Battery health state online evaluation system considering temperature-current coupling effect

Abstract

The invention belongs to the field of battery monitoring, and particularly relates to a battery health state on-line evaluation system considering a temperature-current coupling effect, which comprises an acquisition module, a coupling module, an on-line evaluation module and an early warning module, the acquisition frequency and the window length of current and temperature data are adaptively adjusted through reinforcement learning and a federated frame in the acquisition module; the coupling module constructs a consistency interval based on the current and temperature slope sequence, calculates a segmented current thermal stress factor and a segmented temperature gradient factor, and further generates a temperature-current coupling feature vector; the online evaluation module performs health state evaluation and life decay rate calculation in combination with dynamic weight distribution and principal component analysis; the early warning module gives an alarm in a grading manner according to an evaluation result, and feeds back an attenuation rate to the front-end acquisition module and the online evaluation module to realize closed-loop optimization; according to the method, the accuracy and reliability of battery health state evaluation of the electric two-wheeled vehicle and the tricycle under complex working conditions are effectively improved.

Description

Technical Field

[0001] This invention belongs to the field of battery monitoring, and particularly relates to an online battery health status assessment system that takes into account the temperature-current coupling effect. Background Technology

[0002] Currently, electric two-wheelers and three-wheelers widely use low-cost and technologically mature lead-acid or lithium-ion batteries as their power source. However, due to the general lack of advanced battery management systems (BMS) in these vehicles, the accuracy of online assessment of their State of Health (SOH) is insufficient. In actual operation, the battery is constantly subjected to high-rate, non-stable charging and discharging current surges caused by frequent start-stop cycles, acceleration uphill, and rapid full charging. This causes severe heat generation in the battery cells, forming a strong temperature-current coupling effect with the complex ambient temperature. This effect accelerates internal chemical side reactions in the battery, causing irreversible aging and capacity decay, and even leading to the risk of thermal runaway. Traditional assessment methods often ignore this dynamic coupling relationship, relying only on voltage or a single temperature point for rough estimation. This results in significant deviations in SOH assessment results throughout the vehicle's entire operating lifespan, failing to accurately predict sudden drops in battery capacity and the end of its lifespan, affecting users' range predictions and creating potential safety hazards. Summary of the Invention

[0003] To address the shortcomings of existing technologies, this invention proposes an online battery health status assessment system considering the temperature-current coupling effect. This system includes a data acquisition module, a coupling module, an online assessment module, and an early warning module. The acquisition module adaptively adjusts the acquisition frequency and window length of current and temperature data using reinforcement learning and a federated framework. The coupling module constructs a consistency interval based on the current and temperature slope sequences and calculates piecewise current thermal stress factors and piecewise temperature gradient factors, thereby generating a temperature-current coupling feature vector. The online assessment module combines dynamic weight allocation and principal component analysis to assess the health status and calculate the lifespan decay rate. The early warning module issues tiered alarms based on the assessment results and feeds back the lifespan decay rate to the front-end acquisition module and the online assessment module, achieving closed-loop optimization. This invention effectively improves the accuracy and reliability of battery health status assessment for electric two-wheelers and three-wheelers under complex operating conditions.

[0004] To achieve the above objectives, the present invention provides the following technical solution:

[0005] An online battery health status assessment system considering temperature-current coupling effects includes: a data acquisition module, a coupling module, an online assessment module, and an early warning module;

[0006] The acquisition module is used to acquire the battery's current and temperature data in real time according to a preset enhanced acquisition adjustment model.

[0007] The coupling module is used to obtain a temperature-current coupling feature vector based on the current and temperature data combined with a coupling algorithm.

[0008] The online evaluation module is used to perform real-time evaluation based on the temperature-current coupling feature vector and a preset online evaluation model to obtain the battery health status evaluation value and the battery life decay rate.

[0009] The early warning module is used to issue graded early warnings based on the battery health status assessment value and the set graded early warning thresholds. At the same time, it feeds back the battery life decay rate to the enhanced acquisition adjustment model in the acquisition module and the online evaluation model to adjust the data acquisition strategy and evaluation index weights in real time.

[0010] Specifically, the data acquisition module includes a curve construction unit and a fluctuation analysis unit;

[0011] The curve construction unit is used to combine current and temperature data from different historical scenarios with a dual-axis curve, with time as the horizontal axis, current as the left vertical axis, and temperature as the right vertical axis, to construct current change curves and temperature change curves.

[0012] The fluctuation analysis unit obtains current slope sequences and temperature slope sequences based on current change curves and temperature change curves; and constructs an acquisition adjustment factor based on the ratio of current slope ratio to temperature slope ratio at adjacent time points.

[0013] Simultaneously, based on the current change curve and the temperature change curve, the length of the synchronous time interval between the current slope and the temperature slope is obtained through time series analysis algorithm as the initial acquisition window length, and the data acquisition frequency of the corresponding time interval is used as the initial acquisition frequency.

[0014] Specifically, the acquisition module also includes a model training unit;

[0015] The model training unit, based on the initial acquisition window length, initial acquisition frequency, acquisition adjustment factor, and the reciprocal of the battery life decay rate, combines a distributed reinforcement learning algorithm with corresponding current and temperature data from different historical scenarios to train the acquisition parameters and obtain a trained reinforcement acquisition adjustment model.

[0016] Specifically, the acquisition module also includes an acquisition adjustment unit;

[0017] The acquisition and adjustment unit is used to configure the trained enhanced acquisition and adjustment model into the corresponding sensor in conjunction with the federated framework, and to adjust the length of the differential current and temperature data acquisition window and the acquisition frequency by combining the scene fingerprint constructed by the current slope ratio and temperature slope ratio at adjacent time points with the acquisition adjustment factor.

[0018] The sensor includes a current sensor, a temperature sensor group, and a voltage sensor; the temperature sensor group is configured to acquire the ambient temperature at at least one location on the battery surface and to acquire the temperature at a reference point inside the battery via a temperature sensor attached to the battery terminal.

[0019] Specifically, the coupling module includes a preprocessing unit;

[0020] The preprocessing unit is used to monitor the consistency relationship between current and temperature changes in real time based on a piecewise consistency function constructed by combining the current slope and temperature slope with the convolution kernel function. Based on the consistency relationship between current and temperature changes in different time periods, combined with a preset local filtering model library and the time points and ranges of abnormal consistency fluctuations, the unit calls the local filtering model in the corresponding time period to perform data filtering processing, thereby obtaining the preprocessed current and temperature data for each consistency interval.

[0021] Specifically, the coupling module also includes a current parameter unit and a temperature parameter unit;

[0022] The current parameter unit is used to calculate the integral of the product of the square of the current and the resistance and time per unit time based on the current data of each consistent interval after preprocessing, so as to obtain the segmented current thermal stress factor.

[0023] The temperature parameter unit is used to obtain the piecewise temperature gradient factor based on the fluctuation of the difference between the average ambient temperature and the average internal reference point temperature within the corresponding time length of each consistency interval.

[0024] Specifically, the coupling module further includes a hierarchical coupling unit;

[0025] The hierarchical coupling unit is used to obtain a segmented temperature-current coupling feature vector by combining the ratio of segmented temperature gradient factor to segmented current thermal stress factor, average ambient temperature, average internal reference point temperature, battery life decay rate, and the ratio of temperature slope ratio to current slope ratio with a segmented coupling model. At the same time, it performs real-time calibration of the segmented interval length of the segmented temperature-current coupling feature vector by the battery life decay rate.

[0026] Specifically, the online evaluation module includes a contribution analysis unit and a segmented evaluation unit;

[0027] The contribution analysis unit is used to obtain the main evaluation factor and auxiliary evaluation factor and their corresponding contribution for each segment based on the segmented temperature-current coupling feature vector, with the battery life decay probability as the target, combined with principal component analysis.

[0028] The segmented evaluation unit is used to evaluate the battery health status based on the main evaluation factor and auxiliary evaluation factor and their corresponding contribution in each segment, combined with a comprehensive evaluation algorithm, to obtain the segmented battery health status evaluation value. Based on the fluctuation value of the product of the battery health status evaluation value and the ratio of the current thermal stress factor to the temperature gradient factor, the battery life decay rate is obtained.

[0029] Specifically, the early warning module includes tiered early warning units;

[0030] The graded early warning unit is used to compare the battery health status assessment value with the preset multi-level early warning threshold in real time.

[0031] When the battery health status assessment value is lower than the first threshold, a first-level warning signal is generated and output. The first-level warning signal includes a battery capacity decay prompt and an estimated remaining driving range calculated based on the current battery level and the battery life decay rate.

[0032] When the battery health status assessment value is lower than the preset second threshold, a second-level warning signal is generated and output. The second-level warning signal is a warning of the end of battery life and a replacement reminder. The health status assessment value, the warning signal and the estimated remaining driving range are transmitted to the vehicle display screen or the user's mobile terminal APP for display.

[0033] Specifically, the early warning module also includes a feedback correction unit;

[0034] The feedback correction unit is used to receive the battery life degradation rate from the online evaluation module in real time, and transmit the battery life degradation rate as a feedback signal to the enhanced acquisition adjustment model in the acquisition module to dynamically adjust the acquisition window length and acquisition frequency in the data acquisition strategy; at the same time, the battery life degradation rate is transmitted to the hierarchical coupling unit to adaptively correct the segment interval length of the segmented temperature-current coupling feature vector and adjust the segmented battery health status assessment value in real time.

[0035] Compared with the prior art, the beneficial effects of the present invention are:

[0036] This invention addresses the shortcomings of existing technologies by employing an intelligent acquisition mechanism that combines reinforcement learning and federated learning. This mechanism enables adaptive high-frequency capture and optimized resource allocation of current and temperature data, effectively ensuring the integrity of critical coupled data. Utilizing multi-dimensional slope analysis and consistency interval partitioning techniques, it accurately extracts dynamic feature vectors reflecting the internal electrothermal coupling effect of the battery, providing high-value input for health status assessment. Principal component analysis and dynamic weight allocation algorithms enhance the accuracy and reliability of assessment results under different operating conditions. By establishing a feedback loop for lifespan degradation rate, it achieves adaptive optimization throughout the entire process from data acquisition and feature extraction to assessment and early warning. Ultimately, this significantly improves the accuracy and real-time performance of online battery health status assessment in complex usage scenarios and provides users with precise battery maintenance decision support. Attached Figure Description

[0037] Figure 1 This is a block diagram of the online battery health status assessment system considering the temperature-current coupling effect of the present invention;

[0038] Figure 2 The flowchart shows the unit operation of the online battery health status assessment system that takes into account the temperature-current coupling effect of this invention. Detailed Implementation

[0039] Please see Figure 1 One embodiment of the present invention provides an online battery health status assessment system that considers the temperature-current coupling effect, applied to the assessment of batteries for electric two-wheeled or three-wheeled vehicles, comprising: a data acquisition module, a coupling module, an online assessment module, and an early warning module.

[0040] The acquisition module is used to acquire the battery's current and temperature data in real time according to a preset enhanced acquisition adjustment model.

[0041] It should be further noted that the acquisition module in this embodiment includes a curve construction unit, a fluctuation analysis unit, a model training unit, and an acquisition adjustment unit;

[0042] The curve construction unit is used to combine current and temperature data from different historical scenarios with a dual-axis curve, with time as the horizontal axis, current as the left vertical axis, and temperature as the right vertical axis, to construct current change curves and temperature change curves.

[0043] It should be further explained that, in this embodiment, the electric vehicle operates with multiple physical parameters such as current and temperature, which have different dimensions and large differences in numerical range, making it difficult to compare and analyze the changing trends synchronously on the same scale. Therefore, a dual-axis curve is constructed to achieve coordinated visualization of current and temperature changes.

[0044] It should be further explained that the process of constructing the biaxial curve in this embodiment includes:

[0045] Based on the time series data of various scenarios stored in the historical database, the current sampling value sequence of the current sensor and the temperature sampling value sequence of the temperature sensor are extracted respectively.

[0046] Based on the high-precision timing identifier carried in each sensor data packet, the precise acquisition timestamp of each current data point and temperature data point is parsed and extracted.

[0047] Using the system's global minimum sampling period as the basic time unit, a continuous, equally spaced, unified time axis is generated from the data start time to the end time.

[0048] The original timestamps of current and temperature data points are mapped to this unified time axis. If the original data exists at a specific time point on the unified time axis, the data point is directly included in the synchronized dataset.

[0049] For time points where there are no corresponding original data points on a unified time axis, a data compensation mechanism is activated. An interpolation function is constructed using the Lagrange interpolation algorithm. The values ​​of multiple effective current or temperature data points adjacent to the missing time point are used for interpolation calculation to generate the current interpolation data and temperature interpolation data for that moment.

[0050] By combining the above direct mapping and interpolation compensation, a time-synchronized current data point set and temperature data point set that are fully aligned and complete without any missing data points in the time dimension are finally obtained.

[0051] Based on the preset drawing coordinate system parameters, a dual-axis rectangular coordinate system is established in the drawing space with time as the horizontal axis, current value as the left vertical axis, and temperature value as the right vertical axis.

[0052] The time-synchronized current data point set is mapped to a dual-vertical-axis rectangular coordinate system according to its timestamp and current value. Adjacent data points are connected using linear interpolation to form a continuous curve of current changing with time.

[0053] The time-synchronized temperature data points are mapped to the same coordinate system according to their timestamps and temperature values. Similarly, the linear interpolation method is used to connect adjacent points to form a continuous curve of temperature change over time.

[0054] The completed current change curve and temperature change curve are superimposed and displayed in the same coordinate system view to generate a dual-axis curve that allows for simultaneous comparison and observation of the current and temperature change trends and their corresponding relationships.

[0055] Finally, based on the scene labels marked in the data source, the generated biaxial curves are classified according to scene type and stored in a designated database to form a scene-based current and temperature change curve library that can be used for subsequent analysis and model training.

[0056] The core design motivation of this system stems from a profound understanding of the unique operating conditions of electric two-wheelers and three-wheelers. The driving modes of these vehicles are highly uncertain and transient, with frequent starts and stops, rapid acceleration, hill climbing, and load changes causing the power battery to continuously endure high-rate, non-stable charging and discharging current surges. This dynamic load triggers two interconnected transient effects within the battery: first, the current generates Joule heating through the ohmic internal resistance, intensifying electrochemical polarization and causing a sudden rise in battery temperature; second, intense ion migration generates mechanical stress within the electrode active materials, accelerating material fatigue and aging. Crucially, the coupling strength of these electro-thermal-mechanical effects is not determined by the absolute values ​​of current or temperature, but rather by their dynamic characteristics, namely the rate of current change and the acceleration of its trend. Traditional fixed-frequency acquisition schemes or triggering mechanisms based on single-parameter thresholds have inherent flaws: fixed sampling cannot capture the starting edge of abrupt signal changes, leading to the loss of critical transient data; while simple threshold triggering easily generates a large amount of redundant data during non-stress periods. These two situations respectively create a dual dilemma of unseen risks and wasted resources. Therefore, it should be further explained that the fluctuation analysis unit in this embodiment obtains the current slope sequence and the temperature slope sequence based on the current change curve and the temperature change curve; and constructs the acquisition adjustment factor based on the ratio of the current slope ratio to the temperature slope ratio at adjacent time points.

[0057] Simultaneously, based on the current change curve and the temperature change curve, the length of the synchronous time interval between the current slope and the temperature slope is obtained through time series analysis algorithm as the initial acquisition window length, and the data acquisition frequency of the corresponding time interval is used as the initial acquisition frequency.

[0058] The acquisition adjustment factor in this embodiment is designed based on the deep coupling characteristics of the electro-thermal response of batteries in light electric vehicles such as electric two-wheelers and three-wheelers under dynamic operating conditions. Frequent start-stop and rapid acceleration in these vehicles can easily lead to drastic fluctuations in battery current. The dynamic coupling relationship between current fluctuations and temperature changes directly affects the battery aging rate and performance degradation process. From the perspective of degradation mechanism, the core influencing factors of battery aging are not the instantaneous absolute values ​​of current and temperature, but rather the high-order dynamic correlation between their changing trends: the acceleration characteristic of current change is characterized by the ratio of the current slope at adjacent time points, which reflects the change in the rate of current increase / decrease; the acceleration characteristic of temperature change is characterized by the ratio of the temperature slope at adjacent time points, which reflects the change in the rate of temperature rise / fall. The acquisition adjustment factor is the ratio of these two slopes, essentially quantifying the response inertia of the battery's heat generation effect relative to current excitation, and the intensity of the electro-thermal coupling relationship.

[0059] Specifically, when the acquisition adjustment factor is greater than 1, meaning the acceleration characteristic of current change is greater than the acceleration characteristic of temperature change, it indicates a surge in transient Joule heating of the battery, or abnormal heat generation caused by increased internal resistance in an aging battery. In this case, the battery faces a high risk of irreversible aging, and the system needs to immediately increase the acquisition frequency to capture critical transient data. When this ratio is less than or equal to 1, it indicates that the battery's electro-thermal response is stable, and the system can appropriately reduce the acquisition frequency to optimize resource consumption. Furthermore, this factor is constructed based on the high-order dynamic correlation of the physical quantity itself, without relying on a preset operating condition feature library. By detecting the dynamic correlation between current and temperature, it can adaptively respond to any operating condition, ultimately providing accurate and efficient front-end data support for the entire closed-loop process of online battery health status assessment.

[0060] It should be further explained that the process of obtaining the initial acquisition window length and the initial acquisition frequency in this embodiment is as follows:

[0061] Based on the current change curve obtained by the fluctuation analysis unit, the discrete time points and their corresponding current values ​​are extracted from the curve in chronological order. The ratio of the current difference between two adjacent discrete time points to the time difference is calculated to obtain the current slope value corresponding to each adjacent time point. All current slope values ​​are arranged in chronological order to form a current slope sequence. Abnormal current slope values ​​that exceed the preset reasonable range set based on the current fluctuation characteristics of electric two-wheeled or three-wheeled vehicles are removed from the current slope sequence. The current slope data corresponding to the removed abnormal values ​​is supplemented by linear interpolation to obtain the corrected current slope sequence.

[0062] Based on the temperature change curve obtained by the fluctuation analysis unit, the discrete time points that completely correspond to the discrete time points in the current change curve and their corresponding temperature values ​​are extracted. The ratio of the temperature difference to the time difference between two adjacent discrete time points is calculated to obtain the temperature slope value corresponding to each adjacent time point. All temperature slope values ​​are arranged in time order to form a temperature slope sequence.

[0063] Abnormal temperature slope values ​​that exceed the preset reasonable range based on the temperature fluctuation characteristics of electric two-wheeled or three-wheeled vehicle batteries are removed from the temperature slope sequence. The temperature slope data corresponding to the removed abnormal values ​​is supplemented by linear interpolation to obtain the corrected temperature slope sequence.

[0064] It should be further explained that, in this embodiment, based on the specifications of the battery for the electric two-wheeler or three-wheeler, the maximum allowable continuous discharge current, maximum pulse discharge current, and operating temperature range parameters are obtained, and the initial reasonable range threshold of the current and temperature slope is calculated. Secondly, the threshold is updated based on historical data. Specifically, after system initialization, in the first complete charge-discharge cycle, all current slope values ​​and temperature slope values ​​are recorded; the mean and standard deviation of the sequence are calculated, and the range of the mean plus or minus three times the standard deviation is set as the initial dynamic reasonable range threshold; in subsequent operation, the threshold is dynamically updated using the same statistical method after every M charge-discharge cycles.

[0065] Based on the corrected current slope sequence and the corrected temperature slope sequence, the current slope value and temperature slope value corresponding to any two adjacent time points are selected. The ratio of the current slope value at the next time point to the current slope value at the previous time point is calculated as the current slope ratio. At the same time, the ratio of the temperature slope value at the next time point to the temperature slope value at the previous time point is calculated as the temperature slope ratio. Then, the ratio of the current slope ratio to the temperature slope ratio is calculated to obtain the acquisition adjustment factor corresponding to the adjacent time point. For the acquisition adjustment factors corresponding to all adjacent time points, a sliding average is performed by setting a sliding window length to eliminate the influence of instantaneous fluctuations on the acquisition adjustment factors, resulting in a smoothed acquisition adjustment factor sequence. The sliding window length is set based on the battery operating condition fluctuation frequency of the electric two-wheeler or three-wheeler.

[0066] Based on the corrected current slope sequence and the corrected temperature slope sequence, the cross-correlation analysis method in the time series analysis algorithm is used to calculate the cross-correlation coefficient of the two slope sequences under different time lag values ​​within a set time lag value range; the time lag value range in this embodiment is set based on the battery temperature-current response delay characteristics of electric two-wheeled vehicles or three-wheeled vehicles.

[0067] The time intervals with absolute values ​​of cross-correlation coefficients greater than or equal to a preset synchronization threshold are selected; these time intervals are the synchronization time intervals for the current slope and temperature slope. The start and end times of each synchronization time interval are counted, and the difference between the end and start times is calculated to obtain the length of each synchronization time interval. The frequency of occurrence of all synchronization time interval lengths is counted, and the length of the synchronization time interval with the highest frequency is selected as the initial acquisition window length. If there are multiple synchronization time interval lengths with the same frequency and the highest frequency, the average of these lengths is calculated, and this average is used as the initial acquisition window length.

[0068] Based on all determined synchronization time intervals, the number of raw current data acquisition points and the number of raw temperature data acquisition points within each synchronization time interval are extracted. The total duration of each synchronization time interval is calculated, which is the difference between the end time and the start time of the interval. The acquisition intervals for current data and temperature data within each synchronization time interval are calculated. The minimum value between the current data acquisition interval and the temperature data acquisition interval within each synchronization time interval is taken, and the reciprocal of this minimum value is calculated to obtain the raw data acquisition frequency corresponding to that synchronization time interval. The raw data acquisition frequencies corresponding to all synchronization time intervals are statistically analyzed, and those that exceed the preset reasonable frequency range set based on the sensor hardware performance of electric two-wheeled vehicles or three-wheeled vehicles are removed. The average value of the remaining raw data acquisition frequencies is calculated, and this average value is used as the initial acquisition frequency.

[0069] The model training unit, based on the initial acquisition window length, initial acquisition frequency, acquisition adjustment factor, and the reciprocal of the battery life decay rate, combines a distributed reinforcement learning algorithm with corresponding current and temperature data from different historical scenarios to train the acquisition parameters and obtain a trained reinforcement acquisition adjustment model.

[0070] The scenarios include, but are not limited to, rapid acceleration and high-speed start-up scenarios, continuous uphill climbing and heavy load scenarios, emergency braking and kinetic energy recovery scenarios, frequent start-stop scenarios, rapid full charge scenarios, and recovery scenarios after over-discharge or under-voltage protection.

[0071] It should be further explained that the construction and training process of the enhanced acquisition adjustment model in this embodiment includes:

[0072] Based on the current acquisition window length, current acquisition frequency, real-time acquisition adjustment factor, reciprocal of real-time battery life decay rate, variance of historical current data fluctuation in the current scenario, and variance of historical temperature data fluctuation in the current scenario, a data acquisition adjustment state space is constructed. The variance of historical current data fluctuation and the variance of historical temperature data fluctuation are obtained by calculating the data within 10 historical acquisition cycles through a sliding window to comprehensively characterize the current acquisition parameters and battery operating status.

[0073] Based on the data acquisition and state space adjustment, the action space is divided into three categories of acquisition parameter adjustment actions, specifically including: the first category is acquisition window length adjustment actions, including maintaining the current window length, shortening the current window length to 80% of the original length, and extending the current window length to 120% of the original length; the second category is acquisition frequency adjustment actions, including maintaining the current acquisition frequency, increasing the current acquisition frequency to 150% of the original frequency, and decreasing the current acquisition frequency to 50% of the original frequency; the third category is joint adjustment actions, that is, adjusting the window length and acquisition frequency simultaneously, with the adjustment range being consistent with the first two categories. The upper and lower limits of the range of all adjustment actions are set based on the sensor hardware performance of the electric two-wheeler or three-wheeler.

[0074] It should be further explained that the more detailed implementation process of the adjustment action in this embodiment includes:

[0075] Based on the acquisition adjustment factor in the current state space vector and the reciprocal of the battery life decay rate, the battery's operating condition type is determined in real time. It should be further noted that this embodiment strengthens the operating condition type discrimination logic built into the acquisition adjustment model, specifically as follows:

[0076] Based on the comparison between real-time collected adjustment factor values ​​and preset high-stress and low-stress thresholds, the current stress level of the battery is determined; the process of obtaining the stress level includes:

[0077] Acquire a large amount of historical operational data and extract the acquisition adjustment factor sequence and battery life decay rate sequence under different scenarios;

[0078] The distribution of adjustment factors was statistically collected and divided into three intervals (high, medium, and low) according to percentiles. The boundary values ​​of these intervals were used as thresholds for stress state discrimination. For example, adjustment factors were collected... For high stress, Collection of adjustment factors For moderate stress, an adjustment factor was collected. Low stress; To collect the three-quarters threshold of the adjustment factor, i.e., the high stress threshold, The threshold for collecting adjustment factors is one-quarter of the threshold, i.e., the low stress threshold;

[0079] The current health level of the battery is determined by comparing the reciprocal of the real-time battery life degradation rate with preset high and low health thresholds; the process of obtaining the health level includes:

[0080] Similarly, by statistically analyzing the distribution of battery life degradation rates, their reciprocals are divided into three health state intervals—high, medium, and low—based on percentiles. The boundary values ​​are then used as thresholds for health state discrimination. For example, the reciprocal of the life degradation rate... For high health, Reciprocal of lifespan decay rate For moderate health, the rate of lifespan decline is inverse. For low health; The three-quarters threshold of the health score is considered the high health threshold. The threshold is one-quarter of the health score, i.e., the low health threshold; in this embodiment, the health score is represented by the reciprocal of the lifespan decay rate.

[0081] If the stress level is determined to be high and the health level is low, a high stress-low health condition label is generated.

[0082] If the stress level is determined to be low and the health level is high, a low stress-high health condition label is generated.

[0083] If both the stress level and the health level are determined to be moderate, or if one of them is moderate while the other does not meet the above-mentioned high-low extreme combination conditions, then a moderate stress-moderate health condition label is generated.

[0084] Finally, based on the generated operating condition identifier, the optimal acquisition parameters stored in the mapping relationship are called to adjust the action combination.

[0085] Based on the judgment results, the optimal action is selected from the three types of adjustment actions, specifically:

[0086] When the condition is determined to be high stress and low health, select the combination of shortening the window length to 80% of the original length and increasing the acquisition frequency to 150% of the original frequency in the joint adjustment action;

[0087] When the condition is determined to be low stress-high health, select the combination of extending the window length to 120% of the original length and reducing the acquisition frequency to 50% of the original frequency in the joint adjustment action;

[0088] When the condition is determined to be moderate stress-moderate health, the maintain or slight adjustment command can be selected independently from the first and second types of actions;

[0089] Based on the amplitude constraints set by the sensor hardware performance, the calculated adjustment amplitude is limited: the lower limit of the acquisition window length adjustment is not lower than the window length corresponding to the minimum sampling period, and the upper limit is not higher than the window length supported by the maximum storage capacity; the lower limit of the acquisition frequency adjustment is not lower than the sensor's lowest sampling frequency, and the upper limit is not higher than the sensor's highest sampling frequency.

[0090] Finally, the specific acquisition parameter adjustment instructions, after amplitude constraint processing, are output to the execution unit.

[0091] A multi-objective reward function is constructed with the goals of maximizing the effectiveness of collected data, minimizing resource consumption, and minimizing subsequent SOH assessment error. Specifically, the reward value is calculated as follows: Reward value = α × Data effectiveness score - β × Resource consumption score - γ × Assessment error score, where α, β, and γ are weighting coefficients determined through historical data verification. The data effectiveness score is calculated based on the information entropy of the collected data; higher information entropy indicates richer operating condition characteristics and a higher data effectiveness score. The resource consumption score is calculated based on the product of the acquisition frequency and the window length; a larger product indicates greater sensor energy consumption and data transmission volume, resulting in a higher resource consumption score. The assessment error score is calculated based on the absolute value of the deviation between the calculated SOH assessment value and the actual SOH value of the battery, based on the data acquired under these acquisition parameters; a smaller deviation results in a lower error score.

[0092] The training dataset is divided into multiple subsets according to scene type, with each subset corresponding to a reinforcement learning agent. Each agent trains in parallel on its corresponding scene subset. During training, each agent selects the collected parameters in the action space based on the current state space vector to adjust its action. After executing the action, it obtains the corresponding reward value and updates the action value function based on the temporal difference algorithm. At the same time, a parameter server enables the sharing of model parameters among the agents. The parameters are synchronized once every 100 training iterations to improve the model's generalization ability to different scenes.

[0093] It should be further explained that, during the training process in this embodiment, the model makes adjustment decisions based on the acquisition adjustment factor in the real-time state space vector and the reciprocal of the battery life decay rate. When it is determined that the current operating condition is high stress-low health, the model prioritizes adjusting the acquisition frequency and shortening the window length to enhance the transient data capture capability. When it is determined that the current operating condition is low stress-high health, the model prioritizes adjusting the acquisition frequency and extending the window length to reduce resource consumption. When it is determined that the current operating condition is medium stress-medium health, the model chooses to maintain the current acquisition parameters or make minor adjustments.

[0094] After training, the model is validated using historical scenario data that was not used in the training. The reward value and SOH evaluation error corresponding to the model's selected acquisition parameters are calculated for different scenarios. If the SOH evaluation error is greater than the preset error threshold in a certain scenario, the current, temperature data and acquisition adjustment factor data in that scenario are extracted as supplementary training data to retrain the model until the model's SOH evaluation error in all scenarios is less than the preset error threshold, thus obtaining the final trained enhanced acquisition adjustment model.

[0095] The acquisition and adjustment unit is used to configure the trained enhanced acquisition and adjustment model into the corresponding sensor using the federated framework, and to adjust the length and frequency of the differential current and temperature data acquisition window by combining the scene fingerprint constructed by the current slope ratio and temperature slope ratio at adjacent time points with the acquisition adjustment factor; the sensor includes a current sensor, a temperature sensor group and a voltage sensor; the temperature sensor group is configured to acquire the ambient temperature at at least one location on the battery surface and acquire the temperature of a reference point inside the battery through a temperature sensor attached to the battery terminal.

[0096] The coupling module is used to obtain a temperature-current coupling feature vector based on the current and temperature data combined with a coupling algorithm.

[0097] It should be further noted that the coupling module in this embodiment includes a preprocessing unit, a current parameter unit, a temperature parameter unit, and a hierarchical coupling unit;

[0098] The preprocessing unit is used to monitor the consistency relationship between current and temperature changes in real time based on a piecewise consistency function constructed by combining the current slope and temperature slope with the convolution kernel function. Based on the consistency relationship between current and temperature changes in different time periods, combined with a preset local filtering model library and the time points and ranges of abnormal consistency fluctuations, the unit calls the local filtering model in the corresponding time period to perform data filtering processing, thereby obtaining the preprocessed current and temperature data for each consistency interval.

[0099] It should be further explained that the process of obtaining the current and temperature data for each consistency interval after preprocessing in this embodiment includes:

[0100] Based on the current slope sequence and temperature slope sequence output by the fluctuation analysis unit, smoothing is performed by a preset convolution kernel function to suppress noise interference. The correlation coefficient of the two smoothed sequences in each continuous time period is calculated to obtain the consistency index of current and temperature changes in that time period.

[0101] Based on the time series of battery life decay rate monitored in real time, the rate of change within the sliding window is calculated. When the rate of change exceeds the preset fluctuation threshold, the corresponding time point is determined as the critical point of significant change in battery aging state. By segmenting at the critical point, the dynamically divided time period is obtained.

[0102] Based on each dynamically divided time period, a segmented consistency function is constructed using the consistency index calculated within that time period as the core parameter, thereby obtaining a quantitative characterization of the synergistic relationship between current and temperature changes under different aging stages.

[0103] Based on the piecewise consistency function, the real-time consistency index for the current time period is obtained through real-time calculation.

[0104] Based on historical current slope and historical temperature slope data under different battery life decay rates, the distribution range of historical consistency indicators corresponding to different decay rates is determined through statistical analysis. The distribution range of historical consistency indicators is set as the normal range of consistency relationship under the corresponding decay rate. Based on the current battery life decay rate, the applicable normal range of consistency relationship is determined through correlation analysis algorithm. By comparing the real-time consistency indicators with this normal range, the consistency status of current and temperature changes in the current time period is obtained.

[0105] Based on the comparison between the real-time consistency index and the normal range of the consistency relationship corresponding to the current battery life decay rate, when the real-time consistency index exceeds the normal range, the abnormal time point of the consistency relationship fluctuation is obtained by recording the corresponding time point; in particular, for scenarios with a high battery life decay rate, the sensitivity of identifying consistency fluctuations is improved by narrowing the boundary value of the normal range of the consistency relationship.

[0106] Based on the time points of abnormal fluctuations in the consistency relationship, the starting time point of the fluctuation range is set by tracing back to the time point when the real-time consistency index first enters the normal range corresponding to the current battery life degradation rate; and the ending time point of the fluctuation range is set by tracing back to the time point when the real-time consistency index again stably enters the normal range corresponding to the current battery life degradation rate and remains there for a preset duration. Based on the starting and ending time points, the complete fluctuation range of the abnormal consistency relationship is obtained; where the higher the battery life degradation rate, the shorter the preset duration.

[0107] Based on the battery life degradation rate, the preset duration threshold is dynamically adjusted by multiplying the preset duration threshold by the reciprocal of the degradation rate when the degradation rate is high and keeping the preset duration threshold unchanged when the degradation rate is low.

[0108] Based on historical data on fluctuation ranges at different battery life degradation rates, we statistically analyze the length of the fluctuation range and the fluctuation amplitude of the consistency index within the range corresponding to different degradation rates. Through data fitting, we construct a mathematical correspondence model between the length and amplitude of the fluctuation range and the battery life degradation rate, and determine the length and amplitude thresholds of the fluctuation range at different degradation rates. By comparing the length or amplitude of a certain fluctuation range with the threshold at the corresponding degradation rate, when the threshold is exceeded, the range is determined to be a high-impact abnormal range that needs to be prioritized for handling.

[0109] Based on a pre-set local filtering model library, the library includes dedicated filtering models that adapt to different fluctuation range lengths, fluctuation amplitudes and battery life decay rates, as well as basic filtering models that adapt to periods without abnormal fluctuations.

[0110] For the fluctuation range of abnormal consistency relationship fluctuations, a suitable dedicated filtering model is selected from the local filtering model library based on the length of the range, the fluctuation amplitude, and the current battery life decay rate. The current data and temperature data within the range are input into the dedicated filtering model to obtain the processed data after removing data noise and abnormal fluctuations. The dedicated filtering model is trained by those skilled in the art based on the abnormal range datasets corresponding to different scenarios and combined with filtering algorithms. In this embodiment, the filtering algorithm is obtained by constructing an association mapping index based on the general filtering algorithm library, the data in the corresponding scenario, the evaluation accuracy of the battery health status assessment value, and the association degree obtained by the association analysis algorithm.

[0111] For time periods where no abnormal fluctuations in consistency were observed, the current and temperature data within the interval were input into the basic filtering model for routine processing to obtain the processed current and temperature data.

[0112] Based on the current and temperature data processed by the dedicated filtering model and the basic filtering model, multiple consecutive consistent intervals are divided according to time sequence to obtain the current and temperature data of each consistent interval after preprocessing.

[0113] One implementation process for obtaining the fluctuation range of the consistency relationship fluctuation anomaly in this embodiment includes:

[0114] Based on the real-time consistency index sequence, the query starting point is obtained by setting the time point of abnormal fluctuation in consistency relationship as the query starting point.

[0115] Based on the normal range boundary value of the consistency relationship corresponding to the current battery life decay rate, the historical data traversal result is obtained by traversing the historical data of the real-time consistency index in reverse order.

[0116] Based on the results of historical data traversal, the starting time point of the fluctuation range is obtained by identifying the moment when the real-time consistency index value first falls into the normal range corresponding to the current battery life decay rate during the traversal process.

[0117] Based on the abnormal time points of the aforementioned consistency relationship fluctuations, the real-time data traversal results are obtained by sequentially traversing the real-time consistency index data in reverse order.

[0118] Based on the real-time data traversal results, the termination time point of the fluctuation range is obtained by identifying the moment when the real-time consistency index value falls stably within the normal range corresponding to the current battery life decay rate for multiple consecutive sampling cycles and the duration reaches the preset duration threshold.

[0119] Based on the start and end time points, the complete fluctuation range of the consistency relationship fluctuation anomaly is obtained through integrated processing.

[0120] This embodiment, based on the existing acquisition adjustment factor, still needs to analyze thermal variables, namely the piecewise current thermal stress factor, and the temperature gradient, namely the piecewise temperature gradient factor. The core reason lies in the essential differences in the physical meaning and application goals of the three and the complementarity of the data link function: the acquisition adjustment factor, as a forward-looking and process indicator, only identifies the key monitoring periods through the synergistic relationship between the acceleration of current and temperature changes, such as rapid acceleration and the continuous ramping of aging batteries. The core goal is to optimize the data acquisition strategy, such as triggering high-frequency acquisition, but it cannot quantify the absolute value of the electro-thermal stress and the actual degree of damage to the battery during the period. It only solves the front-end problem of how to efficiently capture key data, but does not involve how to use key data to characterize the back-end core requirement of aging. The heat variable, namely the piecewise current thermal stress factor, is based on Joule's law. It directly quantifies the cumulative energy damage caused by the Joule heating effect in a specific period of time by integrating the product of the square of the current and the internal resistance over time. The internal resistance parameter changes dynamically with the battery's health status, such as increased electrode polarization and electrolyte decomposition, and carries aging information itself, directly relating to the irreversible degradation chemical processes inside the battery. The temperature gradient, namely the piecewise temperature gradient factor, quantifies the heat exchange efficiency and internal temperature distribution uniformity of the battery by the fluctuation of the temperature difference between the ambient temperature and the internal reference point. It directly reflects the degradation of the battery's thermal management performance. For example, the decreased heat dissipation capacity of an aging battery leads to increased temperature fluctuations and thermal management imbalance. These factors are key triggers for accelerating battery aging. As retrospective and outcome indicators, they fill the technical gap where the acquisition adjustment factor cannot be linked to the aging mechanism. The acquisition adjustment factor ensures that the system captures high-quality data, while the thermal variable and temperature gradient use these data to transform the electrothermal change trend into a quantitative damage feature that can be directly linked to lifespan degradation. The two are located at the front and back ends of the data chain, respectively. The former ensures the validity of the data, while the latter supports the accuracy of the assessment. If the thermal variable and temperature gradient are missing, relying solely on the acquisition adjustment factor will lead to subsequent battery health status assessments lacking feature dimensions directly linked to the degradation mechanism, making it impossible to distinguish between high-trend low-damage and high-trend high-damage operating conditions, ultimately resulting in assessment bias.

[0121] The current parameter unit is used to calculate the integral of the product of the square of the current and the resistance and time per unit time based on the current data of each consistent interval after preprocessing, so as to obtain the segmented current thermal stress factor.

[0122] The temperature parameter unit is used to obtain the piecewise temperature gradient factor based on the fluctuation of the difference between the average ambient temperature and the average internal reference point temperature within the corresponding time length of each consistency interval.

[0123] It should be further explained that, in this embodiment, both the segmented current thermal stress factor and the segmented temperature gradient factor are key characteristic factors for quantifying the impact of electro-thermal stress within the battery consistency interval. The segmented current thermal stress factor is calculated by the current parameter unit based on the preprocessed current data of each consistency interval. Its working principle is to convert the instantaneous Joule thermal power into the cumulative energy damage of the interval by calculating the integral of the product of the square of the current and the resistance over time, thereby accurately quantifying the cumulative thermal stress caused by the current within the interval and reflecting the degree of energy damage inside the battery under the action of the current. The segmented temperature gradient factor is obtained by the temperature parameter unit based on the fluctuation of the difference between the average ambient temperature and the average internal reference point temperature within the corresponding time length of each consistency interval. Its working principle is to reflect the heat exchange efficiency inside and outside the battery and the stability of the internal temperature distribution through the fluctuation of the difference, thereby quantifying the actual degree of thermal imbalance inside the battery within the interval and reflecting the gradient impact of the battery thermal management status and temperature distribution on the battery. Together, they provide characteristic support that can directly relate to the electro-thermal damage mechanism for subsequent temperature-current coupling modeling and battery health status assessment.

[0124] The hierarchical coupling unit is used to obtain a segmented temperature-current coupling feature vector by combining the ratio of segmented temperature gradient factor to segmented current thermal stress factor, average ambient temperature, average internal reference point temperature, battery life decay rate, and the ratio of temperature slope ratio to current slope ratio with a segmented coupling model. At the same time, it performs real-time calibration of the segmented interval length of the segmented temperature-current coupling feature vector by the battery life decay rate.

[0125] It should be further explained that the construction process of the segmented temperature-current coupling feature vector in this embodiment includes:

[0126] Based on the min-max normalization algorithm, the five types of input variables are normalized to eliminate dimensional differences and obtain standardized variables. These variables include the ratio of piecewise temperature gradient factor to piecewise current thermal stress factor, battery life decay rate, average ambient temperature, average internal reference point temperature, and the ratio of temperature slope ratio to current slope ratio.

[0127] Based on the Pearson correlation coefficient algorithm, the correlation between the ratio of the standardized piecewise temperature gradient factor to the piecewise current thermal stress factor and the battery life decay rate, the correlation between the difference between the standardized average ambient temperature and the average internal reference point temperature and the ratio of the standardized piecewise temperature gradient factor to the piecewise current thermal stress factor, and the correlation between the ratio of the standardized temperature slope ratio to the current slope ratio and the battery life decay rate were calculated.

[0128] Based on the three correlation values ​​obtained from the calculation, the initial weights of the five types of standardized variables in the coupling mapping are allocated according to the proportion of correlation.

[0129] The temperature difference characteristic is obtained by subtracting the normalized average ambient temperature from the average internal reference point temperature.

[0130] The ratio of the standardized piecewise temperature gradient factor to the piecewise current thermal stress factor is multiplied by the temperature difference feature to obtain the thermal stress-temperature difference interaction feature; the ratio of the standardized temperature slope ratio to the current slope ratio is multiplied by the battery life decay rate to obtain the dynamic trend-degradation rate interaction feature.

[0131] It should be further explained that the motivation for constructing the thermal stress-temperature difference interaction feature in this embodiment is to quantify the dynamic competitive relationship between internal heat generation and heat dissipation environment: the principle is to multiply the ratio of electrothermal factor, which characterizes the cumulative effect of internal heat generation, with the temperature difference between the environment and the internal environment, which reflects the heat dissipation conditions. This aims to capture the high-risk scenario of high heat generation meeting poor heat dissipation, which exacerbates battery thermal aging. This feature value amplifies the risk of thermal runaway. The motivation for constructing the dynamic trend-degradation rate interaction feature is to correlate the external dynamic response of the battery with the internal chemical aging process: the principle is to multiply the ratio of the slope ratio, which characterizes the relative rate of change of current and temperature, with the degradation rate, which directly reflects the rate of aging. This aims to reveal the coupling law that the more intense the charge and discharge dynamics, the faster the lifespan degradation. This feature value directly quantifies the impact intensity of dynamic operating conditions on battery lifespan. Together, these two features transform independent physical variables into composite monitoring signals with clear physical meaning of failure, providing a more sensitive and accurate basis for health status assessment.

[0132] A backpropagation neural network model optimized by particle swarm optimization was established and used as the base model. During model operation, the numerical changes of the thermal stress-temperature difference interaction characteristics and the dynamic trend-degradation rate interaction characteristics were monitored in real time. When the thermal stress-temperature difference interaction characteristics increased, the gradient descent algorithm was used to increase the weight of the ratio of the standardized piecewise temperature gradient factor to the piecewise current thermal stress factor in the model. When the dynamic trend-degradation rate interaction characteristics increased, the gradient descent algorithm was used to increase the weight of the ratio of the standardized temperature slope to the current slope and the battery life degradation rate in the model.

[0133] The standardized five types of variables, two interactive features, and dynamically adjusted weights are input into a convolutional attention network. Local features are extracted through convolutional layers, and feature importance weights are calculated through the attention mechanism. After weighted fusion, a one-dimensional feature vector is output.

[0134] The output one-dimensional feature vector is defined as a piecewise temperature-current coupling feature vector, where each element corresponds to the contribution value of a variable or interaction feature in the coupling effect; based on the training results of historical data, mean squared error is used as the loss function, and the parameters of the convolutional attention network are optimized by the backpropagation algorithm.

[0135] Input historical operating condition data into the trained model to obtain historical coupling feature vectors; calculate the correlation coefficient between the historical coupling feature vectors and the actual battery degradation data; when the correlation coefficient is lower than the set threshold, readjust the model parameters until the accuracy requirements are met.

[0136] The battery life degradation rate is monitored in real time. When the degradation rate exceeds the set change threshold for several consecutive cycles, the length of the corresponding segment interval of the segmented temperature-current coupling feature vector is shortened proportionally. When the degradation rate remains stable, the default segment interval length is restored.

[0137] The online evaluation module is used to perform real-time evaluation based on the temperature-current coupling feature vector and a preset online evaluation model to obtain the battery health status evaluation value and the battery life decay rate.

[0138] It should be further noted that the online evaluation module in this embodiment includes a contribution analysis unit and a segmented evaluation unit;

[0139] The contribution analysis unit is used to obtain the main evaluation factor and auxiliary evaluation factor and their corresponding contribution for each segment based on the segmented temperature-current coupling feature vector, with the battery life decay probability as the target, combined with principal component analysis.

[0140] The segmented evaluation unit is used to evaluate the battery health status based on the main evaluation factor and auxiliary evaluation factor and their corresponding contribution in each segment, combined with a comprehensive evaluation algorithm, to obtain the segmented battery health status evaluation value. Based on the fluctuation value of the product of the battery health status evaluation value and the ratio of the current thermal stress factor to the temperature gradient factor, the battery life decay rate is obtained.

[0141] The preferred online evaluation model is constructed using a comprehensive fuzzy algorithm, the specific process of which includes:

[0142] Based on the main evaluation factor, auxiliary evaluation factor, and their contribution degree corresponding to each segment output by the contribution analysis unit, and combined with the segmented temperature-current coupling feature vector output by the hierarchical coupling unit, a fuzzification module of the comprehensive fuzzy algorithm is constructed. This module maps the main evaluation factor, auxiliary evaluation factor, and their corresponding contribution degrees to fuzzy subsets. The main evaluation factor is the dominant feature variable for assessing battery health status in each segment, specifically a thermal stress-temperature difference interaction feature. The auxiliary evaluation factor is the auxiliary feature variable for each segment, specifically a dynamic trend-degradation rate interaction feature. Simultaneously, it determines... The fuzzy membership functions of each variable are defined, using a triangular membership function, and their parameters are determined through statistical analysis of historical battery health status assessment data and actual degradation data. Secondly, based on the correspondence between the segmented temperature-current coupling feature vectors, the contribution of primary and secondary assessment factors, and the actual battery health status under different historical scenarios, fuzzy inference rules are extracted to form a structured fuzzy rule base. These different historical scenarios specifically include rapid acceleration and high-speed start-up scenarios, continuous hill climbing and heavy load scenarios, emergency braking and kinetic energy recovery scenarios, frequent start-stop scenarios, rapid full charging scenarios, and over-discharge or... In the recovery scenario after undervoltage protection, the actual battery health status is represented by the battery's true SOH value. The fuzzy inference rules are specifically structured rules such as "if the main evaluation factor has a high contribution and the auxiliary evaluation factor has a medium contribution, then the battery health status is good." The fuzzy subsets of the main evaluation factor, auxiliary evaluation factor, and their contribution values ​​for each segment, acquired in real time, are input into the inference engine built using the Mamdani fuzzy inference algorithm. Logical inference is performed based on the rules in the fuzzy rule base to obtain the fuzzy inference results for each segment. These results represent the battery health status of each segment in the form of a fuzzy set. Subsequently, the centroid method is used to convert the fuzzy inference results of each segment into quantitative segmented battery health status assessment values. These segmented battery health status assessment values ​​are specific numerical values ​​of the battery health status under each segment. Finally, based on the battery health status assessment values ​​of all segments, combined with the fluctuation value of the ratio of the current thermal stress factor to the temperature gradient factor corresponding to each segment, the correlation coefficient between this fluctuation value and the segmented battery health status assessment value is calculated using a sliding window algorithm. This correlation coefficient is then combined with a preset decay rate mapping model to finally obtain the battery life decay rate, completing the construction of an online assessment model based on a comprehensive fuzzy algorithm.

[0143] The early warning module is used to issue graded early warnings based on the battery health status assessment value and the set graded early warning thresholds. At the same time, it feeds back the battery life decay rate to the enhanced acquisition adjustment model in the acquisition module and the online evaluation model to adjust the data acquisition strategy and evaluation index weights in real time.

[0144] It should be further noted that the early warning module in this embodiment includes a graded early warning unit and a feedback correction unit;

[0145] The graded early warning unit is used to compare the battery health status assessment value with the preset multi-level early warning threshold in real time.

[0146] When the battery health status assessment value is lower than the first threshold, a first-level warning signal is generated and output. The first-level warning signal includes a battery capacity decay prompt and an estimated remaining driving range calculated based on the current battery level and the battery life decay rate.

[0147] When the battery health status assessment value is lower than the preset second threshold, a second-level warning signal is generated and output. The second-level warning signal is a warning of the end of battery life and a replacement reminder. The health status assessment value, the warning signal and the estimated remaining driving range are transmitted to the vehicle display screen or the user's mobile terminal APP for display.

[0148] The feedback correction unit is used to receive the battery life degradation rate from the online evaluation module in real time, and transmit the battery life degradation rate as a feedback signal to the enhanced acquisition adjustment model in the acquisition module to dynamically adjust the acquisition window length and acquisition frequency in the data acquisition strategy.

[0149] Simultaneously, the battery life degradation rate is transmitted to the hierarchical coupling unit to adaptively correct the segment interval length of the segmented temperature-current coupling feature vector and adjust the segmented battery health status assessment value in real time.

[0150] Further explanation is needed; please refer to [link / reference]. Figure 2 The workflow of the corresponding unit in the online battery health status assessment system considering the temperature-current coupling effect in this embodiment includes:

[0151] Based on time, battery current data, and temperature data, a curve construction unit is used to construct a dual-axis curve with time as the horizontal axis, current data as the left vertical axis, and temperature data as the right vertical axis, to obtain a time-series presentation of current and temperature data.

[0152] Based on the fluctuation analysis unit and the dual-axis curve, the acquisition adjustment factor is obtained; at the same time, the matching degree between the current change trend and the temperature change trend is calculated by the time series correlation analysis algorithm, the length of the interval where the current and temperature changes are synchronized is determined, and the initial acquisition frequency is obtained according to the change frequency of the data within the synchronized time interval.

[0153] Based on the length of the synchronous time interval, initial acquisition frequency, acquisition adjustment factor, and the pre-acquired reciprocal of the battery life decay rate output by the model training unit and the fluctuation analysis unit, historical current and temperature data under different typical operating conditions such as rapid acceleration and high-speed start, continuous climbing and heavy load, emergency braking and kinetic energy recovery, frequent start-stop, rapid full charge, and recovery after over-discharge or undervoltage protection are collected. Distributed reinforcement learning algorithm is used to improve the effectiveness and accuracy of the acquired data in assessing the battery health status. The initial parameters are trained on the acquired historical data to obtain the trained reinforcement acquisition adjustment model.

[0154] Based on the adjustment unit and federated learning framework, the trained enhanced acquisition adjustment model is deployed to the current sensor, the sensor that collects the ambient temperature of the battery surface, and the temperature sensor group that is attached to the battery terminal to collect the internal reference point temperature. By calculating the current slope ratio and temperature slope ratio between the current time point and the previous adjacent time point in real time, a scene fingerprint is generated based on the magnitude, direction and amplitude of the two slope ratios. The enhanced acquisition adjustment model is called according to the generated scene fingerprint to dynamically adjust the acquisition window length and sampling frequency of each sensor node, and the adjusted acquisition window length and sampling frequency are obtained.

[0155] Based on the sensor data output by the preprocessing unit and the acquisition and adjustment unit, a piecewise consistency function is constructed using a convolution kernel function. The trend consistency of the current change and temperature change within the same time period is compared. If the difference in the trend change exceeds a preset range, it is determined to be a consistency anomaly. When a consistency anomaly is detected, the time point and range of the abnormal fluctuation are recorded. A filter model matching the type of abnormal fluctuation is called from the preset local filter model library. Adaptive filtering is performed on the current and temperature data within the abnormal fluctuation range to obtain the current and temperature data for each consistency range.

[0156] Based on the current data of each consistency interval output by the current parameter unit and the preprocessing unit, as well as the pre-stored battery internal resistance data, the integral of the product of the square of the current and the internal resistance within the time range of each consistency interval is calculated, and the integral result is used as the piecewise current thermal stress factor of the corresponding consistency interval to obtain the piecewise current thermal stress factor.

[0157] Based on the ambient temperature data and internal reference point temperature data of each consistency interval collected by the temperature parameter unit and temperature sensor group, the average ambient temperature and average internal reference point temperature in each consistency interval are calculated, and then the continuous fluctuation of the difference between the two average temperatures is calculated. This fluctuation is used as the piecewise temperature gradient factor of the corresponding consistency interval to obtain the piecewise temperature gradient factor.

[0158] Based on the segmented current thermal stress factor output by the hierarchical coupling unit and the current parameter unit, the segmented temperature gradient factor output by the temperature parameter unit, the battery life decay rate fed back by the subsequent segmented evaluation unit, the average ambient temperature and average internal reference point temperature of each consistency interval, and the ratio of the current slope ratio and the temperature slope ratio output by the fluctuation analysis unit, a segmented temperature-current coupling feature vector is generated through fusion processing.

[0159] Based on the segmented temperature-current coupling feature vector output by the contribution analysis unit and the hierarchical coupling unit, and taking the battery life degradation probability as the analysis target, principal component analysis is used to extract the parameter with the greatest impact on battery life degradation from the segmented temperature-current coupling feature vector as the main evaluation factor, and the remaining parameters as auxiliary evaluation factors. By analyzing the correlation between each evaluation factor and the actual historical battery life degradation data, the contribution of the main evaluation factor and each auxiliary evaluation factor to the battery health status assessment is calculated, and the main evaluation factor, auxiliary evaluation factor and their corresponding contribution are obtained.

[0160] Based on the main evaluation factors, auxiliary evaluation factors and their contribution degrees output by the segmented evaluation unit and the contribution analysis unit, and combined with the voltage data output by the preprocessing unit, the segmented battery health status evaluation value corresponding to each consistency interval is calculated by a comprehensive evaluation algorithm. At the same time, by calculating the fluctuation value of the product of the segmented battery health status evaluation value and the ratio of the segmented current thermal stress factor and the segmented temperature gradient factor, the battery life decay rate is determined according to the magnitude of the fluctuation value, and the segmented battery health status evaluation value and battery life decay rate are obtained.

[0161] Based on the segmented battery health status assessment values ​​and battery life degradation rates output by the hierarchical warning unit and segmented evaluation unit, the system compares the segmented battery health status assessment values ​​with preset two-level warning thresholds: if the segmented battery health status assessment value is higher than the first threshold, no warning signal is generated; if the segmented battery health status assessment value is less than or equal to the first threshold but higher than the second threshold, a battery capacity degradation warning signal is generated, and the remaining driving range is estimated using a range estimation model by combining the current battery power data and the segmented battery health status assessment value; if the segmented battery health status assessment value is less than or equal to the second threshold, or the battery life degradation rate is higher than the preset degradation rate threshold, a battery life end warning signal is generated and the battery charge and discharge protection threshold adjustment is triggered simultaneously; the segmented battery health status assessment values, warning signal types, and remaining driving range estimates are transmitted to the vehicle display and user mobile APP via the CAN bus to obtain the warning signal, remaining driving range estimates, and information transmission results.

[0162] Based on the actual range data within a preset number of usage cycles after the feedback correction unit and the graded early warning unit trigger an early warning, the actual range data is compared with the estimated remaining range value estimated by the graded early warning unit. If the deviation between the two exceeds the preset deviation threshold, the segmented temperature-current coupling feature vector of the corresponding usage cycle is extracted and transmitted to the contribution analysis unit to adjust the contribution weight of the main evaluation factor and the auxiliary evaluation factor.

[0163] Simultaneously, the battery life degradation rate output by the segmented evaluation unit is fed back to the enhanced acquisition and adjustment model of the acquisition and adjustment unit in real time. If the battery life degradation rate accelerates, the acquisition window length and sampling frequency are optimized through the model. In addition, the battery life degradation rate is fed back to the hierarchical coupling unit. If the battery life degradation rate accelerates, the segment interval length of the segmented temperature-current coupling feature vector is adjusted to obtain the adjusted contribution weight, acquisition window length, sampling frequency, and segment interval length, forming a fully closed-loop adaptive processing flow from data acquisition, feature extraction, state assessment to early warning output and feedback optimization.

[0164] The online battery health status assessment system provided in this application achieves high-precision, real-time monitoring and early warning of the battery health status of electric two-wheelers and three-wheelers by constructing a fully closed-loop adaptive processing flow, resulting in significant beneficial effects. The system first uses a curve construction unit to transform current and temperature data with different dimensions and large numerical ranges of multiple physical quantity parameters into co-visible dual-axis time-series curves, laying the foundation for subsequent analysis. The fluctuation analysis unit extracts the slope sequence based on the current and temperature change curves and constructs an acquisition adjustment factor by calculating the ratio of slopes at adjacent time points. This factor, as a dimensionless index characterizing the coordinated acceleration of the battery's electrothermal response, can keenly capture the relative acceleration relationship between current and temperature change trends. It can automatically trigger high-frequency acquisition to capture transient details when the battery experiences severe electrothermal stress changes, and can reduce the acquisition frequency to save resources in stable states. More importantly, it can achieve adaptive acquisition adjustment without relying on a preset scene feature library, effectively solving the pain point problem that traditional fixed-frequency acquisition or single-parameter threshold triggering methods cannot effectively capture dynamic coupling processes. The model training unit, based on initial acquisition parameters and the reciprocal of the battery life decay rate, employs a distributed reinforcement learning algorithm to train a reinforced acquisition adjustment model on historical data under different typical operating conditions. This model, optimized through a multi-objective reward function, intelligently balances data validity, resource consumption, and evaluation error, achieving optimal acquisition strategy. The acquisition adjustment unit deploys the trained reinforced acquisition adjustment model to each sensor node using a federated learning framework, dynamically adjusting the acquisition window length and sampling frequency based on real-time generated scene fingerprints to ensure complete capture of key data. The preprocessing unit monitors the consistency of electrothermal changes by constructing a piecewise consistency function and uses a local filtering model library to adaptively filter abnormal fluctuation intervals, obtaining high-quality consistent interval data. The current parameter unit and temperature parameter unit calculate the piecewise current thermal stress factor and piecewise temperature gradient factor, respectively. These two posterior indicators directly quantify the absolute value of the actual electrothermal stress borne by the battery and its specific impact, complementing the forward-looking acquisition adjustment factor and providing key feature dimensions for accurate evaluation. The hierarchical coupling unit transforms five types of input variables into segmented temperature-current coupling feature vectors through a multivariate fusion algorithm, effectively characterizing the impact of temperature-current coupling on battery health. The contribution analysis unit and segmented evaluation unit use principal component analysis and comprehensive evaluation algorithms, combined with real-time voltage data, to calculate segmented battery health status evaluation values ​​and life decay rates, ensuring the accuracy of the evaluation results. The hierarchical early warning unit generates hierarchical early warning signals and estimated driving range values ​​based on the evaluation values ​​and preset thresholds, providing users with intuitive decision support. The feedback correction unit feeds back the life decay rate to the front-end acquisition module and online evaluation module in real time, dynamically adjusting the acquisition strategy and the segmented interval length of the feature vector to form a complete optimization closed loop.Ultimately, the system achieved end-to-end optimization from data acquisition, feature extraction, status assessment to early warning output, significantly improving the accuracy and reliability of battery health status assessment, extending battery life, and providing effective battery maintenance and safety assurance for electric vehicle users.

[0165] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments under the guidance of the present invention without departing from the spirit and scope of the claims. All of these variations are within the protection scope of the present invention.

Claims

1. A battery health status online assessment system considering temperature-current coupling effect, characterized in that, include: The system comprises a data acquisition module, a coupling module, an online evaluation module, and an early warning module. The acquisition module is used to acquire the battery's current and temperature data in real time according to a preset enhanced acquisition adjustment model. The coupling module is used to obtain a temperature-current coupling feature vector based on the current and temperature data combined with a coupling algorithm. The online evaluation module is used to perform real-time evaluation based on the temperature-current coupling feature vector and a preset online evaluation model to obtain the battery health status evaluation value and the battery life decay rate. The early warning module is used to issue graded early warnings based on the battery health status assessment value and the set graded early warning thresholds. At the same time, it feeds back the battery life decay rate to the enhanced acquisition adjustment model in the acquisition module and the online evaluation model to adjust the data acquisition strategy and evaluation index weights in real time.

2. The online battery health status assessment system considering temperature-current coupling effect as described in claim 1, characterized in that, The acquisition module includes a curve construction unit and a fluctuation analysis unit; The curve construction unit is used to combine current and temperature data from different historical scenarios with a dual-axis curve, with time as the horizontal axis, current as the left vertical axis, and temperature as the right vertical axis, to construct current change curves and temperature change curves. The fluctuation analysis unit obtains current slope sequences and temperature slope sequences based on current change curves and temperature change curves; and constructs an acquisition adjustment factor based on the ratio of current slope ratio to temperature slope ratio at adjacent time points. Simultaneously, based on the current change curve and the temperature change curve, the length of the synchronous time interval between the current slope and the temperature slope is obtained through time series analysis algorithm as the initial acquisition window length, and the data acquisition frequency of the corresponding time interval is used as the initial acquisition frequency.

3. The online battery health status assessment system considering temperature-current coupling effect as described in claim 2, characterized in that, The acquisition module also includes a model training unit; The model training unit, based on the initial acquisition window length, initial acquisition frequency, acquisition adjustment factor, and the reciprocal of the battery life decay rate, combines a distributed reinforcement learning algorithm with corresponding current and temperature data from different historical scenarios to train the acquisition parameters and obtain a trained reinforcement acquisition adjustment model.

4. The online battery health status assessment system considering temperature-current coupling effect as described in claim 3, characterized in that, The acquisition module also includes an acquisition adjustment unit; The acquisition and adjustment unit is used to configure the trained enhanced acquisition and adjustment model into the corresponding sensor in conjunction with the federated framework, and to adjust the length of the differential current and temperature data acquisition window and the acquisition frequency by combining the scene fingerprint constructed by the current slope ratio and temperature slope ratio at adjacent time points with the acquisition adjustment factor. The sensor includes a current sensor, a temperature sensor group, and a voltage sensor; the temperature sensor group is configured to acquire the ambient temperature at at least one location on the battery surface and to acquire the temperature at a reference point inside the battery via a temperature sensor attached to the battery terminal.

5. The online battery health status assessment system considering temperature-current coupling effect as described in claim 4, characterized in that, The coupling module includes a preprocessing unit; The preprocessing unit is used to monitor the consistency relationship between current and temperature changes in real time based on a piecewise consistency function constructed by combining the current slope and temperature slope with the convolution kernel function. Based on the consistency relationship between current and temperature changes in different time periods, combined with a preset local filtering model library and the time points and ranges of abnormal consistency fluctuations, the unit calls the local filtering model in the corresponding time period to perform data filtering processing, thereby obtaining the preprocessed current and temperature data for each consistency interval.

6. The online battery health status assessment system considering temperature-current coupling effect as described in claim 5, characterized in that, The coupling module further includes a current parameter unit and a temperature parameter unit; The current parameter unit is used to calculate the integral of the product of the square of the current and the resistance and time per unit time based on the current data of each consistent interval after preprocessing, so as to obtain the segmented current thermal stress factor. The temperature parameter unit is used to obtain the piecewise temperature gradient factor based on the fluctuation of the difference between the average ambient temperature and the average internal reference point temperature within the corresponding time length of each consistency interval.

7. The online battery health status assessment system considering temperature-current coupling effect as described in claim 6, characterized in that, The coupling module further includes a hierarchical coupling unit; The hierarchical coupling unit is used to obtain a segmented temperature-current coupling feature vector by combining the ratio of segmented temperature gradient factor to segmented current thermal stress factor, average ambient temperature, average internal reference point temperature, battery life decay rate, and the ratio of temperature slope ratio to current slope ratio with a segmented coupling model. At the same time, it performs real-time calibration of the segmented interval length of the segmented temperature-current coupling feature vector by the battery life decay rate.

8. The online battery health status assessment system considering temperature-current coupling effect as described in claim 7, characterized in that, The online evaluation module includes a contribution analysis unit and a segmented evaluation unit; The contribution analysis unit is used to obtain the main evaluation factor and auxiliary evaluation factor and their corresponding contribution for each segment based on the segmented temperature-current coupling feature vector, with the battery life decay probability as the target, combined with principal component analysis. The segmented evaluation unit is used to evaluate the battery health status based on the main evaluation factor and auxiliary evaluation factor and their corresponding contribution in each segment, combined with a comprehensive evaluation algorithm, to obtain the segmented battery health status evaluation value. Based on the fluctuation value of the product of the battery health status evaluation value and the ratio of the current thermal stress factor to the temperature gradient factor, the battery life decay rate is obtained.

9. The online battery health status assessment system considering temperature-current coupling effect as described in claim 8, characterized in that, The early warning module includes a tiered early warning unit; The graded early warning unit is used to compare the battery health status assessment value with the preset multi-level early warning threshold in real time. When the battery health status assessment value is lower than the first threshold, a first-level warning signal is generated and output. The first-level warning signal includes a battery capacity decay prompt and an estimated remaining driving range calculated based on the current battery level and the battery life decay rate. When the battery health status assessment value is lower than the preset second threshold, a second-level warning signal is generated and output. The second-level warning signal is a warning of the end of battery life and a replacement reminder. The health status assessment value, the warning signal and the estimated remaining driving range are transmitted to the vehicle display screen or the user's mobile terminal APP for display.

10. The online battery health status assessment system considering temperature-current coupling effect as described in claim 9, characterized in that, The early warning module also includes a feedback correction unit; The feedback correction unit is used to receive the battery life degradation rate from the online evaluation module in real time, and transmit the battery life degradation rate as a feedback signal to the enhanced acquisition adjustment model in the acquisition module to dynamically adjust the acquisition window length and acquisition frequency in the data acquisition strategy. Simultaneously, the battery life degradation rate is transmitted to the hierarchical coupling unit to adaptively correct the segment interval length of the segmented temperature-current coupling feature vector and adjust the segmented battery health status assessment value in real time.

Images