Hybrid battery pack aging state online identification and life prediction method

Abstract

The application discloses a hybrid battery pack aging state online identification and life prediction method and belongs to the technical field of battery management. The method comprises the following steps: acquiring an end voltage response sequence and a current excitation sequence of a hybrid battery pack under charge and discharge excitation; extracting voltage change rate features at each moment by time domain difference on the end voltage response sequence, and calculating cumulative charge amount features at each moment by cumulative charge amount calculation on the current excitation sequence; constructing a voltage-charge amount response trajectory and extracting trajectory curvature values in each voltage interval; obtaining an aging sensitive gradient field by gradient operation of the trajectory curvature values along the voltage direction and spatially statistically obtaining aging feature distribution parameters; comparing the aging feature distribution parameters with initial feature distribution parameters to obtain an aging state identification result; and updating a recursive equation and extrapolating to obtain a life prediction result. The application realizes joint evaluation of battery aging state identification and life prediction, and effectively improves prediction accuracy.

Description

Technical Field

[0001] This invention relates to the field of battery management technology, specifically to a method for online identification of aging status and lifespan prediction of hybrid power battery packs. Background Technology

[0002] During long-term use, the evolution of internal material structure in hybrid battery packs leads to capacity degradation and increased internal resistance, necessitating online identification of aging conditions and prediction of remaining service life. Existing aging condition identification schemes are mostly based on equivalent circuit model parameter identification or electrochemical impedance spectroscopy (EIS) analysis. The EIS method fits offline calibrated battery model parameters with actual operating data to extract parameters such as ohmic internal resistance or polarization internal resistance, indirectly inferring the degree of aging based on parameter change trends. This method relies on highly accurate battery models, whose parameters are sensitive to the coupling of temperature, state of charge, and operating conditions. Under complex dynamic excitations, the model exhibits poor adaptability, and the identified parameters are severely affected by noise, leading to easy drift in aging condition estimates. The EIS method requires the application of small AC perturbations within a specific frequency range, resulting in stringent testing conditions that are difficult to implement online in actual vehicle operation. It is typically only suitable for periodic laboratory testing.

[0003] Data-driven methods utilize extensive historical data for training to establish a mapping relationship between charge / discharge curve features and capacity degradation. However, these methods require sufficient samples covering the entire lifecycle. In real-world use, hybrid battery packs exhibit incomplete charge / discharge segments and discrete operating conditions, leading to unstable feature extraction and limited model generalization ability. Existing technologies also employ differential voltage analysis or incremental capacity analysis to extract peak features of capacity changes with voltage to characterize aging. However, these methods typically require long-term charge / discharge cycles with low current to obtain quasi-static curves, which cannot adapt to the short-term dynamic segments of frequent start-stop cycles and drastic current fluctuations in hybrid battery packs.

[0004] Existing technologies struggle to capture subtle morphological changes reflecting gradual aging evolution during dynamic, incomplete charge-discharge cycles in aging feature extraction. Changes in the local curvature of the voltage-charge curve in hybrid battery packs under shallow charge-discharge conditions reveal information about phase transitions in electrode active materials and limited lithium-ion diffusion. Directly analyzing the entire curve globally obscures these local deformation features. Furthermore, when converting aging features into lifetime predictions, existing recursive models often employ simple empirical degradation models or constant-rate assumptions, failing to establish closed-loop correction with online-identified aging features. This results in the predicted trajectory failing to track the nonlinear drift of the actual aging rate. Summary of the Invention

[0005] This invention provides an online identification and lifetime prediction method for the aging state of hybrid battery packs, aiming to solve the problems of insufficient extraction of local aging morphology features of dynamic excitation segments and the disconnect between the aging recursive model and online identification.

[0006] To achieve the above objectives, the present invention provides the following technical solution: The present invention provides a method for online identification of aging status and lifespan prediction of hybrid power battery packs, comprising: The process involves acquiring the terminal voltage response sequence and current excitation sequence of a hybrid battery pack under charge and discharge excitation; performing time-domain difference processing on the terminal voltage response sequence to extract the voltage change rate features at each sampling time, and calculating the cumulative charge at each sampling time on the current excitation sequence; constructing a voltage-charge response trajectory using the voltage change rate features and the cumulative charge features, and extracting the trajectory curvature values ​​of the response trajectory in each preset voltage range; performing gradient calculation along the voltage direction on the trajectory curvature values ​​to obtain an aging-sensitive gradient field; performing spatial distribution statistics on the aging-sensitive gradient field to obtain aging feature distribution parameters; calculating the distribution morphology offset based on the aging feature distribution parameters and pre-stored initial feature distribution parameters to obtain aging state identification results; updating the aging state recursive equation using the aging state identification results, and performing extrapolation calculations based on the aging state recursive equation to obtain lifetime prediction results.

[0007] As a technical solution of the present invention, when performing time-domain differential processing on the terminal voltage response sequence, the difference between the terminal voltage value at the i-th sampling time and the terminal voltage value at the (i-1)-th sampling time is calculated to obtain a voltage difference value. This voltage difference value is then compared with the current value at the corresponding time to obtain the instantaneous polarization impedance value. The instantaneous polarization impedance values ​​at each sampling time are arranged into an instantaneous polarization impedance sequence, which serves as the voltage change rate feature. This feature can reflect the speed of response of the battery's internal polarization state to excitation in real time, effectively amplifying the minute impedance changes caused by aging.

[0008] Preferably, when calculating the accumulated charge characteristics, the product of the current value from the initial time to the i-th sampling time and the sampling time interval is accumulated to obtain the accumulated charge value. The difference between adjacent instantaneous polarization impedance values ​​is calculated as the impedance fluctuation value. The accumulated charge value and the impedance fluctuation value constitute a feature pair, which are arranged according to the sampling time sequence. This formed feature pair simultaneously contains the charge accumulation history and the instantaneous fluctuation information of the polarization impedance, providing multi-dimensional sensitive parameters for subsequent trajectory construction.

[0009] As a further improvement of this invention, when constructing the voltage-charge response trajectory, the cumulative charge value at each sampling moment is used as the abscissa, and the corresponding instantaneous polarization impedance value is used as the ordinate to establish the trajectory in a two-dimensional plane, and multiple preset voltage intervals are divided according to the charge range. Within each voltage interval, a quadratic polynomial local fitting is performed on the discrete trajectory points, and the ratio of the second derivative value to the first derivative value of the fitted curve is determined as the trajectory curvature value of that interval. This curvature value is extremely sensitive to the inflection characteristics of the internal chemical reaction of the battery, and can highlight the nonlinear changes in polarization characteristics caused by aging, significantly improving the resolution of aging identification.

[0010] After obtaining the trajectory curvature values, they are arranged in order of voltage intervals to form a curvature value sequence. The difference between adjacent curvature values ​​is then calculated to obtain a curvature gradient sequence. Next, the difference between adjacent curvature gradient sequences is calculated to obtain a second-order curvature gradient sequence. Gradient positions in the curvature gradient sequence that exceed a preset positive threshold are marked as positive gradient positions, and second-order gradient positions in the second-order curvature gradient sequence that cross zero are marked as inflection point gradient positions. The trajectory curvature values ​​corresponding to the positive gradient positions and inflection point gradient positions are extracted to form an aging-sensitive gradient field. This processing method can automatically select the curvature change regions with the most severe aging response, effectively suppressing interference from measurement noise and individual differences, and ensuring high repeatability of the extracted aging-sensitive features.

[0011] Preferably, when performing spatial distribution statistics on the aging-sensitive gradient field, the first number of positive gradient positions and the second number of inflection point gradient positions are counted. The arithmetic mean of the trajectory curvature values ​​corresponding to the positive gradient positions and inflection point gradient positions are calculated respectively to obtain the mean of the positive gradient and the mean of the inflection point gradient. Then, the weighted value is obtained by multiplying the mean by the corresponding number, and the sum of the weighted values ​​is divided by the total number to obtain the aging characteristic distribution parameters. Through the statistical method of quantity weighting, the distribution parameters not only reflect the magnitude of the curvature amplitude, but also reflect the density of the sensitive gradient distribution in the voltage range, thus more comprehensively depicting the spatial distribution pattern of battery aging.

[0012] After obtaining the aging characteristic distribution parameters, the initial characteristic distribution parameters calculated using the same method under the new factory condition are retrieved. The difference between the mean positive gradient in the aging characteristic distribution parameters and the mean initial positive gradient is calculated to obtain the positive gradient mean offset. The difference between the mean gradient at the inflection point and the mean gradient at the initial inflection point is calculated to obtain the inflection point gradient mean offset. The two are then weighted and summed to obtain the distribution morphology offset. Mapping the distribution morphology offset to a preset aging level range yields the aging state identification result at the current moment. Through dual-path offset fusion, both the overall upward shift of polarization impedance and the migration of curvature inflection points can be captured simultaneously, ensuring that the identification results remain accurate for different types of battery aging.

[0013] When updating the aging state recursive equation using the aging state identification results, the current identification result is used as the observed value, and the aging state estimate from the previous moment is used as the predicted value. The state residual is calculated, and the predicted value is corrected using a pre-configured Kalman gain to obtain the aging state estimate for the current moment. This estimate is then substituted into the coefficient matrix of the recursive equation to complete the equation update. This closed-loop recursive method enables the state estimation to track the actual aging process of the battery in real time, eliminates possible fluctuations in the identification results, and provides a stable and continuous recursive basis for extrapolation prediction.

[0014] In the lifetime prediction phase, the current moment is used as the extrapolation starting point. The extrapolation time step and total number of steps are set, and the current aging state estimate is used as the initial state vector of the recursive equation. The aging state prediction value for each extrapolation step is iteratively solved. During the iteration process, the state prediction value, covariance prediction value, and Kalman gain update value for the current moment are calculated using the state prediction value and covariance matrix from the previous moment. These are then combined with preset virtual measurements to obtain the aging state prediction value for the current step. When the predicted value reaches the preset aging failure threshold, the corresponding extrapolation step number is recorded, and its product with the time step is used as the remaining lifetime prediction result. This prediction method inherits the uncertainty quantification capability of the Kalman filter framework, enabling the lifetime prediction results to maintain a fast response while possessing high confidence, providing a reliable decision-making basis for the health management of hybrid battery packs.

[0015] The technical effects and advantages provided by the present invention in the above technical solution are as follows: By constructing a voltage-charge response trajectory and extracting the trajectory curvature values ​​for each preset voltage range, the local electrochemical characteristics of short-term dynamic charge-discharge segments are transformed into geometric morphological parameters. Hybrid battery packs experience severe current fluctuations under actual operating conditions, and the trajectory curvature values ​​are extremely sensitive to the microscale bending of the voltage-charge curve, capturing the phase transition peak shifts and shape distortions caused by the reduction of available lithium sites in the electrode active material and the limitation of solid-phase diffusion. Based on the trajectory curvature values, gradient calculations are performed along the voltage direction to obtain curvature gradient sequences and second-order curvature gradient sequences. Positive gradient positions and inflection point gradient positions are then selected to construct an aging-sensitive gradient field. This gradient field directly locks onto the voltage region in the curve where the aging reaction is most severe, filtering out spurious morphological fluctuations caused by sampling noise and operating condition disturbances. This ensures that the extracted features focus on structural phase transition signals strongly correlated with capacity decay, rather than the overall trend changes of the curve. Spatial distribution statistics of the aging-sensitive gradient field are performed, and by combining the number and mean of positive gradient positions and inflection point gradient positions, quantitative parameters that can stably characterize the degree of aging are obtained, overcoming the limitations of conventional methods that rely solely on a single peak or mean value and lose distribution information.

[0016] By calculating the distribution morphology offset, the current aging characteristic distribution parameters are compared with the initial characteristic distribution parameters under new factory conditions, extracting the positive gradient mean offset and the inflection point gradient mean offset. This dual-morphology offset takes into account two independent aging modes: the high decay of the phase transition peak and the inflection point migration, enabling online identification of the dominant degradation mechanism of the hybrid battery pack at different stages of use. The aging state recursive equation is updated using this identification result, with the observed distribution morphology offset used as feedback. The state estimate of the forward recursion is corrected by Kalman gain, ensuring that the coefficient matrix and state vector of the recursive equation continuously adjust to the actual aging process. When performing extrapolation calculations based on the updated recursive equation, the propagation of state covariance and process noise is considered, and the aging state prediction value for each extrapolation step is obtained through a step-by-step iterative approach. This closed-loop recursive extrapolation mechanism prevents the asymptotic deviation between the predicted trajectory and the actual degradation path caused by model solidification in open-loop prediction, enabling the lifetime prediction results to adapt to the nonlinear time-varying characteristics of the aging rate. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0018] Figure 1 This is a flowchart of a method for online identification of aging status and life prediction of hybrid battery packs. Figure 2 This is a flowchart for extracting the voltage change rate characteristics and cumulative charge characteristics of hybrid battery packs; Figure 3 It is the response trajectory of instantaneous polarization impedance and accumulated charge during the charging and discharging process; Figure 4 These are curves showing the change in the gradient of the trajectory curvature value with the number of cycles for different preset voltage ranges; Figure 5 It is a curve showing the change in the aging distribution pattern offset of a hybrid battery pack with the number of cycles. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, 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 scope of protection of the present invention.

[0020] See Figure 1This invention provides an online identification and lifetime prediction method for the aging state of a hybrid power battery pack. The method first acquires the terminal voltage response sequence and current excitation sequence of the hybrid power battery pack under charge and discharge excitation. Time-domain difference processing is performed on the terminal voltage response sequence to extract the voltage change rate features at each sampling time, and the cumulative charge is calculated on the current excitation sequence to obtain the cumulative charge features at each sampling time. A voltage-charge response trajectory is constructed using the voltage change rate features and the cumulative charge features, and the trajectory curvature values ​​in each preset voltage range are extracted. Gradient calculation along the voltage direction is performed on the trajectory curvature values ​​to obtain the aging-sensitive gradient field. Spatial distribution statistics are performed on the aging-sensitive gradient field to obtain aging feature distribution parameters. Based on the aging feature distribution parameters and pre-stored initial feature distribution parameters, the distribution shape offset is calculated to obtain the aging state identification result. The aging state recursive equation is updated using the aging state identification result, and extrapolation calculation is performed based on this aging state recursive equation to obtain the lifetime prediction result.

[0021] In specific implementation, please refer to Figure 2 After obtaining the terminal voltage response sequence and current excitation sequence of the hybrid battery pack under charge and discharge excitation, time-domain difference processing is performed on the terminal voltage response sequence. For the i-th sampling time in the terminal voltage response sequence, where i is a positive integer greater than 1, the terminal voltage value V at the i-th sampling time is obtained. i and the terminal voltage value V at the (i-1)th sampling time i-1 Calculate the terminal voltage value V i With terminal voltage value V i-1 The difference is used to obtain the voltage difference value at the i-th sampling time. The current value I at the i-th sampling time in the current excitation sequence is obtained. i Calculate the voltage difference and current value I. i The ratio of the two values ​​is used to obtain the instantaneous polarization impedance value R at the i-th sampling time. p,i The formula for calculating the instantaneous polarization impedance is as follows: ; Among them, R p,i V represents the instantaneous polarization impedance value at the i-th sampling time. i V represents the terminal voltage value at the i-th sampling time. i-1 I represents the terminal voltage value at the (i-1)th sampling time. i Let I represent the current value in the current excitation sequence at the i-th sampling time, where i is the sampling time number and is an integer greater than 1. When the current value I... i When the absolute value of is zero or less than a preset small value, the instantaneous polarization impedance value R at the i-th sampling time is... p,i Set as the instantaneous polarization impedance value R at the (i-1)th sampling time. p,i-1For sampling time i=1, the instantaneous polarization impedance value of the first sampling time is set to zero. After traversing all sampling times in the above manner, the obtained instantaneous polarization impedance values ​​R are... p,i Arranged in ascending order of sampling time number, an instantaneous polarization impedance sequence is formed, and the instantaneous polarization impedance sequence is used as a characteristic of voltage change rate.

[0022] In practice, the accumulated charge is calculated for the current excitation sequence. The sampling time interval is set to a fixed value Δt. For the i-th sampling moment, starting from the initial sampling moment of the current excitation sequence, the current value I is sequentially taken at each sampling moment. k Where k is an integer from 1 to i, calculate each current value I. k The cumulative charge value Q at the i-th sampling time is obtained by multiplying the product of the sampling time interval Δt and summing all the products. i Using the obtained instantaneous polarization impedance sequence, the instantaneous polarization impedance value R at the i-th sampling time is obtained. p,i and the instantaneous polarization impedance value R at the (i-1)th sampling time p,i-1 Calculate the instantaneous polarization impedance value R p,i With instantaneous polarization impedance value R p,i-1 The difference is used to obtain the impedance fluctuation value R at the i-th sampling time. p,i For sampling time i=1, the impedance fluctuation value R p,i Set it to zero. Set the accumulated charge value Q at the i-th sampling time step. i With impedance fluctuation value R p,i Form a feature pair. Perform the above operation on all sampling times, arrange the resulting feature pairs in ascending order of sampling time number to form a feature pair sequence, and use the feature pair sequence as the feature of accumulated charge.

[0023] See Figure 3 The horizontal axis of the graph represents the accumulated charge in ampere-hours (Ah), ranging from 0 to approximately 50 Ah, while the vertical axis represents the instantaneous polarization impedance in milliohms (mΩ), ranging from 0.5 to 3.0 mΩ. The graph contains two trajectory curves: the solid blue line represents the charging trajectory, and the dashed red line represents the discharging trajectory. The legend clearly indicates the correspondence between the two.

[0024] As the accumulated charge increases, the instantaneous polarization impedance of the charging trajectory exhibits significant fluctuations, with peaks appearing at approximately 11 Ah, 28 Ah, and 43 Ah, reaching a maximum peak value of nearly 2.9 mΩ and a minimum valley value of approximately 0.8 mΩ. The discharging trajectory also shows fluctuations, with relative peaks appearing at approximately 7 Ah, 22 Ah, and 38 Ah, reaching a maximum peak value of approximately 2.7 mΩ and a minimum valley value of approximately 0.7 mΩ. The two trajectories show a certain degree of offset and overlap at the peaks and troughs, reflecting the different dynamic response characteristics of polarization impedance with changes in accumulated charge during charging and discharging.

[0025] Several sampling points, marked with blue dots and red squares, are evenly distributed on the trajectory curve, indicating that the discrete sampling and time-domain differential processing of the instantaneous polarization impedance values ​​cover the entire accumulated charge range. Overall, the polarization impedance fluctuation amplitude of the trajectory during charging is slightly larger than that during discharging, and a more significant peak change is observed in the middle accumulated charge range (approximately 20Ah to 30Ah), reflecting that the battery polarization characteristics are more sensitive to changes in this range.

[0026] In practical implementation, a voltage-charge response trajectory is constructed using voltage rate of change characteristics and cumulative charge characteristics. The voltage rate of change characteristic is a sequence of instantaneous polarization impedance, and the cumulative charge characteristic is a sequence of feature pairs consisting of the cumulative charge value and impedance fluctuation value at each sampling time. For each sampling time, the cumulative charge value is extracted from the cumulative charge characteristic, and the corresponding instantaneous polarization impedance value is extracted from the voltage rate of change characteristic. A discrete point is marked on a two-dimensional coordinate plane with the cumulative charge value as the x-axis and the instantaneous polarization impedance value as the y-axis. After traversing all sampling times, all discrete points are connected in ascending order of sampling time number to form the voltage-charge response trajectory.

[0027] In specific implementation, the voltage-charge response trajectory is divided into N preset voltage intervals based on the charge value. During division, the range of the horizontal axis of the voltage-charge response trajectory, i.e., the minimum and maximum values ​​of the accumulated charge, is obtained. This horizontal axis range is then divided into N equal segments, each corresponding to a preset voltage interval. The value of N is determined by the voltage platform characteristics of the hybrid battery pack; for hybrid battery packs with a nominal voltage of 300V to 400V, the value of N ranges from eight to fifteen. Each preset voltage interval contains at least three discrete points. When the number of discrete points in a preset voltage interval is less than three, that preset voltage interval is merged with adjacent preset voltage intervals. Within each preset voltage interval, a quadratic polynomial local fitting is performed on the discrete points within that interval. For the m-th preset voltage interval, where m is an integer from one to N, the set of discrete points within that interval is (q... m1 ,r m1 ),(q m2 ,r m2),...,(q mL ,r mL ), where L is the number of discrete points within the interval, and q m1 to q mL r represents the cumulative charge value at each discrete point. m1 to r mL Let f be the instantaneous polarization impedance value at each discrete point. A quadratic polynomial is fitted using the least squares method to obtain the local fitting curve function f for the m-th preset voltage interval. m (q)=a m q 2 +b m q+c m , where a m b is the fitting coefficient for the quadratic term. m c is the fitting coefficient for the first-order term. m Let be the constant term fitting coefficients, and q be the cumulative charge variable. The solution process for the quadratic polynomial fitting is as follows: Construct a Vandermonde matrix composed of the cumulative charge values ​​at each discrete point; multiply the pseudo-inverse of the Vandermonde matrix by the column vector composed of the instantaneous polarization impedance values ​​at each discrete point to obtain the fitting coefficient vector, which contains the quadratic term fitting coefficients α. m The fitting coefficient b of the first-order term m and the fitting coefficient c of the constant term m .

[0028] In practical implementation, the trajectory curvature value of the local fitting curve corresponding to each preset voltage range is calculated. The formula for calculating the trajectory curvature value is: ; Among them, C m This represents the trajectory curvature value for the m-th preset voltage interval; This indicates the local fitting curve of the m-th preset voltage range at the reference point. The second derivative value at; This indicates the local fitting curve of the m-th preset voltage range at the reference point. The first derivative value at; This represents the reference point for the m-th preset voltage range. The value is the arithmetic mean of the cumulative charge values ​​at all discrete points within the m-th preset voltage interval; m is the index of the preset voltage interval, and the value of m ranges from one to N. The first derivative value of the locally fitted curve. By reference point Substituting the first derivative function The second derivative value of the locally fitted curve was calculated. By reference point Substituting the second derivative function Calculations show that a mb is the fitting coefficient for the quadratic term. m These are the fitting coefficients for the linear term. The calculated ratios... The trajectory curvature value C for the m-th preset voltage range m The above calculation is performed on all N preset voltage ranges to obtain N trajectory curvature values.

[0029] In practical implementation, gradient calculations are performed along the voltage direction on the trajectory curvature values ​​to obtain an aging-sensitive gradient field. The N trajectory curvature values ​​corresponding to N preset voltage intervals are arranged in ascending order of voltage value within the preset voltage intervals, forming a curvature value sequence. The u-th element in the curvature value sequence is denoted as C. u Where u takes the value of an integer from one to N. The difference between two adjacent trajectory curvature values ​​in the curvature value sequence is calculated; that is, for u decreasing by one from one to N, the curvature value C of the u-th trajectory in the curvature value sequence is calculated. u+1 With the curvature value C of the u-th trajectory u The difference is used to obtain the u-th curvature gradient value G. u The formula for calculating the curvature gradient value is G. u =C u+1 -C u All calculated curvature gradient values ​​are arranged in ascending order of u value to form a curvature gradient sequence, which contains N minus one curvature gradient value.

[0030] In practice, the difference between two adjacent curvature gradient values ​​in the curvature gradient sequence is calculated. That is, for u from one to N minus two, the u-th curvature gradient value G in the curvature gradient sequence is calculated. u+1 With the u-th curvature gradient value G u The difference is used to obtain the second-order gradient value H of the u-th curvature. u All calculated second-order curvature gradient values ​​are arranged in ascending order of u value to form a second-order curvature gradient sequence, which contains N minus two second-order curvature gradient values.

[0031] In practical implementation, a preset positive threshold is set as θ. The preset positive threshold θ is the median of the absolute values ​​of all curvature gradient values ​​in the curvature gradient sequence multiplied by a scaling factor α, where the scaling factor α ranges from 0.3 to 0.6. The curvature gradient values ​​G in the curvature gradient sequence are then... u Gradient positions greater than a preset positive threshold θ are marked as positive gradient positions, and their corresponding indexes in the curvature value sequence are recorded. For the second-order curvature gradient sequence, the second-order curvature gradient values ​​H in the second-order curvature gradient sequence are... u The sign of the adjacent second-order gradient value of curvature H u-1The positions of the second-order gradients with opposite signs are marked as inflection point gradient positions. The index of each inflection point gradient position in the curvature value sequence is recorded, where the index is u plus one. The trajectory curvature values ​​corresponding to the positive gradient positions are extracted, i.e., the corresponding trajectory curvature values ​​are obtained from the curvature value sequence according to the index of the positive gradient position. The trajectory curvature values ​​corresponding to the inflection point gradient positions are extracted, i.e., the corresponding trajectory curvature values ​​are obtained from the curvature value sequence according to the index of the inflection point gradient position. The trajectory curvature values ​​corresponding to all positive gradient positions and all inflection point gradient positions are summarized to form the aging-sensitive gradient field.

[0032] In specific implementation, spatial distribution statistics are performed on the aging-sensitive gradient field to obtain aging characteristic distribution parameters. The first number of positive gradient positions in the aging-sensitive gradient field is counted, denoted as S1. The second number of inflection point gradient positions in the aging-sensitive gradient field is counted, denoted as S2. The trajectory curvature values ​​corresponding to all positive gradient positions in the aging-sensitive gradient field are obtained, and the arithmetic mean of the trajectory curvature values ​​corresponding to all positive gradient positions is calculated, i.e., the sum of the trajectory curvature values ​​corresponding to all positive gradient positions is divided by the first number S1 to obtain the positive gradient mean M1. The trajectory curvature values ​​corresponding to all inflection point gradient positions in the aging-sensitive gradient field are obtained, and the arithmetic mean of the trajectory curvature values ​​corresponding to all inflection point gradient positions is calculated, i.e., the sum of the trajectory curvature values ​​corresponding to all inflection point gradient positions is divided by the second number S2 to obtain the inflection point gradient mean M2. The product of the positive gradient mean M1 and the first number S1 is calculated to obtain the first weighted value W1, and the formula for calculating the first weighted value is W1 = M1 × S1. The product of the inflection point gradient mean M2 and the second quantity S2 is used to obtain the second weighted value W2. The formula for calculating the second weighted value is W2 = M2 × S2.

[0033] In practical implementation, the aging characteristic distribution parameters are obtained by dividing the sum of the first weighted value W1 and the second weighted value W2 by the sum of the first quantity S1 and the second quantity S2. These aging characteristic distribution parameters are denoted as... The formula for calculating the aging characteristic distribution parameters is expressed as follows: ; in, Let M1 represent the aging characteristic distribution parameters, W1 represent the first weighting value, W2 represent the second weighting value, S1 represent the first number of positive gradient positions in the aging-sensitive gradient field, and S2 represent the second number of inflection point gradient positions in the aging-sensitive gradient field. W1 is obtained by multiplying the positive gradient mean M1 by the first number S1, and W2 is obtained by multiplying the inflection point gradient mean M2 by the second number S2. The positive gradient mean M1 is the arithmetic mean of the trajectory curvature values ​​corresponding to all positive gradient positions, and the inflection point gradient mean M2 is the arithmetic mean of the trajectory curvature values ​​corresponding to all inflection point gradient positions.

[0034] See Figure 4In the figure, the vertical axis represents the trajectory curvature gradient value, and the horizontal axis represents the number of cycles. The curves illustrate the trend of curvature gradient changes with the number of cycles after performing gradient calculations along the voltage direction on the trajectory curvature values ​​of N preset voltage intervals (intervals 1 to 11 shown in the figure) in the above embodiment. Specifically, the curvature gradient values ​​of all preset voltage intervals gradually increase with the number of cycles, exhibiting the characteristics of the aging-sensitive gradient field evolving over time. The curvature gradient curves of different voltage intervals differ in numerical amplitude and rate of change. Among them, the curvature gradient values ​​of preset voltage intervals 10 and 11 are the largest, close to 0.8, and the overall curves are relatively smooth and steadily rising, indicating that the trajectory curvature changes in these intervals are more sensitive to the aging state. Relatively low curvature gradient values ​​appear in preset voltage intervals 1 and 2, and the values ​​start from near 0 and gradually increase with the number of cycles, showing obvious characteristics of the aging start-up stage. Around the 60th cycle, the curvature gradient curves of each voltage interval show obvious abrupt jumps, reflecting that the hybrid battery pack has undergone significant structural or performance changes at this stage, leading to a sharp increase in the trajectory curvature gradient.

[0035] In practice, based on the aging characteristic distribution parameters and the pre-stored initial characteristic distribution parameters, the distribution morphology offset is calculated to obtain the aging state identification result. The pre-stored initial characteristic distribution parameters are calculated by the hybrid battery pack in its new factory condition using the same method as obtaining the aging characteristic distribution parameters. The initial characteristic distribution parameters include the initial positive gradient mean, the initial inflection point gradient mean, the initial first quantity, and the initial second quantity. The initial positive gradient mean is the arithmetic mean of the trajectory curvature values ​​corresponding to all positive gradient positions in the aging-sensitive gradient field in the new factory condition. The initial inflection point gradient mean is the arithmetic mean of the trajectory curvature values ​​corresponding to all inflection point gradient positions in the aging-sensitive gradient field in the new factory condition. The initial first quantity is the statistical count of positive gradient positions in the aging-sensitive gradient field in the new factory condition. The initial second quantity is the statistical count of inflection point gradient positions in the aging-sensitive gradient field in the new factory condition. The initial characteristic distribution parameters are stored in the non-volatile storage unit of the hybrid battery pack's battery management system and are retrieved each time online aging state identification is performed.

[0036] In practical implementation, after calling the pre-stored initial feature distribution parameters, the mean of the positive gradient, the mean of the inflection point gradient, the first quantity, and the second quantity are obtained from the aging feature distribution parameters. The difference between the mean of the positive gradient in the aging feature distribution parameters and the initial mean of the positive gradient is calculated to obtain the positive gradient mean offset, denoted as ΔM1. The difference between the mean of the inflection point gradient in the aging feature distribution parameters and the initial mean of the inflection point gradient is calculated to obtain the inflection point gradient mean offset, denoted as ΔM2. The distribution shape offset is calculated by weighting the positive gradient mean offset ΔM1 and the inflection point gradient mean offset ΔM2. The formula for calculating the distribution shape offset is expressed as: ; in, The values ​​represent the distribution pattern shift: ΔM1 represents the positive gradient mean shift, ΔM2 represents the inflection point gradient mean shift, β1 represents the weighting coefficient of the positive gradient mean shift, and β2 represents the weighting coefficient of the inflection point gradient mean shift. The weighting coefficients β1 and β2 are configured based on the aging sensitivity characteristics of the hybrid battery pack. The values ​​of the weighting coefficients β1 and β2 satisfy β1 + β2 = 1. For hybrid battery packs where the aging mechanism is primarily based on the loss of active lithium ions, the weighting coefficient β1 is set to 0.6 to 0.7, and the weighting coefficient β2 is set to 0.3 to 0.4. For hybrid battery packs where the aging mechanism is primarily based on the loss of positive electrode active material, the weighting coefficient β1 is set to 0.3 to 0.4, and the weighting coefficient β2 is set to 0.6 to 0.7. The specific values ​​of weighting coefficients β1 and β2 were determined by analyzing the correlation between the positive gradient mean offset and the inflection point gradient mean offset and the capacity decay after the hybrid battery pack under laboratory cyclic aging test conditions. The combination of weighting coefficients that maximizes the linear correlation coefficient between the distribution shape offset and the capacity decay was selected as the configuration value.

[0037] In practical implementation, the distribution morphology offset is obtained. Then, the distribution pattern offset is... The aging status is mapped to a preset aging level range to determine the aging status identification result. The preset aging level range is divided into five levels: healthy state, slightly aged state, moderately aged state, severely aged state, and failure state. The division of the preset aging level range is based on the distribution pattern offset. The numerical range corresponds to the actual usable capacity retention rate of the hybrid battery pack. When the distribution pattern shifts... When the value is less than the first threshold, the aging state identification result is considered healthy. The first threshold is set to three standard deviations of the baseline value of the distribution morphology offset under new factory conditions. When the distribution morphology offset... When the value is greater than or equal to the first threshold and less than the second threshold, the aging state identification result is a slight aging state. The second threshold is the distribution morphology offset value corresponding to the capacity decay of the hybrid battery pack to 90% of its nominal capacity. When the distribution morphology offset... When the value is greater than or equal to the second threshold and less than the third threshold, the aging state identification result is a moderate aging state. The third threshold is the distribution morphology offset value corresponding to the capacity decay of the hybrid battery pack to 80% of its nominal capacity. When the distribution morphology offset... When the value is greater than or equal to the third threshold and less than the fourth threshold, the aging state identification result is a severe aging state. The fourth threshold is the distribution morphology offset value corresponding to the capacity decay of the hybrid battery pack to 70% of its nominal capacity. When the distribution morphology offset... When the threshold is greater than or equal to the fourth threshold, the aging state identification result is a failure state. The first, second, third, and fourth thresholds are all calibrated by simultaneously collecting the terminal voltage response sequence and current excitation sequence during the cyclic aging test of the hybrid battery pack under laboratory conditions, calculating the distribution pattern offset corresponding to each aging stage, and pre-stored in the battery management system.

[0038] See Figure 5 In the graph, the horizontal axis represents the number of cycles of the hybrid battery pack, ranging from 0 to 600, and the vertical axis represents the distribution morphology shift, ranging from 0 to 0.6. The blue solid line curve shows the trend of the calculated distribution morphology shift as the number of cycles increases. The graph also marks four key threshold lines: the green dashed line is the first threshold, with a value of approximately 0.02, representing the boundary between a healthy state and a slightly aged state; the orange dashed line is the second threshold, with a value of approximately 0.1, representing the boundary between a slightly aged state and a moderately aged state; the red dashed line is the third threshold, with a value of approximately 0.2, representing the boundary between a moderately aged state and a severely aged state; and the purple dashed line is the fourth threshold, with a value of approximately 0.35, representing the boundary between a severely aged state and a failure state. According to the trend of the blue curve in the figure, the distribution pattern offset starts from close to zero and gradually increases with the number of cycles. It crosses the second threshold of 0.1 at about 300 cycles and enters the moderate aging state range. It then continues to rise and exceeds the third threshold of 0.2 at about 460 cycles, entering the severe aging state range. At about 550 cycles, it reaches and exceeds the fourth threshold of 0.35 and enters the failure state range.

[0039] In practical implementation, the aging state recursive equation is updated using the aging state identification results. The aging state recursive equation adopts a linear time-varying state-space model, which takes the form of a first-order discrete state equation. At time t, the aging state recursive equation is represented as the aging state vector x. t The recurrence relation is expressed as: xt =A t x t-1 +B t u t +w t ; Where, x t Let x be the aging state vector at the current time t. t-1 Let A be the aging state vector of the previous time step t-1. t Let B be the state transition matrix at time t. t Let u be the input matrix at the current time t. t Let w be the input vector at the current time t. t Let A be the process noise vector at the current time t. The aging state vector contains two state variables: the capacity degradation rate and the internal resistance growth rate of the hybrid battery pack. The state transition matrix A is... t and input matrix B t The elements in the matrix are associated with the aging level represented by the aging state identification result at the current moment, and different aging levels correspond to different matrix element values. The matrix element values ​​are obtained offline by using a system identification method after the hybrid battery pack has undergone cyclic testing at different aging levels under laboratory conditions, and are stored in the memory of the battery management system in a lookup table manner.

[0040] In practice, the aging state identification result obtained by mapping the distribution morphology offset at the current moment is used as the observation value z. t Observation z t As a scalar, its value is the capacity decay rate corresponding to the aging state identification result. Obtain the estimated aging state value 'a' from the previous time step t-1. t-1|t-1 This aging state estimate is set to a zero vector during battery management system initialization and is updated and stored after each identification. The previous aging state estimate, a... t-1|t-1 As the predicted value, the corresponding predicted observed value is calculated, that is, the predicted value is mapped to the observation space using the observation matrix H. The observation matrix H takes values ​​of [1,0], indicating that the observed value directly corresponds to the capacity decay rate. The observed value z is then calculated. t Compared with the predicted observation Ha t-1|t-1 The difference is used to obtain the state residual e. t State residual e t The formula for calculation is: e t =z t -Ha t-1|t-1 ; Among them, e t Represents the state residual; z t Represents the observed value, a scalar, which is the numerical value of the capacity decay rate; H represents the observation matrix, with values ​​in the range [1,0]; a t-1|t-1This represents the estimated aging state value at the previous time t-1, expressed as a vector.

[0041] In the specific implementation, a pre-configured Kalman gain K is obtained. Kalman gain K is a two-dimensional column vector, and its value is obtained through offline calibration. During offline calibration, the terminal voltage response sequence and current excitation sequence of the hybrid battery pack under known aging conditions are collected. Using the Monte Carlo simulation method, the two element values ​​of Kalman gain K are iteratively optimized with the objective of minimizing the mean square value of the aging state estimation error. The calibrated Kalman gain K is stored in the non-volatile memory of the battery management system and directly called during online runtime. The state residual e is then utilized. t The product of the Kalman gain K and the predicted value a t-1|t-1 Perform the correction to obtain the aging state estimate a at the current time t. t|t The revised calculation formula is as follows: a t|t =a t-1|t-1 +Ke t ; Among them, a t|t Represents the estimated aging state at current time t, a vector; a t-1|t-1 Represents the estimated aging state value at the previous moment, a vector; K represents the Kalman gain, a two-dimensional column vector; e t Represents the state residual, a scalar.

[0042] In practical implementation, the estimated aging state value a at the current time t will be... t|t The first element, the estimated capacity decay rate, is substituted into the lookup function of the coefficient matrix of the aging state recursive equation. The lookup function outputs the corresponding state transition matrix A based on the numerical range of the estimated capacity decay rate. t and input matrix B t This completes the update of the aging state recurrence equation. The updated aging state recurrence equation is used for the recurrence calculation of the next time step and subsequent extrapolation calculations. Input vector u t During recursive updates, the value is taken as the rate of change of the cumulative charge at the current moment. This rate of change of the cumulative charge is obtained by dividing the difference between the cumulative charge value at the current moment and the cumulative charge value at the previous moment by the sampling time interval.

[0043] In practice, extrapolation calculations are performed based on the updated aging state recursive equation to obtain the lifetime prediction results. The current time t is taken as the starting time for extrapolation, and the extrapolation time step ΔT is set. The extrapolation time step ΔT is an integer multiple of the typical charge-discharge cycle of the hybrid battery pack, for example, the duration of one standard charge-discharge cycle. The total number of extrapolation steps is set. , The value is a positive integer sufficient to cause the predicted aging state value to reach the preset aging failure threshold, for example, a value of two thousand. The estimated aging state value a at the current time t is... t|t The initial state vector, as used in the aging state recursive equation, is denoted as... .

[0044] In practice, the aging state recursive equation is solved iteratively step by step according to the extrapolation time step ΔT. During the iterative solution process, no new actual measurements are introduced; instead, preset virtual measurements are used to replace the actual observations. The preset virtual measurements are set as the capacity decay rate corresponding to the aging state identification result at the current moment, and remain constant during the extrapolation process. For the k-th extrapolation iteration, the value of k ranges from 1 to... The integer represents the predicted aging state value at step k-1. Substituting the state transition matrix A into the aging state recurrence equation t With input matrix B t The lookup table function retrieves the state transition matrix A corresponding to the k-th step. k and input matrix B k The input vector takes a constant value during the iteration process; this constant value is the average rate of change of accumulated charge of the hybrid battery pack under standard charge-discharge cycles. The expression for calculating the state prediction at step k is: ; in, Let A represent the estimated state value at step k. k This represents the state transition matrix at step k. B represents the predicted aging state at step k-1. k This represents the input matrix at step k. This is the vector of the average cumulative charge change rate under standard charge-discharge cycles.

[0045] In practical implementation, the predicted aging state value at step k-1 is obtained. The corresponding covariance matrix Initial values ​​of the covariance matrix Set it to a unit diagonal matrix. Then set the covariance matrix of step (k-1) to... Substitute into state transition matrix A k The expression for calculating the covariance estimate at step k is: ; in, Let A represent the covariance estimate at step k. k This represents the state transition matrix at step k. Let the covariance matrix at step k-1 be denoted as . A represents kThe transpose of the matrix is ​​given, where Q is the process noise covariance matrix. The process noise covariance matrix Q is a diagonal matrix, and its diagonal elements are determined by analyzing the noise statistical characteristics of the hybrid battery pack's cycle aging data. These values ​​are fixed and stored in the battery management system. The covariance prediction value from step k is then used. Calculate the Kalman gain update value K at step k using the measurement noise covariance matrix R. k The calculation formula is: ; Among them, K k This represents the Kalman gain update value at step k. H represents the covariance prediction at step k, where H is the observation matrix with values ​​in the range [1,0]. This represents the transpose of the observation matrix H, and R is the measurement noise covariance, which is a scalar value determined through statistical analysis of the measurement noise under laboratory conditions.

[0046] In practice, the state prediction value at step k is used. The Kalman gain update value K at step k. k and preset virtual measurement values Calculate the predicted aging state value at step k. The formula for calculating the predicted aging status is expressed as: ; in, This represents the predicted aging state value obtained from the k-th extrapolation iteration; This represents the state prediction value at step k, derived from the aging state prediction value at step (k-1). K is obtained through recursion using the aging state recursive equation. k The updated Kalman gain value at step k is calculated from the covariance estimate at step k and the measurement noise covariance matrix. represents the preset virtual measurement value, which is the capacity decay rate corresponding to the aging state identification result at the current moment; H represents the observation matrix, with a value of [1,0]; k is the extrapolation iteration step number, and the value of k ranges from 1 to 1. The integer. After calculating the aging state prediction value in step k, the covariance matrix in step k is updated simultaneously. , where I is the identity matrix, used for the next iteration.

[0047] In practice, after each extrapolation iteration, the predicted aging state value at step k is checked. The system checks whether the capacity degradation rate component reaches the preset aging failure threshold. The preset aging failure threshold is set at 20% of the nominal capacity of the hybrid battery pack; that is, when the capacity degradation rate component exceeds 0.2%, aging failure is considered to have occurred. When the predicted aging state value reaches the preset aging failure threshold, the corresponding extrapolation step number is recorded. Extrapolate the number of steps. The product of the product and the extrapolated time step ΔT is used as the predicted remaining useful life. The formula for calculating the predicted remaining useful life is as follows: ; RUL represents the predicted remaining useful life. This indicates the number of extrapolation steps when the aging failure threshold is reached, and ΔT represents the extrapolation time step. The remaining service life prediction value is output as the service life prediction result.

[0048] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method for online identification of aging status and lifespan prediction of hybrid power battery packs, characterized in that, The method includes: Obtain the terminal voltage response sequence and current excitation sequence of the hybrid battery pack under charge and discharge excitation; The terminal voltage response sequence is subjected to time-domain differential processing to extract the voltage change rate features at each sampling time, and the cumulative charge is calculated on the current excitation sequence to obtain the cumulative charge features at each sampling time. A voltage-charge response trajectory is constructed using the voltage change rate feature and the accumulated charge feature, and the trajectory curvature value of the response trajectory in each preset voltage range is extracted; Gradient calculations are performed along the voltage direction on the trajectory curvature value to obtain an aging-sensitive gradient field; Spatial distribution statistics are performed on the aging-sensitive gradient field to obtain aging characteristic distribution parameters; Based on the aging feature distribution parameters and the pre-stored initial feature distribution parameters, the distribution morphology offset is calculated to obtain the aging state identification result. The aging state identification results are used to update the aging state recursive equation, and extrapolation calculations are performed based on the aging state recursive equation to obtain the lifetime prediction results.

2. The method for online identification of aging status and lifespan prediction of hybrid power battery packs according to claim 1, characterized in that, The step of performing time-domain differential processing on the terminal voltage response sequence to extract the voltage change rate features at each sampling time includes: The difference between the terminal voltage value at the i-th sampling time and the terminal voltage value at the (i-1)-th sampling time in the terminal voltage response sequence is calculated to obtain the voltage difference value at the i-th sampling time. The instantaneous polarization impedance value at the i-th sampling moment is obtained by calculating the ratio of the voltage difference value at the i-th sampling moment to the current value at the i-th sampling moment in the current excitation sequence. The instantaneous polarization impedance values ​​at each sampling time are arranged according to the sampling time sequence to form an instantaneous polarization impedance sequence, and the instantaneous polarization impedance sequence is used as the voltage change rate feature.

3. The method for online identification of aging status and lifespan prediction of hybrid power battery packs according to claim 2, characterized in that, The step of performing cumulative charge calculation on the current excitation sequence to obtain the cumulative charge characteristics at each sampling time includes: The product of the current value and the sampling time interval at each sampling time from the initial sampling time to the i-th sampling time in the current excitation sequence is summed to obtain the cumulative charge value at the i-th sampling time. The difference between the instantaneous polarization impedance value at the i-th sampling time and the instantaneous polarization impedance value at the (i-1)-th sampling time is calculated to obtain the impedance fluctuation value at the i-th sampling time. The cumulative charge value and impedance fluctuation value at the i-th sampling time are used to form a feature pair at the i-th sampling time. All feature pairs at all sampling times are arranged in time sequence as the cumulative charge feature.

4. The method for online identification of aging status and lifespan prediction of hybrid power battery packs according to claim 3, characterized in that, The step of constructing a voltage-charge response trajectory using the voltage change rate feature and the accumulated charge feature, and extracting the trajectory curvature value of the response trajectory in each preset voltage range, includes: The voltage-charge response trajectory is constructed in a two-dimensional coordinate plane with the cumulative charge value at each sampling time as the abscissa and the instantaneous polarization impedance value at each sampling time as the ordinate. The voltage-charge response trajectory is divided into N preset voltage intervals according to the charge value; Within each preset voltage range, a quadratic polynomial local fitting is performed on the discrete points of the voltage-charge response trajectory to obtain the local fitting curve corresponding to each preset voltage range. Calculate the ratio of the second derivative to the first derivative of the local fitting curve corresponding to each preset voltage range, and use the ratio as the trajectory curvature value of each preset voltage range.

5. The method for online identification of aging status and lifespan prediction of hybrid power battery packs according to claim 4, characterized in that, The step of performing gradient calculation along the voltage direction on the trajectory curvature value to obtain the aging-sensitive gradient field includes: The trajectory curvature values ​​of each preset voltage range are arranged into a curvature value sequence according to the voltage range order; Perform difference calculation on two adjacent curvature values ​​in the curvature value sequence to obtain a curvature gradient sequence; Perform difference calculation on two adjacent curvature gradient values ​​in the curvature gradient sequence to obtain the second-order curvature gradient sequence; The gradient positions in the curvature gradient sequence that are greater than a preset positive threshold are marked as positive gradient positions, and the second-order gradient positions in the curvature second-order gradient sequence that cross zero are marked as inflection point gradient positions; The trajectory curvature values ​​corresponding to the positive gradient positions and the trajectory curvature values ​​corresponding to the inflection point gradient positions are extracted to form an aging-sensitive gradient field.

6. The method for online identification of aging status and lifespan prediction of hybrid power battery packs according to claim 5, characterized in that, The spatial distribution statistics of the aging-sensitive gradient field are performed to obtain aging characteristic distribution parameters, including: The first number of positive gradient positions in the aging-sensitive gradient field is counted, and the second number of inflection point gradient positions in the aging-sensitive gradient field is counted. Calculate the arithmetic mean of the trajectory curvature values ​​corresponding to each positive gradient position to obtain the mean of the positive gradient; Calculate the arithmetic mean of the trajectory curvature values ​​corresponding to each inflection point gradient position to obtain the mean gradient of the inflection point; Calculate the product of the mean of the positive gradient and the first quantity to obtain the first weighted value, and calculate the product of the mean of the inflection point gradient and the second quantity to obtain the second weighted value; The aging characteristic distribution parameters are obtained by dividing the sum of the first weighted value and the second weighted value by the sum of the first quantity and the second quantity.

7. The method for online identification of aging status and lifespan prediction of hybrid power battery packs according to claim 1, characterized in that, The step of calculating the distribution morphology offset based on the aging feature distribution parameters and the pre-stored initial feature distribution parameters to obtain the aging state identification result includes: The pre-stored initial feature distribution parameters are called, wherein the initial positive gradient mean, initial inflection point gradient mean, initial first quantity and initial second quantity are calculated by the method described in claim 6 under the condition of being newly manufactured; The difference between the mean positive gradient in the aging characteristic distribution parameters and the initial mean positive gradient is calculated to obtain the offset of the mean positive gradient. The difference between the mean gradient at the inflection point and the mean gradient at the initial inflection point in the aging characteristic distribution parameters is calculated to obtain the offset of the mean gradient at the inflection point. The weighted sum of the positive gradient mean offset and the inflection point gradient mean offset is used as the distribution pattern offset; The distribution pattern offset is mapped to a preset aging level range to determine the aging state identification result.

8. The method for online identification of aging status and lifespan prediction of hybrid power battery packs according to claim 7, characterized in that, The step of updating the aging state recursive equation using the aging state identification result includes: The current aging state identification result is used as the observation value, and the aging state estimate of the previous moment is used as the prediction value. The difference between the observed value and the predicted value is calculated to obtain the state residual; The predicted value is corrected by multiplying the state residual with the pre-configured Kalman gain to obtain the aging state estimate at the current moment. The aging state estimate at the current moment is substituted into the coefficient matrix of the aging state recursive equation to complete the update of the aging state recursive equation.

9. The method for online identification of aging status and lifespan prediction of hybrid power battery packs according to claim 8, characterized in that, The extrapolation calculation based on the aging state recursive equation to obtain the lifetime prediction result includes: Set the extrapolation time step and the total number of extrapolation steps, with the current time as the starting time of the extrapolation. The current estimated aging state is used as the initial state vector of the aging state recursive equation. The aging state recursive equation is solved iteratively according to the extrapolation time step to obtain the aging state prediction value corresponding to each extrapolation step. When the predicted aging state value reaches the preset aging failure threshold, the corresponding extrapolation step number is recorded. The product of the extrapolation step number and the extrapolation time step is used as the remaining useful life prediction value, which is the useful life prediction result.

10. The method for online identification of aging status and lifespan prediction of hybrid power battery packs according to claim 9, characterized in that, The step of iteratively solving the aging state recursive equation according to the extrapolation time step to obtain the aging state prediction value corresponding to each extrapolation step includes: Substitute the aging state prediction value of step k-1 into the state transition matrix and input matrix of the aging state recursive equation to calculate the state prediction value of step k. Substitute the covariance matrix of the aging state prediction value at step k-1 into the state transition matrix and the process noise covariance matrix to calculate the covariance prediction value at step k. The updated Kalman gain is calculated using the covariance prediction at step k and the measurement noise covariance matrix. The aging state prediction value for step k is calculated using the state prediction value for step k, the Kalman gain update value for step k, and the preset virtual measurement value.

Images