadly, and may for example be fixedly connected, detachably connected, or integrally connected, either mechanically or electrically. They may be connected directly or indirectly through intervening media, or they may be interconnected between two elements. The specific meaning of the above terms in the present invention can be understood in specific cases to those skilled in the art.
Modifications to embodiments and combinations of different embodiments of the present disclosure described in the foregoing are possible without departing from the scope of the present disclosure as defined by the accompanying claims. Expressions such as “including”, “comprising”, “incorporating”, “have”, and “is” used to describe and claim the present disclosure are intended to be construed in a non-exclusive manner, namely allowing for items, components or elements not explicitly described also to be present. Reference to the singular is also to be construed to relate to the plural where appropriate.
Although embodiments have been described with reference to a number of illustrative embodiments thereof, it should be understood that numerous other modifications and embodiments can be devised by those skilled in the art that will fall within the scope of the principles of this disclosure. More particularly, various variations and modifications are possible in the component parts and/or arrangements of the subject combination arrangement within the scope of the present disclosure, the drawings, and the appended claims. In addition to variations and modifications in the component parts and/or arrangements, alternative uses will also be apparent to those skilled in the art.
,CLAIMS:WE CLAIM:
1. A system (100) for generating an operational profile for a battery pack (102) of an electric vehicle (104), the system (100) comprising:
- a sensing module (106) operatively coupled to a plurality of vehicle subsystems (108) and the battery pack (102), and configured to acquire real-time vehicle parameters from the plurality of vehicle subsystems (108) and the battery pack (102);
- a memory module (110) operatively coupled to the sensing module (106) and configured to store real-time vehicle parameters acquired from the sensing module (106) and historical vehicle parameters collected from past operational cycles of the electric vehicle (104); and
- a processing unit (112) operatively coupled to the memory module (110) and configured to execute at least one data processing routine, the data processing routine comprising:
- quantizing real-time vehicle parameters received from the memory module (110);
- applying a temporal consistency filtering to the quantized real-time vehicle parameters; and
- extracting a plurality of frequency-domain attributes from the filtered real-time vehicle parameters,
wherein the processing unit (112) iteratively retrains the data processing routines within the memory module (110) based on the plurality of frequency-domain attributes extracted from the real-time vehicle parameters and the historical vehicle parameters.

2. The system (100) as claimed in claim 1, wherein the processing unit (112) is configured to convert the real-time vehicle parameters into a plurality of reduced-bit fixed-point representations.

3. The system (100) as claimed in claim 1, wherein the processing unit (112) is configured to determine a plurality of quantization parameters, wherein each quantization parameters comprise the reduced-bit fixed-point representations and quantization intervals.

4. The system (100) as claimed in claim 1, wherein the processing unit (112) is configured to apply adaptive quantization by selecting the quantization parameters based on real-time operational state variables of the vehicle.

5. The system (100) as claimed in claim 1, wherein the processing unit (112) is configured to apply the temporal consistency filtering by employing a sliding-window temporal averaging based on a deviation of the quantized parameters with a predefined temporal threshold.

6. The system (100) as claimed in claim 5, wherein the processing unit (112) is configured to compute the plurality of frequency-domain attributes based on the temporal consistency filtered vehicle parameters.

7. The system (100) as claimed in claim 6, wherein the processing unit (112) is configured to employ computed frequency-domain attributes of the filtered vehicle parameter sequences as input training data for iterative retraining of the data processing routines.

8. The system (100) as claimed in claim 7, wherein the processing unit (112) iteratively updates data processing routine parameters via batches of computed frequency-domain attributes with dynamically adjusted step sizes.

9. A method (200) for generating an operational profile for a battery pack (102) of an electric vehicle (104), the method (200) comprising:
- acquiring real-time vehicle parameters from a plurality of vehicle subsystems (108), via a sensing module (106);
- executing at least one data processing routine, via a processing unit (112);
- applying adaptive quantization by selecting the quantization parameters based on real-time operational state variables of the vehicle (104), via the processing unit (112);
- applying temporal consistency filtering by employing a sliding-window temporal averaging, via the processing unit (112); and
- extracting a plurality of frequency-domain attributes from the filtered real-time vehicle parameters, via the processing unit (112).