A system and method for cascade utilization of retired power battery

Abstract

The present application relates to the field of power battery detection and cascade utilization, in particular to a retired power battery cascade utilization system and method; comprising a detection starting step, connecting the retired power battery to the detection station, applying a mixed pulse sequence and collecting voltage response and surface temperature change data; a feature extraction step, calculating transient impedance characteristic data, extracting intrinsic mode function through Hilbert-Huang transform; a vector solving step, combining the thermodynamic entropy production model to solve the attenuation vector through mapping transformation; a clustering grouping step, performing clustering algorithm analysis based on the attenuation vector to generate grouping instructions containing vector parallelism information; a dynamic recombination step, controlling the dynamic topology recombination mechanism to perform complementary assembly according to the grouping instructions; the present application breaks through the limitation of static screening, identifies the micro aging mechanism by using entropy production mapping, and reduces the module wood barrel effect.

Description

Technical Field

[0001] This invention relates to the field of power battery testing and secondary utilization, specifically to a system and method for the secondary utilization of retired power batteries. Background Technology

[0002] In the scenario of cascade utilization of retired power batteries, it is necessary to accurately test and reassemble cells of mixed origins and varying degrees of aging. Existing sorting schemes generally rely on static voltage, internal resistance indicators, or long-cycle charge-discharge aging tests, which can only characterize the current static health of the battery. Although this scheme has certain feasibility in the initial screening, it is difficult to capture the nonlinear decay trend caused by the thickening of the SEI film or the loss of active materials because it lacks in-depth analysis of the microscopic electrochemical reaction kinetics and thermodynamic entropy production mechanism inside the battery.

[0003] This limitation based on current state snapshots causes seemingly similar battery cells to exhibit divergent degradation paths in subsequent cycles after assembly, leading to a severe "weakest link" effect. Furthermore, long-cycle testing severely restricts production efficiency and fails to meet the needs of large-scale industrial applications.

[0004] Therefore, how to overcome the limitations of static indicators while ensuring testing efficiency, accurately predict the future aging trajectory of battery cells through short-term excitation, and achieve highly consistent reorganization throughout the entire life cycle has become an urgent technical problem to be solved. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for the tiered utilization of retired power batteries, avoiding the limitations of traditional static index-based screening methods that cannot predict future aging trends and have low detection efficiency. Furthermore, through electrochemical entropy production mapping-based decay vector analysis and complementary assembly, it is easier to achieve precise grouping and long-life tiered utilization of retired power batteries. Specifically, the technical solution of this invention is as follows:

[0006] A method for the cascade utilization of retired power batteries includes:

[0007] Connect the retired power battery to the testing station and initialize the testing module;

[0008] In response to the detection initiation command, an electrochemical state dynamic fitting process is executed, including:

[0009] Step 1: The control detection module applies a mixed pulse sequence to the retired power battery and simultaneously collects voltage response data and surface temperature change data;

[0010] Step 2: Based on the voltage response data and surface temperature change data, calculate the transient impedance characteristic data, and process the transient impedance characteristic data through Hilbert-Huang transform to extract the intrinsic mode functions;

[0011] Step 3: Combining the intrinsic mode function with the preset thermodynamic entropy production model, the decay vector characterizing the state of the retired power battery is obtained through mapping transformation;

[0012] Step 4: Based on the attenuation vectors of multiple sets of retired power batteries, a clustering algorithm is used to analyze and generate grouping instructions containing vector parallelism information.

[0013] Step 5: According to the grouping instructions, control the dynamic topology reconfiguration mechanism to perform complementary assembly of retired power batteries to generate cascade utilization modules.

[0014] Preferably, the mixed pulse sequence includes:

[0015] High-frequency, high-power pulses with frequencies between preset low-frequency thresholds and preset high-frequency thresholds;

[0016] The acquired voltage response data includes:

[0017] The second derivative features of voltage response data are captured using a high-precision voltage acquisition card and used as a transient pulse feature identifier.

[0018] Preferably, the preset low-frequency threshold is 0.1Hz and the preset high-frequency threshold is 1kHz;

[0019] The high-precision voltage acquisition card has a sampling rate of over 10kHz.

[0020] Preferably, the process involves Hilbert-Huang transform, including:

[0021] Empirical mode decomposition is performed on the real and imaginary parts of the transient impedance spectrum trajectory;

[0022] Obtain the intrinsic mode function components that reflect the kinetics of electrochemical reactions.

[0023] Preferably, the attenuation vector includes:

[0024] Directional characteristics are used to characterize the nonlinear decay rate trend of retired power batteries in future charge-discharge cycles.

[0025] Modulus length features are used to characterize the current health status of retired power batteries.

[0026] Preferably, clustering analysis is performed, including:

[0027] Calculate the cosine of the angle between the attenuation vectors of any two retired power batteries;

[0028] If the cosine of the included angle is greater than the preset parallelism threshold, the corresponding retired power battery is determined to belong to the same degradation trend group, and a grouping instruction is generated; otherwise, the corresponding retired power battery is determined to belong to different degradation trend groups.

[0029] Preferably, the controlled dynamic topology reconfiguration mechanism performs complementary assembly of retired power batteries, including:

[0030] Within the same degradation trend group, identify the first type of cell with an internal resistance growth rate higher than the preset internal resistance benchmark and a capacity degradation rate lower than the preset capacity benchmark, and the second type of cell with an internal resistance growth rate lower than the preset internal resistance benchmark and a capacity degradation rate higher than the preset capacity benchmark.

[0031] The first type of battery cell and the second type of battery cell are connected in parallel for compensation.

[0032] A system for the secondary use of retired power batteries, comprising:

[0033] The detection module is used to generate mixed pulse sequences and acquire data;

[0034] The data processing module includes:

[0035] The feature extraction unit is used to calculate transient impedance characteristic data and extract intrinsic mode functions based on voltage response data and surface temperature change data;

[0036] The vector mapping unit is used to combine the intrinsic mode function with the preset thermodynamic entropy production model to calculate the decay vector;

[0037] The clustering analysis unit is used to generate grouping instructions based on the attenuation vector and the clustering algorithm.

[0038] A dynamic topology reconfiguration mechanism for performing complementary assembly based on grouping instructions.

[0039] Compared with the prior art, the present invention has the following beneficial effects:

[0040] 1. This invention introduces the concept of electrochemical entropy production and executes a dynamic fitting process for electrochemical states. It uses a mixed pulse sequence to excite reactions at different time scales inside the battery and combines Hilbert-Huang transform to extract intrinsic mode functions to construct a transient entropy production characteristic model. This model can map the transient response of retired batteries in a very short time into physical quantities characterizing irreversible thermodynamic losses. This method replaces the traditional full charge-discharge cycle aging test that lasts for several weeks with a 30-second pulse entropy production mapping. While ensuring the identification of micro-aging mechanisms such as SEI film thickening, it significantly solves the deadlock contradiction between detection efficiency and prediction accuracy in tiered utilization and greatly improves the efficiency of industrial screening.

[0041] 2. This invention constructs an attenuation vector containing directional and modulus features, maps the battery state to a high-dimensional feature space, and uses a clustering algorithm to analyze the parallelism between vectors to generate grouping instructions. This overcomes the limitations of traditional methods that rely solely on current static indicators or health status for screening. This solution enables the prediction of the future nonlinear attenuation rate trend of the battery and electrochemical gene matching, ensuring that cells belonging to the same attenuation trend group maintain a highly consistent aging path throughout their subsequent life cycle, effectively avoiding premature module performance failure caused by the divergence of individual cell attenuation.

[0042] 3. This invention further identifies complementary cells within the same attenuation trend group whose internal resistance growth rate and capacity attenuation rate exhibit asynchronous evolution patterns, and controls a dynamic topology reconfiguration mechanism to implement parallel compensation connections for them, thereby forming a passive physical equilibrium mechanism within the module. Utilizing the current-sharing characteristics of parallel circuits, cells with lower internal resistance share the current surge, while cells with higher capacity retention replenish the power deficit. As a result, the capacity variance of the reconfigured module shows a convergence trend with the increase of cycle number, effectively overcoming the barrel effect and significantly extending the cycle life of the cascade utilization module.

[0043] 4. This invention utilizes the second derivative characteristics of voltage response data captured by a high-precision acquisition card as a transient pulse feature identifier. Combined with depolarization processing and analytical signal method to calculate instantaneous complex impedance, it can detect potential safety hazards such as micro-short circuit trends inside the battery or micro-cracks in electrode materials with extremely high sensitivity. In addition, the Hilbert-Huang transform's ability to process nonlinear and non-stationary signals solves the problem that traditional Fourier transform cannot characterize the energy of coupled physical fields, ensuring that energy calculation strictly follows the physical definition. This effectively eliminates dud cells with potential safety hazards and improves the safety and reliability of cascade utilization products. Attached Figure Description

[0044] The present invention will be further explained below with reference to the accompanying drawings and embodiments:

[0045] Figure 1 This is a flowchart of the method of the present invention;

[0046] Figure 2 This is a structural diagram of the system of the present invention. Detailed Implementation

[0047] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0048] Example 1:

[0049] Please see Figure 1 Example 1:

[0050] A method for the secondary utilization of retired power batteries includes: connecting the retired power battery to a testing station and initializing the testing module; responding to the testing start command, executing an electrochemical state dynamic fitting process, including: step 1, controlling the testing module to apply a mixed pulse sequence to the retired power battery, and simultaneously collecting voltage response data and surface temperature change data;

[0051] Step 2: Based on the voltage response data and surface temperature change data, calculate the transient impedance characteristic data, and process the transient impedance characteristic data through Hilbert-Huang transform to extract the intrinsic mode functions;

[0052] Step 3: Combining the intrinsic mode function with the preset thermodynamic entropy production model, the decay vector characterizing the state of the retired power battery is obtained through mapping transformation;

[0053] Step 4: Based on the attenuation vectors of multiple sets of retired power batteries, a clustering algorithm is used to analyze and generate grouping instructions containing vector parallelism information.

[0054] Step 5: According to the grouping instructions, control the dynamic topology reconfiguration mechanism to perform complementary assembly of retired power batteries to generate cascade utilization modules.

[0055] This embodiment discloses a system and method for the cascade utilization of retired power batteries. The core logic is to break through the limitations of traditional screening based on static indicators. By introducing the concept of electrochemical entropy production, the transient response of retired batteries under pulse excitation is mapped into a decay vector in a high-dimensional space, thereby realizing the prediction and grouping of the future aging trend of the batteries. The retired power batteries to be tested are connected to the testing station, the system initializes the testing module, calibrates the voltage, current and temperature sensors, and ensures that the background noise is lower than the preset limit.

[0056] In this step, the system will perform a 1kHz AC impedance measurement to obtain the initial ohmic internal resistance of the battery cell as the system reference impedance. The unit is This is used for subsequent signal normalization processing; in response to the start command, the detection module controls the bidirectional high-power power supply to apply a mixed pulse sequence to the battery cell. This sequence is designed to excite electrochemical reactions at different time scales inside the battery in a very short time. During this period, the system synchronously acquires raw voltage response data. and surface temperature change data The system calculates transient impedance characteristic data based on the collected data. Specifically, it uses the analytical signal method to depolarize the original voltage data to obtain the overpotential. Calculate the instantaneous complex impedance, and then use the formula:

[0057]

[0058] in, The imaginary unit satisfies ; This represents the Hilbert transform operator, used to construct analytic signals and process transient impedance characteristic data through the Hilbert-Huang transform. It addresses nonlinear and non-stationary transient impedance data by decomposing the complex impedance spectrum trajectory into several eigenmode functions through empirical mode decomposition. Based on this, and combining the extracted eigenmode functions with a pre-defined thermodynamic entropy production model, the attenuation vector characterizing the cell state is calculated through a mapping transformation. To quantify the irreversible thermodynamic losses within the battery cell, this embodiment constructs the following transient entropy production characteristic model:

[0059]

[0060] in, The calculated transient entropy production characteristic value, physically representing the generalized thermodynamic loss of a decommissioned power battery under pulse excitation, is expressed in units of... ; The preset pulse sequence duration for the system, in units of ; The instantaneous overpotential after baseline removal is calculated using the following formula:

[0061]

[0062] in, This is the measured terminal voltage. The static open-circuit voltage before the pulse sequence is applied, in units of ;

[0063] The instantaneous excitation current is collected in real time, and the unit is 1. ;

[0064] The real-time surface temperature rise collected by the infrared sensor, i.e. The unit is ;

[0065] This is the ambient reference absolute temperature, in units of... ;

[0066] For the first The dimensionless entropy production sensitivity weights of each component are derived from training on a historical database.

[0067] The total number of intrinsic mode function (IMF) components obtained from empirical mode decomposition (EMD) is dimensionless. For the first Energy density of each IMF component, in units of The specific calculation formula is as follows:

[0068]

[0069] The physical meaning is the first Energy density of each IMF component, in units of ; The source is signal transformation calculation. To address the problem in traditional signal processing where a single variable cannot characterize the energy of coupled physical fields, and to ensure that energy calculations strictly adhere to the Joule thermodynamic definition, the specific formula is as follows:

[0070]

[0071] in, To obtain the first step of EMD decomposition of the real part trajectory of the transient impedance characteristic data calculated in step 2, i.e., the transient resistance part,... Each intrinsic mode function component, in units of The physical derivation logic for introducing this parameter here is as follows: the instantaneous heat generation power of the battery is determined by Joule's law. The standard algorithm for energy spectral analysis using the Hilbert-Huang transform processes the square of the signal amplitude as the energy density.

[0072] In order to convert the impedance component To map the contribution to heat production into energy spectrum calculations, an equivalent signal must be constructed. This makes its squared magnitude equal to the instantaneous power contribution of that component, i.e. Therefore, it is defined as:

[0073]

[0074] Among them, here Units are The unit after taking the arithmetic square root is , with current After multiplication, The dimensional derivation is as follows Its modulus squared The dimension of is This strictly corresponds to the instantaneous power definition in physics; although this dimension is rarely used alone in conventional physics, in the signal space of this technical solution, its physical meaning is rigorously defined as the equivalent amplitude of the transient thermal effect, that is, characterizing the relationship between current excitation and the first... The square root of the power signal generated after layer impedance fluctuation coupling, and its magnitude squared. That is, transient power The integral over time corresponds to the energy. This eliminates the ambiguity in parameter definition;

[0075] The source is obtained through training on a historical database, and the physical meaning corresponds to the first... The dimensionless entropy production sensitivity weight of each component is used to characterize the proportion of the energy in this frequency band to the total entropy production characteristics.

[0076] The specific operation involves constructing a training set containing multiple sets of impulse response and capacity decay data for the entire life cycle of batteries of the same model, and employing a multiple linear regression algorithm. To prevent overfitting caused by collinearity among features, this embodiment specifically uses ridge regression to solve for the weights and construct a loss function. The calculation formula is as follows:

[0077]

[0078] in, For the first The true capacity decay rate of a training sample measured in accelerated aging tests in the laboratory, dimensionless, is defined as... , The predicted value is based on energy characteristics. To address the dimension mismatch issue between the capacity decay rate (dimensionless) and the energy entropy characteristic (unit: J / K), this embodiment explicitly defines the regression prediction model as follows:

[0079]

[0080] in, The global mapping constant is calibrated as The calibration process involves performing a pulse heating experiment on a standard resistor in an adiabatic calorimeter to ensure that the total energy calculated by the model is equal to the entropy production increment measured by the sensor, thereby determining the unit alignment coefficient. The weighting coefficient is then used to determine this. It is a dimensionless value, thus ensuring The physical units of the two terms in the formula are unified as follows: ;

[0081] To clarify the boundary between physical models and prediction algorithms, and to define Physical meaning in attenuation vector calculation: Defined in this embodiment For generalized equivalent entropy production; although The value is obtained by fitting the capacity decay rate. It is obtained, but its physical essence is defined as the structural dissipation coefficient, that is, it characterizes a specific frequency band. energy This transforms into irreversible structural damage, leading to capacity decay and reduced efficiency; therefore, Substitute return After the formula, the second term Physically, it represents latent structural entropy production, which, together with the first term, namely explicit Joule thermal entropy production, constitutes a complete measure of electrochemical irreversibility.

[0082] It should be noted that, although It originates from the regression training of capacity decay rate, but in this physical model, it is substantially given the physical meaning of "the degree of structural disorder caused by unit energy", that is, it is assumed that the macroscopic capacity decay of the battery is a linear cumulative manifestation of the microscopic structural entropy increase, thus realizing the logical closed loop from the data space to the thermodynamic space.

[0083] This processing method reverses the data-driven prediction target, namely capacity decay, back to the physical space, i.e., entropy production, making... Still maintain Thermodynamic dimensions are thus used for subsequent calculations of the attenuation vector. Length of the module It provides a legitimate physical basis, rather than merely a mathematical proxy variable; The regularization coefficients are determined through 5-fold cross-validation. The specific steps are as follows: to Fifty candidate values ​​are set in the logarithmic space, the mean square error of the validation set is calculated, and the value corresponding to the minimum MSE is selected. The value, typically 0.01, is used to obtain the weight coefficients with generalization ability by minimizing this loss function. ;

[0084] Based on the calculation And the energy distribution of each frequency band, the system constructs attenuation vectors This vector not only contains current health status information, but also implies the guidance of the battery's internal microstructure on the future degradation path; the system collects the degradation vectors of multiple battery cells, uses a clustering algorithm to analyze the spatial relationship between the vectors, generates a grouping instruction containing vector parallelism information, and controls the dynamic topology reassembly mechanism to perform complementary assembly of the battery cells according to the instruction, generating a cascade utilization module.

[0085] In the scenario of power battery recycling, this embodiment replaces the traditional long-cycle aging test of several weeks with a pulse entropy generation mapping at the 30-second level, which solves the deadlock contradiction between detection efficiency and prediction accuracy in cascade utilization; the electrochemical gene matching achieved by the decay vector effectively identifies the micro-aging mechanisms inside the battery, such as the thickening of the SEI film, so that the recombined module maintains a high degree of consistency in subsequent cycles, and greatly reduces the barrel effect caused by individual differences.

[0086] Example 2:

[0087] The mixed pulse sequence includes: high-frequency high-power pulses with frequencies between preset low-frequency thresholds and preset high-frequency thresholds; voltage response data is acquired, including: capturing the second derivative features of the voltage response data using a high-precision voltage acquisition card as a transient pulse feature identifier; the preset low-frequency threshold is 0.1Hz, and the preset high-frequency threshold is 1kHz; the sampling rate of the high-precision voltage acquisition card is higher than 10kHz.

[0088] This embodiment specifies the hybrid pulse sequence and data acquisition to cover the key attenuation mechanism frequency band in the secondary use of power batteries; the hybrid pulse sequence is designed to include frequencies between a preset low-frequency threshold. With preset high frequency threshold High-frequency, high-power pulses between; wherein, a preset low-frequency threshold The frequency is set to 0.1Hz, which corresponds to the timescale of the solid-phase diffusion process inside the battery. A preset high-frequency threshold is also provided. The frequency is set to 1kHz, which corresponds to the time scale of the battery's ohmic impedance and part of the charge transfer process; the pulse rate is set to 1C to 3C to ensure that significant thermal effects and voltage overshoot can be generated, thereby improving the signal-to-noise ratio.

[0089] To capture minute electrochemical kinetic anomalies, this embodiment utilizes a high-precision voltage acquisition card with a sampling rate exceeding 10 kHz to capture the second derivative characteristics of the voltage response data, i.e., acquiring more than 10,000 points per second. To avoid high-frequency quantization noise introduced by direct differentiation operations masking the true signal, the system applies a Savitzky-Golay digital filter before calculating the derivative, and sets the filter window length. The timeframe is 11 points, with a window of approximately 1ms. A third-order polynomial is used for fitting. The specific formula for calculating the smoothed second-order derivative is as follows:

[0090]

[0091] in, The sampling time interval is numerically equal to the reciprocal of the sampling rate, i.e. For example, when the sampling rate is 10kHz, ;parameter The pre-stored convolution coefficients are determined using a standard mathematical library, such as... In the library Function to set window length polynomial order Derivative order A set of normalized convolution coefficient vectors is directly calculated; based on the Savitzky-Golay filter principle, for a length of A third-order polynomial least squares fit is performed on the discrete data points within a local window. The linear weight coefficients corresponding to the second derivative of the fitted polynomial at the center point are then calculated. Based on this, the transient pulse characteristic identifier is calculated. The definition is as follows:

[0092]

[0093] in, The source is calculated, and its physical meaning is a transient pulse characteristic identifier, with units of... ;

[0094] The source is voltage response data after SG filtering, in units of ;

[0095] The source is the system clock, and its physical meaning is the sampling time, with the unit being... ;

[0096] The source is a preset value, and its physical meaning is the relaxation window time after the pulse rise edge, set to 10% of the pulse width, for example, 100ms, to focus on the transient behavior in the early stage of electrochemical polarization establishment. The unit is... ;

[0097] In this embodiment, in the battery screening scenario, by limiting the pulse frequency range from 0.1Hz to 1kHz, the key frequency band of the electrochemical reaction is precisely covered; by using the second derivative as a feature identifier, it can detect micro-short circuit trends inside the battery or micro-cracks in the electrode material with extremely high sensitivity. These tiny defects are not visible on the conventional voltage curve, thereby effectively eliminating dud cells with safety hazards and improving the safety of the cascade utilization products.

[0098] Example 3:

[0099] The process involves Hilbert-Huang transform, including empirical mode decomposition of the real and imaginary parts of the transient impedance spectrum trajectory, and obtaining the eigenmode function components that reflect the electrochemical reaction kinetics.

[0100] This embodiment details the specific processing steps of the Hilbert-Huang transform, aiming to solve the problem that the Fourier transform cannot handle non-stationary signals; the system processes the instantaneous polarization voltage data after baseline removal. ,Right now To eliminate the masking effect of DC open-circuit voltage on transient impedance calculations and current data. Apply the Hilbert transform to construct the corresponding analytic signals. and To avoid numerical singularities at current zero-crossing points or minimum values, this embodiment employs a regularized complex division method to calculate the instantaneous complex impedance. The formula is as follows:

[0101]

[0102] in, The conjugate of the current analytic signal, The regularization factor takes a value of The calculation result is decomposed into the real part trajectory. And the trajectory of the imaginary part For the real part of the trajectory And the trajectory of the imaginary part Empirical mode decomposition is performed on each signal. This decomposition process involves repeatedly filtering out extreme points in the signal and fitting an envelope to extract fluctuation components at different time scales. The decomposition results are expressed as follows:

[0103]

[0104] in, The source is calculated, and its physical meaning is the imaginary part locus of the impedance spectrum, with units of... ;

[0105] The source is adaptively determined, and its physical meaning is the total number of components of the intrinsic mode function, that is, the total number of modes decomposition layers. It is consistent with the notation in the aforementioned embodiments and is dimensionless.

[0106] The source is EMD decomposition, and the physical meaning is the first Each intrinsic mode function component represents a polarization fluctuation in different frequency bands, with units of . Subscript here Specifically used to identify the modal sequence after the imaginary part trajectory decomposition, to distinguish it from the subscripts used in the subsequent real part trajectory decomposition. ;

[0107] The source is EMD decomposition, and its physical meaning is the residual trend term, which characterizes the reference impedance drift of the battery, with units of... ;

[0108] To clarify the key inputs required for the aforementioned entropy production model, this embodiment focuses on the real part trajectory. Performing the same decomposition operation, the decomposition result is expressed as follows:

[0109]

[0110] in, The first real part of the trajectory Each intrinsic mode function component, in units of ; Subscript here Specifically designed to identify the modal sequence after real part trajectory decomposition; high-frequency IMF components are selected to reflect changes in SEI film impedance, while low-frequency IMF components are selected to reflect changes in the diffusion coefficient of lithium ions in the active material, wherein the IMF components obtained from real part trajectory decomposition... The energy density is fed into the entropy production model in the aforementioned embodiments for calculation. Thus, these components directly map the electrochemical reaction kinetics of the battery;

[0111] In this embodiment, under the non-steady-state signal processing scenario, the transient change signal generated by the battery under pulse excitation was successfully processed by utilizing the characteristic that HHT does not require the assumption of linearity and stationarity of the signal. By extracting the transient impedance by analytical signal method and combining it with the IMF component, the ohmic internal resistance and polarization internal resistance of the battery were physically separated in the time domain, providing a pure and high-fidelity data source for subsequent accurate identification of the attenuation mechanism.

[0112] Example 4:

[0113] The attenuation vector includes: directional features, used to characterize the nonlinear attenuation rate trend of the retired power battery in future charge-discharge cycles; and modulus features, used to characterize the current health status of the retired power battery. Analysis is performed using a clustering algorithm, including: calculating the cosine of the angle between the attenuation vectors of any two retired power batteries; if the cosine of the angle is greater than a preset parallelism threshold, the corresponding retired power batteries are determined to belong to the same attenuation trend group, and a grouping instruction is generated; otherwise, the corresponding retired power batteries are determined to belong to different attenuation trend groups.

[0114] This embodiment details the definition of the decay vector and the clustering algorithm based on this vector, aiming to predict the future aging trajectory of the battery; defining the decay vector. For a feature containing modulus length and directional features The vector; where the magnitude feature It is negatively correlated with the transient entropy production characteristic value, and the specific calculation model is as follows:

[0115]

[0116] in, The source is experimental calibration, specifically the average entropy production characteristic value of new batteries of the same model under standard operating conditions. Its physical meaning is the entropy production baseline constant under healthy conditions, and the unit is... ;

[0117] The source is a system preset; its physical meaning is a numerical stability factor to prevent the denominator from being zero; the unit is... The modulus length feature is used to characterize the current health status of the battery cell; that is, the lower the entropy, the larger the modulus length, and the higher the health status. Directional features... The ratio of high-frequency IMF energy to low-frequency IMF energy is determined by the specific calculation model:

[0118]

[0119] Among them, the subscript of the molecule part The index of intrinsic mode components representing high-frequency bands, from order 1 to order 2. Order; subscript of the denominator The index of intrinsic mode components representing the low-frequency band, starting from the first... to the first Step; here and Both refer to the energy density defined above. ; The cutoff index for distinguishing high-frequency and low-frequency components is determined as follows: For each eigenmode function obtained from the decomposition... The Hilbert transform is applied to obtain the analytic signal, and the derivative is used to calculate the instantaneous frequency function. The calculation formula is as follows:

[0120]

[0121] in, , Operations are used to eliminate phase in The truncation transition at the point ensures the continuity and numerical stability of the differential operation; its average frequency over the pulse duration is calculated. ; Defined as satisfying average frequency Maximum index value The 10Hz threshold is derived from the statistical analysis of electrochemical impedance spectroscopy tests conducted on the same batch of retired batteries. This frequency point is the statistical boundary between the charge transfer impedance semicircle and the diffusion impedance tail.

[0122] This directional feature Physically, it reflects the relative distribution relationship between high-frequency interface impedance energy and low-frequency diffusion impedance energy in the transient response of the battery, characterizes the current electrochemical kinetics or impedance spectrum morphology of the battery, and ensures that batteries in the same group have the same polarization time constant distribution characteristics.

[0123] In order to perform efficient vector space operations, the system will attenuate the vector. Mapped to a two-dimensional Cartesian feature space, the specific construction is as follows: The system executes a graph-based clustering algorithm for analysis, with the following steps: Constructing a similarity adjacency matrix: For the data to be grouped... For a retired power battery, calculate any two cells. and attenuation vector and The cosine of the angle between them is calculated using the following formula:

[0124]

[0125] like Greater than the preset parallelism threshold Then in the adjacency matrix Middle Mark This indicates that both belong to the same dynamic characteristic group; otherwise, it is marked as such. Regarding the preset parallelism threshold Method for determining: In this embodiment These are not arbitrarily set empirical values, but are calculated based on the statistical distribution of historical data;

[0126] Specifically, select items from the historical database that are known to belong to the same aging batch, i.e., have the same aging mechanism label. For each battery cell sample, calculate the cosine of the angle between each pair of cells to form a similarity set. ;calculate average and standard deviation To ensure high purity of the grouping and eliminate outlier interference, we define... Under the typical experimental conditions of this embodiment, calculations were performed. Therefore, set This threshold setting ensures that similar cells within a 95% confidence interval can be correctly grouped, while strictly distinguishing dissimilar cells.

[0127] Extracting connected components: Traverse the adjacency matrix using either depth-first search or breadth-first search algorithms. The system identifies all connected components in the graph and generates grouping instructions: each independent connected component constitutes a decay trend group, ensuring that there is a direct or indirect high similarity path between any two cells within the group; the system assigns a unique group ID to each connected component and generates grouping instructions.

[0128] In this embodiment, in the battery pack scenario, the pairwise similarity judgment is extended to a global grouping strategy through the above algorithm, which breaks through the limitation of traditional sorting that only considers the current state and realizes the prediction of the possibility of future degradation trajectory. This grouping strategy based on the parallelism of degradation vector ensures that the cells after packing maintain a high degree of consistency in dynamic response characteristics throughout the entire life cycle, significantly reduces the balancing pressure of the battery management system, and extends the service life of the module.

[0129] Example 5:

[0130] The control dynamic topology reconfiguration mechanism performs complementary assembly of retired power batteries, including: identifying a first type of cell with an internal resistance growth rate higher than a preset internal resistance benchmark and a capacity decay rate lower than a preset capacity benchmark, and a second type of cell with an internal resistance growth rate lower than a preset internal resistance benchmark and a capacity decay rate higher than a preset capacity benchmark within the same degradation trend group; and connecting the first type of cell and the second type of cell in parallel for compensation.

[0131] This embodiment describes a strategy of complementary assembly within the same attenuation trend group, which is a key step in achieving module self-balancing. Within the same group determined by the aforementioned steps, the system further identifies two types of complementary cells. Although cells within the same group have similar dynamic vector directions... That is, they have the same decay mechanism path, but due to the discreteness of production batches or microstructures, their current health status... The specific aging rate exhibits non-linear stage characteristics; the first type of cell identified is characterized by the rate of internal resistance growth. Higher than the preset internal resistance reference However, the capacity decay rate Below the preset capacity benchmark That is, the impedance aging bias type;

[0132] The second type of battery cell identified was characterized by its internal resistance growth rate. Below However, the capacity decay rate Higher than This refers to capacity aging bias; the definitions and physical meanings of the aforementioned key parameters are as follows:

[0133]

[0134] in, The physical meaning of Ω is the ohmic internal resistance of a retired power battery, and the unit is Ω. ; The physical meaning is the current capacity of retired power batteries, in units of... ; The physical meaning of "charge-discharge cycle number" is expressed in units of 1000 kcal / kg. ;

[0135] The source is derived from decay vector mapping. The specific process involves constructing a support vector regression model, using the radial basis function as the kernel function, and defining the regression decision function as follows:

[0136]

[0137] in, This represents the total number of support vectors. and These are the Lagrange multipliers obtained during model training using the Lagrange multiplier method. This represents the bias term in the regression model; This is the kernel parameter, preset to 0.5. To train support vectors, input feature vectors The attenuation vector after Z-Score normalization The polar coordinate components, i.e. , These are the mean and standard deviation of the training set, respectively, to eliminate the influence of differences in the dimensions of the modulus and angle on the RBF kernel function;

[0138] This model, once trained, can capture: in a fixed decay direction Below, with health The change in internal resistance rate, i.e., the battery is in different aging stages, is due to the change in the rate of internal resistance growth. and capacity decay rate It is not a synchronous linear change, but rather an asynchronous evolution pattern in which one side increases while the other decreases, such as the SEI film thickening-dominant period. High, while the period dominated by loss of active materials High, which makes it based on A vector can uniquely map the current rate combination; during model training, a grid search method is used. Within the range, the optimal penalty factor is found by minimizing the root mean square error of the prediction. and insensitive loss coefficient In this embodiment, the optimal value determined by grid search is and This ensures the generalization accuracy of the model, thereby establishing a nonlinear mapping from the vector space to the physical parameter space.

[0139] The source is dynamically calculated, specifically taking the median of the predicted internal resistance growth rate of all cells within the currently identified group of cells exhibiting the same degradation trend. Using the median as a benchmark, the cells in the current group can be adaptively divided into two subgroups: one with relatively high impedance growth and the other with relatively low impedance growth.

[0140] The source is obtained based on decay vector mapping, and it also utilizes the above SVR model architecture, using independent training weights, with decay vector... As input, the predicted capacity decay rate is output. ;

[0141] The source is dynamic calculation, and the specific value is the median of the predicted capacity decay rate of all cells in the same decay trend group currently identified;

[0142] The system executes complementary matching logic: classification and filtering: traverses the cells within the group and marks those that meet the criteria. These are Class I cells, meeting the requirements. The cells that do not meet the above two conditions are classified as the second type of cells. For cells that do not meet the above two conditions, such as dual-differential or dual-superior cells, they are marked as non-complementary cells and temporarily removed from the current assembly sequence to enter the subsequent processing flow. Parallel compensation: The dynamic topology reorganization mechanism is controlled to connect the selected first type of cells and the second type of cells in parallel for compensation.

[0143] This complementary assembly utilizes the current shunting characteristics of parallel circuits. In the early stage of discharge, the second type of cell with lower internal resistance carries more current, protecting the first type of cell from overheating. In the late stage of discharge, the first type of cell with higher capacity retention releases the remaining charge to make up for the deficit of the second type of cell.

[0144] In this embodiment, under the scenario of module manufacturing, by clarifying the SVR kernel function mapping relationship and strictly complementary screening logic, the rate statistical deviation generated by different aging stages within the same kinetic group is used for pairing, forming a passive physical equilibrium mechanism within the module. Compared with random assembly or single-dimensional sorting assembly, the cycle life of the cascaded utilization module using this complementary assembly strategy is significantly improved, and the overall capacity variance of the module shows a convergent trend with the increase of the number of cycles, effectively solving the problem of short service life of old batteries due to characteristic divergence.

[0145] Example 6:

[0146] Please see Figure 2 A system for the cascade utilization of retired power batteries includes: a detection module for generating a mixed pulse sequence and acquiring data; a data processing module including: a feature extraction unit for calculating transient impedance characteristic data and extracting intrinsic mode functions based on voltage response data and surface temperature change data; a vector mapping unit for combining the intrinsic mode functions with a preset thermodynamic entropy production model to obtain the decay vector; a clustering analysis unit for generating grouping instructions based on the decay vector through clustering algorithm analysis; and a dynamic topology reconfiguration mechanism for performing complementary assembly according to the grouping instructions.

[0147] This embodiment provides a hardware system architecture for executing the above method, realizing a closed-loop integrated hardware and software system from signal excitation and data processing algorithms to mechanical execution. The system includes a detection module, a data processing module, and a dynamic topology reassembly mechanism. The detection module includes a programmable bidirectional pulse power supply with a response time of less than 1ms, a high-frequency voltage acquisition card, and an infrared thermal imager, used to generate a mixed pulse sequence and acquire voltage data. Current and temperature Data; the feature extraction unit in the data processing module has a built-in DSP or FPGA chip, runs the HHT algorithm, calculates transient impedance in real time and extracts IMF; the vector mapping unit stores the thermodynamic entropy production model and maps the IMF to a decay vector. The clustering analysis unit runs the cosine similarity clustering algorithm and outputs grouping instructions.

[0148] The dynamic topology reconfiguration mechanism includes a vision-guided six-axis robotic arm, an automatic barcode scanner, and laser welding equipment. This mechanism directly receives grouping instructions and picks up cells with specific IDs to specific module trays for parallel welding. In this embodiment, under industrial production scenarios, the hardware implementation of the feature extraction unit, i.e., DSP or FPGA, ensures that complex HHT transformation and entropy production calculations can still be completed even at a high-speed production cycle of inspecting multiple cells per minute. Combined with the automated dynamic topology reconfiguration mechanism, it realizes a fully automated operation from battery detection to reconfiguration for cascade utilization, possessing extremely high engineering application value and production efficiency.

[0149] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for the cascade utilization of retired power batteries, characterized in that, include: Connect the retired power battery to the testing station and initialize the testing module; In response to the detection initiation command, an electrochemical state dynamic fitting process is executed, including: Step 1: The control detection module applies a mixed pulse sequence to the retired power battery and simultaneously collects voltage response data and surface temperature change data; Step 2: Based on the voltage response data and surface temperature change data, calculate the transient impedance characteristic data, and process the transient impedance characteristic data through Hilbert-Huang transform to extract the intrinsic mode functions; Step 3: Combining the intrinsic mode function with the preset thermodynamic entropy production model, the decay vector characterizing the state of the retired power battery is obtained through mapping transformation; Step 4: Based on the attenuation vectors of multiple sets of retired power batteries, a clustering algorithm is used to analyze and generate grouping instructions containing vector parallelism information. Step 5: According to the grouping instructions, control the dynamic topology reconfiguration mechanism to perform complementary assembly of retired power batteries to generate cascade utilization modules. Attenuation vector, including: Directional characteristics are used to characterize the nonlinear decay rate trend of retired power batteries in future charge-discharge cycles, and are determined by the ratio of high-frequency intrinsic mode function energy to low-frequency intrinsic mode function energy. Modulus length feature, used to characterize the current health status of retired power batteries, is negatively correlated with transient entropy production feature value; Clustering algorithm analysis, including: Calculate the cosine of the angle between the attenuation vectors of any two retired power batteries; If the cosine of the included angle is greater than the preset parallelism threshold, the corresponding retired power battery is determined to belong to the same degradation trend group, and a grouping instruction is generated; otherwise, the corresponding retired power battery is determined to belong to different degradation trend groups. The control mechanism for dynamic topology reconfiguration performs complementary assembly of retired power batteries, including: Within the same degradation trend group, identify the first type of cell with an internal resistance growth rate higher than the preset internal resistance benchmark and a capacity degradation rate lower than the preset capacity benchmark, and the second type of cell with an internal resistance growth rate lower than the preset internal resistance benchmark and a capacity degradation rate higher than the preset capacity benchmark. The first type of battery cell and the second type of battery cell are connected in parallel with compensation. Both the internal resistance growth rate and the capacity decay rate are obtained based on decay vector mapping.

2. The method for the cascade utilization of retired power batteries according to claim 1, characterized in that, Mixed pulse sequences, including: A pulse with a frequency between a preset low-frequency threshold and a preset high-frequency threshold; The acquired voltage response data includes: The second derivative features of voltage response data are captured using a high-precision voltage acquisition card and used as a transient pulse feature identifier. The preset low-frequency threshold is 0.1Hz, and the preset high-frequency threshold is 1kHz; The high-precision voltage acquisition card has a sampling rate of over 10kHz.

3. The method for the cascade utilization of retired power batteries according to claim 1, characterized in that, Processed using the Hilbert-Huang transform, including: Empirical mode decomposition is performed on the real and imaginary parts of the transient impedance spectrum trajectory; Obtain the intrinsic mode function components that reflect the kinetics of electrochemical reactions.

4. A system for the cascade utilization of retired power batteries, characterized in that, The system is configured to perform a method for the cascade utilization of retired power batteries as described in any one of claims 1-3, the system comprising: The detection module is used to generate mixed pulse sequences and acquire data; The data processing module includes: The feature extraction unit is used to calculate transient impedance characteristic data and extract intrinsic mode functions based on voltage response data and surface temperature change data; The vector mapping unit is used to combine the intrinsic mode function with the preset thermodynamic entropy production model to calculate the decay vector; The clustering analysis unit is used to generate grouping instructions based on the attenuation vector and the clustering algorithm. A dynamic topology reconfiguration mechanism for performing complementary assembly based on grouping instructions.

Images