An adaptive communication-free electric vehicle battery pack health detection method and system

Abstract

The application provides a self-adaptive non-communication electric vehicle battery pack health detection method and system, relates to the technical field of battery health detection, and has the technical scheme as follows: time series data of voltage, current and temperature in the charging process are collected in a non-invasive manner, a plurality of aging characteristic quantities representing different electrochemical aging mechanisms are extracted, a corresponding reference characteristic set is matched according to the battery material type, a health degree value is obtained through normalized characteristic ratio weighted summation, and thus, in the case of not depending on battery communication and complete charging process, self-adaptive and accurate detection of the health degree of different types of battery packs is realized, the universality and reliability of detection are effectively improved, and the method has the advantages that accurate and self-adaptive detection of the health degree of battery packs of different material types and voltage platforms can be realized without depending on battery management system communication and complete charging data, the universality and reliability of detection are improved, and thermal runaway accidents are effectively prevented.

Description

Technical Field

[0001] This application relates to the field of battery health detection technology, and more specifically, to an adaptive method and system for detecting the health of a battery pack in a non-communication electric vehicle. Background Technology

[0002] With the widespread adoption of electric vehicles, they have become an important means of daily transportation. However, during long-term use, electric vehicle battery packs are prone to fires and thermal runaway during charging due to factors such as battery aging, unauthorized modifications, improper charging, and abnormal internal health conditions, posing a serious threat to public safety. To effectively prevent such accidents, it is necessary to accurately monitor the State of Health (SOH) of electric vehicle battery packs to promptly identify batteries with potential safety hazards.

[0003] In practical civilian applications, conducting health testing on electric vehicle battery packs faces numerous objective challenges. First, the types of electric vehicle batteries on the market are extremely diverse. Common battery materials include ternary lithium batteries, lithium iron phosphate batteries, lithium manganese oxide batteries, and lead-acid batteries. Different material systems exhibit vastly different electrochemical characteristics during charging, resulting in varying voltage change patterns. Second, battery nominal voltage platforms vary widely, such as 36V, 48V, 60V, and 72V, and a large number of non-standard batteries are privately modified by users, making it difficult for standardized testing equipment to be effectively adapted. More critically, the vast majority of civilian electric vehicle battery packs on the market do not have a pre-existing communication interface for the Battery Management System (BMS) that can be accessed by external devices. Testing equipment cannot directly read the battery's internal state parameters and historical data via communication. Traditional methods relying on BMS data for capacity degradation assessment or internal resistance calculation are essentially ineffective without communication.

[0004] In the absence of internal communication methods, existing technologies typically attempt to collect operating parameters such as voltage, current, and temperature during the charging process using non-invasive external sensors, and then combine this data with the battery's factory-specified capacity or a complete charge-discharge curve to estimate its health. However, these methods are highly dependent on obtaining standard battery parameters and a complete charging process. In real-world scenarios, user charging behavior is often fragmented, such as temporary short-term charging, making it difficult to collect a complete charging data curve from low to full charge. Furthermore, even if partial charging data can be collected, how to convert the collected absolute physical quantities such as voltage and current into relative indicators that reflect the degree of battery aging, while adapting to the differences in chemical characteristics among different battery types, remains a problem that current technologies have not yet adequately solved, especially in the absence of reference information such as nominal capacity.

[0005] Furthermore, in non-invasive testing without communication, battery health assessment typically relies on a series of aging characteristic parameters extracted from the charging or resting phases, such as the self-discharge voltage drop characteristic after charging, which characterizes the battery's self-discharge level and internal micro-short circuit risk. Under normal operating conditions at room temperature, the polarization voltage accumulated inside the battery after charging can dissipate rapidly within a preset fixed resting window, and the subsequently measured static terminal voltage drop can relatively purely reflect the self-discharge characteristics. However, when the ambient temperature decreases or the battery's internal damping increases due to aging, the dissipation and relaxation process of polarization charge slows down significantly, and the electrochemical time constant is extended exponentially. In this case, if sampling is still performed according to a fixed resting time, the measured terminal voltage drop will be deeply mixed with residual polarization voltage that has not completely dissipated, resulting in an abnormally high extracted self-discharge voltage drop characteristic parameter, which in turn distorts the health calculation results and may even lead to misjudgment of battery micro-short circuit faults and unnecessary charging safety interception. This problem of unstable self-discharge feature extraction due to changes in polarization dissipation rate is common in real-world operating environments with large temperature fluctuations or poor battery consistency, posing new challenges to the accuracy and stability of non-communication health detection.

[0006] As the above analysis shows, in the context of diverse types of electric vehicle battery packs, a large number of battery packs lacking BMS communication interfaces, fragmented charging processes, and complex and variable operating environments, how to accurately and stably calculate the health status of battery packs of different types and voltage platforms using only externally collected non-invasive parameters such as charging voltage, current, and temperature, without relying on the battery's nominal capacity and the complete charging process, and to accurately identify battery aging and safety hazards, is a technical problem that urgently needs to be solved in this field.

[0007] To address the aforementioned issues, existing technologies urgently need improvement. Summary of the Invention

[0008] The purpose of this application is to provide an adaptive non-communication electric vehicle battery pack health detection method and system, which can accurately and adaptively detect the health of battery packs of different material types and voltage platforms without relying on the communication and complete charging data of the battery management system, thereby improving the universality and reliability of the detection and effectively preventing thermal runaway accidents.

[0009] This application provides an adaptive health detection method for battery packs in non-communication electric vehicles, the technical solution of which is as follows:

[0010] An adaptive health detection method for battery packs in non-communication electric vehicles, the method comprising:

[0011] During the charging process of the battery pack, the voltage, current and temperature timing data of the battery pack are collected non-invasively by the external sensors of the charger.

[0012] Multiple aging characteristic quantities are extracted from time-series data to characterize different electrochemical aging mechanisms of the battery pack, and the material type of the battery pack is determined based on the time-series data.

[0013] Obtain a pre-stored set of benchmark features corresponding to the material type, obtained from a brand-new battery pack of the same type under standard charging conditions;

[0014] The normalized feature ratio is obtained by comparing each aging feature with the corresponding benchmark value in the benchmark feature set. Weights are assigned to each normalized feature ratio according to the material type. All weighted normalized feature ratios are summed to obtain a health value that characterizes the health status of the battery pack.

[0015] Furthermore, this application also proposes that the aging characteristic quantities include:

[0016] The voltage step characteristics at the moment of power-on are extracted based on the jump difference of the battery terminal voltage at the moment of power-on; the short-time voltage rise rate characteristics are extracted based on the slope of the change of the battery terminal voltage over time in the initial stage of charging; the constant voltage current drop rate characteristics are extracted based on the decay slope of the output current decreasing over time in the constant voltage charging stage; and the self-discharge voltage drop characteristics are extracted based on the static voltage drop of the battery terminal voltage in the resting stage after charging.

[0017] Furthermore, this application also proposes assigning weights to each normalized characteristic ratio based on the material type, including:

[0018] Obtain the preset weight value corresponding to the material type, and weight each normalized feature ratio according to the preset weight value.

[0019] Furthermore, this application proposes that, when a constant-voltage charging phase is detected to be missing in the current charging process, weights are assigned to each normalized characteristic ratio based on the material type, including:

[0020] The weight corresponding to the normalized characteristic ratio obtained based on the constant voltage section current decrease rate characteristic is reduced, and the reduced weight share is allocated to the weight of the normalized characteristic ratio corresponding to at least one other aging characteristic quantity.

[0021] Furthermore, this application also proposes that, when the collected ambient temperature meets preset temperature anomaly conditions, weights be assigned to each normalized characteristic ratio based on the material type, including:

[0022] The weight of the normalized feature ratio corresponding to the first preset aging feature quantity is reduced, and the weight reduction share generated by the reduction is allocated to the weight of the normalized feature ratio corresponding to at least one second preset aging feature quantity.

[0023] The stability of the first preset aging characteristic quantity under abnormal temperature conditions is lower than that of the second preset aging characteristic quantity under abnormal temperature conditions.

[0024] Furthermore, this application also proposes that the self-discharge voltage drop characteristics after charging are determined by the release state, which characterizes the dissipation of polarization voltage during the resting phase after charging, and the release state is determined based on the charging phase characteristic information extracted during the charging phase.

[0025] Furthermore, this application also proposes that determining the extraction method based on the release status includes:

[0026] The static voltage timing data during the static phase is divided into multiple consecutive time periods;

[0027] Determine the slope of the voltage change at the battery terminals during each time period;

[0028] In multiple time periods, based on the convergence condition of the decay of the voltage change slope between adjacent time periods, the slow-changing region in the resting stage, dominated by the self-discharge process, is determined.

[0029] Based on the voltage drop at the battery terminals in the slow-changing region, the characteristics of the self-discharge voltage drop after charging are extracted.

[0030] Furthermore, this application also proposes that the self-discharge voltage drop characteristics after charging be extracted in the following manner:

[0031] The residual polarization component is estimated based on the termination slope information of the battery terminal voltage at the end or termination of the resting phase.

[0032] Based on the total voltage drop during the resting phase, the residual polarization component is subtracted to obtain the self-discharge voltage drop characteristics after charging.

[0033] Furthermore, this application also proposes that, before extracting aging features from time-series data, the following steps are also included:

[0034] Standard cleaning is performed on time series data. The standard cleaning process determines the cleaning interval based on the mean and standard deviation of the time series data and removes outlier data that exceeds the cleaning interval.

[0035] When the collected ambient temperature meets the preset temperature anomaly conditions, or the initial charging voltage of the battery pack meets the preset low voltage conditions, the boundary range of the cleaning interval is expanded, or the time series data is temperature compensated before the standard cleaning process is performed, so as to retain the normal fluctuation data in the time series data under the conditions.

[0036] In addition, this application also provides an adaptive non-communication electric vehicle battery pack health detection system, the technical solution of which is as follows:

[0037] An adaptive non-communication electric vehicle battery pack health detection system for performing the above method includes: a data acquisition module for non-invasively acquiring timing data of the battery pack's voltage, current, and temperature from external sensors of the charger during the battery pack charging process;

[0038] The feature extraction and material determination module is used to extract multiple aging feature quantities from time-series data to characterize different electrochemical aging mechanisms of the battery pack, and to determine the material type of the battery pack based on the time-series data.

[0039] The benchmark acquisition module is used to acquire a pre-stored benchmark feature set corresponding to the material type, obtained from a brand-new battery pack of the same type under standard charging conditions;

[0040] The health calculation module is used to calculate the ratio of each aging characteristic quantity to the corresponding benchmark value in the benchmark feature set to obtain a normalized characteristic ratio. It also assigns weights to each normalized characteristic ratio according to the material type, and sums up all the weighted normalized characteristic ratios to obtain a health value that characterizes the health status of the battery pack.

[0041] As can be seen from the above, the adaptive non-communication electric vehicle battery pack health detection method and system provided in this application non-invasively collects voltage, current and temperature time-series data during the charging process, extracts multiple aging characteristic quantities characterizing different electrochemical aging mechanisms, and matches corresponding benchmark feature sets according to the battery material type. The health value is obtained by weighted summation of normalized feature ratios. Thus, without relying on battery communication and the complete charging process, it achieves adaptive and accurate detection of the health of different types of battery packs, effectively improving the universality and reliability of the detection. It has the advantages of being able to accurately and adaptively detect the health of battery packs with different material types and voltage platforms without relying on battery management system communication and complete charging data, improving the universality and reliability of the detection, and effectively preventing thermal runaway accidents. Attached Figure Description

[0042] Figure 1 A schematic diagram of the method flow provided in this application.

[0043] Figure 2 A schematic diagram of the overall process provided for this application. Detailed Implementation

[0044] The following description, in conjunction with the technical solution of this application, provides a clearer and more complete explanation of the relevant content. It should be noted that the embodiments described herein are only a part of the implementation methods of this application, and not all of them. Other implementation methods obtained by those skilled in the art based on the embodiments of this application without creative effort should also fall within the protection scope of this application. Furthermore, in the description of this application, the terms "first," "second," etc., are mainly for distinction and should not be construed as indicating relative importance.

[0045] Reference Figure 1 This application proposes an adaptive health detection method for battery packs in non-communication electric vehicles, the method comprising:

[0046] S1. During the battery pack charging process, the voltage, current and temperature timing data of the battery pack are collected non-invasively by the charger's external sensors.

[0047] S2. Extract multiple aging characteristic quantities from the time series data to characterize different electrochemical aging mechanisms of the battery pack, and determine the material type of the battery pack based on the time series data.

[0048] S3. Obtain the pre-stored reference feature set corresponding to the material type, obtained from a brand new battery pack of the same type under standard charging conditions.

[0049] S4. Calculate the ratio of each aging characteristic quantity to the corresponding benchmark value in the benchmark characteristic set to obtain the normalized characteristic ratio. Assign weights to each normalized characteristic ratio according to the material type. Sum all the weighted normalized characteristic ratios to obtain the health value that characterizes the health status of the battery pack.

[0050] When an electric vehicle battery pack is connected to a smart charger to begin charging, the external sensors inside the charger do not need to communicate or interact with the battery management system inside the battery pack. They can directly and non-intrusively collect the voltage, current, and temperature timing data of the battery pack from the charging circuit. This timing data can be acquired in real time at a set sampling frequency and uploaded to a backend server for persistent storage. This data acquisition method differs from traditional solutions that rely on the battery's internal communication interface, making this method directly applicable to the large number of civilian electric vehicle battery packs on the market that do not have external communication interfaces.

[0051] After retrieving the time-series data stored in the database for this charging process, the backend server extracts several aging characteristic quantities to characterize different electrochemical aging mechanisms of the battery pack. These aging characteristic quantities correspond to different degradation paths within the battery. For example, the voltage step feature at the moment of power-on can be extracted from the voltage jump difference at the moment of power-on to characterize the ohmic internal resistance; the short-time voltage rise rate feature can be extracted from the slope of the battery terminal voltage change over time in the initial stage of charging to characterize the remaining state of the effective active material; the constant-voltage current drop rate feature can be extracted from the decay slope of the output current decreasing over time in the constant-voltage charging stage to characterize the polarization degree and cell consistency; and the self-discharge voltage drop feature after charging can be extracted from the static voltage drop of the battery terminal voltage in the resting stage after charging to characterize self-discharge and internal micro-short circuit characteristics. At the same time, based on the voltage, current, and other physical quantities in the initial stage of the time-series data, combined with preset classification logic, the material type of the current battery pack is automatically determined, such as whether the battery pack is a ternary lithium battery, lithium iron phosphate battery, lithium manganese oxide battery, or lead-acid battery.

[0052] After determining the material type, the backend server retrieves a pre-stored benchmark feature set corresponding to that material type. This benchmark feature set is obtained by testing brand-new battery packs with the same material type and voltage platform under standard charging conditions in a standard laboratory environment. It includes new state benchmark values ​​corresponding to various aging characteristics, such as standard power-on step voltage benchmark value, standard short-time voltage rise rate benchmark value, standard constant voltage current drop rate benchmark value, and standard self-discharge voltage drop benchmark value. These benchmark values ​​characterize the normal electrochemical response level of brand-new batteries of the same type in a healthy, non-deteriorated state.

[0053] After obtaining the aging characteristics and the baseline feature set, the backend server performs a ratio calculation between the actual measured value of each aging characteristic and the corresponding baseline value in the baseline feature set to obtain a normalized feature ratio. For example, for the voltage step characteristic at the moment of power-on, its baseline value is divided by the actual measured value to obtain a ratio reflecting the degree of battery degradation relative to its brand-new state under that feature dimension. Through this ratio normalization process, the absolute physical quantities that were originally strongly correlated with the battery's nominal capacity and specific charging conditions are transformed into relative degradation ratios, thereby eliminating the dependence on the battery's specific nominal capacity data and complete charging curves.

[0054] After obtaining the normalized characteristic ratios for each dimension, the backend server adaptively assigns weights to each ratio based on the previously determined battery pack material type. Different battery pack material types exhibit varying degrees of manifestation and risk weights for different electrochemical mechanisms during aging. For example, the increased ohmic internal resistance and loss of active material in ternary lithium batteries have a significant impact on safety, thus requiring higher weights for the voltage step characteristic and short-term voltage rise rate characteristic. Conversely, self-discharge and micro-short circuit issues are key safety concerns for lithium iron phosphate batteries, necessitating a higher weight for the self-discharge characteristic. The weighted normalized characteristic ratios are then summed to obtain a health score characterizing the overall health of the battery pack, typically ranging from 0 to 1 or expressed as a percentage.

[0055] Specifically, suppose a user illegally modifies an electric vehicle that originally used a 48-volt lead-acid battery to a 60-volt high-capacity ternary lithium battery pack, and this modified battery pack lacks an interface for external communication with the battery management system. When the user charges the vehicle using a regular smart charger, traditional detection methods become completely ineffective because they cannot obtain internal battery data and cannot determine the actual nominal capacity of the modified battery. Therefore, this solution performs a health check:

[0056] In the S1, the smart charger can directly collect the timing data of voltage, current, and temperature of the battery pack during the charging process from the charging output port through its external voltage, current, and temperature sensors the moment it is powered on, without requiring any communication data from the battery pack. Even if the capacity and voltage platform of the modified battery are unknown, the sensors can still objectively record the change curves of external physical quantities throughout the entire process from the start to the end of charging.

[0057] In S2, the backend server extracts four core aging characteristics from these time-series data. Specifically, within 0 to 3 seconds of charging, the maximum voltage jump difference at the battery terminals is collected as a voltage step characteristic at power-on. The more pronounced the step, the greater the ohmic internal resistance of the battery, reflecting potential thickening of the negative electrode SEI film or electrolyte degradation in the modified battery. During the initial charging phase (0 to 60 seconds), the average slope of the battery terminal voltage change over time is statistically analyzed as a short-term voltage rise rate characteristic. If the battery has experienced active material shedding due to long-term use, the voltage will rise faster under the same charging current. During the constant-voltage charging phase, the decay slope of the output current over time is statistically analyzed as a constant-voltage current decrease rate characteristic. If the modified battery suffers from severe polarization due to poor cell consistency or lithium deposition, the current decrease rate will be abnormally rapid. After charging is completed and the battery has been left to rest for a period of time, the static voltage drop at the battery terminals is collected as a self-discharge voltage drop characteristic, reflecting the degree of chronic leakage and micro-short circuits within the battery. Meanwhile, based on the voltage amplitude and charging current response characteristics collected during the initial charging stage, the server automatically identifies the battery pack's material type as ternary lithium and its voltage platform as 60 volts by comparing it with electrochemical feature templates of different material systems in the background database.

[0058] In S3, the backend server retrieves a set of benchmark features for a new ternary lithium battery under standard charging conditions from a pre-stored benchmark feature database, based on the identified ternary lithium material type and 60-volt voltage platform. This set of benchmark features records the electrochemical response parameters that a new battery of the same type should have under standard on-state step voltage, standard short-time voltage rise rate, standard constant-voltage current drop rate, and standard self-discharge voltage drop.

[0059] In S4, the backend server calculates the ratios of the measured values ​​of the four aging characteristics of the modified battery to their corresponding benchmark values. For example, dividing the standard power-on step voltage benchmark value by the measured power-on step voltage value yields a normalized characteristic ratio less than 1. The closer this ratio is to 1, the less severe the aging in that dimension. Because the ratio calculation eliminates the numerical differences caused by the battery's specific nominal capacity and charging conditions, even when the battery capacity has changed significantly through modification, a stable ratio reflecting the relative degradation of each dimension can still be obtained. Subsequently, the system assigns matching weights to the four normalized characteristic ratios according to the material type characteristics of ternary lithium batteries. For example, the power-on voltage step characteristic, which reflects ohmic internal resistance, is assigned a higher weight; the short-time voltage rise rate characteristic, which reflects the active material, is assigned a second-highest weight; and the constant voltage current drop rate characteristic and self-discharge characteristic are assigned moderate weights. After summing the weighted ratios of the four normalized features, the overall health value of the modified battery pack was obtained. It was ultimately determined that the battery had been severely aged due to long-term modification and use, and its health value was below the safety threshold, which was considered a dangerous state.

[0060] In some specific implementations, refer to Figure 2 This method demonstrates the specific process of adaptive non-communication electric vehicle battery pack health detection. First, while the battery pack is charging, external sensors on the charger non-invasively collect real-time time-series data on the battery pack's voltage, current, and temperature. After data acquisition, the time-series data is analyzed to extract multiple aging characteristic quantities that can characterize different electrochemical aging mechanisms of the battery pack, and the material type of the battery pack is determined based on the time-series data. Next, a pre-stored benchmark feature set is acquired. This benchmark feature set corresponds to the determined material type and is obtained from testing brand-new battery packs of the same type under standard charging conditions. Finally, the ratio of each extracted aging characteristic quantity to the corresponding benchmark value in the benchmark feature set is calculated to obtain a normalized feature ratio. Simultaneously, corresponding weights are assigned to each normalized feature ratio according to the determined material type. All weighted normalized feature ratios are summed to obtain a health value reflecting the current health status of the battery pack.

[0061] Relying solely on externally acquired, non-invasive time-series data of voltage, current, and temperature, extracting only a single feature is insufficient to comprehensively reflect the different aging mechanisms of a battery, easily leading to misjudgments or omissions in battery health assessment. To address this, this application proposes that aging features include: a voltage step feature extracted from the voltage jump difference at the moment of charging; a short-term voltage rise rate feature extracted from the slope of the battery terminal voltage change over time during the initial charging phase; a constant-voltage current decrease rate feature extracted from the decay slope of the output current decreasing over time during the constant-voltage charging phase; and a self-discharge voltage drop feature extracted from the static voltage drop of the battery terminal voltage during the resting phase after charging.

[0062] Specifically, the rapid voltage jump at the moment of charging is initiated is primarily governed by the ohmic internal resistance. The maximum voltage jump difference at the moment of charging is initiated can be collected as a voltage step characteristic. This jump difference can be extracted within 0 to 3 seconds after the start of charging. This characteristic can characterize the ohmic internal resistance of the battery; a larger voltage jump indicates a higher ohmic internal resistance, corresponding to aging issues such as thickening of the negative electrode solid electrolyte interface film or electrolyte degradation.

[0063] During the initial charging phase, the rate of voltage rise at the battery terminals is closely related to the number of electrochemical active sites. The average slope of the voltage change over time during the initial charging phase can be statistically analyzed as a short-term voltage rise rate characteristic. This average slope can be calculated within 0 to 60 seconds after the start of charging. Batteries with active material shedding or capacity decay have fewer electrochemical active sites and will exhibit a faster voltage rise under the same charging current; therefore, this characteristic can characterize the remaining state of the battery's effective active material.

[0064] During the constant-voltage charging phase, the rate of current decay reflects the degree of polarization and cell consistency. The rate of current decay over time during the constant-voltage charging phase can be statistically analyzed as a characteristic of the current decrease rate in the constant-voltage segment. When battery polarization is severe or cell consistency is poor, the rate of current decay in the constant-voltage segment will be abnormally accelerated; this characteristic can characterize the deviation in cell polarization and consistency.

[0065] After charging is complete and the battery has been allowed to rest, the slow decrease in battery terminal voltage is mainly due to internal self-discharge and micro-short circuits. The static voltage drop at the battery terminal during the resting period after charging can be used as a characteristic of the self-discharge voltage drop after charging. This static voltage drop can be collected 30 minutes after charging is complete and the battery has been allowed to rest. Resting allows most of the rapidly relaxing polarization voltage to dissipate; the voltage drop measured at this time is mainly caused by slow self-discharge behavior and internal micro-short circuits. A larger voltage drop indicates more severe internal self-discharge or a risk of internal micro-short circuits.

[0066] In addition, while extracting the above features, the voltage difference during the constant current stage can be calculated in chronological order and divided by the corresponding charging current to calculate the dynamic internal resistance value. After data cleaning, the average value of the calculated dynamic internal resistance value is obtained to get the average dynamic internal resistance difference value, which is used to help evaluate the degree of degradation of the battery internal resistance. The average dynamic internal resistance difference value is an optional aging characteristic quantity.

[0067] By extracting the four features corresponding to different aging mechanisms from different stages of a single charging process, a multi-dimensional electrochemical analysis of battery health status is achieved. The changing trend of each feature is directly related to the specific electrochemical aging mechanism, enabling the subsequently calculated health value to analyze the specific cause of aging and provide clear fault indications for safety warnings.

[0068] After obtaining the normalized eigenvalues, it is necessary to consider how to assign eigenvalue weights to batteries of different material types. Because batteries of different material systems exhibit differences in electrochemical properties, the contribution of different aging characteristics to battery health varies across different material systems. If a uniform eigenvalue weight is applied to all types of batteries, the calculated health value may not accurately reflect the true aging state of each battery type.

[0069] Furthermore, in some preferred methods, weights are assigned to each normalized feature ratio based on the material type, including: obtaining a preset weight value corresponding to the material type, and weighting each normalized feature ratio according to the preset weight value.

[0070] Pre-defined weight values ​​can be stored in the standard benchmark database on the backend server. Once the material type of the battery pack under test is determined, the pre-defined weight value table associated with that material type can be retrieved from the standard benchmark database. This pre-defined weight value table defines the weight coefficients of each normalized characteristic ratio under that material system. After obtaining this set of values, they can be assigned to the corresponding normalized characteristic ratios as multiplier factors in the subsequent linear weighted summation calculation of health, and the sum of all weight coefficients satisfies α+β+γ+δ+σ=1.

[0071] The preset weight values ​​can differ depending on the material type. When the material type is ternary lithium battery, due to the significant internal resistance and thermal effect of ternary lithium batteries under aging conditions, thermal runaway is easily triggered by the increase in ohmic internal resistance, and the loss of active material affects the lifespan. Therefore, a higher weight can be assigned to the voltage step characteristic at the moment of power-on. For example, the weight of the voltage step characteristic at the moment of power-on can be set to 0.35, the weight of the short-term voltage rise rate characteristic to 0.30, the weight of the constant voltage current drop rate characteristic to 0.20, the weight of the self-discharge voltage drop characteristic after charging to 0.10, and the weight of the average dynamic internal resistance difference to 0.05.

[0072] When the material type is lithium iron phosphate battery, due to the flat charging and discharging voltage plateau of lithium iron phosphate batteries, conventional voltage characteristics may not be able to capture its subtle aging. However, micro-short circuits and self-discharge anomalies are key areas for safety prevention, so the weight of the self-discharge voltage drop characteristic after charging can be increased. For example, the weight of the voltage step characteristic at the moment of power-on can be set to 0.30, the weight of the short-term voltage rise rate characteristic to 0.25, the weight of the constant voltage current drop rate characteristic to 0.20, the weight of the self-discharge voltage drop characteristic after charging to 0.20, and the weight of the average dynamic internal resistance difference to 0.05.

[0073] When the material type is lead-acid battery or lithium manganese oxide battery, the aging of lead-acid battery is mainly manifested as plate sulfation and grid corrosion, which will lead to an increase in ohmic internal resistance. The weight of the voltage step characteristic at the moment of power-on can be set to 0.40, the weight of the short-time voltage rise rate characteristic can be set to 0.25, the weight of the constant voltage current drop rate characteristic can be set to 0.15, the weight of the self-discharge voltage drop characteristic after charging can be set to 0.15, and the weight of the average dynamic internal resistance difference can be set to 0.05.

[0074] By assigning different base weights to batteries with different material systems, aging characteristics sensitive to specific material systems can be amplified while reducing interference from insensitive characteristics. This allows health calculations to more accurately reflect the true degradation level of different battery types. Furthermore, these preset weight values ​​can serve as benchmarks for subsequent dynamic adjustments under specific operating conditions, ensuring the adaptive capability of the entire testing process is built on a solid foundation.

[0075] In everyday civilian use, users' charging behavior often exhibits a distinctly fragmented characteristic. For example, a user might unplug the charger before the battery is fully charged due to a short trip. This fragmented charging behavior causes the entire charging process to end prematurely, thus missing the constant-voltage charging phase. Since the constant-voltage current drop rate characteristic depends on the current decay data during the constant-voltage charging phase, when this phase is missing, this characteristic cannot be effectively extracted, or the extracted data is completely unreliable. If, under such circumstances, this characteristic is still included in the health calculation according to a fixed weight set based on material type, the invalid data will severely interfere with the final health value, causing the calculation result to deviate from the battery's true state.

[0076] In response, this application proposes that when a constant voltage charging stage is detected to be missing in the current charging process, the weights allocated to each normalized characteristic ratio according to the material type include: reducing the weights corresponding to the normalized characteristic ratios obtained based on the current decrease rate characteristics of the constant voltage stage, and allocating the reduced weight share to the weights of the normalized characteristic ratios corresponding to at least one other aging characteristic quantity.

[0077] Specifically, to accurately determine whether the current charging process is missing a constant-voltage charging phase, this can be identified by monitoring the charging phase status indicators reported by the charger. If the charger's status indicators do not include a constant-voltage phase indicator, it can be determined that the phase is missing. Alternatively, this can be determined by analyzing the characteristics of changes in the collected voltage and current timing data. For example, if the charging current drops to zero before the battery terminal voltage reaches the preset constant-voltage threshold, and there is no interval in the entire timing data where the voltage remains constant while the current continuously decreases, it can be confirmed that the current charging process has not undergone a constant-voltage charging phase.

[0078] It is worth noting that the reason this application dynamically adjusts the weights when a missing constant-voltage charging stage is detected is because the current drop rate characteristic in the constant-voltage stage loses its physical basis for characterizing the battery polarization and cell consistency under this operating condition. To avoid this failure characteristic negatively impacting the health calculation, its corresponding weight needs to be reduced.

[0079] In practical applications, the reduction margin can be flexibly set according to the battery type and the fragmentation of charging conditions. Taking a common ternary lithium battery as an example, under complete charging conditions, the weight corresponding to the constant voltage stage current drop rate characteristic can usually be set to 0.20. When a missing constant voltage charging stage is detected, this weight can be reduced from 0.20 to 0.05. This significant reduction can eliminate the undue influence of failure characteristics on the weighted summation result.

[0080] After reducing the weights, to ensure that the sum of all feature weights still meets the calculation requirements, the resulting weight share needs to be allocated to the weights of at least one other aging feature corresponding to the normalized feature ratio. When selecting the allocation targets, priority should be given to features that can still be stably extracted under constant pressure conditions and contribute significantly to the aging characterization of the current battery material system.

[0081] Continuing with the example of ternary lithium batteries, the reduced weight of 0.15 can be equally redistributed to the weights corresponding to the voltage step characteristic at power-on and the self-discharge voltage drop characteristic after charging. Specifically, the weight of the voltage step characteristic at power-on can be increased from 0.35 to 0.425, while the weight of the self-discharge voltage drop characteristic after charging can be increased from 0.10 to 0.175. Through this equal redistribution, the focus of health calculation shifts to aging characteristics that remain reliable in the current fragmented charging scenario.

[0082] This dynamic adjustment mechanism applies not only to ternary lithium batteries but also to battery packs made of other materials, such as lithium iron phosphate and lead-acid batteries. Although the basic weight configurations differ for different material types, when faced with non-constant voltage operating conditions, the stability of the health calculation can be maintained by reducing the weight of the constant voltage current drop rate characteristic and transferring its share to other stable characteristics.

[0083] Through the above processing, the weighting system transforms from a fixed preset configuration based on material type to a dynamic configuration adapted to non-constant voltage charging conditions. This shift ensures that the health status value primarily reflects the electrochemical aging state characterized by actually measurable features, thus avoiding misjudgments caused by the failure of a single feature. In civilian scenarios with numerous fragmented charging behaviors, this effectively guarantees the stability and accuracy of health status detection, allowing the entire detection method to no longer require a complete charging process. Consequently, it enables reliable assessment of the health status of various battery packs without relying on nominal capacity and a complete charging process.

[0084] In practical use, the electrochemical reaction kinetics inside the battery change when the ambient temperature is extremely low or high. For example, at low temperatures, the lithium-ion migration rate decreases, leading to an increase in polarization resistance. This causes fluctuations in the values ​​of some aging characteristics extracted based on the principle of electrochemical polarization, and their normalized characteristic ratios deviate from the true aging state. If the fixed weighting configuration under normal temperature conditions is continued, distorted characteristics will be given higher weights, resulting in calculated health values ​​that cannot accurately reflect the true state of the battery and may lead to misjudgments.

[0085] Furthermore, in some preferred methods, when the collected ambient temperature meets the preset temperature anomaly conditions, weights are assigned to each normalized characteristic ratio according to the material type, including: reducing the weight of the normalized characteristic ratio corresponding to the first preset aging characteristic quantity, and allocating the weight reduction share generated by the reduction to the weight of the normalized characteristic ratio corresponding to at least one second preset aging characteristic quantity; wherein, the stability of the first preset aging characteristic quantity under temperature anomaly conditions is lower than the stability of the second preset aging characteristic quantity under temperature anomaly conditions.

[0086] Anomaly conditions can be set to ambient temperatures below 0 degrees Celsius or above 40 degrees Celsius. When the ambient temperature falls within this range, a dynamic weight adjustment mechanism is triggered. The first preset aging characteristic can be configured as the constant voltage range current drop rate characteristic. This characteristic mainly reflects the polarization degree of the battery and is greatly affected by changes in ion migration rate under abnormal temperature conditions, resulting in relatively low stability. The second preset aging characteristic can be configured as the self-discharge voltage drop characteristic or the average dynamic internal resistance difference characteristic. These characteristics mainly reflect the physical internal resistance or static leakage state and are less affected by temperature fluctuations, resulting in relatively high stability.

[0087] After the adjustment mechanism is triggered, the initial weight value corresponding to the first preset aging characteristic is obtained and reduced to a preset lower threshold. The difference between the initial weight value and the lower threshold is calculated as the weight reduction share, i.e., weight reduction share = initial weight value - lower threshold. Subsequently, this weight reduction share is added to the initial weight value corresponding to the second preset aging characteristic to generate an updated weight combination. For example, the weight corresponding to the constant voltage section current drop rate characteristic can be reduced from 0.20 to 0.05, resulting in a weight reduction share of 0.15. This 0.15 share is then allocated to the self-discharge voltage drop characteristic weight and the average dynamic internal resistance difference characteristic weight.

[0088] By dynamically adjusting the weights as described above, the contribution of distorted features to health assessment is reduced under abnormal temperature conditions, while the contribution of stable features is enhanced. This neutralizes the interference of temperature on some electrochemical features, avoids large jumps in health assessment values, and ensures detection stability under extreme conditions.

[0089] Under normal operating conditions at room temperature, the relaxation time constant of the internal polarization voltage dissipation of the battery differs by orders of magnitude from the time constant of the slow electrochemical reaction of self-discharge and micro-short circuits. After the preset resting window ends, most of the polarization voltage has been released, and the measured static terminal voltage drop at this time can relatively purely characterize the battery's true self-discharge characteristics. However, in actual daily operation, as the ambient temperature fluctuates downwards, or due to the cumulative increase in the internal electrolyte damping caused by multiple rounds of continuous charging and discharging, the relaxation rate of polarization charge dissipation slows down, and the corresponding electrochemical time constant drifts and lengthens. This causes the cooperative relationship between the preset fixed resting time window and the slower polarization relaxation rate to loosen. At the moment of sampling after the resting period ends, the residual polarization charge that should have been completely released still remains at the electrode interface. The measured static voltage drop of the terminal voltage is mixed with the residual polarization voltage that has not been completely dissipated. This causes the extracted self-discharge voltage drop characteristic parameter to have an excessively high amplitude when the battery has not experienced an internal micro-short circuit, making the detection results unable to reflect the self-discharge state of the battery.

[0090] In response, this application proposes that the self-discharge voltage drop characteristics after charging are determined by the release state, which characterizes the dissipation of polarization voltage during the resting phase after charging. The release state is determined based on the charging phase characteristic information extracted during the charging phase.

[0091] After charging is complete, the characteristic information extracted during the charging phase can be obtained. This information can include the average dynamic internal resistance difference, the duration of the constant-voltage charging phase, the current drop slope during the constant-voltage phase, and the real-time battery terminal temperature at the end of charging. This information reflects the amount of polarization charge accumulated in the battery during the charging phase and the resistance to dissipation during subsequent resting. For example, a large average dynamic internal resistance difference indicates high ohmic and polarization resistance; a short duration of the constant-voltage charging phase or an abnormal current drop slope during the constant-voltage phase indicates significant polarization accumulation; and a low end-of-charge temperature indicates significant ion migration resistance.

[0092] Based on the aforementioned charging stage characteristics, a polarization severity index can be calculated to assess the polarization dissipation of the battery during the resting stage. The formula for calculating the polarization severity index can be expressed as:

[0093] P_index = 0.4 * avg_dR + 0.3 * 60 / t_cv + 0.2 * dI_dt - 0.1 * T_end;

[0094] Wherein, P_index represents the polarization severity index, avg_dR represents the average dynamic internal resistance difference, t_cv represents the duration of the constant voltage charging stage, dI_dt represents the current drop slope in the constant voltage stage, and T_end represents the real-time battery temperature at the end of charging. All physical quantities have been processed by preset normalization coefficients to eliminate the influence of dimensions.

[0095] Based on the calculated polarization severity index and real-time temperature, the release state characterizing the polarization voltage dissipation can be determined. The release state can be categorized into normal release state, delayed release state, and abnormal slow release state. When the polarization severity index is less than or equal to a set threshold and the real-time temperature is above 0 degrees Celsius, the release state can be determined as normal release state, indicating that the polarization voltage dissipates relatively quickly within a preset resting window. When the polarization severity index is greater than the set threshold or the real-time temperature is below 0 degrees Celsius, but the resting voltage drop trend is still continuously trackable, the release state can be determined as delayed release state, indicating that polarization dissipation is significantly slower. When the polarization severity index is much greater than the set threshold and the resting voltage drop is still mainly controlled by polarization relaxation within the observable time, the release state can be determined as abnormal slow release state.

[0096] After determining the release state, the extraction method for the self-discharge voltage drop characteristics can be dynamically determined based on this release state. The extraction method is used to determine the specific value path of the self-discharge voltage drop characteristics after charging from the voltage data during the resting phase.

[0097] When the release state is a normal release state, a conventional extraction method can be used. Specifically, within a preset static window, after the polarization voltage has basically dissipated and entered the slow-changing region, the voltage difference within the slow-changing region can be extracted as a self-discharge voltage drop characteristic, thereby isolating polarization interference.

[0098] When the release state is a delayed release state, a delayed extraction method can be used. Specifically, the observation period can be extended, waiting for the polarization voltage to fully dissipate and enter the slow-changing region before extracting the voltage difference from the start of the slow-changing region to the end of the extended observation period as the self-discharge voltage drop characteristic. By delaying the value acquisition, the use of fixed-duration sampling point data can be avoided when the polarization has not fully dissipated.

[0099] When the release state is an abnormal slow-release state, or when the resting period is terminated due to external intervention, a correction extraction method can be used. Specifically, the residual undissipated polarization voltage component at the current time point can be estimated based on the voltage slope at the end of the resting period, and this residual polarization component can be subtracted from the total voltage drop during resting to obtain the corrected self-discharge voltage drop characteristics. The estimation of the residual polarization component can be calculated based on the polarization decay time constant, which can be linearly estimated based on the temperature at the end of charging and the average dynamic internal resistance difference.

[0100] Through the above processing method, the extracted self-discharge voltage drop characteristic is transformed from a mixed total voltage drop that may include residual polarization interference to a characteristic value dynamically obtained based on the actual polarization dissipation status of the battery. This state change blocks the path of residual polarization voltage propagation that could lead to distortion in health calculations. Under conditions where polarization relaxation is slowed down, such as at low temperatures or with increased internal battery damping, this mechanism can isolate the interference of residual polarization voltage on the self-discharge characteristic, enabling the self-discharge characteristic to reflect the internal micro-short circuit or self-discharge degree of the battery. This avoids misjudgment of health status and unnecessary charging safety misinterpretations caused by distortion of self-discharge characteristics, ensuring the accuracy and reliability of the health detection method under various practical conditions.

[0101] Furthermore, in some preferred methods, determining the extraction method based on the release state includes: dividing the static voltage time series data during the resting phase into multiple consecutive time periods; determining the voltage change slope of the battery terminal voltage in each time period; in multiple time periods, determining the slow-changing region in the resting phase where the voltage drop is dominated by the self-discharge process based on the convergence condition of the decay of the voltage change slope of adjacent time periods; and extracting the self-discharge voltage drop characteristics after charging based on the voltage drop of the battery terminal voltage in the slow-changing region.

[0102] In scenarios where low temperatures or battery aging slow down polarization relaxation, residual polarization voltage may remain in the early stages of resting. If the total voltage drop during the entire resting period is directly obtained, the extracted voltage drop is prone to being mixed with residual polarization components. To accurately extract the true self-discharge characteristics, the stability can be determined and the extraction range can be located based on the actual trend of the resting voltage.

[0103] Specifically, the stationary voltage time series data during the settling period can be divided into multiple consecutive time intervals. For example, in the basic 30-minute observation mode, the stationary period can be divided into six consecutive 5-minute intervals. In the extended 50-minute observation mode, the stationary period can be divided into ten consecutive 5-minute intervals. This segmented processing can capture the rate of voltage change and avoid the randomness of single-point values.

[0104] For each time period, the slope of the battery terminal voltage change can be determined. The voltage change slope is used to quantify the rate of voltage drop within each time period, and its specific calculation method can be expressed as follows:

[0105] S_i = V_start_i / 5 - V_end_i / 5;

[0106] Where S_i represents the voltage change slope in the i-th time period, V_start_i represents the terminal voltage value at the beginning of the time period, and V_end_i represents the terminal voltage value at the end of the time period. The unit of S_i is millivolts per minute.

[0107] After determining the voltage change slope for each time period, the slow-changing region dominated by the self-discharge process during the resting phase can be identified by considering the convergence condition based on the decay of the voltage change slope between adjacent time periods. Polarization dissipation typically exhibits an exponential decay characteristic of being fast at first and then slow, while the self-discharge voltage drop changes approximately linearly and slowly over a short period. When the ratio of the slopes between adjacent time periods decreases significantly and the absolute value of the slope drops to a low level, it indicates that the rapid phase of polarization dissipation has essentially ended. The specific convergence condition can be set to simultaneously satisfy two sub-conditions. The first sub-condition is that the ratio of the current time period slope S_i to the previous time period slope S_prev is less than a convergence threshold, which can be set to 0.3. The second sub-condition is that the absolute value of the current time period slope S_i is less than a stability threshold, which can be set to 0.1 mV / min.

[0108] If both of the above sub-conditions are met simultaneously starting from a certain time period, and the slope change remains stable in subsequent time periods, it can be determined that the resting voltage has entered the slow-changing region at the beginning of that time period, at which point the polarization voltage has been basically released. The beginning of this time period can be defined as the starting point of the slow-changing region.

[0109] After identifying the slow-varying region, the self-discharge voltage drop characteristic after charging can be extracted based on the voltage drop across the battery terminals within this region. Specifically, this can be achieved by obtaining the battery voltage at the start of the slow-varying region and the battery voltage at the end of the observation period, using the difference as the self-discharge voltage drop characteristic. For example, in a normal release state, if the slow-varying region is determined to begin at the 15th minute, the battery voltage at the 15th minute and the battery voltage at the end of the 30-minute observation period are obtained, and the difference is calculated. In a delayed release state, if the slow-varying region is determined to begin at the 35th minute, the battery voltage at the 35th minute and the battery voltage at the end of the 50-minute observation period are obtained, and the difference is calculated. By discarding the rapid voltage drop before the start of the slow-varying region, primarily caused by polarization dissipation, polarization interference can be isolated.

[0110] As a complete working example, when the real-time battery temperature at the end of charging is 5 degrees Celsius and the calculated polarization severity index is greater than the set threshold, the static observation mode can be adjusted to a 50-minute extended observation mode. In the extended observation mode, the 50 minutes are divided into ten 5-minute intervals. When calculating the slope of each interval and comparing adjacent intervals, if the slope ratio of the first few intervals is greater than 0.3, it indicates that polarization dissipation is still proceeding rapidly. When calculating the 7th interval, if the slope ratio of the 7th interval to the 6th interval is less than 0.3, and the absolute value of the slope of the 7th interval is less than 0.1 mV / min, then the 35th minute is determined to be the starting point of the slow-change region. Subsequently, the voltage drop between the 35th and 50th minutes is extracted as a self-discharge characteristic.

[0111] Through the aforementioned segmented stabilization and slow-varying-region positioning mechanism, the value range of the self-discharge voltage drop characteristic is changed from the entire range or a fixed later stage range that may contain components of rapid polarization dissipation, to a local range of the slow-varying region mainly contributed by the self-discharge process, as confirmed by trend stabilization. The voltage drop value of the self-discharge characteristic is changed from a mixed voltage drop of polarization and self-discharge to a pure, slow voltage drop that is closer to the actual self-discharge state. This avoids underestimation of health and misjudgment of micro-short-circuit faults caused by residual polarization voltage contaminating the self-discharge characteristic, thus improving the accuracy and anti-interference capability of self-discharge characteristic extraction. In other embodiments, other methods can also be used to determine the extraction method based on the release state.

[0112] In some cases, such as when battery polarization is severe or the resting process is prematurely terminated, the aforementioned stability criteria cannot be met, and the resting voltage drop fails to enter the stable slow-changing region. To address this, this application further proposes the following solution. Furthermore, in some preferred methods, the self-discharge voltage drop characteristics after charging are extracted as follows: Based on the termination slope information of the battery terminal voltage at the end or termination of the resting phase, the residual polarization component is estimated; based on the total resting voltage drop during the resting phase, the residual polarization component is subtracted to obtain the self-discharge voltage drop characteristics after charging.

[0113] When the static release state is determined to be an abnormal slow release, or when the static process is detected to be prematurely terminated, the voltage change during the static phase has not yet entered the slow-changing region dominated by the self-discharge process. At this time, the total static voltage drop contains residual polarization voltage that has not been completely dissipated. To obtain self-discharge characteristics that are close to the actual electrochemical state, a residual polarization subtraction mechanism can be used for correction and extraction.

[0114] Specifically, the termination slope information is obtained at the end or termination of the resting phase. The termination slope information can be taken as the average slope of the voltage change during the last resting period to characterize the rate of voltage drop. At the same time, combined with polarization decay information that reflects the battery's polarization dissipation capability, the voltage value of the currently undissipated residual polarization component is estimated.

[0115] In one specific implementation, the residual polarization component can be estimated in the following way:

[0116] First, calculate the polarization decay time constant tau.

[0117] Introduce a first attenuation factor factor1 and a second attenuation factor factor2.

[0118] The formula for calculating the first attenuation factor factor1 is: factor1 = 1 + 0.05 * avg_dR;

[0119] Where avg_dR is the average dynamic internal resistance difference extracted during the charging phase.

[0120] The formula for calculating the second attenuation factor factor2 is: factor2 = 1 - 0.02 * T_end;

[0121] Where T_end is the temperature at the end of the charging process.

[0122] The formula for calculating the polarization decay time constant tau is: tau = 600 * factor1 * factor2;

[0123] Subsequently, the residual polarization component V_polar_residual is calculated using the termination slope information S_end and the polarization decay time constant tau. The calculation formula is: V_polar_residual = S_end * tau;

[0124] After obtaining the residual polarization component, calculate the total voltage drop at rest;

[0125] The total voltage drop during rest can be obtained by subtracting the voltage of the last frame at the end or termination of rest from the battery terminal voltage V_start at the start of rest.

[0126] Next, the residual polarization component is subtracted from the static total voltage drop to obtain the corrected self-discharge voltage drop characteristic dU_sd_corrected, calculated using the following formula:

[0127] dU_sd_corrected = V_start - V_end - V_polar_residual;

[0128] If the calculated corrected self-discharge voltage drop characteristic dU_sd_corrected is greater than 0, this value can be used as the self-discharge voltage drop characteristic after charging and assigned a high confidence level. If the calculation result is less than or equal to 0, it can be marked as low confidence and its use intensity can be limited in subsequent calculations.

[0129] Through the above estimation and subtraction processes, the original static total voltage drop is decoupled. The self-discharge characteristic parameters used for subsequent calculations are free from the interference of residual polarization voltage, and their values ​​are corrected from being too high to be physical quantities that are close to the actual self-discharge state inside the battery. This processing method can avoid incorrectly judging a normal battery as having severe self-discharge or internal micro-short circuits due to polarization tailing phenomenon under certain operating conditions, thereby avoiding false health alarms and false charging safety interceptions triggered by this, and ensuring that the detection method maintains the accuracy of battery self-discharge state judgment under specific conditions.

[0130] The raw time-series data collected during charging contains abnormal noise such as transient jitter and pulse interference. Directly using this data for feature extraction can affect the accuracy of calculations, thus requiring data cleaning. Conventional data cleaning typically employs a fixed cleaning interval. Under normal ambient temperature and moderate battery charge conditions, this fixed interval can distinguish normal data from interference noise. However, when the ambient temperature decreases or the battery is in a deep discharge state resulting in extremely low initial voltage, the electrochemical reaction kinetics within the battery change. The normal voltage response during charging may exceed the boundaries of the fixed cleaning interval in both shape and amplitude. This causes valid data containing battery aging information to be misjudged as abnormal noise and discarded, resulting in the loss of crucial weak features in the dataset used for subsequent feature extraction.

[0131] Furthermore, in some preferred methods, before extracting aging characteristics from time-series data, the following steps are also included: performing standard cleaning on the time-series data, whereby the standard cleaning process determines the cleaning interval based on the mean and standard deviation of the time-series data and removes abnormal data that exceeds the cleaning interval; when the collected ambient temperature meets the preset abnormal temperature conditions, or when the initial charging voltage of the battery pack meets the preset low voltage conditions, the boundary range of the cleaning interval is expanded, or the time-series data is temperature compensated before performing standard cleaning to retain normal fluctuation data in the time-series data under the conditions.

[0132] To address this, standard cleaning is performed on the time-series data before extracting aging features. Standard cleaning calculates the mean μ and standard deviation σ of the time-series data and constructs a cleaning interval. Under standard conditions, the cleaning interval can be set to a range greater than μ - 3*σ and less than μ + 3*σ, removing outlier data that falls outside this range.

[0133] Before or during cleaning, monitor the current ambient temperature and the initial charging voltage of the battery pack. The preset abnormal temperature condition can be set to an ambient temperature below zero degrees Celsius or above forty degrees Celsius. The preset low voltage condition can be set to an initial charging voltage below the low charge threshold corresponding to the nominal voltage platform of the battery pack, such as a single cell voltage below the total voltage corresponding to 2.5V. At this time, the battery is determined to be in a deep discharge state.

[0134] When the ambient temperature is detected to meet the preset abnormal temperature condition, or the initial charging voltage meets the preset low voltage condition, the cleaning boundary fault tolerance mechanism is triggered.

[0135] Specifically, the boundary range of the cleaning interval can be expanded. For example, a scaling factor k can be introduced to adjust the cleaning interval to a range greater than μ-k*σ and less than μ+k*σ. The scaling factor k can be linearly or non-linearly amplified according to the temperature or voltage offset; for example, adjusting the k value from 3.0 to 4.5 allows a wider range of non-instantaneous voltage fluctuations to pass through the cleaning process.

[0136] Alternatively, temperature compensation can be applied to the time-series data before standard cleaning. Specifically, a preset temperature correction coefficient Kt is used to compensate for the amplitude of the voltage data at low temperatures. The measured voltage value is multiplied by a temperature-related compensation factor to correct it to the equivalent value at room temperature, eliminating the interference of temperature on the amplitude. Then, cleaning is performed using a standard cleaning interval greater than μ-3*σ and less than μ+3*σ. The temperature correction coefficient Kt can be obtained through a lookup table or a fitting function.

[0137] At low temperatures, electrolyte viscosity increases and lithium-ion migration rate decreases, leading to a slower voltage response during charging. The voltage fluctuations become smoother but their amplitude changes. In deep discharge states, initial battery polarization is severe, making the voltage reconstruction process complex and increasing data volatility during the initial charging phase. Through the aforementioned boundary tolerance processing, while removing instantaneous pulse spikes and abnormal jumps, the subtle voltage changes containing aging information under low-temperature polarization and deep discharge states can be preserved. This results in a more complete and effective dataset after cleaning, preventing distortion of subsequently extracted aging features due to missing data sources and ensuring the stability and accuracy of health calculations under harsh conditions.

[0138] This application also proposes an adaptive non-communication electric vehicle battery pack health detection system for performing any of the above method steps, which may include:

[0139] The data acquisition module is used to non-invasively acquire the voltage, current, and temperature timing data of the battery pack by external sensors of the charger during the battery pack charging process.

[0140] The feature extraction and material determination module is used to extract multiple aging feature quantities from time-series data to characterize different electrochemical aging mechanisms of the battery pack, and to determine the material type of the battery pack based on the time-series data.

[0141] The benchmark acquisition module is used to acquire a pre-stored benchmark feature set corresponding to the material type, obtained from a brand-new battery pack of the same type under standard charging conditions;

[0142] The health calculation module is used to calculate the ratio of each aging characteristic quantity to the corresponding benchmark value in the benchmark feature set to obtain a normalized characteristic ratio. It also assigns weights to each normalized characteristic ratio according to the material type, and sums up all the weighted normalized characteristic ratios to obtain a health value that characterizes the health status of the battery pack.

[0143] The system described above can adapt to various battery types and voltage platforms without communicating with the battery pack, calculate the health of the battery pack based on non-intrusive charging data, and identify battery aging and safety hazards.

[0144] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. All modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. An adaptive method for detecting the health of a battery pack in a non-communication electric vehicle, characterized in that, The method includes: During the charging process of the battery pack, the voltage, current and temperature timing data of the battery pack are collected non-invasively by external sensors of the charger. Multiple aging characteristic quantities are extracted from the time series data to characterize different electrochemical aging mechanisms of the battery pack, and the material type of the battery pack is determined based on the time series data. Obtain a pre-stored set of reference features corresponding to the material type, obtained from a brand-new battery pack of the same type under standard charging conditions; The normalized feature ratio is obtained by performing a ratio calculation between each of the aging feature quantities and the corresponding benchmark value in the benchmark feature set. Weights are assigned to each normalized feature ratio according to the material type. All weighted normalized feature ratios are summed to obtain a health value that characterizes the health status of the battery pack. The aging characteristics include: The voltage step characteristics at the moment of power-on are extracted based on the jump difference of the battery terminal voltage at the moment of power-on; the short-time voltage rise rate characteristics are extracted based on the slope of the change of the battery terminal voltage over time in the initial stage of charging; the constant voltage current drop rate characteristics are extracted based on the decay slope of the output current decreasing over time in the constant voltage charging stage; and the self-discharge voltage drop characteristics are extracted based on the static voltage drop of the battery terminal voltage in the resting stage after charging.

2. The adaptive non-communication electric vehicle battery pack health detection method according to claim 1, characterized in that, The step of assigning weights to each normalized feature ratio based on the material type includes: Obtain a preset weight value corresponding to the material type, and weight each normalized feature ratio according to the preset weight value.

3. The adaptive non-communication electric vehicle battery pack health detection method according to claim 2, characterized in that, When a constant-voltage charging phase is detected as missing in the current charging process, the weighting of each normalized characteristic ratio based on the material type includes: The weight corresponding to the normalized characteristic ratio obtained based on the constant voltage section current decrease rate characteristic is reduced, and the reduced weight share is allocated to the weight of the normalized characteristic ratio corresponding to at least one other aging characteristic quantity.

4. The adaptive non-communication electric vehicle battery pack health detection method according to claim 2, characterized in that, When the collected ambient temperature meets the preset temperature anomaly conditions, the process of assigning weights to each normalized feature ratio based on the material type includes: The weight of the normalized feature ratio corresponding to the first preset aging feature quantity is reduced, and the weight reduction share generated by the reduction is allocated to the weight of the normalized feature ratio corresponding to at least one second preset aging feature quantity. The stability of the first preset aging characteristic under the abnormal temperature conditions is lower than the stability of the second preset aging characteristic under the abnormal temperature conditions.

5. The adaptive non-communication electric vehicle battery pack health detection method according to claim 1, characterized in that, The self-discharge voltage drop characteristic after charging is extracted based on the release state, which characterizes the dissipation of polarization voltage during the resting phase after charging. The release state is determined based on the charging phase characteristic information extracted during the charging phase.

6. The adaptive non-communication electric vehicle battery pack health detection method according to claim 5, characterized in that, The step of determining the extraction method based on the release state includes: The static voltage timing data during the static phase is divided into multiple consecutive time periods; Determine the slope of the voltage change of the battery terminal voltage within each of the aforementioned time periods; In the multiple time periods, based on the convergence condition of the decay of the voltage change slope in adjacent time periods, the slow-changing region in the resting phase dominated by the self-discharge process is determined. Based on the voltage drop at the battery terminals within the slow-changing region, the self-discharge voltage drop characteristics after charging are extracted.

7. The adaptive non-communication electric vehicle battery pack health detection method according to claim 5, characterized in that, The self-discharge voltage drop characteristic after charging is extracted according to the following method: The residual polarization component is estimated based on the termination slope information of the battery terminal voltage at the end or termination of the resting phase. Based on the total voltage drop during the resting phase, the residual polarization component is subtracted to obtain the self-discharge voltage drop characteristics after charging.

8. The adaptive non-communication electric vehicle battery pack health detection method according to claim 1, characterized in that, Before extracting the aging feature from the time-series data, the method further includes: The time series data is subjected to standard cleaning processing, wherein the standard cleaning processing determines the cleaning interval based on the mean and standard deviation of the time series data, and removes abnormal data that exceeds the cleaning interval; When the collected ambient temperature meets the preset temperature anomaly condition, or when the initial charging voltage of the battery pack meets the preset low voltage condition, the boundary range of the cleaning interval is expanded, or the time series data is temperature compensated before the standard cleaning process is performed, so as to retain the normal fluctuation data in the time series data under the conditions.

9. An adaptive non-communication electric vehicle battery pack health detection system, used to perform the method according to any one of claims 1 to 8, characterized in that, include: The data acquisition module is used to non-invasively acquire the voltage, current, and temperature timing data of the battery pack by external sensors of the charger during the battery pack charging process. The feature extraction and material determination module is used to extract multiple aging feature quantities from the time series data, which are used to characterize different electrochemical aging mechanisms of the battery pack, and to determine the material type of the battery pack based on the time series data. The benchmark acquisition module is used to acquire a pre-stored benchmark feature set corresponding to the material type, obtained from a brand-new battery pack of the same type under standard charging conditions; The health calculation module is used to perform a ratio calculation between each of the aging characteristic quantities and the corresponding benchmark value in the benchmark feature set to obtain a normalized characteristic ratio, and to assign weights to each normalized characteristic ratio according to the material type. The module then sums up all the weighted normalized characteristic ratios to obtain a health value that characterizes the health status of the battery pack. The aging characteristics include: The voltage step characteristics at the moment of power-on are extracted based on the jump difference of the battery terminal voltage at the moment of power-on; the short-time voltage rise rate characteristics are extracted based on the slope of the change of the battery terminal voltage over time in the initial stage of charging; the constant voltage current drop rate characteristics are extracted based on the decay slope of the output current decreasing over time in the constant voltage charging stage; and the self-discharge voltage drop characteristics are extracted based on the static voltage drop of the battery terminal voltage in the resting stage after charging.

Images