Lithium battery health state estimation and correction method and system based on hierarchical fusion

Abstract

The invention discloses a lithium battery health state online estimation and correction method and system based on hierarchical fusion, and belongs to the technical field of battery management systems. The method comprises the steps that an off-line prior layer inquires a life table according to historical operation data of a battery to calculate the initial maximum available capacity; the on-line estimation layer adopts a main-auxiliary double extended Kalman filter to cooperatively estimate the state of charge (SOC) and the maximum available capacity, and outputs an on-line health state estimation value; the strong correction layer calculates the actual capacity through a two-point method when a high confidence coefficient condition is met, and outputs a strong correction health state value; and the fusion decision-making layer performs weighted fusion according to the confidence coefficient to output a final health state value, and feeds back the actual capacity to the auxiliary extended Kalman filter for state resetting. The method solves the problems that the coupling estimation of the state of charge and the maximum available capacity is difficult, and the online estimation long-term drifting and working condition adaptability is insufficient, and has high precision, high robustness and good engineering practicability.

Description

Technical Field

[0001] This invention relates to the field of battery management system technology, specifically a method and system for estimating and correcting the state of health of lithium batteries based on hierarchical fusion. Background Technology

[0002] Accurate estimation of the State of Health (SOH) of lithium batteries is a key technological foundation for Battery Management Systems (BMS) to achieve state monitoring, lifespan prediction, and safety management. State of health is typically defined as the ratio of the battery's current maximum usable capacity to its factory rated capacity, reflecting the degree of battery aging and remaining lifespan. With the rapid development of electric vehicles and energy storage systems, accurate estimation of battery state of health is of great significance for ensuring safe system operation, optimizing energy management strategies, and realizing the secondary use of batteries.

[0003] Existing methods for estimating state of health (SOC) face the following main technical challenges. First, the relationship between the battery's SOC and maximum usable capacity is problematic. The high degree of coupling in electrical characteristics makes traditional single-model filtering methods prone to mutual interference when simultaneously estimating these two parameters, leading to slow convergence or even divergence in the estimation results. Secondly, filtering-based estimation methods heavily rely on accurate battery models and precise initial states; initial value deviations or model parameter mismatches directly affect estimation accuracy and system convergence. Thirdly, purely online filtering algorithms lack absolute benchmark correction mechanisms; estimation errors caused by sensor noise and model mismatch accumulate over time, resulting in long-term drift in the estimation results. Furthermore, many high-precision algorithms rely on complete standard charge-discharge curves, making them difficult to trigger or effectively apply under the intermittent and dynamic load conditions of actual vehicle operation.

[0004] While some existing technologies have attempted to employ a dual Kalman filter architecture, most solutions only apply it to the joint estimation of the state of charge and model parameters. They lack a comprehensive technical solution for systematically and hierarchically fusing online estimation with offline benchmarks and high-precision anchor events, thus failing to fundamentally solve the aforementioned problems of coupled estimation, long-term drift, and operational adaptability. Therefore, there is an urgent need for a health state estimation scheme that can achieve complementary advantages, possess adaptive capabilities, and balance high accuracy with high robustness. Summary of the Invention

[0005] This invention proposes a method and system for estimating and correcting the state of health of lithium batteries based on hierarchical fusion, in order to solve the technical problems in the prior art, such as the difficulty in estimating the coupled state of charge and maximum available capacity, the lack of an absolute benchmark in online estimation leading to long-term drift, and the insufficient adaptability of the algorithm to actual dynamic operating conditions.

[0006] To address the aforementioned technical problems, this invention provides a method for estimating and correcting the state of health of lithium batteries based on hierarchical fusion, comprising the following steps: Step S1: The offline prior layer queries the pre-stored lifetime table based on the battery's historical operating data to calculate the initial maximum usable capacity. The initial maximum available capacity Passed to the online estimation layer; Step S2: Online estimation of the secondary extended Kalman filter of the layer with the initial maximum available capacity. Using initial values, estimate the maximum available capacity. The main extended Kalman filter utilizes the maximum available capacity. Estimate the State of Charge (SOC); the sub-extended Kalman filter updates the maximum available capacity using the SOC. According to the maximum available capacity With rated capacity The ratio of the two values ​​is used to output an online health status estimate. ; Step S3: The strong calibration layer monitors the battery's operating conditions. When the high confidence condition is met, the actual capacity is calculated using the two-point method. According to the actual capacity With the rated capacity The ratio of the output to the strongly corrected health status value. ; Step S4: The fusion decision layer bases its decisions on the online health status estimate. and the strongly corrected health status value The confidence levels are weighted and fused to output the final health status value. ; the actual capacity Feedback is sent to the secondary extended Kalman filter to reset the maximum available capacity. .

[0007] Preferably, in step S1, the battery historical operating data includes historical average temperature, historical average state of charge range, and cumulative equivalent cycle count, and the life table includes a calendar life table and a cycle life table.

[0008] Preferably, the state equation of the main extended Kalman filter is: ; ; in, Let SOC be the value at time k. Indicates the sampling period. Indicates Coulomb efficiency. This represents the current at time k. This represents the maximum capacity at time k-1. Let be the terminal voltage of the RC circuit at time k. The polarization resistance is represented by τ, and the time constant is represented by τ. This indicates process noise.

[0009] Preferably, in step S2, the secondary extended Kalman filter reduces the maximum available capacity. Modeled as a random walk process, its state equation is: ; in, The process noise is represented; the observation equation of the secondary extended Kalman filter is based on the ampere-hour integral principle, and the virtual observation is constructed using the state of charge (SOC) output by the primary extended Kalman filter.

[0010] Preferably, the observation equation of the sub-extended Kalman filter is: ; in, and These are the state-of-charge estimates output by the main extended Kalman filter at time k and time k-1, respectively. Let be the current value at time k. Let η be the sampling period and η be the coulomb efficiency. This represents the maximum capacity at time k-1. The observation noise is related to the observation noise covariance of the secondary extended Kalman filter and the state-of-charge estimation error covariance of the primary extended Kalman filter.

[0011] Preferably, in step S3, the high confidence condition includes at least one of the following conditions: Condition A: During the constant current charging phase, the initial state of charge is lower than a preset low threshold. Charge to the cutoff voltage ; Condition B: During the constant current discharge phase, the initial state of charge is fully charged, and the discharge continues until the cutoff voltage is reached. Or the state of charge is lower than the preset low threshold. .

[0012] Preferably, in step S3, the two-point method is used to calculate the actual capacity. The formula is: ; in, This indicates the total throughput of the charging or discharging segment. and These represent the charge states at the start and end times of the segment, respectively.

[0013] Preferably, in step S4, the formula for weighted fusion is: ; ; in, This represents the final health status value. Indicates the fusion weight. Indicates strongly corrected health status value The confidence coefficient, Indicates online health status estimate The confidence coefficient.

[0014] Preferably, in step S4, the reset includes: changing the state variable of the sub-extended Kalman filter from the current maximum available capacity. Reset to the actual capacity The error covariance matrix of the secondary extended Kalman filter is then reset to a preset minimum value.

[0015] This invention also provides a lithium battery health state estimation and correction system based on hierarchical fusion, applicable to the above-mentioned method, comprising: The offline prior layer module is used to query the pre-stored life table based on the battery's historical operating data and calculate the initial maximum usable capacity. and the initial maximum available capacity Passed to the online estimation layer module; The online estimation layer module includes a main extended Kalman filter and a secondary extended Kalman filter, wherein the secondary extended Kalman filter is configured with the initial maximum available capacity. Estimate the maximum available capacity for the initial values. The main extended Kalman filter utilizes the maximum available capacity. The sub-extended Kalman filter estimates the state of charge (SOC) and updates the maximum available capacity using the SOC. According to the maximum available capacity With rated capacity The ratio outputs the online health status estimate. ; The strong calibration layer module is used to monitor the battery's operating conditions and calculates the actual capacity using a two-point method when high confidence conditions are met. According to the actual capacity With the rated capacity The ratio of the output strongly corrected health status value ; The fusion decision layer module is used to determine the online health status estimate. and the strongly corrected health status value The confidence levels are weighted and fused to output the final health status value. and the actual capacity The secondary extended Kalman filter, fed back to the online estimation layer module, resets the maximum available capacity. .

[0016] The beneficial effects of the present invention include at least the following: First, by periodically introducing high-precision, strongly corrected health state values ​​based on physical calculations as absolute benchmark anchors, this invention fundamentally eliminates the long-term drift problem that may occur in pure model online estimation algorithms, thus ensuring the accuracy and reliability of health state estimation throughout the battery's entire life cycle.

[0017] Second, the present invention uses an offline prior layer to provide good initial values ​​for the sub-extended Kalman filter, and combines a strong correction mechanism to periodically provide an absolute truth reference, enabling the system to quickly track sudden changes in capacity or correct small deviations accumulated over a long period of time, significantly enhancing its adaptive capability and improving its convergence speed.

[0018] Third, this invention fully utilizes the voltage, current, and temperature data collected by existing sensors in the battery management system without incurring additional hardware costs. The computational load adopts a hierarchical strategy, with strong correction only performed when triggered by specific operating conditions. The online computational load of the extended Kalman filter is optimized to meet the real-time requirements of automotive-grade chips, facilitating implementation and integration on existing battery management system hardware platforms.

[0019] Fourth, the three-layer structure of this invention forms a multi-layered protection mechanism: even if the offline lifetime table data has deviations, the dual extended Kalman filter in the online estimation layer can gradually correct them during operation; even if the online estimation temporarily shows a divergent trend, the strong correction layer can bring it back on track; even if the strong correction is not triggered for a long time, the online estimation layer can still provide continuous health status estimates. This redundant design significantly improves the overall fault tolerance and reliability of the system. Attached Figure Description

[0020] Figure 1 A schematic diagram of the system architecture of the online estimation and correction method for lithium battery health status based on hierarchical fusion provided in an embodiment of the present invention; Figure 2 A block diagram of the main-sub dual extended Kalman filter collaborative estimation provided in an embodiment of the present invention; Figure 3 The strong correction trigger and state reset timing logic diagram provided for the embodiments of the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the protection scope of the present invention.

[0022] This invention provides a hierarchical fusion-based method for estimating and correcting the health status of lithium batteries. Its core idea is to construct a four-layer fusion architecture consisting of an offline prior layer, an online estimation layer, a strong correction layer, and a fusion decision layer, achieving an organic unity of rapid initialization, dynamic tracking, error elimination, and reliable output. Figure 1 As shown, the method includes the following steps.

[0023] Step S1: The offline prior layer queries the pre-stored lifetime table based on the battery's historical operating data to calculate the initial maximum usable capacity. The initial maximum available capacity It is then passed to the online estimation layer.

[0024] The main function of the offline prior layer is to quickly obtain a reasonable initial health state value based on the battery's historical operating statistics when the system starts up or wakes up, avoiding the need for online estimation of the convergence process starting from the rated value. Specifically, the battery management system reads the battery pack's historical operating statistics from non-volatile memory, which includes information on three dimensions: historical average temperature, historical average state of charge range, and cumulative equivalent cycle count.

[0025] Based on the aforementioned historical operational statistics, the system queries the lifespan table pre-stored in read-only memory. This lifespan table comprises two parts: a calendar lifespan table and a cycle lifespan table. The calendar lifespan table reflects the battery's calendar aging pattern under different temperature and state of charge ranges, while the cycle lifespan table reflects the battery's cycle aging pattern under different cycle depths and cycle counts. After obtaining the calendar aging coefficient and cycle aging coefficient from the tables, the initial maximum usable capacity is calculated. The offline prior layer will set the initial maximum available capacity. The values ​​are passed to the online estimation layer as the initial state values ​​for the sub-extended Kalman filter.

[0026] Step S2: Online estimation of the secondary extended Kalman filter of the layer with initial maximum available capacity Using initial values, estimate the maximum available capacity. The main extended Kalman filter utilizes the maximum available capacity. Estimate the State of Charge (SOC); the sub-extended Kalman filter updates the maximum available capacity using the SOC. Based on maximum available capacity With rated capacity The ratio of the two values ​​is used to output an online health status estimate. .

[0027] The online estimation layer is the core module of this invention, employing a dual-filter architecture with a main extended Kalman filter and a secondary extended Kalman filter operating in parallel. The core idea of ​​this architecture is to separately estimate fast-changing and slow-changing states. The main extended Kalman filter is responsible for estimating the rapidly changing state of charge (SOC), while the secondary extended Kalman filter is responsible for estimating the slowly changing maximum available capacity. The two filters work collaboratively through parameter exchange, effectively avoiding the mutual interference problem caused by state coupling in traditional single-filter architectures. Figure 2 As shown, the collaborative estimation process of the master-slave dual extended Kalman filter includes the following steps.

[0028] The main extended Kalman filter is established based on the Thevenin equivalent circuit model. This model consists of an ideal voltage source, an ohmic internal resistance, and an RC parallel network, which can well describe the static and dynamic response characteristics of the battery. A discretized state-space model is established. The state equation of the main extended Kalman filter is: ; ; in, Let SOC be the value at time k. Indicates the sampling period. Indicates Coulomb efficiency. This represents the current at time k. This represents the maximum capacity at time k-1. Let be the terminal voltage of the RC circuit at time k. The polarization resistance is represented by τ, and the time constant is represented by τ. This represents process noise. The observation equation for the main extended Kalman filter is: ; in, This is the measured value of the terminal voltage at time k. Let be the open-circuit voltage, a function of the state of charge. For ohmic internal resistance, For observation noise. It is particularly important to note the maximum available capacity in the state equation. It is not a fixed constant, but an estimate provided in real time by the secondary extended Kalman filter, which is the key link to enable the two filters to work together.

[0029] The secondary extended Kalman filter has the maximum usable capacity. Let be the state variable. Considering that the battery capacity changes extremely slowly over a short timescale, it is modeled as a random walk process, with the state equation as follows: ; in, This represents process noise, reflecting the uncertainty of the capacity's slow decay over time. The observation equations of the secondary extended Kalman filter are based on the ampere-hour integral principle, using the state of charge (SOC) output of the primary extended Kalman filter to construct virtual observations. Specifically, according to the definition of the ampere-hour integral, there is a deterministic relationship between the change in SOC and the current integral and the capacity, thus constructing virtual observations: ; in, and These are the state-of-charge estimates output by the main extended Kalman filter at time k and time k-1, respectively. Let be the current value at time k. Let η be the sampling period and η be the coulomb efficiency. This represents the maximum capacity at time k-1. This is to account for observation noise. It's worth noting that, due to inherent errors in the state of charge estimation itself, the observation noise covariance of the secondary extended Kalman filter needs dynamic adjustment, and its value is related to the state of charge estimation error covariance of the primary extended Kalman filter. This dynamic coupling mechanism allows the two filters to mutually perceive each other's estimation confidence levels, achieving effective information fusion.

[0030] The collaborative workflow of the master-slave dual extended Kalman filter is as follows. During the initialization phase, the state vector of the master extended Kalman filter is initialized as follows: The state initialization of the secondary extended Kalman filter is provided by the offline prior layer. Simultaneously initialize the error covariance matrices of the two filters.

[0031] During the time update phase, within each sampling period, the main extended Kalman filter and the secondary extended Kalman filter perform one-step state prediction and covariance prediction according to their respective state equations. During the measurement update phase, the main extended Kalman filter first calculates innovation using the acquired terminal voltage measurements and the maximum available capacity prediction provided by the secondary extended Kalman filter, then calculates the Kalman gain and updates the state of charge (SOC) estimate. Subsequently, the secondary extended Kalman filter constructs virtual observations using the SOC estimate sequence updated by the main extended Kalman filter, and updates the maximum available capacity estimate by combining this with the dynamically calculated observation noise covariance.

[0032] During the output and iteration phase, the system outputs the estimated state of charge at the current moment. and online health status estimates ,in = / The secondary extended Kalman filter will update the maximum available capacity. The information is passed to the main extended Kalman filter for collaborative estimation in the next sampling period, forming a two-way flow of information.

[0033] Step S3: The strong calibration layer monitors the battery's operating conditions. When the high confidence condition is met, the actual capacity is calculated using the two-point method. According to actual capacity With rated capacity The ratio of the output to the strongly corrected health status value. .

[0034] The purpose of the strong correction layer is to directly calculate the actual capacity of the battery using the two-point method under specific high-confidence operating conditions, obtaining a high-precision absolute benchmark value of the battery's health state, which is then used to periodically eliminate the accumulated error of online estimation. For example... Figure 3 As shown, the workflow of the strong correction layer includes three stages: trigger condition monitoring, two-point capacity calculation, and confidence management.

[0035] In the trigger condition monitoring stage, the system monitors the battery's operating condition in real time to determine whether the high-confidence capacity calculation condition is met. The high-confidence condition defined in this invention includes at least one of the following two typical scenarios.

[0036] Condition A is a complete constant current CC charging phase, starting from the initial SOC (and triggering SOC static correction) being lower than the threshold. (e.g., 20%) Start charging until the charging cutoff condition (e.g., reaching the cutoff voltage) is met. trigger).

[0037] Condition B is a complete constant current CC discharge phase, starting from high SOC (full charge correction has been triggered) until the discharge cutoff condition (such as reaching the cutoff voltage). or below the threshold (And triggers SOC static correction).

[0038] When any of the above high-confidence conditions are met, the two-point capacity calculation is initiated for strong correction. The system records the start time of the charging or discharging segment. and end time By performing precise ampere-hour integration on the current within this time interval, the total throughput of this segment can be obtained. Simultaneously, by consulting the open-circuit voltage-state-of-charge relationship curve using the resting voltage at the start and end times, the initial state of charge can be obtained. and end of state of charge According to the two-point method principle, this segment reflects the actual capacity. The calculation formula is: ; This leads to the strongly corrected SOH value for the healthy state: = / .

[0039] In the confidence management phase, the system provides strong correction for health status values. Assign the highest confidence coefficient This indicates a high degree of confidence in the value. Simultaneously, based on the error covariance of the sub-extended Kalman filter or the online health status estimate... The stability of recent outputs is used to dynamically calculate the confidence coefficient. The value range is [0,1].

[0040] Step S4: The fusion decision layer bases its decisions on the online health status estimates. Strongly corrected health status value The confidence levels are weighted and fused to output the final health status value. ; actual capacity Feedback is sent to the secondary extended Kalman filter to reset the maximum available capacity. .

[0041] The fusion decision layer is the output hub of this invention, responsible for receiving the initial health status value from the offline prior layer and the health status estimate value from the online estimation layer. Health status values ​​of the strong correction layer Based on the information on its confidence level and the comprehensive decision-making process, the final health status value is output. The integrated decision-making layer has two operating modes.

[0042] In normal operation mode, when the strong correction layer is not triggered, the fusion decision layer directly outputs the health status estimate of the online estimation layer. As the final health status value .

[0043] In strong correction trigger mode, when the strong correction layer calculates Subsequently, the fusion decision layer performs two operations: weighted fusion and state reset. The weighted fusion algorithm calculates the fusion weights based on the confidence coefficients and then performs a weighted average of the online estimates and strong correction values. ; Among them, the fusion weight Determined by the confidence coefficient: ; Due to the strong correction confidence coefficient Typically set to a higher value, the fusion result will quickly shift towards a strongly corrected health state value. To move closer.

[0044] The state reset operation is a crucial step in the strong correction mechanism. The fusion decision layer resets the state variables of the sub-extended Kalman filter from the current maximum available capacity. Force reset to actual capacity Simultaneously, the error covariance matrix of the secondary extended Kalman filter is reset to a preset minimum value, indicating a high degree of confidence in the current capacity estimate. This operation is equivalent to zeroing the online capacity tracker, fundamentally eliminating the slow-varying drift error that may accumulate in the secondary extended Kalman filter. The calibrated secondary extended Kalman filter will provide a more accurate maximum available capacity to the primary extended Kalman filter, further improving the state-of-charge estimation accuracy of the primary extended Kalman filter, forming a positive feedback loop.

[0045] Taking a certain electric vehicle power battery as an example, its rated capacity =60Ah, sampling period Δt=1s. After the system is powered on or woken up, the health status is estimated according to the method of this invention. The workflow is as follows.

[0046] During system initialization, the battery management system reads historical statistics of the battery pack from non-volatile memory, calculating the historical average temperature, historical average SOC, and updated cumulative equivalent cycle count since the last critical data update. Using this data as an index, it queries the three-dimensional (temperature, SOC, cycle count) calendar cycle capacity decay table stored in read-only memory to obtain the initial state of health value. The value is 0.98, corresponding to the initial maximum available capacity. The initial state of charge (SFC) of the primary extended Kalman filter is 58.8 Ah. The SFC of the secondary extended Kalman filter is estimated at 50% based on the open-circuit voltage, and initialized to 58.8 Ah.

[0047] In the collaborative estimation phase of the main-sub dual extended Kalman filters, it is assumed that the current at time k is 10A (discharge) and the measured terminal voltage is 3.65V. The main extended Kalman filter uses the state-of-charge (SOC) estimate from the previous time step and the maximum available capacity estimate provided by the sub extended Kalman filter to perform SOC prediction, and then updates the measurement using the terminal voltage measurement, assuming that the updated SOC estimate is 48%. The sub extended Kalman filter uses the SOC estimates from the main extended Kalman filter at the current time step and the previous time step. and Through constructed virtual observations Perform a measurement update, assuming the updated estimate of the maximum available capacity is obtained. The system outputs an estimated online health status of 58.75 Ah. It is 0.979.

[0048] During the strong correction event triggering phase, it is assumed that the vehicle completes a deep discharge and is then connected to a slow charging station for charging. The battery management system detects that the charging current is in a stable constant current mode, and the state of charge corresponding to the initial resting voltage is approximately 5%, which is lower than the preset low threshold of 10%, thus determining that the conditions for complete constant current charging are met. When the charging current cuts off and enters the constant voltage phase, the end time is recorded. The total throughput is obtained by integrating the constant current charging segment in ampere-hours. The capacity is 55.3 Ah. By consulting the open-circuit voltage-state-of-charge curve using the resting voltage after charging, the final state of charge is approximately 100%. The actual capacity is calculated using the two-point method. The result is 55.3Ah / 0.95=58.2Ah, corresponding to a strongly corrected health status value. It is 0.97.

[0049] During the fusion correction and output phase, the fusion decision layer receives the strongly corrected health status value. The confidence coefficient is set to 0.97. The value is 1.0. Assume the current online health status estimate is... The confidence coefficient is 0.7, and the fusion weight is calculated as 1.0 / (1.0+0.7)≈0.58. The final health state value is calculated as 0.59×0.97+0.41×0.979≈0.974. Simultaneously, the state variables of the secondary extended Kalman filter are forcibly reset to 58.2Ah, and the error covariance matrix is ​​reset to its minimum value. Afterward, the primary-secondary dual extended Kalman filter will continue to operate from the corrected state, and the system will continuously output high-precision estimates of the state of charge and health state.

[0050] This invention also provides a lithium battery health state estimation and correction system based on hierarchical fusion, applicable to the above-described method embodiments. The system comprises four functional modules: an offline prior layer module, an online estimation layer module, a strong correction layer module, and a fusion decision layer module. These modules interact with each other through a data interface.

[0051] The offline prior layer module is used to query the pre-stored lifetime table based on the battery's historical operating data and calculate the initial maximum usable capacity. This initial maximum available capacity is then passed to the online estimation layer module. This module operates during system startup or wake-up, providing a good initial value for online estimation.

[0052] The online estimation layer module consists of two sub-modules: a main extended Kalman filter and a secondary extended Kalman filter. The secondary extended Kalman filter uses the initial maximum available capacity. Estimate the maximum available capacity for the initial values. The main extended Kalman filter utilizes the maximum available capacity. The sub-extended Kalman filter estimates the State of Charge (SOC) and updates the maximum available capacity using the SOC. Based on the maximum available capacity With rated capacity The ratio outputs the online health status estimate. This module operates continuously during system operation, providing real-time health status estimates.

[0053] The strong calibration layer module is used to monitor the battery's operating conditions. When high confidence conditions are met, the actual capacity is calculated using the two-point method. Output a strong correction health status value based on the ratio of actual capacity to rated capacity. This module operates only when triggered by specific operating events, providing a high-precision absolute reference for health status.

[0054] The fusion decision layer module is used to calculate the online health status estimate. Strongly corrected health status value The confidence levels are weighted and fused to output the final health status value. and the actual capacity The secondary extended Kalman filter, fed back to the online estimation layer module, resets the maximum available capacity. This module serves as the system's output hub, responsible for comprehensive decision-making and feedback control.

[0055] The four modules work together to form a complete hierarchical fusion health status estimation system. The offline prior layer module provides initial values ​​for the online estimation layer module, the online estimation layer module continuously outputs health status estimates, the strong correction layer module provides high-precision benchmark values ​​under specific conditions, and the fusion decision layer module integrates information from all layers to output the final result and achieves feedback correction. This hierarchical architecture design enables the system to have comprehensive capabilities such as rapid convergence, dynamic tracking, error elimination, and fault-tolerant operation.

[0056] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described; only preferred embodiments of the present invention are illustrated. The descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the present invention. As long as the combination of these technical features does not contradict each other, it should be considered within the scope of this specification.

[0057] It should be noted that those skilled in the art can make various modifications and improvements without departing from the inventive concept, and these all fall within the scope of protection of this invention. Therefore, the scope of protection of this invention should be determined by the appended claims.

Claims

1. A method for estimating and correcting the state of health of lithium batteries based on hierarchical fusion, characterized in that, Includes the following steps: Step S1: The offline prior layer queries the pre-stored lifetime table based on the battery's historical operating data to calculate the initial maximum usable capacity. The initial maximum available capacity Passed to the online estimation layer; Step S2: Online estimation of the secondary extended Kalman filter of the layer with the initial maximum available capacity. Using initial values, estimate the maximum available capacity. The main extended Kalman filter utilizes the maximum available capacity. Estimate the State of Charge (SOC); the sub-extended Kalman filter updates the maximum available capacity using the SOC. ; Based on the maximum available capacity With rated capacity The ratio of the two values ​​is used to output an online health status estimate. ; Step S3: The strong calibration layer monitors the battery's operating conditions. When the high confidence condition is met, the actual capacity is calculated using the two-point method. According to the actual capacity With the rated capacity The ratio of the output to the strongly corrected health status value. ; Step S4: The fusion decision layer bases its decisions on the online health status estimate. and the strongly corrected health status value The confidence levels are weighted and fused to output the final health status value. ; the actual capacity Feedback is sent to the secondary extended Kalman filter to reset the maximum available capacity. .

2. The method according to claim 1, characterized in that, In step S1, the battery historical operating data includes historical average temperature, historical average state of charge range, and cumulative equivalent cycle count, and the life table includes a calendar life table and a cycle life table.

3. The method according to claim 1, characterized in that, The state equation of the main extended Kalman filter is: ； ； in, Let SOC be the value at time k. Indicates the sampling period. Indicates Coulomb efficiency. This represents the current at time k. The maximum capacity at time k-1. Let be the terminal voltage of the RC circuit at time k. τ represents the polarization resistance, and τ represents the time constant. This indicates process noise.

4. The method according to claim 1, characterized in that, In step S2, the secondary extended Kalman filter will reduce the maximum available capacity. Modeled as a random walk process, its state equation is: ； in, The process noise is represented; the observation equation of the secondary extended Kalman filter is based on the ampere-hour integral principle, and the virtual observation is constructed using the state of charge (SOC) output by the primary extended Kalman filter.

5. The method according to claim 4, characterized in that, The observation equation of the sub-extended Kalman filter is: ； in, and These are the state-of-charge estimates output by the main extended Kalman filter at time k and time k-1, respectively. Let be the current value at time k. Let η be the sampling period and η be the coulomb efficiency. The maximum capacity at time k-1. The observation noise is related to the observation noise covariance of the secondary extended Kalman filter and the state-of-charge estimation error covariance of the primary extended Kalman filter.

6. The method according to claim 1, characterized in that, In step S3, the high confidence condition includes at least one of the following conditions: Condition A: During the constant current charging phase, the initial state of charge is lower than a preset low threshold. Charge to the cutoff voltage ; Condition B: During the constant current discharge phase, the initial state of charge is fully charged, and the discharge continues until the cutoff voltage is reached. Or the state of charge is lower than the preset low threshold. .

7. The method according to claim 6, characterized in that, In step S3, the two-point method is used to calculate the actual capacity. The formula is: ； in, This indicates the total throughput of the charging or discharging segment. and These represent the charge states at the start and end times of the segment, respectively.

8. The method according to claim 1, characterized in that, In step S4, the formula for weighted fusion is: ； ； in, This represents the final health status value. Indicates the fusion weight. Indicates strongly corrected health status value The confidence coefficient, Indicates online health status estimate The confidence coefficient.

9. The method according to claim 1, characterized in that, In step S4, the reset includes: changing the state variables of the sub-extended Kalman filter from the current maximum available capacity. Reset to the actual capacity The error covariance matrix of the secondary extended Kalman filter is then reset to a preset minimum value.

10. A lithium battery health state estimation and correction system based on hierarchical fusion, applicable to the method described in any one of claims 1 to 9, characterized in that, include: The offline prior layer module is used to query the pre-stored life table based on the battery's historical operating data and calculate the initial maximum usable capacity. and the initial maximum available capacity Passed to the online estimation layer module; The online estimation layer module includes a main extended Kalman filter and a secondary extended Kalman filter, wherein the secondary extended Kalman filter is configured with the initial maximum available capacity. Estimate the maximum available capacity for the initial values. The main extended Kalman filter utilizes the maximum available capacity. The sub-extended Kalman filter estimates the state of charge (SOC) and updates the maximum available capacity using the SOC. According to the maximum available capacity With rated capacity The ratio outputs the online health status estimate. ; The strong calibration layer module is used to monitor the battery's operating conditions and calculates the actual capacity using a two-point method when high confidence conditions are met. According to the actual capacity With the rated capacity The ratio of the output strongly corrected health status value ; The fusion decision layer module is used to determine the online health status estimate. and the strongly corrected health status value The confidence levels are weighted and fused to output the final health status value. and the actual capacity The secondary extended Kalman filter, fed back to the online estimation layer module, resets the maximum available capacity. .

Images