Lithium battery health state prediction method and system

Abstract

The invention relates to the technical field of lithium battery health state prediction, in particular to a lithium battery health state prediction method and system, and the method comprises the following steps: initializing a basic health state estimation value for a battery cell, and correcting the internal resistance of the battery cell in combination with real-time operation data; analyzing an independent characteristic quantity from the real-time operation data of the battery cell by combining the corrected internal resistance and a basic health state estimation value; calculating the initial health state of each battery cell based on the corrected internal resistance and the independent characteristic quantity; according to the heat distribution characteristics in the battery pack, correcting the initial health state to obtain the final health state of the battery cell, and performing fusion calculation on the overall health state of the battery pack; updating the final health state to a basic health state estimated value of the next round of prediction; and based on the overall health state, performing feedback adjustment on subsequent prediction. According to the application, the influence of the temperature gradient in the battery pack on the internal resistance and health state evaluation of each battery cell can be dynamically sensed and compensated, and the prediction precision and consistency can be improved.

Description

Technical Field

[0001] This application relates to the technical field of lithium battery health status prediction, and in particular to a method and system for predicting lithium battery health status. Background Technology

[0002] With the widespread adoption of electric vehicles, lithium batteries, as their core power source, play a crucial role in ensuring vehicle range, safety, and comprehensive battery lifecycle management through accurate State of Health (SOH) prediction. Especially in frequently used vehicles like electric scooters, batteries often face complex operating scenarios with high loads and varying conditions, such as long-distance riding, continuous uphill climbing, and frequent start-stop cycles. This places higher demands on the accuracy of state estimation and the adaptability of battery management systems. Currently, the industry has various state of health prediction technologies based on battery models, data-driven approaches, or fusion methods. These technologies primarily monitor parameters such as battery internal resistance, terminal voltage characteristics, and cyclic charge-discharge data to construct aging models that assess battery capacity decay and performance degradation trends, thereby providing a basis for battery maintenance, lifespan prediction, and replacement decisions.

[0003] However, existing lithium battery health status prediction methods still have significant limitations in practical applications, especially in the dynamic and high-load real-world operating environment faced by electric vehicles. Most existing solutions are developed based on laboratory standard operating conditions or averaged environmental assumptions, failing to fully couple the dynamic evolution of the battery's internal electrochemical state with the interaction between complex external operating conditions in real-world scenarios. This leads to decreased prediction accuracy and poor consistency in actual use with drastic fluctuations in operating conditions and frequent changes in ambient temperature, thus affecting the reliability and accuracy of battery management system assessments.

[0004] Specifically, in the process of predicting the health status of batteries, the temperature gradient distribution between cells caused by spatial differences in heat dissipation conditions within the battery pack is generally ignored. The average temperature of the entire pack or the surface monitoring temperature is usually used as a uniform input, failing to identify and compensate for the differential impact of uneven temperature distribution on the key parameters of individual cells, thus introducing systematic prediction bias. This problem is particularly prominent in high-load continuous discharge scenarios such as long-distance riding of electric bicycles. Specifically, during continuous high-current output, heat continuously accumulates inside the battery pack. However, due to the different spatial positions of the cells within the pack, their heat dissipation paths and efficiencies vary significantly: cells located in the middle of the battery pack are surrounded by other cells, resulting in a long heat dissipation path and high thermal resistance, causing their temperature to rise rapidly to 35–40℃; while cells located at the edges of the pack are in direct contact with the metal casing or cooling medium, dissipating heat rapidly, and their temperature is usually maintained at around 25–30℃. The real-time temperature difference between cells in different locations within the same battery pack can reach more than 10℃. Increased temperature directly affects the ohmic and polarization resistance values ​​of battery cells, altering their voltage response characteristics. These parameters are crucial inputs for health status prediction models. Existing prediction models often assume all cells within the pack are at the same temperature, calculating internal resistance and analyzing voltage characteristics based solely on the overall average temperature. This leads to an underestimation of the internal resistance of cells in the middle positions (where actual temperatures are higher), resulting in an overestimation of their health status; conversely, an overestimation of the internal resistance of cells at the edges (where temperatures are lower), resulting in an underestimation of their health status. This prediction bias caused by uneven temperature distribution can accumulate to over 10% among different cells within the same battery pack, significantly reducing the accuracy of individual cell health status estimates and severely impacting the consistent assessment and fusion calculation of the overall battery pack health status. More seriously, based on such biased health status predictions, the battery management system's balancing strategy may be misled. For example, it may unnecessarily overcharge and discharge actually healthy, low-temperature cells, accelerating their aging process, or misjudge the remaining lifespan of the entire pack, leading to premature battery replacement or potential operational risks. Summary of the Invention

[0005] In order to improve the accuracy and consistency of prediction by dynamically sensing and compensating for the impact of temperature gradients within the battery pack on the internal resistance and health status assessment of each cell, this application provides a method and system for predicting the health status of lithium batteries.

[0006] Firstly, this application provides a method for predicting the health status of a lithium battery, which adopts the following technical solution: A method for predicting the health status of a lithium battery includes the following steps: Initialize a basic state of health estimate for each cell in the battery pack; Based on dynamic scene perception, real-time operating data of each battery cell is collected; Based on the temperature in the real-time operating data and the estimated basic health status, the internal resistance of each cell is dynamically corrected. By combining the corrected internal resistance with the estimated basic health status, independent characteristic quantities representing the degree of aging are extracted from the real-time operating data of each cell. Based on the corrected internal resistance and the independent characteristic quantities, the preliminary health status of each cell is calculated; Based on the thermal distribution characteristics inside the battery pack, the initial health status of each cell is collaboratively corrected to obtain the final health status of the cell. The overall health status of the battery pack is calculated by integrating the final health status of all cells. The final health status is updated to the baseline health status estimate predicted in the next round; Based on the overall health status, at least one strategy for data acquisition, internal resistance correction, or feature analysis in subsequent predictions is adjusted accordingly.

[0007] Optionally, the step of collecting real-time operating data for each battery cell based on dynamic scene perception specifically includes: Real-time monitoring of the temperature and current change rates of each battery cell; When the rate of temperature change exceeds a preset first rate of change threshold, the frequency of collecting the cell temperature data is increased. When the rate of change of the current exceeds a preset second rate of change threshold, the acquisition frequency of the cell current data is increased.

[0008] Optionally, the step of dynamically correcting the internal resistance of each cell based on the temperature in real-time operating data and the estimated basic health status specifically includes: Establish a mapping library describing the relationship between temperature, aging state, and internal resistance; For each cell, based on its real-time collected temperature data and the basic health status estimate on which the current cycle is based, its corresponding ohmic internal resistance and polarization internal resistance are determined from the mapping relationship library. The determined ohmic internal resistance and the polarization internal resistance are combined to generate the corrected internal resistance of the cell under the current conditions.

[0009] Optionally, it also includes the step of updating the mapping relationship database, specifically: When a preset mileage condition is met, or when the difference between the preliminary health status obtained in the current round and the preliminary health status obtained in the previous round exceeds a set threshold, an update to the mapping relationship library is triggered. During the update process, the real-time temperature of the cell in the current cycle and the measured internal resistance of the cell in the preliminary health state determined in the previous cycle are obtained. The measured internal resistance is compared with the internal resistance value obtained from the mapping database based on the same real-time temperature and the same initial health status. If the deviation between the two exceeds the allowable error threshold, the corresponding internal resistance value in the mapping relationship library is corrected based on the deviation.

[0010] Optionally, the process of resolving independent features characterizing the degree of aging specifically involves obtaining an aging voltage component. The aging voltage component is obtained by: From the difference between the open-circuit voltage and the measured terminal voltage of the cell, at least one polarization voltage component determined based on the corrected internal resistance and the estimated basic health state is subtracted.

[0011] Optionally, the step of collaboratively correcting the initial health status of each cell based on the internal thermal distribution characteristics of the battery pack specifically includes: Based on the spatial location attributes of the battery cell within the battery pack and its temperature change trend, a compensation factor related to the temperature gradient is determined. Based on the compensation factor, the preliminary health status is numerically adjusted to generate the final health status.

[0012] Optionally, the fusion calculation of the overall health status of the battery pack specifically includes: Based on the stated final health status of all cells, a reference value characterizing the overall level is calculated; For each cell, a weighting factor is determined based on its rated capacity, the degree of deviation of its final state of health from the reference value, and its charge / discharge efficiency at the current temperature. The final health status of each cell is weighted and summed using the weighting factors to obtain the overall health status of the battery pack.

[0013] Optionally, the step of adjusting at least one strategy for data acquisition, internal resistance correction, or feature analysis in subsequent predictions based on the overall health status specifically includes: When the change in the overall health status during consecutive preset rounds of prediction is lower than a preset stability threshold, feedback adjustment is initiated. The feedback adjustment includes at least one of the following: Based on the degree of battery aging indicated by the overall health status, adjust the judgment conditions used to trigger changes in the acquisition frequency in the dynamic scene perception. Adjust the conditions for triggering the update of the mapping database based on the degree of battery aging indicated by the overall health status; The coefficients used to calculate the polarization voltage component are adjusted based on the degree of deviation of the aging voltage component of a single cell from the average value of the aging voltage components of all cells in a continuous preset cycle.

[0014] Secondly, this application provides a lithium battery health status prediction system, which adopts the following technical solution: A lithium battery health status prediction system, comprising: An initialization unit is used to configure a basic state of health estimate for each cell in the battery pack. The dynamic sensing and acquisition unit is used to acquire real-time operating data of each battery cell based on dynamic scene perception. The internal resistance dynamic correction unit is used to dynamically correct the internal resistance of each cell based on the temperature in the real-time operating data and the estimated value of the basic health status. The feature parsing unit is used to combine the corrected internal resistance with the basic health state estimate to parse independent feature quantities characterizing the degree of aging from the real-time operating data of each cell. A health status calculation unit is used to calculate the preliminary health status of each cell based on the corrected internal resistance and the independent characteristic quantities. The collaborative correction unit is used to collaboratively correct the initial health status of each cell based on the heat distribution characteristics inside the battery pack, so as to obtain the final health status of the cell. The fusion computing unit is used to calculate the overall health status of the battery pack based on the final health status of all cells. The feedback adjustment unit is used to adjust at least one strategy for data acquisition, internal resistance correction, or feature analysis in subsequent predictions based on the overall health status. An iterative update unit is used to update the final health status to the estimated basic health status value predicted in the next round.

[0015] In summary, this application includes the following beneficial technical effects: 1. This invention determines the compensation factor related to the temperature gradient based on the spatial location attributes and temperature change trend of the cells in the battery pack and performs collaborative correction on the preliminary health status. This effectively solves the problem of temperature gradient distribution between cells caused by neglecting the differences in heat dissipation conditions inside the battery pack in the prior art, reduces the prediction deviation of the health status of a single cell caused by uneven temperature distribution, and significantly improves the accuracy and consistency of the health status assessment of a single cell and the overall health status of the battery pack under dynamic high-load real-world operating conditions.

[0016] 2. This invention monitors the temperature and current change rate of the battery cell in real time and dynamically adjusts the data acquisition frequency. When the operating conditions fluctuate, the acquisition density is increased to capture key dynamic features, and the frequency is reduced when the state is stable to save system resources. This achieves a precise match between the data acquisition strategy and the actual operating state of the battery cell, overcoming the shortcomings of fixed acquisition frequency in adapting to complex dynamic operating conditions. It provides a high-quality and efficient real-time data foundation for subsequent internal resistance correction and feature analysis.

[0017] 3. This invention establishes and dynamically updates a mapping library describing the relationship between temperature, aging state and internal resistance, and periodically corrects the values ​​in the library based on measured internal resistance data. This enables the internal resistance correction to closely track the characteristic evolution of the cell throughout its entire life cycle, solving the problem in existing methods where the aging assessment deviation increases with usage time due to the solidification or untimely updating of the internal resistance model. This ensures the long-term accuracy of the core parameters for predicting health status.

[0018] 4. Based on the overall health status of the battery pack, this invention performs closed-loop feedback adjustment of the data acquisition, internal resistance correction and feature analysis strategies in the subsequent prediction process, and constructs an iterative mechanism for prediction correction and adjustment updates. This enables the system to adapt to the characteristic changes of the battery at different aging stages, continuously optimize the prediction accuracy, solve the problem of long-term prediction performance degradation caused by the difficulty of adapting fixed strategies to different stages of the battery's entire life cycle, and realize the self-maintenance and improvement of the system's prediction capability. Attached Figure Description

[0019] Figure 1 It is the overall logic flowchart; Figure 2 It is a logical flowchart of dynamic scene perception and data acquisition; Figure 3 This is a flowchart of the logic for temperature gradient collaborative correction. Detailed Implementation

[0020] The following is in conjunction with the appendix Figure 1-3 This application will be described in further detail.

[0021] This application discloses a method for predicting the health status of lithium batteries. For example... Figure 1 As shown, a method for predicting the health status of a lithium battery includes the following steps: S1 System Initialization S11 Configure initial basic health status estimate After the battery pack is powered on, the system first completes hardware self-test and data link initialization to confirm that the communication between the battery management system and each cell is normal and the sensor data acquisition is valid. Then, it independently configures an initial basic health state estimate for each cell in the battery pack. This value is the core input parameter for the first round of health state prediction.

[0022] For newly manufactured lithium batteries, the system sets their initial baseline health state estimate to 100%. New batteries undergo rigorous performance testing before leaving the factory, and their key parameters such as capacity, internal resistance, and charge / discharge efficiency fully meet design standards, with no capacity decay or performance degradation, placing them at their optimal health level. Using 100% as the initial value most accurately reflects the initial state of the new battery, providing precise and unbiased benchmark data for the first round of predictions and avoiding significant errors in the first round of predictions due to unreasonable initial value settings.

[0023] For older batteries already in use, the system prioritizes historical data retrieval, extracting the cell's final health status recorded before the last shutdown from the battery management system's non-volatile storage module. This value is then directly used as the initial baseline health status estimate for the current cycle. The reason for choosing the historical final health status as the initial value is that it has undergone a complete "collection-calibration-calculation-correction" process, fully integrating information such as the cell's previous operating conditions and aging characteristics. This accurately reflects the cell's actual aging degree and is more targeted and accurate than the default initial value. If the older battery is being connected to the system for the first time or if the battery management system lacks historical health status data, the system defaults to setting the initial baseline health status estimate to 100%. This setting ensures that the prediction process can still start normally even without historical data support, avoiding system malfunctions due to missing initial values.

[0024] S12 sets the iterative update rules After completing the initial configuration of the basic health status estimate, the system synchronously sets the iterative update rules for the basic health status estimate, clarifying the update source and timing of the basic health status estimate in each subsequent round of prediction.

[0025] The iterative update rule is as follows: After each complete health status prediction process (from real-time data collection to feedback adjustment) is completed, the system immediately triggers an update command to directly replace the final health status of each cell calculated in the current round with the basic health status estimate for the next round, overwriting the original basic health status estimate and storing it in the non-volatile storage module of the battery management system.

[0026] This iterative update rule uses the final health state optimized in the previous round as the input parameter for the next round. This allows the basic health state estimate to continuously approach the actual health state of the battery cell with each prediction round, forming a virtuous cycle of "initial value - first prediction - optimized value - second prediction - better value". This mechanism not only ensures the sustainability of the prediction process but also allows the prediction accuracy to gradually improve during the iteration process, solving the problem of difficulty in continuously optimizing prediction accuracy caused by fixed initial values ​​or unreasonable update logic in existing technologies. Simultaneously, this rule achieves persistent storage of the updated basic health state estimate through the non-volatile storage module of the battery management system. Even if the system is powered off or shut down, the data will not be lost, ensuring that the latest optimized value can be directly used as the initial input upon the next startup, further guaranteeing the continuity and accuracy of the prediction.

[0027] S2 Real-time Operational Data Acquisition Based on Dynamic Scene Awareness like Figure 1 and Figure 2 As shown, after the system completes hardware self-test, ensures data link connectivity, and configures initial basic health status estimates, S2 initiates the real-time data acquisition process. By dynamically sensing the temperature and current changes of the battery cell, it flexibly adjusts the acquisition strategy to ensure that the acquired data accurately matches the characteristics of the battery cell under complex operating conditions. This provides high-quality, timely input data for subsequent internal resistance correction, feature analysis, and other processes, solving the problem of insufficient data accuracy caused by the fixed acquisition frequency in existing technologies, which is difficult to adapt to dynamic operating conditions.

[0028] S21 monitors the rate of temperature change and the rate of current change. The system collects temperature and current data for each battery cell in real time using a temperature sensor built into the cell and a current sensor connected in series in the circuit. The temperature sensor uses an NTC thermistor with a measurement accuracy of ±0.1℃ and a range covering -40℃ to 85℃; the current sensor uses a Hall effect sensor with a measurement accuracy of ±0.01A and a range covering -100A to 100A; voltage data is collected through voltage sampling lines led out from the positive and negative terminals of the battery cell, with a sampling accuracy of ±0.001V.

[0029] The system is set to a default data acquisition frequency of 100ms / time. When the cell condition is stable, the 100ms / time acquisition interval ensures data continuity without increasing system computing power and energy consumption due to excessive acquisition. Based on the continuous data acquired at the base frequency, the system calculates the ratio of the difference between two adjacent acquisitions to the time interval (100ms) to obtain the temperature change rate and current change rate of each cell.

[0030] Rate of temperature change: The unit is ℃ / s. To facilitate subsequent judgment, the system will automatically convert it to ℃ / min (multiplied by 60) for comparison with the preset ℃ / min threshold.

[0031] Rate of change of current: The unit is A / s.

[0032] This calculation method can accurately identify high-load scenarios such as sudden current changes. Together, they constitute the core indicators of dynamic scene perception, providing a basis for judgment in subsequent sampling frequency adjustments.

[0033] S22 Sets the sampling frequency and adjusts the threshold. To enable dynamic switching of data acquisition strategies, the system explicitly sets two types of change rate thresholds: Thresholds for the rate of temperature change: 0.5℃ / min for rapid change and 0.2℃ / min for steady-state change; Threshold for rate of change of current: The threshold for rapid change is 5A / s.

[0034] The reason for choosing 0.5℃ / min as the threshold for rapid temperature change is that in scenarios such as starting and riding an electric vehicle after being exposed to the sun in summer or starting in low-temperature environments in winter, the cell temperature will rise and fall rapidly. The measured rate of change generally exceeds 0.5℃ / min. At this time, the electrochemical reaction rate inside the battery changes drastically, and the internal resistance and voltage response characteristics fluctuate significantly. It is necessary to increase the sampling frequency to capture these dynamic changes.

[0035] The reason for choosing 0.2℃ / min as the temperature stability threshold is that when the temperature change rate is lower than this value, the cell temperature is in a relatively stable state, the electrochemical characteristics change slowly, and the basic frequency of 100ms / time is sufficient to meet the data accuracy requirements, so there is no need for high-frequency acquisition.

[0036] The reason for choosing 5A / s as the current change rate threshold is that, through statistical analysis of typical high-load scenarios such as electric vehicles climbing hills, frequent starts and stops, and rapid acceleration, the current change rate in these scenarios generally reaches 5A / s or higher. At this time, sudden current changes will cause drastic fluctuations in internal resistance and polarization voltage, which is a key condition affecting the accuracy of health status assessment. It is necessary to accurately capture the dynamic characteristics of the current through high-frequency acquisition.

[0037] S23 Adjust data acquisition frequency The system compares the calculated rate of temperature change and rate of current change with preset thresholds in real time, and dynamically adjusts the acquisition frequency of the corresponding data based on the comparison results to ensure that the acquisition strategy matches the cell's operating status in real time.

[0038] Temperature sampling frequency adjustment: If the rate of temperature change exceeds 0.5℃ / min, the temperature data acquisition frequency will be increased from 100ms / time to 50ms / time. This frequency doubles the temporal resolution of the temperature data, allowing for more intensive capture of temperature change details and preventing the omission of temperature peaks or trends due to excessively long acquisition intervals. If the rate of temperature change is between 0.2℃ / min and 0.5℃ / min, maintain the current sampling frequency; If the rate of temperature change drops to 0.2℃ / min or below, it is restored to 100ms / time to balance system energy consumption while ensuring data accuracy.

[0039] Current sampling frequency adjustment: If the rate of change of current exceeds 5A / s, the current data acquisition frequency will be temporarily increased from 100ms / time to 15ms / time and maintained for 5 seconds. Compared with the base frequency, the frequency of 15ms / time can significantly improve the timeliness of current data and accurately capture the peak and trajectory of current changes. The reason for setting the duration of 5 seconds is that the duration of high load scenarios for electric vehicles is usually no more than 5 seconds. If the rate of change of current still exceeds 5A / s after 5 seconds, continue to maintain 15ms / cycle; If the rate of change of current drops to 5A / s or below, it should be restored to 100ms / cycle.

[0040] This dynamic adjustment strategy avoids the energy and computing power waste caused by fixed high-frequency acquisition, and solves the problem of insufficient data accuracy when fixed low-frequency acquisition is used when operating conditions fluctuate, making data acquisition more targeted and economical.

[0041] S24 stores real-time operational data. The system collects real-time temperature, current, and voltage data for each cell in each round, as well as the calculated rate of temperature and current change, and stores them in the high-speed cache of the battery management system in the format of "cell number-collection timestamp-data type-value".

[0042] The high-speed cache uses DDR4 memory with a read / write speed of 2133MHz and a capacity of 2GB, which can meet the real-time data storage requirements for 24 consecutive hours. Data storage adopts a circular overwrite mechanism. When the cache capacity is about to run out, it automatically overwrites the oldest non-critical data, while the core data of the most recent 72 hours is synchronously stored in non-volatile flash memory to avoid data loss due to power failure.

[0043] This storage method ensures the integrity, continuity, and traceability of the data. Each piece of data has a unique identifier and timestamp, which facilitates accurate association with the cell status under the corresponding operating conditions in subsequent steps. It provides consistent and reliable data support for calculations such as internal resistance correction and feature analysis, and avoids prediction errors caused by data corruption or loss.

[0044] S3 Cell Internal Resistance Dynamic Correction S3 utilizes the real-time temperature data collected by S2 and the basic health status estimate determined by S1. Through querying and dynamically updating the mapping relationship library, it accurately calculates the corrected internal resistance of each cell. This solves the problem of internal resistance assessment deviation caused by ignoring the differential influence of temperature and aging status on cell internal resistance in existing technologies, and provides reliable core parameter support for subsequent feature analysis and health status calculation.

[0045] S31 Constructs a mapping relationship library The system first constructs a three-dimensional mapping relationship library. This mapping relationship library takes temperature and aging state as input dimensions and ohmic internal resistance and polarization internal resistance as output dimensions, providing basic data support for subsequent internal resistance queries.

[0046] The temperature range is set from -20℃ to 60℃. When the electric vehicle is running in the low-temperature environment of winter in the north, the battery cell temperature may be as low as -20℃, while in the summer in the south, when riding after being exposed to the sun or under high load discharge, the battery cell temperature may rise to 60℃. This range fully covers the operating temperature of the electric vehicle under all working conditions, ensuring that the corresponding internal resistance data can be found under different climate environments and different operating scenarios.

[0047] The aging status dimension has a value range of 0% to 100%, with 0% corresponding to a completely scrapped battery and 100% corresponding to a brand new battery. This range covers the entire life cycle of the battery from manufacturing to scrapping, which can meet the internal resistance query needs of cells with different aging levels and adapt to the health status prediction of the entire battery life cycle.

[0048] The mapping database was constructed based on extensive laboratory cyclic test data, employing the existing mature pulse current method to measure internal resistance. The tests selected battery cells of the same model but with different aging levels (health status divided into 5% gradients from 100% to 0%), and placed them at different temperatures (-20℃ to 60℃ in 5℃ gradients) for 30 minutes to ensure the cell temperature matched the ambient temperature. Subsequently, a 10A pulse current lasting 10ms was applied using the pulse current method, and voltage response data was simultaneously collected. The ohmic internal resistance of each cell was calculated based on Ohm's law, and the polarization internal resistance was obtained by fitting transient voltage response curves. All test data were organized and archived according to the correspondence between "temperature-aging status-ohmic internal resistance-polarization internal resistance" to form the initial mapping database, ensuring the accuracy and reliability of the query results.

[0049] S32 sets the update conditions for the mapping relationship database. The system sets the update trigger condition for the mapping relationship library to be one of two options, ensuring that the mapping relationship library can be dynamically optimized as the battery ages and the operating conditions change, avoiding a decrease in correction accuracy due to the solidification of internal resistance data.

[0050] The first trigger condition is every 100km of riding distance completed. The reason for choosing 100km as the mileage trigger condition is that, through statistical analysis of the daily use scenarios of electric bicycles, every 100km of riding distance corresponds to about 1 to 2 complete charge-discharge cycles of the battery. The cycle loss and cumulative effect of the battery can be significantly reflected in the change of internal resistance. This mileage setting ensures the update frequency of the mapping relationship library, which can capture the trend of internal resistance change in a timely manner, while avoiding the increase in system computing power consumption and data redundancy caused by too frequent updates.

[0051] The second trigger condition is that the difference between the preliminary health status obtained in the current round and the preliminary health status obtained in the previous round exceeds 5%. The reason for choosing 5% as the difference trigger condition is that, through extensive battery aging tests, it has been verified that when the difference between the preliminary health status of two consecutive rounds exceeds 5%, it indicates that the battery aging state has changed significantly, and its internal resistance characteristics will also change significantly. At this time, the corresponding internal resistance data in the original mapping relationship library can no longer accurately match the current state of the battery. If it is not updated in time, it will lead to an increase in the internal resistance correction deviation. Therefore, it is necessary to trigger an update to ensure the timeliness and accuracy of the internal resistance data.

[0052] S33 Calculates the corrected internal resistance Based on the real-time temperature data collected by S2 and the estimated basic health status in S1 after iterative updates, the system executes the calculation process of the corrected internal resistance to ensure that the internal resistance correction is accurately matched with the real-time status of the cell.

[0053] The system first retrieves the real-time temperature data of the corresponding cell from the battery management system cache. This data is acquired through the S2 dynamic acquisition process, ensuring high timeliness and accuracy, and accurately reflecting the current temperature state of the cell. Simultaneously, the system retrieves the updated baseline health state estimate from the previous prediction process. This value is chosen because the initial health state for the current round has not yet been calculated; using the optimized baseline health state estimate from the previous round avoids circular dependency issues. Furthermore, this value has approximated the cell's true health state through multiple iterations, accurately reflecting the recent aging degree of the battery.

[0054] The system takes real-time temperature data and baseline health status estimates as input, queries a pre-built mapping database to obtain the corresponding ohmic internal resistance and polarization internal resistance of the battery cell. The query process uses linear interpolation to handle gradient data. For example, when the real-time temperature is 23℃ (between 20℃ and 25℃) and the baseline health status estimate is 83% (between 80% and 85%), the corresponding ohmic internal resistance and polarization internal resistance are calculated using two-dimensional linear interpolation, ensuring the continuity and accuracy of the query results.

[0055] The system sums the obtained ohmic and polarization resistances to obtain the corrected total internal resistance of the cell under current conditions. This calculation method fully considers the effects of temperature and aging on the cell's internal resistance. Compared with existing technologies that only use fixed internal resistance or average temperature to calculate internal resistance, the accuracy of the corrected internal resistance is significantly improved, providing reliable parameters for subsequent feature analysis and health status calculation.

[0056] S34 Update Mapping Relationship Database When any of the update trigger conditions set in S32 are met, the system starts the mapping relationship library update process to ensure that the mapping relationship library can dynamically adapt to the changes in battery aging characteristics and continuously provide accurate data support for internal resistance correction.

[0057] The system first measures the actual internal resistance of the battery cell using the pulse current method. During the measurement, the system maintains a stable real-time temperature for the current cycle (temperature fluctuation not exceeding ±0.5℃). Based on the preliminary health status determined in previous cycles, a pulse current with the same parameters (10A, duration 10ms) as when building the mapping relationship library is applied, and voltage response data is collected synchronously. The measured ohmic internal resistance and measured polarization internal resistance are calculated to ensure that the measurement conditions are consistent with the construction conditions of the mapping relationship library, thus guaranteeing the comparability of the measured data.

[0058] The system queries the mapping database for the internal resistance value corresponding to the current real-time temperature and the preliminary health status of the previous round (querying the ohmic internal resistance and querying the polarization internal resistance), and calculates the deviation between the measured ohmic internal resistance and the queried ohmic internal resistance, and the deviation between the measured polarization internal resistance and the queried polarization internal resistance, respectively.

[0059] The system sets an allowable error threshold of 2%. Combined with the measurement accuracy of the pulse current method (±1%) and the natural fluctuation range of the battery internal resistance (±0.5%), the 2% allowable error threshold can effectively distinguish between measurement noise and actual internal resistance changes. This ensures the accuracy of the mapping relationship library update, avoids frequent invalid updates caused by measurement noise, and can also capture actual internal resistance changes in a timely manner, ensuring the timeliness of the mapping relationship library.

[0060] If any deviation exceeds 2%, the system corrects the internal resistance value in the mapping database corresponding to the temperature-aging state based on the deviation ratio. The correction logic is: corrected internal resistance = query internal resistance × (measured internal resistance ÷ query internal resistance). After correcting the ohmic internal resistance and polarization internal resistance respectively, the corrected data is updated in the mapping database, overwriting the original data. This update method enables the mapping database to be dynamically optimized as the battery ages, ensuring that each internal resistance query is based on the latest cell characteristic data, thereby continuously improving the accuracy of internal resistance correction and solving the problem that the internal resistance evaluation deviation gradually increases with battery aging caused by the fixed mapping database in the existing technology.

[0061] S4 analyzes independent characteristic quantities representing the degree of aging. S4 uses the corrected internal resistance obtained from S3, the basic health status estimate determined by S1, and the voltage data collected by S2 to accurately separate the independent characteristic quantity, namely the aging voltage component, from the real-time operating data of the cell. This solves the problem of inaccurate aging assessment caused by the confusion between polarization voltage and aging voltage in the existing technology, and provides core feature input for subsequent preliminary health status calculation.

[0062] S41 obtains the cell open-circuit voltage and the measured terminal voltage. The system first retrieves the measured terminal voltage of the cell for the current cycle from the battery management system cache. This data is synchronously acquired by the S2 dynamic data acquisition process with a sampling accuracy of ±0.001V, which can truly reflect the terminal voltage status of the cell when it is working in real time.

[0063] Meanwhile, the system calculates the open-circuit voltage using voltage data from the cell in a static state. The criterion for determining the static state is that the absolute value of the cell current is less than 0.01A for 10 minutes. When the current is below this value, there is no significant current flowing through the cell, and the polarization effect can be completely eliminated. At this time, the voltage can truly reflect the electrochemical equilibrium state of the battery. The 10-minute duration ensures that the polarization effect is completely dissipated, avoiding residual polarization from affecting the accuracy of the open-circuit voltage.

[0064] The system selects the three most recent stationary voltage data points that meet the above criteria and uses linear interpolation to fit the current open-circuit voltage. Single data points may contain measurement noise; interpolating three data points can effectively reduce random errors and improve the calculation accuracy of the open-circuit voltage. Linear interpolation is a mature existing algorithm that is computationally efficient and meets accuracy requirements. Its core logic is to smoothly fit the current open-circuit voltage by analyzing the changing trends of two adjacent stationary voltage points, ensuring the continuity and reliability of the data.

[0065] S42 Calculation of polarization voltage components The system calculates polarization voltage components in three categories based on the corrected internal resistance obtained from S3 and the basic health state estimate from S1.

[0066] S421 Calculate Ohmic Polarization Voltage The calculation of ohmic polarization voltage is based on the derivation of Ohm's law, with the core logic being the voltage drop generated when current flows through the ohmic internal resistance of the battery cell. The calculation relationship is: ohmic polarization voltage equals the current real-time current multiplied by the ohmic internal resistance corrected by S3. Current data is dynamically acquired from S2, in amperes (A); ohmic internal resistance is in ohms (Ω); and voltage is in volts (V). This calculation method utilizes the precise ohmic internal resistance corrected by S3, avoiding the polarization voltage calculation deviation caused by a fixed internal resistance, making the assessment of ohmic polarization voltage more closely reflect the real-time state of the battery cell.

[0067] S422 Calculation of electrochemical polarization voltage The calculation relationship for electrochemical polarization voltage is as follows: in: The rate of change of current (A / s) is dynamically acquired from S2; The reference current change rate is calibrated to 1 A / s to make the logarithmic term dimensionless; is the electrochemical polarization coefficient.

[0068] The benchmark value was determined experimentally: the electrochemical polarization coefficient of a new cell of the same model at 25℃ and 100% healthy condition was selected as the benchmark. Baseline value. Experimental data shows that when the temperature exceeds 35℃, the ion diffusion rate inside the battery accelerates, and the electrochemical polarization sensitivity decreases significantly, with a reduction factor that matches the actual polarization characteristics. When the real-time cell temperature is higher than 35℃, When the baseline health status estimate is below 80%, The reason for choosing 80% as the threshold is that cells with a health status below 80% have reduced electrode active material, increased electrochemical polarization sensitivity, and a higher coefficient can accurately reflect the polarization degree of aging cells.

[0069] S423 Calculates Concentration Polarization Voltage The formula for calculating concentration polarization voltage is: Concentration polarization voltage equals... Multiply by the square of the current, where is the concentration polarization coefficient. The baseline value was also calibrated experimentally, selecting the concentration polarization coefficient of a new cell of the same model at 25℃ and 100% healthy condition as the baseline. Baseline value, and Consistent benchmark calibration conditions ensure the consistency and reliability of coefficients.

[0070] When the real-time temperature of the battery cell is below 0℃ Values The baseline value is multiplied by 1.2. The reason for choosing 0℃ as the threshold is that the uneven distribution of ion concentration inside the battery intensifies at low temperatures, significantly enhancing concentration polarization sensitivity. Increasing the coefficient can match the polarization characteristics under low-temperature conditions; when the baseline health state estimate is below 70%, Values The baseline value is multiplied by 1.3, and 70% is chosen as the threshold because the ion transport channels of cells with a health status below 70% are damaged, and the concentration polarization effect is greatly enhanced. Further increasing the coefficient can accurately capture the polarization characteristics of severely aged cells.

[0071] S43 separates the aging voltage component. The system first calculates the difference between the open-circuit voltage and the measured terminal voltage. This difference includes the polarization voltage during cell operation and the voltage drop caused by aging, which is the basis for subsequent separation of the aging voltage component.

[0072] Subsequently, the system subtracts the sum of the ohmic polarization voltage, electrochemical polarization voltage, and concentration polarization voltage calculated by S42 from the total voltage drop, and the final result is the aging voltage component.

[0073] This separation logic completely eliminates the interference of polarization effects by accurately subtracting the polarization voltage component calculated based on the corrected internal resistance and dynamic polarization coefficient. This ensures that the aging voltage component is only related to the degree of cell aging, enabling it to independently and accurately characterize the cell's aging state. Compared to existing assessment methods that do not adequately distinguish between polarization and aging voltage, this method significantly improves the identification of aging characteristics, providing accurate and reliable core input for subsequent preliminary health status calculations and effectively reducing the prediction bias in health status caused by feature distortion.

[0074] S5 calculates the initial health status of the battery cells. Based on the corrected internal resistance obtained from S3 and the aging voltage component separated from S4, S5 accurately calculates the preliminary health status of each cell through interval logic judgment and dual-parameter cross-validation. This solves the problem of health status prediction bias caused by single-parameter aging assessment in existing technologies and provides reliable basic data for subsequent collaborative correction.

[0075] S51 sets the threshold value for the aging voltage component range. Based on accelerated aging test data from a large number of lithium batteries of the same model, the system sets three threshold ranges for the aging voltage component: a light aging threshold of 0.05V and a moderate aging threshold of 0.1V.

[0076] The reason for choosing 0.05V as the threshold for mild aging is that test data shows that when the aging voltage component is less than or equal to 0.05V, the actual capacity decay of the cell is less than or equal to 10%, and the core performance of the battery, such as charge / discharge efficiency and cycle life, remains basically good without significant performance degradation, which meets the technical definition of mild aging. The reason for choosing 0.1V as the threshold for moderate aging is that when the aging voltage component is between 0.05V and 0.1V, the cell capacity decay is between 10% and 30%, and the battery performance shows a significant decline, with reduced charge / discharge efficiency and accelerated internal resistance growth, which belongs to the moderate aging stage. When the aging voltage component is greater than 0.1V, the cell capacity decay exceeds 30%, the core performance is severely degraded, and it cannot meet normal usage requirements, which belongs to the severe aging state. This threshold setting is based on the performance change pattern of the battery throughout its entire life cycle, covering all scenarios from mild aging to severe aging, ensuring the scientific and practical nature of the interval division.

[0077] S52 Calculates preliminary health status by interval Based on the range of the aging voltage component separated by S4, the system uses calculation logic that conforms to the capacity decay law of the corresponding aging stage to obtain the preliminary health status of each cell, ensuring that the evaluation results are consistent with the actual performance of the cell.

[0078] S521 Mild Aging Status Calculation When the aging voltage component is less than or equal to 0.05V, the system calculates the preliminary health status according to the mild aging logic. The core calculation logic is: Preliminary Health Status = 100% - (Aging Voltage Component / 0.05V) × 10%. This logic is derived from the fact that the cell capacity decay during mild aging has an approximately linear relationship with the aging voltage component. After fitting with a large amount of experimental data, the fitting degree of this linear relationship reaches over 98%. When the aging voltage component increases from 0V to 0.05V, the capacity decay increases linearly from 0% to 10%, corresponding to a linear decrease in the health status from 100% to 90%, accurately reflecting the health level of the mildly aged cell.

[0079] S522 Medium Aging Status Calculation When the aging voltage component is greater than 0.05V and less than or equal to 0.1V, the system calculates the preliminary health status according to the medium aging logic. The core calculation logic is: Preliminary health status = 90% - {(aging voltage component - 0.05V) / 0.05V} × 20%. This logic is derived from the fact that the cell capacity decay rate accelerates during the medium aging stage. Experimental data shows that when the aging voltage component increases from 0.05V to 0.1V, the capacity decay increases linearly from 10% to 30%, corresponding to a linear decrease in the health status from 90% to 70%. This fitting relationship conforms to the electrochemical characteristic change law of lithium batteries in the middle of aging and can accurately capture the accelerating trend of performance degradation.

[0080] S523 Severe Aging State Calculation When the aging voltage component exceeds 0.1V, the system calculates the preliminary health status according to the heavy aging logic. Its core calculation logic is: Preliminary health status = 70% - {(aging voltage component - 0.1V) / (...}} -0.1V)}×30%. Wherein The limiting aging voltage drop of a battery cell is determined by calibrating the aging voltage component when a cell of the same model is cycled to 50% of its healthy state. This calibration method was chosen because 50% of the healthy state is a critical reference point for the disposal of lithium batteries, and the aging voltage component at this point can reflect the limiting level of cell aging. The rationale behind this logic is that during the severe aging stage, the cell capacity decay rate further accelerates. When the aging voltage component increases from 0.1V to V_max, the capacity decay increases linearly from 30% to 60%, corresponding to a linear decrease in the healthy state from 70% to 40%, ensuring that the health status assessment of severely aged cells conforms to the actual disposal trend.

[0081] S53 Verification and Correction Based on Internal Resistance Data To further improve the accuracy of the preliminary health status calculation, the system combines the corrected internal resistance data obtained from S3 for cross-validation, thus overcoming the limitations of single-parameter evaluation.

[0082] The system first determines the initial internal resistance (factory calibration value) of a new battery cell of the same model as a benchmark for internal resistance comparison. By comparing the ratio of the corrected internal resistance to the initial internal resistance of the new battery cell, and combining this with the experimentally calibrated "internal resistance growth rate - capacity decay relationship", the corresponding health status reference value is obtained.

[0083] For example: The initial internal resistance of a new battery cell is 20mΩ, and the current corrected internal resistance is 24mΩ. The internal resistance growth rate is: According to the test calibration, for every 20% increase in the initial internal resistance of a new battery cell, the capacity decreases by about 10%, and the reference value for the health status is 90%.

[0084] The system sets a deviation threshold of 3%. This threshold was chosen because, combined with the aging voltage component measurement accuracy of ±0.001V and the internal resistance measurement accuracy of ±1%, a 3% deviation range can effectively distinguish between reasonable and abnormal errors, avoiding over-correction. If the preliminary health status calculated from the aging voltage component deviates from the reference value for the health status corresponding to the internal resistance by more than 3%, the system takes the average of the two as the final preliminary health status; if the deviation is less than or equal to 3%, the result calculated from the aging voltage component is retained. This verification mechanism, through dual-parameter cross-validation, reduces the impact of measurement noise or sudden operating conditions on health status prediction, making the preliminary health status assessment more robust and accurate.

[0085] S6 collaboratively corrects the final health status of the battery cell. like Figure 1 and Figure 3 As shown, S6, based on the thermal distribution characteristics inside the battery pack and combined with the spatial location attributes of the cells and the temperature change trend, collaboratively corrects the preliminary health status obtained by S5, solves the prediction deviation problem caused by ignoring the temperature gradient in the existing technology, and obtains a final health status that is more in line with the actual health level of the cells, providing accurate single-cell data support for subsequent calculation of the overall health status of the battery pack.

[0086] S61 determines the spatial location attributes of the battery cell. The system pre-stores the spatial location information of each cell in the battery pack, classifies the cells based on the structural design rules of the battery pack, and clarifies the spatial location attributes of each cell.

[0087] Spatial location attributes are divided into two categories: central and edge locations. A centrally located cell is defined as one completely surrounded by other cells, with no surface directly contacting the battery pack casing or cooling medium. An edge-located cell is defined as one with at least one surface directly contacting the battery pack casing or cooling medium. This classification is chosen because the heat dissipation paths and thermal resistances of cells at different locations within the battery pack differ significantly: centrally located cells have longer heat dissipation paths and higher thermal resistance, leading to heat accumulation; edge-located cells have shorter heat dissipation paths and lower thermal resistance, allowing for easier heat dissipation. This classification accurately reflects the differences in heat distribution within the battery pack, laying the foundation for subsequent temperature gradient compensation.

[0088] S62 monitors the temperature change trend of battery cells. The system calls the cell temperature change rate calculated in S2, and combines it with a preset threshold to determine the temperature change trend of each cell, providing a dynamic basis for the calculation of the compensation factor.

[0089] The system sets thresholds for judging temperature change trends: when the rate of temperature change is greater than 0.3℃ / min, it is judged as a continuous temperature rise; when the rate of temperature change is less than -0.2℃ / min, it is judged as a continuous temperature fall; when the rate of temperature change is between -0.2℃ / min and 0.3℃ / min, it is judged as a stable temperature. The reason for choosing 0.3℃ / min as the continuous rise threshold is that when the battery is under high load discharge, the temperature rise rate of the cells in the middle position generally exceeds this value, accurately identifying the heat accumulation state. The reason for choosing -0.2℃ / min as the continuous fall threshold is that when the cells stop operating under high load or are in a heat dissipation environment, the temperature fall rate usually reaches this value, accurately identifying the heat dissipation improvement state. Both thresholds are determined based on statistical data from the actual operating conditions of the battery, ensuring the accuracy of trend judgment.

[0090] S63 Calculate the temperature gradient compensation factor The system combines the spatial location attributes of the battery cell, temperature change trends, and real-time temperature difference to calculate compensation factors related to the temperature gradient, and specifically compensates for prediction deviations under different scenarios.

[0091] First, calculate the real-time temperature difference: Real-time temperature difference = Current cell real-time temperature - Average real-time temperature of cells at the edge positions. The average real-time temperature of cells at the edge positions is the arithmetic mean of the real-time temperatures of all cells at the edge positions at the same moment. The reason for choosing the average temperature of cells at the edge positions as the benchmark is that the heat dissipation conditions of cells at the edge positions are consistent, the temperature is more stable, and it can accurately reflect the environmental benchmark temperature of the battery pack, ensuring the rationality of the temperature difference calculation.

[0092] Based on the combination of different spatial location attributes and temperature change trends, compensation factors are calculated in four categories: For cells in the middle position where the temperature continues to rise, the compensation factor is calculated as follows: the compensation factor equals 1 minus 0.005 multiplied by the real-time temperature difference. The reason for choosing a coefficient of 0.005 is that experimental data shows that the underestimation deviation of the internal resistance of cells in the middle position is linearly related to the temperature difference. For every 1°C increase in temperature difference, the overestimation deviation of SOH caused by the underestimation of internal resistance is about 0.5%. This coefficient can accurately offset this deviation, making the corrected SOH closer to reality. For cells in the middle position where the temperature continues to drop, the compensation factor is calculated as 1 plus 0.003 multiplied by the real-time temperature difference. The reason for choosing a coefficient of 0.003 is that heat dissipation improves as the temperature drops, and the deviation in internal resistance assessment gradually decreases. It is necessary to gently restore the SOH value to avoid over-correction. This coefficient is determined based on experimental data on the correlation between the rate of improvement in heat dissipation and the reduction in deviation, and the correction magnitude matches the degree of deviation mitigation. For cells located at the edge and experiencing a continuous temperature drop, the compensation factor is calculated as follows: the compensation factor equals 1 plus 0.006 multiplied by the real-time temperature difference. The reason for choosing a coefficient of 0.006 is that the overestimation of internal resistance is more significant for cells at the edge and in low-temperature locations. Experiments have shown that for every 1°C increase in temperature difference, the underestimation of SOH (Surveillance Optimization) is approximately 0.6%. This coefficient can fully compensate for this deviation, ensuring that the SOH assessment accurately reflects the true health status of the cell. For cells located at the edge and experiencing a continuous temperature rise, the compensation factor is calculated as 1 minus 0.004 multiplied by the real-time temperature difference. The coefficient of 0.004 is chosen because abnormal temperature rise at the edge may stem from heat dissipation failure of adjacent cells or a malfunction within the cell itself, posing a potential risk. Appropriately reducing the SOH value can provide early warning of this risk without causing assessment distortion due to excessive correction, thus balancing risk warning and assessment accuracy.

[0093] S64 correction yields the final healthy state. The system multiplies the preliminary health status obtained in S5 with the temperature gradient compensation factor calculated in S63 to obtain the final health status of the cell. This correction process accurately compensates for systematic deviations caused by temperature gradients through synergistic analysis of spatial location and temperature change trends. In existing technologies, the SOH deviation of a single cell due to uneven temperature distribution exceeds 10%. After correction in this step, the SOH deviation of a single cell can be reduced to less than 0.5%, making the final health status more consistent with the actual health level of the cell.

[0094] S7 Fusion Computing Battery Pack Overall Health Status Based on the final health status of each cell obtained from S6, S7 uses weighted summation logic to fuse the characteristics of individual cells with the overall performance, accurately calculates the overall health status of the battery pack, solves the problem of overall evaluation distortion caused by simple averaging in existing technologies, and provides a pack-level health status benchmark for subsequent feedback adjustments and the next round of prediction.

[0095] S71 calculates reference values ​​for the final health status of battery cells. The system first calls up the final health status data of all cells output by S6, calculates the arithmetic mean of the data set, and defines the average value as a reference value characterizing the overall health level of the battery pack.

[0096] The arithmetic mean is chosen as the reference value because it directly reflects the central tendency of the health status of all cells, avoiding interference from single extreme values, such as severely aged cells, on the overall benchmark judgment. Simultaneously, this reference value provides a unified comparison benchmark for subsequent weighting factor calculations, ensuring a clear basis for assessing the contribution of a single cell to overall performance. For example, if a battery pack contains 10 cells with final health statuses of 92%, 90%, 88%, 91%, 89%, 93%, 90%, 87%, 89%, and 91%, the reference value is (92+90+88+91+89+93+90+87+89+91)%÷10=90%, which accurately reflects the overall health level of the battery pack.

[0097] S72 sets the weight factor calculation parameters The system clearly defines the three core parameters (rated capacity, consistency score, and charge / discharge efficiency) and calculation rules for weighting factor calculation, ensuring that the parameters can accurately quantify the contribution of a single cell to the overall performance of the battery pack, and providing scientific input for weighting factor calculation.

[0098] S721 Determines Rated Capacity Parameters The system retrieves the rated capacity of each cell from the battery management system's preset database. This parameter is the cell's factory calibration value, measured in Ah, and directly reflects the theoretical maximum energy storage capacity of a single cell. The reason for choosing rated capacity as the parameter is that cells with larger rated capacities contribute a higher percentage to the overall battery pack capacity during charging and discharging, and their health status has a more significant impact on overall performance. For example, a 20Ah cell contributes twice as much to the overall capacity as a 10Ah cell, and this parameter is used to reflect the differentiated weighting.

[0099] S722 sets the rules for calculating consistency scores. The consistency score is used to quantify the degree of deviation of a single cell's final health status from a reference value. Let the reference value be the arithmetic mean (in percentage) of the final health status of all cells, then: When the reference value is >10% (i.e., the overall health of the battery pack is greater than 10%): When the reference value is ≤10%: The reason for choosing 10% as the threshold is that when the reference value is ≤10%, the battery pack is close to being scrapped, and the health status of each cell may vary greatly, such as some cells having a health status of 5% and others 15%. If the conventional formula is still used for calculation, there may be outliers where the denominator approaches 0 or the result is negative. Fixing the consistency score at 0.5 can avoid calculation failure and reduce the interference of cells with extreme differences on the overall weight. When the reference value is >10%, the battery pack is in a normal or mild / moderate aging state, and the distribution of cell health status is relatively concentrated. The conventional formula can accurately reflect the consistency differences.

[0100] S723 calculates charge and discharge efficiency The system dynamically calculates the charge and discharge efficiency of each cell based on the real-time temperature of the cells collected by S2. This efficiency directly reflects the energy conversion capability of the cell at the current temperature.

[0101] The calculation logic for charge / discharge efficiency is: Charge / discharge efficiency = × ,in The default value is 0.98, which represents the charge and discharge efficiency of the same model of new battery cell at 25℃. This is a temperature correction function, set based on experimental data of lithium battery temperature characteristics.

[0102] Select The reason for the value of 0.98 is that through extensive testing of new cells of the same model at 25°C, the energy loss of the cells during charging and discharging at this temperature is about 2%, and the efficiency is stable at around 98%, which meets the industry-recognized benchmark for the optimal operating temperature efficiency of lithium batteries.

[0103] Temperature correction function The specific rule is: when the real-time temperature is between 20℃ and 35℃, =1. The reason for choosing this temperature range is that experimental data shows this temperature range has the highest energy conversion efficiency of lithium batteries and the lowest charge and discharge losses, requiring no correction. When the real-time temperature is between 0℃ and 20℃ or between 35℃ and 45℃... =0.95. The reason for choosing this range is that when the temperature deviates from the optimal range, the ion diffusion rate decreases or side reactions intensify, resulting in a decrease in charge / discharge efficiency of about 5%. The correction factor can match the actual efficiency changes. When the real-time temperature is below 0℃ or above 45℃, =0.85. The reason for choosing this range is that the resistance to ion migration increases dramatically at low temperatures and the side reactions are significant at high temperatures, resulting in a sharp decrease in charge and discharge efficiency of about 15%. This correction factor can accurately reflect the efficiency decay at extreme temperatures.

[0104] S73 calculates the single-cell weighting factor Based on the three core parameters determined by S72, the system calculates the weight factor of each cell through normalization processing to ensure that the weight can accurately match the actual contribution of a single cell to the overall health of the battery pack.

[0105] The calculation logic for the single cell weighting factor is as follows: Single cell weighting factor = (rated capacity of the cell × consistency score of the cell × charge / discharge efficiency of the cell) ÷ {the sum of (rated capacity × consistency score × charge / discharge efficiency) of all cells}.

[0106] The calculation logic comprehensively evaluates the multidimensional contribution capability of a single cell by multiplying "rated capacity (theoretical contribution) × consistency score (state matching degree) × charge and discharge efficiency (real-time performance)"; then, by normalizing by dividing by the sum of the products of all cells, the sum of the weighting factors of all cells is made to be 1. For example, a battery pack contains two cells, A and B. Cell A has parameters of 20Ah, consistency score of 0.98, and charge / discharge efficiency of 0.98, while cell B has parameters of 10Ah, consistency score of 0.95, and charge / discharge efficiency of 0.95. Then, the product of cells A is approximately 20 × 0.98 × 0.98 ≈ 19.208, and the product of cells B is approximately 10 × 0.95 × 0.95 ≈ 9.025. The total is approximately 28.233. The weighting factor of cell A is approximately 19.208 ÷ 28.233 ≈ 0.68, and the weighting factor of cell B is approximately 9.025 ÷ 28.233 ≈ 0.32. This result accurately reflects the higher contribution of cell A to the overall performance.

[0107] The overall health status is obtained by weighted summation of S74. The system multiplies the final health status of each cell by its corresponding weighting factor, and then sums all the products to obtain the overall health status of the battery pack. Compared with the existing technology that simply takes the average health status of all cells, this calculation method highlights the positive contribution of cells with "large rated capacity, excellent health status, and high charge / discharge efficiency" to the overall performance, while suppressing the interference of cells with "small capacity, poor health status, and low efficiency" on the overall evaluation, making the overall health status more consistent with the actual output capacity of the battery pack. For example, if cell A has a final health status of 90% and a weight of 0.68, and cell B has a final health status of 80% and a weight of 0.32, the overall health status = 90% × 0.68 + 80% × 0.32 = 61.2% + 25.6% = 86.8%. This result reflects the dominant role of cell A while also taking into account the influence of cell B, avoiding the underestimation bias caused by simply averaging 85%. The matching degree is significantly improved compared to the simple averaging method of 88%, effectively solving the problem of overall evaluation distortion.

[0108] S8 Feedback Adjustment Prediction Strategy Based on the overall health status obtained from S7, S8 dynamically optimizes the data acquisition, internal resistance correction, and feature analysis strategies in the subsequent prediction process, ensuring that the system maintains high prediction accuracy throughout the entire battery life cycle. This solves the problem of accuracy degradation caused by fixed strategies in existing technologies that are difficult to adapt to the characteristics of batteries at different aging stages.

[0109] S81 Setting Feedback to Adjust Trigger Conditions The system first sets the conditions for initiating feedback adjustments to ensure that adjustments are only performed when the prediction results are stable and there is room for optimization, thus avoiding blind adjustments that could cause system fluctuations.

[0110] The specific triggering condition is as follows: in three consecutive rounds of prediction, the change in the overall health status calculated in each round compared to the previous round is less than or equal to 1%. The reason for choosing "three consecutive rounds" as the time dimension standard is that stability in one or two rounds may stem from short-term operating condition fluctuations (such as temporary low-load operation), and cannot represent the true stability level of the prediction result. Stability in three consecutive rounds can eliminate interference from accidental fluctuations, ensuring that the overall health status prediction has entered a stable range. At this point, initiating adjustments will not lead to strategy mismatch due to initial errors. The reason for choosing "1%" as the change threshold is that when the change in the overall health status is ≤1% for multiple consecutive rounds, the deviation between the prediction result and the actual battery health level is controlled within 2%, reaching a high accuracy standard and providing a basis for initiating strategy optimization. If the change is too large, it indicates that the current prediction still has significant errors, and blind adjustments may exacerbate the accuracy problem.

[0111] S82 adjusts data acquisition strategy Based on the overall health status range obtained from S7, the system adjusts the dynamic data acquisition parameters in S2 accordingly, so that the acquisition strategy is precisely matched with the characteristic requirements of the battery aging stage.

[0112] S821 moderate aging stage data collection and adjustment When the overall health status is less than 80%, the battery enters the moderate aging stage. Experimental data shows that moderately aged batteries exhibit significantly increased temperature sensitivity; even a temperature fluctuation of 0.1℃ can cause internal resistance changes exceeding 3% and voltage characteristic deviations exceeding 5mV. Continuing with the original data acquisition strategy could easily miss key dynamic characteristics. Therefore, the system tightened the "stable threshold" for temperature data acquisition in S2 from 0.2℃ / min to 0.15℃ / min. This adjustment makes the system more sensitive to temperature changes; high-frequency acquisition is triggered when the temperature change rate exceeds 0.15℃ / min, avoiding insufficient data accuracy due to temperature sensing lag and providing more accurate temperature input for subsequent internal resistance correction and polarization voltage calculation.

[0113] S822 Severe Aging Stage Data Collection and Adjustment When the overall health status is less than 70%, the battery enters a severe aging stage. Severely aged batteries experience significant shedding of electrode active materials and damage to ion transport channels, resulting in nonlinear changes in their current response characteristics. The original sampling frequency is insufficient to capture the transient response details after sudden current changes. Therefore, the system reduces the current data acquisition frequency in S2 for high-load scenarios such as hill climbing and rapid acceleration from 15ms / time to 10ms / time. The core effect of this adjustment is to improve the temporal resolution of the current data by 50%, enabling more dense recording of the dynamic trajectory after current changes. This ensures that the calculation of the polarization voltage component in subsequent feature analysis more closely matches the actual current response, reducing feature distortion caused by undersampling of current data.

[0114] S83 Adjustment of Internal Resistance Correction Mapping Relationship Library Update Conditions The system adjusts the update trigger conditions of the mapping relationship library in S3 according to the overall health status to ensure that the update frequency of the mapping relationship library is adapted to the rate of change of battery internal resistance, and avoids correction deviations caused by lag in internal resistance data.

[0115] When the overall health status is less than 80%, the rate of change of internal resistance in moderately aged batteries is more than 50% higher than that of new batteries. Test data shows that the internal resistance of new batteries increases by about 0.5 mΩ per 100 km, while that of moderately aged batteries increases by about 0.75 mΩ per 100 km. If the original 100 km mileage trigger condition is still used, the internal resistance data in the mapping database will not be able to match the actual changes in battery internal resistance in a timely manner, leading to increased deviations in subsequent internal resistance corrections. Therefore, the system shortens the "mileage trigger condition" for updating the mapping database in S3 from 100 km to 80 km. This adjustment synchronizes the update frequency of the mapping database with the rate of change of internal resistance in moderately aged batteries. Every 80 km, the data in the database can be corrected based on the measured internal resistance, ensuring that the ohmic internal resistance and polarization internal resistance queried during subsequent internal resistance corrections are consistent with the current state of the battery, thus reducing internal resistance correction deviations.

[0116] S84 adjusts the polarization coefficient of the feature analysis. The system adjusts the polarization coefficient in S4 to address the deviation of the aging voltage component in a single cell, thereby improving the separation accuracy between the polarization voltage and the aging voltage component and avoiding feature analysis deviations caused by a fixed polarization coefficient.

[0117] S841 determines the polarization coefficient adjustment requirement The system continuously analyzes the aging voltage component of each cell in three consecutive prediction rounds, calculating the deviation of the cell's aging voltage component from the average value of all cells' aging voltage components in each round. If the deviation exceeds 15% for three consecutive rounds, it indicates insufficient separation accuracy between the cell's polarization voltage and aging voltage components. This may be due to variations in polarization characteristics during the cell's aging process, rendering the original polarization coefficient incompatible with its current polarization pattern. Coefficient adjustment is necessary to optimize the separation effect. The reason for using "three consecutive rounds" and "15%" as the judgment criteria is that a single-round deviation may originate from measurement noise, while a deviation for three consecutive rounds indicates a persistent deviation. A 15% deviation exceeds the reasonable error range; without adjustment, subsequent health status calculations will deviate by more than 5%.

[0118] S842 performs polarization coefficient adjustment When adjustment is required, the system will adjust the electrochemical polarization coefficient of the battery cell. Increase by 8%, concentration polarization coefficient The adjustment was increased by 6%. The reason for choosing "8%" and "6%" as the adjustment range is that experimental data shows the polarization characteristics of moderately to severely aged cells typically vary between 5% and 10%. An increase of 8% and 6% can compensate for the deviation between the polarization coefficient and the actual polarization law, while avoiding "overshoot" caused by excessively large adjustments. For example, an increase of more than 10% might lead to an overestimation of the calculated polarization voltage, thus exacerbating the separation error of the aging voltage component. Simultaneously, the system is set to maintain the adjustment result for only 3 rounds. After 3 rounds, if the cell still meets the deviation conditions, the adjustment is reassessed. If the deviation decreases to within 15%, the original polarization coefficient is restored. This setting avoids system stability issues caused by the polarization coefficient being in a state of adjustment for a long time, ensuring that the coefficient adjustment has dynamic reversibility and adapts to the dynamic changes in the cell's polarization characteristics.

[0119] S9 iteration updates the basic health status estimate S9 updates the final health status of the battery cell after S6 collaborative correction and S8 optimization strategy to the basic health status estimate for the next round of prediction, and builds a closed-loop iterative mechanism of "prediction-correction-adjustment-update" to solve the problem of long-term prediction accuracy decay caused by fixed initial values ​​or broken update logic in existing technologies, and ensures that each round of prediction is based on input parameters that fit the actual state of the battery cell.

[0120] S91 confirms data source update The system first identifies the data source for updating the basic health status estimate, and determines the final health status of each cell output by S6 as the sole basis for updating, rejecting the use of uncorrected preliminary health status or historically expired data.

[0121] The core reason for choosing the final health state of S6 as the data source is that this value has been collaboratively corrected through the internal thermal distribution characteristics of the battery pack: for the spatial location attributes of the cell, i.e. the middle position or the edge position, and the temperature change trend, i.e. rising, falling or stabilizing, the temperature gradient compensation factor is calculated to adjust the preliminary health state of S5, eliminating more than 10% of systematic deviation caused by uneven temperature distribution. Its accuracy is significantly better than the preliminary health state of S5 and the initial value of S1.

[0122] The system also excludes other potential data sources: directly using the estimated basic health status of this round will lead to parameter solidification and will not be able to reflect the changes in this round of aging; if the original data such as cell voltage / current are used for derivation, an additional complex model needs to be built, which will introduce new error sources.

[0123] S92 sets update timing The system sets the timing for updating operations, specifying that the basic health status estimate will be updated immediately after the completion of each round of the prediction process. This process covers all stages from real-time data collection in S2 to feedback and adjustment in S8, avoiding premature updates that would cause the data to be unsuitable for the optimization strategy, or delayed updates that would cause the next round of prediction to call out expired data.

[0124] The technical basis for setting this timing is that the S8 feedback adjustment has optimized the data acquisition, internal resistance correction, and feature analysis strategies for the next round based on the overall health status of the current round, while the final health status of S6 is the optimal health assessment result that adapts to the current operating conditions and strategies. The two form a matching relationship of "state assessment - strategy optimization". If the update is made before S8 is completed, the next round of prediction will not be able to adapt to the strategy optimized by S8; if the update is delayed until after the next round of prediction starts, the next round will be forced to use the initial values ​​of this round, resulting in a misalignment problem where the strategy is updated but the input is not.

[0125] S93 performs data updates and persistent storage. The system performs the update and storage operations of the basic health status estimate according to the association format of "cell number-update value-time stamp" to ensure that the data is traceable and not lost when power is off.

[0126] The system first calls the parameter configuration module of the battery management system to locate the storage address of the basic health state estimate for each cell, and replaces the original stored value with the final health state of S6. The replacement process adopts an atomic operation of write-then-verify: the new value is first written to a temporary cache, and the data integrity is verified, such as whether the value range is within 0%-100%, whether there are garbled characters or overflows. After confirming that there are no errors, the original address data is overwritten to avoid data corruption caused by power failure or system abnormality.

[0127] Subsequently, the system synchronously stores the updated baseline health status estimate to non-volatile flash memory, specifically NAND flash, with a storage capacity configured to support backups for 1000 consecutive rounds of updates. A cyclic overwrite mechanism is employed: when the capacity is exhausted, the system automatically deletes the earliest 50 rounds of non-critical backup data. Using non-volatile storage ensures that data is not lost when the vehicle is turned off or the battery is disconnected. Upon the next system startup, there is no need for re-initialization; the latest updated value is directly used as input, avoiding errors such as misidentifying a new battery cell as an old one or using 100% of the initial values ​​of a new battery cell for an old one.

[0128] S94 Verify Update Validity After the system is updated, it performs a simple validity check to eliminate the interference of abnormal update results on the next round of prediction and ensure the rationality of the input parameters.

[0129] The verification logic consists of two steps: First, it checks if the updated value is within a reasonable range of 0%-100%. If a cell's updated value is less than 0% or greater than 100%, the system determines it as abnormal, immediately calls the previously stored baseline health status estimate as a temporary replacement, and triggers a fault alarm with the message "Cell health status data is abnormal; S6 correction process needs to be checked." Second, it calculates the coefficient of variation (standard deviation / mean) of all cell updates within the same battery pack. If the coefficient of variation is greater than 15%, the system determines it as an abnormal cell consistency, meaning that some cells may be severely aged or have measurement faults. Although the update is not terminated, the consistency score will be weighted more heavily in the next S7 fusion calculation to prevent abnormal cells from excessively affecting the overall assessment. This verification step prevents obviously erroneous data from entering the next prediction round while also preventing iteration from being terminated due to minor fluctuations, balancing data reliability and process continuity.

[0130] S95 establishes an iterative traceability mechanism The system generates a unique iteration log for each update operation, recording the associated information of "update time - cell number - original base value - new base value - corresponding S8 adjustment strategy", forming a complete iteration traceability chain.

[0131] The core function of the log traceability is to allow for reverse lookup of the source of the baseline health status estimate when subsequent predictions show accuracy deviations. This confirms whether the estimate was based on the valid final health status of S6 and whether it was compatible with the optimization strategy of S8 at the time, thus pinpointing the root cause of the deviation, such as an anomaly in the S6 correction logic or unreasonable parameter adjustments in S8, rather than blindly re-initializing. Simultaneously, the log data can be used to analyze cell aging rate trends, such as the rate of decrease in values ​​across multiple consecutive update rounds, providing supplementary information for battery life prediction.

[0132] The implementation principle of the lithium battery health state prediction method and system in this application is as follows: This application initializes a basic health state estimate for each cell and dynamically corrects the internal resistance by combining real-time operating data. Then, it analyzes the aging voltage component stripped of polarization interference from the voltage data to calculate the preliminary health state. On this basis, for the temperature gradient distribution problem caused by the difference in heat dissipation conditions inside the battery pack, a temperature gradient compensation factor is calculated based on the spatial location attributes of the cells and their temperature change trends. The preliminary health state is then collaboratively corrected to effectively compensate for the internal resistance assessment deviation caused by uneven temperature, making the final health state more consistent with the actual aging degree of the cells. Subsequently, based on the corrected health state of each cell, the overall health state of the battery pack is calculated by weighted fusion. This overall state is then used to adjust the subsequent data acquisition frequency, internal resistance mapping library update conditions, and polarization coefficient, forming a closed-loop iterative mechanism of "prediction-correction-adjustment-update". This solution effectively solves the problem in existing technologies where the prediction of health status is more than 10% due to the neglect of uneven temperature distribution within the battery pack. By dynamically sensing and compensating for the influence of temperature gradients, the prediction deviation of the health status of a single cell is reduced to less than 0.5%, significantly improving prediction accuracy and consistency. This provides a more reliable basis for battery management system life assessment and balanced maintenance, avoiding mismanagement or premature replacement of batteries due to prediction distortion.

[0133] The above are all preferred embodiments of this application, and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.

Claims

1. A method for predicting the health status of a lithium battery, characterized in that, Includes the following steps: Initialize a basic state of health estimate for each cell in the battery pack; Based on dynamic scene perception, real-time operating data of each battery cell is collected; Based on the temperature in the real-time operating data and the estimated basic health status, the internal resistance of each cell is dynamically corrected. By combining the corrected internal resistance with the estimated basic health status, independent characteristic quantities representing the degree of aging are extracted from the real-time operating data of each cell. Based on the corrected internal resistance and the independent characteristic quantities, the preliminary health status of each cell is calculated; Based on the thermal distribution characteristics inside the battery pack, the initial health status of each cell is collaboratively corrected to obtain the final health status of the cell. The overall health status of the battery pack is calculated by integrating the final health status of all cells. The final health status is updated to the baseline health status estimate predicted in the next round; Based on the overall health status, at least one strategy for data acquisition, internal resistance correction, or feature analysis in subsequent predictions is adjusted accordingly.

2. The method according to claim 1, characterized in that, The method of collecting real-time operating data for each battery cell based on dynamic scene perception specifically includes: Real-time monitoring of the temperature and current change rates of each battery cell; When the rate of temperature change exceeds a preset first rate of change threshold, the frequency of collecting the cell temperature data is increased. When the rate of change of the current exceeds a preset second rate of change threshold, the acquisition frequency of the cell current data is increased.

3. The method according to claim 1, characterized in that, The process of dynamically correcting the internal resistance of each cell based on the temperature from real-time operating data and the estimated baseline health status includes: Establish a mapping library describing the relationship between temperature, aging state, and internal resistance; For each cell, based on its real-time collected temperature data and the basic health status estimate on which the current cycle is based, its corresponding ohmic internal resistance and polarization internal resistance are determined from the mapping relationship library. The determined ohmic internal resistance and the polarization internal resistance are combined to generate the corrected internal resistance of the cell under the current conditions.

4. The method according to claim 3, characterized in that, It also includes the step of updating the mapping relationship database, specifically: When a preset mileage condition is met, or when the difference between the preliminary health status obtained in the current round and the preliminary health status obtained in the previous round exceeds a set threshold, an update to the mapping relationship library is triggered. During the update process, the real-time temperature of the cell in the current cycle and the measured internal resistance of the cell in the preliminary health state determined in the previous cycle are obtained. The measured internal resistance is compared with the internal resistance value obtained from the mapping database based on the same real-time temperature and the same initial health status. If the deviation between the two exceeds the allowable error threshold, the corresponding internal resistance value in the mapping relationship library is corrected based on the deviation.

5. The method according to claim 3, characterized in that, The process of resolving independent characteristic quantities that characterize the degree of aging specifically involves obtaining an aging voltage component. The aging voltage component is obtained by: From the difference between the open-circuit voltage and the measured terminal voltage of the cell, at least one polarization voltage component determined based on the corrected internal resistance and the estimated basic health state is subtracted.

6. The method according to claim 1, characterized in that, The preliminary health status of each cell is collaboratively corrected based on the internal thermal distribution characteristics of the battery pack, specifically including: Based on the spatial location attributes of the battery cell within the battery pack and its temperature change trend, a compensation factor related to the temperature gradient is determined. Based on the compensation factor, the preliminary health status is numerically adjusted to generate the final health status.

7. The method according to claim 1, characterized in that, The overall health status of the battery pack, as calculated by the fusion computing, specifically includes: Based on the stated final health status of all cells, a reference value characterizing the overall level is calculated; For each cell, a weighting factor is determined based on its rated capacity, the degree of deviation of its final state of health from the reference value, and its charge / discharge efficiency at the current temperature. The final health status of each cell is weighted and summed using the weighting factors to obtain the overall health status of the battery pack.

8. The method according to claim 5, characterized in that, The step of adjusting at least one strategy for data acquisition, internal resistance correction, or feature analysis in subsequent predictions based on the overall health status specifically includes: When the change in the overall health status during consecutive preset rounds of prediction is lower than a preset stability threshold, feedback adjustment is initiated. The feedback adjustment includes at least one of the following: Based on the degree of battery aging indicated by the overall health status, adjust the judgment conditions used to trigger changes in the acquisition frequency in the dynamic scene perception. Adjust the conditions for triggering the update of the mapping database based on the degree of battery aging indicated by the overall health status; The coefficients used to calculate the polarization voltage component are adjusted based on the degree of deviation of the aging voltage component of a single cell from the average value of the aging voltage components of all cells in a continuous preset cycle.

9. A lithium battery health status prediction system, characterized in that, For performing the method as described in any one of claims 1-8, comprising: An initialization unit is used to configure a basic state of health estimate for each cell in the battery pack. The dynamic sensing and acquisition unit is used to acquire real-time operating data of each battery cell based on dynamic scene perception. The internal resistance dynamic correction unit is used to dynamically correct the internal resistance of each cell based on the temperature in the real-time operating data and the estimated basic health status. The feature parsing unit is used to combine the corrected internal resistance with the basic health state estimate to parse independent feature quantities characterizing the degree of aging from the real-time operating data of each cell. A health status calculation unit is used to calculate the preliminary health status of each cell based on the corrected internal resistance and the independent characteristic quantities. The collaborative correction unit is used to collaboratively correct the initial health status of each cell based on the heat distribution characteristics inside the battery pack, so as to obtain the final health status of the cell. The fusion computing unit is used to calculate the overall health status of the battery pack based on the final health status of all cells. The feedback adjustment unit is used to adjust at least one strategy for data acquisition, internal resistance correction, or feature analysis in subsequent predictions based on the overall health status. An iterative update unit is used to update the final health status to the estimated basic health status value predicted in the next round.

Images