A capacity analysis method and system of an energy storage power supply

Abstract

The application discloses a capacity analysis method and system of an energy storage power supply, and the method comprises the following steps: obtaining an end voltage time sequence signal and constructing an original response manifold trajectory; extracting a geodesic curvature sequence and a geodesic torsion sequence of a smooth trajectory through Gaussian regularization convolution; mapping to a curvature-torsion phase space to obtain a point cloud distribution diagram; performing a coherence analysis to extract a set of persistent feature points; calculating the persistent entropy value and the life value of each feature point and constructing a two-dimensional persistent contribution vector, and obtaining a capacity degradation topological tensor through weighted tensor aggregation; calculating the current unit power topological distortion rate and comparing it with the initial value to obtain a distortion rate offset, and correcting the rated capacity to obtain the residual capacity. The application converts the capacity estimation into a manifold geometric topological analysis problem, does not depend on historical data and experience mapping, and has physical interpretability and working condition adaptability.

Description

Technical Field

[0001] This invention belongs to the field of energy storage power capacity analysis technology, and particularly relates to a method and system for energy storage power capacity analysis. Background Technology

[0002] During long-term cyclic use, energy storage power sources (including lithium-ion batteries, lead-acid batteries, sodium-ion batteries, etc.) undergo structural degradation phenomena such as lattice volume changes, microcrack proliferation, active material pulverization, and conductive network breakage in electrode materials, leading to a gradual decrease in usable capacity. Accurately estimating the remaining capacity of energy storage power sources under current conditions is of great significance for power system dispatch management, electric vehicle range forecasting, and operation and maintenance decisions for energy storage power stations.

[0003] Currently, energy storage power capacity estimation methods mainly include the following categories: The first type is based on ampere-hour integration, which estimates capacity by accumulating charge and discharge amounts and combining this with coulombic efficiency correction. This type of method is simple and easy to implement, but it suffers from cumulative errors and cannot reflect the impact of battery aging on capacity. The estimation accuracy decreases significantly after long-term operation.

[0004] The second category is based on equivalent circuit model parameter identification. This method establishes an equivalent circuit model including components such as resistors and capacitors, and uses algorithms such as recursive least squares or Kalman filtering to identify model parameters online, thereby estimating the capacity. This type of method relies on the accuracy of the model, but the electrochemical characteristics of batteries change complexly during aging, and equivalent circuits with fixed topologies are difficult to adapt to parameter changes throughout the entire battery life cycle.

[0005] The third category is data-driven methods, including support vector regression, Gaussian process regression, and deep neural networks, which train capacity estimation models using a large amount of historical data. These methods require a large amount of labeled data, the model's generalization ability is limited by the coverage of the training data, and the black-box nature of the methods leads to a lack of interpretability in the estimation results.

[0006] The fourth category is based on voltage curve feature extraction methods, which indirectly infer capacity by analyzing changes in the shape of the discharge voltage curve. Existing technologies employ precise subsequence matching of the discharge current sequence, estimating capacity by comparing the similarity between the current current sequence and historical current sequences. This type of method relies on the accumulation of a large amount of historical discharge cycle data and utilizes only information from a single current dimension, failing to fully leverage the rich electrochemical information contained in the voltage response signal, and thus has limited adaptability under varying operating conditions.

[0007] In summary, existing technologies generally suffer from the following shortcomings: reliance on large amounts of historical data or prior calibration, lack of clear physical interpretability in estimation results, failure to effectively utilize the geometric information of voltage response signals, and insufficient adaptability throughout the entire life cycle of energy storage power sources. Summary of the Invention

[0008] This invention provides a capacity analysis method and system for energy storage power sources, which solves the technical problems in the prior art where energy storage power source capacity estimation relies on historical data, has poor interpretability, and fails to fully utilize the geometric information of voltage response signals.

[0009] In a first aspect, the present invention provides a capacity analysis method for energy storage power sources, comprising: The terminal voltage timing signal obtained by continuous sampling of the energy storage power supply during the current discharge cycle is acquired. The original response manifold trajectory of the terminal voltage timing signal is constructed in a two-dimensional coordinate system with sampling time as the horizontal axis and terminal voltage value as the vertical axis. The original response manifold trajectory is subjected to Gaussian regularized convolution to obtain a smoothed regularized response manifold trajectory. The geometric invariant sequence of the regularized response manifold trajectory is extracted according to a preset resampling rule. The geometric invariant sequence includes at least the geodesic curvature sequence and geodesic torsion sequence of the regularized response manifold trajectory at each sampling point. Using the geodesic curvature in the geodesic curvature sequence and the geodesic torsion in the geodesic torsion sequence as coordinate axes, the regular response manifold trajectory is mapped to the curvature-torsion phase space to obtain the phase space point cloud distribution map; Homology analysis is performed on the phase space point cloud distribution map to calculate the change trajectory of the phase space point cloud distribution map under different scale parameters, and the persistent feature point set of the phase space point cloud distribution map is extracted based on the change trajectory. The persistent feature point set contains multiple persistent feature points, and each persistent feature point corresponds to the generation scale and the disappearance scale of a topological feature. The persistence entropy and lifetime values ​​of each persistent feature point are calculated based on the generation and extinction scales of each persistent feature point. A two-dimensional persistent contribution vector of each persistent feature point is constructed using the persistence entropy and lifetime values ​​as dimensions. Weighted tensor aggregation is performed on the two-dimensional persistent contribution vectors of all persistent feature points to obtain the capacity degradation topology tensor of the current discharge cycle. Using the distortion rate offset of the current discharge cycle as the capacity decay driving factor of the capacity degradation topology tensor in the current discharge cycle, the remaining capacity of the energy storage power source at the current moment is calculated based on the product relationship between the capacity decay driving factor and the rated capacity of the energy storage power source.

[0010] Secondly, the present invention provides a capacity analysis system for energy storage power sources, comprising: The acquisition module is configured to acquire the terminal voltage timing signal continuously sampled by the energy storage power supply during the current discharge cycle, and construct the original response manifold trajectory of the terminal voltage timing signal in a two-dimensional coordinate system with the sampling time as the horizontal axis and the terminal voltage value as the vertical axis. The extraction module is configured to perform Gaussian regularized convolution processing on the original response manifold trajectory to obtain a smoothed regularized response manifold trajectory, and extract the geometric invariant sequence of the regularized response manifold trajectory according to a preset resampling rule. The geometric invariant sequence includes at least the geodesic curvature sequence and geodesic torsion sequence of the regularized response manifold trajectory at each sampling point. The mapping module is configured to map the regular response manifold trajectory to the curvature-torsion phase space using the geodesic curvature in the geodesic curvature sequence and the geodesic torsion in the geodesic torsion sequence as coordinate axes, thereby obtaining a phase space point cloud distribution map. The calculation module is configured to perform homology analysis on the phase space point cloud distribution map, calculate the change trajectory of the phase space point cloud distribution map under different scale parameters, and extract the persistent feature point set of the phase space point cloud distribution map based on the change trajectory. The persistent feature point set contains multiple persistent feature points, and each persistent feature point corresponds to the generation scale and the disappearance scale of a topological feature. The aggregation module is configured to calculate the persistence entropy and lifetime values ​​of each persistent feature point based on the generation and extinction scales of each persistent feature point, construct a two-dimensional persistent contribution vector of each persistent feature point using the persistence entropy and lifetime values ​​as dimensions, and perform weighted tensor aggregation on the two-dimensional persistent contribution vectors of all persistent feature points to obtain the capacity degradation topology tensor of the current discharge cycle. The output module is configured to use the distortion rate offset of the current discharge cycle as the capacity decay driving factor of the capacity degradation topology tensor in the current discharge cycle, and calculate the remaining capacity of the energy storage power supply at the current moment based on the product relationship between the capacity decay driving factor and the rated capacity of the energy storage power supply.

[0011] Thirdly, an electronic device is provided, comprising: at least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the steps of the capacity analysis method for an energy storage power source according to any embodiment of the present invention.

[0012] Fourthly, the present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein when the program instructions are executed by a processor, the processor performs the steps of the capacity analysis method for an energy storage power source according to any embodiment of the present invention.

[0013] The capacity analysis method and system for energy storage power sources disclosed in this application converts the terminal voltage time-series signal of the energy storage power source into a two-dimensional manifold trajectory. It constructs a curvature-torsion phase space point cloud distribution map by extracting geodesic curvature and geodesic torsion. Then, it uses homology analysis to detect the generation and disappearance of closed loops in the point cloud connection map, extracts persistent feature point sets, calculates the persistent entropy and lifetime values ​​of each closed loop, and constructs a two-dimensional persistent contribution vector. After weighted tensor aggregation, it obtains a capacity degradation topology tensor. Finally, it calculates the topology distortion rate per unit charge and compares it with the initial factory value to obtain the distortion rate offset, thereby correcting the rated capacity to obtain the remaining capacity. Compared with existing technologies, this invention has the following advantages: geodesic curvature and geodesic torsion, as intrinsic geometric invariants of the manifold trajectory, fundamentally eliminate the influence of discharge rate and sampling frequency. The influence of the rate on the analysis results ensures good consistency of capacity analysis under different operating conditions. The coherence analysis and closed-loop detection driven by the distribution density of point cloud data completely avoid subjective scale parameter settings. The lifetime value and persistence entropy value of the closed loop quantify the kinetic consistency and polarization distribution uniformity of the electrochemical process, respectively, and have complete physical interpretability. Weighted tensor aggregation makes the capacity degradation topology tensor more sensitive to the capacity decay mechanism related to bulk diffusion, while effectively suppressing surface polarization noise interference. The entire calculation process only requires the initial factory reference value, without accumulating historical discharge data, fitting empirical mapping curves, or relying on black box models, realizing online, in-situ, and calibration-free capacity estimation, which significantly reduces the data threshold and computational complexity of engineering deployment. Attached Figure Description

[0014] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0015] Figure 1 A flowchart illustrating a capacity analysis method for an energy storage power source according to an embodiment of the present invention; Figure 2 This is a structural block diagram of a capacity analysis system for an energy storage power source provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0016] 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.

[0017] Please see Figure 1 The flowchart illustrates a capacity analysis method for an energy storage power source according to this application.

[0018] like Figure 1 As shown, the capacity analysis method for energy storage power sources specifically includes the following steps: Step S101: Obtain the terminal voltage timing signal continuously sampled by the energy storage power supply during the current discharge cycle, and construct the original response manifold trajectory of the terminal voltage timing signal in a two-dimensional coordinate system with the sampling time as the horizontal axis and the terminal voltage value as the vertical axis.

[0019] In this step, the terminal voltage values ​​at each sampling time in the terminal voltage timing signal are arranged sequentially according to the sampling time. Each sampling time is used as the horizontal axis value, and the terminal voltage value corresponding to each sampling time is used as the vertical axis value to generate a discrete voltage sampling point column in the two-dimensional coordinate system. Interpolation fitting is performed on the discrete voltage sampling point sequence to generate a continuously differentiable two-dimensional curve, which is then used as the original response manifold trajectory.

[0020] In one specific embodiment, the terminal voltage timing signal of the energy storage power source is continuously sampled during the current discharge cycle. The energy storage power source includes rechargeable batteries such as lithium-ion batteries, sodium-ion batteries, lead-acid batteries, or nickel-metal hydride batteries. The terminal voltage timing signal is acquired by a voltage sensor at a fixed sampling frequency (e.g., 10Hz to 1000Hz). The start time of the acquisition is the discharge start time of the current discharge cycle, and the end time of the acquisition is the current time.

[0021] Using sampling time as the horizontal axis and terminal voltage value as the vertical axis, the original response manifold trajectory of the terminal voltage time series signal is constructed in a two-dimensional coordinate system. Specifically, the terminal voltage value at each sampling time in the terminal voltage time series signal is arranged sequentially according to the order of sampling time, with each sampling time as the horizontal axis value and the corresponding terminal voltage value as the vertical axis value, to generate a discrete voltage sampling point series in the two-dimensional coordinate system.

[0022] Interpolation fitting is performed on the discrete voltage sampling point sequence to generate a continuously differentiable two-dimensional curve. This two-dimensional curve is used as the original response manifold trajectory. The interpolation fitting method can adopt the piecewise cubic spline interpolation algorithm, where the interpolation nodes are each sampling time. The function value at each interpolation node is equal to the terminal voltage value at the corresponding sampling time, and the first and second derivatives at each interpolation node are continuous. Piecewise cubic spline interpolation can ensure that the generated trajectory curve has second-order continuous differentiability, providing a mathematical basis for subsequent geometric invariant extraction.

[0023] Step S102: Perform Gaussian regularized convolution on the original response manifold trajectory to obtain a smoothed regularized response manifold trajectory, and extract the geometric invariant sequence of the regularized response manifold trajectory according to a preset resampling rule. The geometric invariant sequence includes at least the geodesic curvature sequence and geodesic torsion sequence of the regularized response manifold trajectory at each sampling point.

[0024] In this step, a first-order differential operation is performed on the terminal voltage timing signal to obtain a voltage differential sequence. All local extreme points are detected on the voltage differential sequence, and the voltage values ​​corresponding to each local extreme point are arranged in descending order. The voltage values ​​corresponding to the first few local extreme points are selected as feature voltage values. Centered on each characteristic voltage value, a voltage window interval is determined with a preset voltage window radius. The union of all voltage window intervals is taken as the set of characteristic voltage intervals. The voltage window radius is a preset percentage of the total voltage change amplitude of the terminal voltage timing signal. On the regular response manifold trajectory, all trajectory segments whose ordinate values ​​fall within the set of characteristic voltage intervals are determined. A first resampling density is set in each of the trajectory segments according to the equal arc length method. A second resampling density is set on the remaining trajectory segments on the regular response manifold trajectory other than the set of characteristic voltage intervals. The first resampling density is greater than the second resampling density. Multiple resampling positions are determined on the regular response manifold trajectory according to the first resampling density and the second resampling density, and the first-order directional derivative vector and the second-order directional derivative vector of the regular response manifold trajectory are calculated at each resampling position; At each resampling location, the curvature component of the resampling location is calculated based on the first-order directional derivative vector and the second-order directional derivative vector, and the curvature components are arranged in the order of the resampling locations on the regular response manifold trajectory to form the geodesic curvature sequence. At each resampling location, the torsion component of the resampling location is calculated based on the first-order directional derivative vector and the second-order directional derivative vector, and the torsion components are arranged in the order of the resampling locations on the regular response manifold trajectory to form the geodesic torsion sequence.

[0025] In one specific embodiment, the original response manifold trajectory is subjected to Gaussian regularized convolution processing to obtain a smoothed regularized response manifold trajectory. The Gaussian regularized convolution processing is used to filter out high-frequency noise components in the terminal voltage time series signal while retaining the main geometric features of the trajectory curve. A Gaussian kernel function with a preset standard deviation parameter is selected, and the Gaussian kernel function is slid along the original response manifold trajectory in the time domain. Convolution operations are performed on the horizontal coordinate sequence and the vertical coordinate sequence of the original response manifold trajectory respectively to obtain the horizontal coordinate smooth sequence and the vertical coordinate smooth sequence. The regularized response manifold trajectory is composed of the horizontal coordinate smooth sequence and the vertical coordinate smooth sequence. The standard deviation parameter of the Gaussian kernel function is determined according to the sampling frequency and noise level of the terminal voltage time series signal, and the value range is generally 2 to 10 times the sampling interval.

[0026] The geometric invariant sequence of the regular response manifold trajectory is extracted according to the preset resampling rules. The geometric invariant sequence includes at least the geodesic curvature sequence and geodesic torsion sequence of the regular response manifold trajectory at each sampling point.

[0027] In a preferred embodiment, the resampling rule adopts a differentiated resampling method based on characteristic voltage ranges, and the specific steps are as follows: A first-order differential operation is performed on the terminal voltage time sequence signal to obtain a voltage differential sequence. Each differential value in the voltage differential sequence represents the change in terminal voltage between two adjacent sampling times. All local extrema are detected on the voltage differential sequence (i.e., points where the differential value is greater than the two adjacent differential values ​​are local maxima, and points where the differential value is less than the two adjacent differential values ​​are local minima). The voltage values ​​corresponding to each local extrema are arranged in descending order. The voltage values ​​corresponding to the top local extrema are selected as characteristic voltage values. The number of selected local extrema can be 3 to 10. These extrema correspond to the characteristic voltage positions with the largest voltage change rate during the discharge process, i.e., the regions with the most intense electrochemical phase transitions.

[0028] Centered on each characteristic voltage value, a voltage window interval is determined with a preset voltage window radius. The union of all voltage window intervals is taken as the characteristic voltage interval set. The voltage window radius is a preset percentage of the total voltage change amplitude of the terminal voltage timing signal. The preset percentage ranges from 5% to 15%. The total voltage change amplitude is the difference between the maximum and minimum voltage values ​​in the terminal voltage timing signal.

[0029] On the regular response manifold trajectory, identify all trajectory segments whose ordinate values ​​fall within the characteristic voltage interval set. Set the first resampling density in each trajectory segment according to the equal arc length method. Set the second resampling density on the remaining trajectory segments on the regular response manifold trajectory other than the characteristic voltage interval set. The first resampling density is greater than the second resampling density. For example, the first resampling density is 10 resampling points per unit arc length, and the second resampling density is 2 resampling points per unit arc length.

[0030] Multiple resampling positions are determined on the regular response manifold trajectory according to the first and second resampling densities. At each resampling position, the first-order directional derivative vector and the second-order directional derivative vector of the regular response manifold trajectory are calculated. The first-order directional derivative vector represents the tangent direction of the manifold trajectory at that point, and the second-order directional derivative vector represents the curvature direction of the manifold trajectory at that point.

[0031] At each resampling location, the curvature and torsion components are calculated based on the first and second directional derivative vectors. The formula for calculating the curvature component is as follows: , In the formula, For curvature components, It is the first-order directional derivative vector. It is the second-order directional derivative vector. For vector cross product, Let be the magnitude of the vector; The formula for calculating the torsion component is: , In the formula, It is the third-order directional derivative vector. It is the dot product of vectors; The curvature components of all resampled locations are arranged in order of sampling time to form a geodesic curvature sequence, and the torsion components of all resampled locations are arranged in order of sampling time to form a geodesic torsion sequence.

[0032] In this embodiment, the terminal voltage time-series signal of the energy storage power supply is constructed as a continuous differentiable manifold trajectory on a two-dimensional plane. From the perspective of differential geometry, the geodesic curvature and geodesic torsion of the manifold trajectory are extracted as geometric invariants. The geodesic curvature quantifies the bending degree of the voltage response manifold at each sampling point, corresponding to the rate of change of voltage decay rate during the discharge process of the energy storage power supply. The geodesic torsion quantifies the twisting rate of the voltage response manifold at each sampling point, corresponding to the rate of change of voltage decay non-uniformity during the discharge process of the energy storage power supply. The point cloud distribution map in the curvature-torsion phase space completely records the evolution trajectory of the voltage response manifold geometry during the entire discharge process. Compared with the existing method of directly analyzing the original voltage or current numerical sequence, this invention can eliminate the influence of discharge rate and sampling frequency on the analysis results by extracting the intrinsic geometric invariants of the manifold trajectory, so that the capacity analysis results have better consistency under different discharge conditions.

[0033] Step S103: Using the geodesic curvature in the geodesic curvature sequence and the geodesic torsion in the geodesic torsion sequence as coordinate axes, the regular response manifold trajectory is mapped to the curvature-torsion phase space to obtain a phase space point cloud distribution map.

[0034] In this step, each resampling position on the regular response manifold trajectory is mapped to a point cloud in the curvature-torsion phase space, with each geodesic curvature value in the geodesic curvature sequence as the abscissa and each geodesic torsion value in the geodesic torsion sequence as the ordinate. The point clouds corresponding to all resampling positions together constitute the phase space point cloud distribution map.

[0035] The phase space point cloud distribution map reflects the distribution characteristics of the voltage response manifold in the geometric space. For healthy batteries, the discharge voltage curve decreases smoothly and monotonically, and the curvature and torsion values ​​at each sampling position are relatively stable, with the point cloud distribution concentrated in a small area. For aged batteries, the discharge voltage curve shows more local fluctuations and oscillations, and the curvature and torsion values ​​at each sampling position change drastically, with the point cloud distribution diverging to a wider area.

[0036] Step S104: Perform homology analysis on the phase space point cloud distribution map, calculate the change trajectory of the phase space point cloud distribution map under different scale parameters, and extract the persistent feature point set of the phase space point cloud distribution map according to the change trajectory. The persistent feature point set contains multiple persistent feature points, and each persistent feature point corresponds to the generation scale and the disappearance scale of a topological feature.

[0037] In this step, the pairwise Euclidean distances between all point clouds in the phase space point cloud distribution map are calculated. All pairwise Euclidean distances are sorted in ascending order of value to obtain an ordered distance sequence. Pairwise Euclidean distances greater than a preset distance threshold are removed from the ordered distance sequence. The remaining pairwise Euclidean distances are then arranged in ascending order to obtain a candidate scale parameter sequence. The preset distance threshold is the sum of the average value and standard deviation of all pairwise Euclidean distances. This preset distance threshold can be automatically adjusted according to the overall dispersion of the point cloud distribution to effectively filter out non-physical noise topological features. Each candidate scale parameter value in the candidate scale parameter sequence is taken out as the current distance threshold. Under each current distance threshold, any two point clouds in the phase space point cloud distribution map whose Euclidean distance is less than the current distance threshold are connected by a straight line segment to obtain the connection graph under the current distance threshold. As the distance threshold increases from small to large, the number of edges in the connection graph increases continuously, and the connection relationship between point clouds gradually expands from local to global. In the connection graph at the current distance threshold, all closed loops formed by the line segments are detected, and the number of closed loops is counted to obtain the trajectory curve of the number of closed loops as the distance threshold changes. The perimeter of the closed loop is equal to the sum of the lengths of the line segments that make up the closed loop. The detection of closed loops can be implemented using a depth-first search algorithm or a disjoint-set data structure algorithm. As the distance threshold increases, the trajectory of the number of closed loops shows a trend of first rising and then falling: when the distance threshold is small, only a few point clouds with close distances are connected, and the number of closed loops is small; when the distance threshold is moderate, enough point clouds are connected to form closed loops, and the number of closed loops reaches its peak; when the distance threshold is large, new connecting lines continuously fill the existing closed loops, and the number of closed loops gradually decreases to zero. All local extreme points are detected on the closed loop number trajectory curve. The distance threshold corresponding to the starting point of the rising edge of the closed loop number trajectory curve where the number of closed loops first changes from zero to a positive integer is used as the generation scale of the corresponding closed loop. The distance threshold corresponding to the ending point of the falling edge of the closed loop number trajectory curve where the number of closed loops changes from a positive integer to zero is used as the extinction scale of the corresponding closed loop. The generation and extinction scales of each closed loop are grouped into pairs, and the pairs corresponding to all closed loops are collected into a set of pairs, which yields the set of persistent feature points.

[0038] In this embodiment, homology analysis is performed on the curvature-torsion phase space point cloud distribution map. The pairwise Euclidean distance between point clouds is used as the candidate scale parameter. By constructing connection graphs under different distance thresholds and detecting the generation and disappearance of closed loops, a persistent feature point set is extracted. This homology analysis process is entirely driven by the distribution density of the point cloud data itself and does not depend on the scale step size or criterion threshold set by humans. The generation of a closed loop represents that the connection relationship between point clouds in the phase space has reached a critical state, reflecting that the voltage response manifold forms a closed loop structure in the curvature-torsion combination space - that is, the energy storage power supply has experienced a complete voltage rise and fall cycle. The disappearance of a closed loop represents that a new connection line fills the closed loop region, reflecting that the voltage fluctuation mode is covered by a larger scale topology. The span between the generation scale and the disappearance scale (i.e., the lifetime value) quantifies the stable existence range of the corresponding closed loop structure in the phase space, directly mapping the dynamic consistency of a specific electrochemical process in the energy storage power supply.

[0039] Step S105: Calculate the persistence entropy and lifetime values ​​of each persistent feature point based on the generation and extinction scales of each persistent feature point. Construct a two-dimensional persistent contribution vector for each persistent feature point using the persistence entropy and lifetime values ​​as dimensions. Perform weighted tensor aggregation on the two-dimensional persistent contribution vectors of all persistent feature points to obtain the capacity degradation topology tensor of the current discharge cycle.

[0040] In this step, for each persistent feature point, the extinction scale of the persistent feature point is subtracted from the generation scale of the persistent feature point to obtain the lifetime value of the persistent feature point. The lifetime value represents the span of the corresponding closed loop that persists throughout the entire range of the distance threshold from the generation scale to the extinction scale. The larger the lifetime value, the wider the feature scale range occupied by the corresponding voltage fluctuation mode in the phase space, that is, the wider the range of the fluctuation mode that persists stably during the discharge process. The long-lifetime closed loop corresponds to the bulk diffusion process inside the electrode material, is most sensitive to the structural integrity of the electrode material, and carries the richest capacity degradation information.

[0041] For each persistent feature point, the generation scale of the persistent feature point is divided by the disappearance scale of the persistent feature point to obtain a scale ratio. The persistent entropy value of the persistent feature point is determined based on the scale ratio. The persistent entropy value characterizes the uniformity of the distribution of the area enclosed by the corresponding closed loop in the connectivity graph from generation to disappearance as a function of distance threshold. The persistent entropy value can be calculated using a two-component entropy function: the first entropy component is obtained by calculating an exponential function value with the natural constant as the base and the negative of the scale ratio as the exponent; the first difference is obtained by subtracting the scale ratio from the first entropy component; the second entropy component is obtained by calculating an exponential function value with the natural constant as the base and the negative of the first difference as the exponent; the first entropy component is multiplied by the scale ratio to obtain a first product term; the second entropy component is multiplied by the first difference to obtain a second product term; the first product term and the second product term are added together to obtain the persistent entropy value. For each persistent feature point, the persistent entropy value of the persistent feature point is used as the horizontal axis value, and the lifetime value of the persistent feature point is used as the vertical axis value to form a two-dimensional persistent contribution vector of the persistent feature point on a two-dimensional plane.

[0042] Furthermore, all persistent feature points are arranged in descending order of their lifetime values ​​to obtain a sorted sequence of persistent feature points. The persistent feature points ranked from the first to the Mth in the sequence are assigned to the first weight level, and the persistent feature points ranked from the M+1th to the Nth are assigned to the second weight level. Here, M is a positive integer less than N, and N is the total number of all persistent feature points. The value of M can be set according to actual application requirements. For example, M is equal to one-third of N or one-half of N. Configure a first weight coefficient for the first weight level and a second weight coefficient for the second weight level. The first weight coefficient is greater than the second weight coefficient, and the ratio of the first weight coefficient to the second weight coefficient is not less than 2. For example, the first weight coefficient is 2 and the second weight coefficient is 1. For each persistent feature point belonging to the first weight level, the two-dimensional persistent contribution vector of the persistent feature point is externally productted with the transpose of the two-dimensional persistent contribution vector to obtain the second-order contribution matrix of the persistent feature point. The second-order contribution matrix is ​​then multiplied by the first weight coefficient to obtain the first weighted contribution matrix of the persistent feature point. For each persistent feature point belonging to the second weight level, the two-dimensional persistent contribution vector of the persistent feature point is externally producted with the transpose of the two-dimensional persistent contribution vector to obtain the second-order contribution matrix of the persistent feature point. The second-order contribution matrix is ​​then multiplied by the second weight coefficient to obtain the second weighted contribution matrix of the persistent feature point. The first weighted contribution matrix of all persistent feature points and the second weighted contribution matrix of all persistent feature points are summed by matrix addition to obtain a 2×2 dimensional summation matrix. The summation matrix is ​​used as the capacity degradation topology tensor of the current discharge cycle.

[0043] In this embodiment, by calculating the persistent entropy value of each persistent feature point, the uniformity of the distribution of the area enclosed by each closed loop as a function of the distance threshold is quantified. Closed loops with large persistent entropy values ​​correspond to voltage fluctuations that exhibit a distribution pattern of "slow development followed by sharp intensification" on the time axis, representing the cumulative effect of the polarization process of the electrode material. Closed loops with small persistent entropy values ​​correspond to voltage fluctuations that exhibit a distribution pattern of "sudden appearance and rapid disappearance" on the time axis, representing instantaneous local events inside the electrode material. By combining the persistent entropy value and the lifetime value into a two-dimensional persistent contribution vector, a multi-dimensional joint characterization of the closed loop structural features is achieved, providing rich feature information for subsequent capacity degradation analysis. By employing a weighted tensor aggregation approach, higher weight coefficients are assigned to closed loops with larger lifetime values. This makes the capacity degradation topology tensor more sensitive to the capacity decay mechanism related to bulk diffusion of electrode materials, while effectively suppressing interference from non-structural degradation factors such as surface transient polarization and measurement noise. The capacity degradation topology tensor is a 2×2 matrix, whose diagonal and off-diagonal elements respectively store the distribution structure information and cooperative change information of all closed loops in the two-dimensional space of continuous entropy-lifetime. This fully preserves the cross-coupling relationship between closed loop features and avoids the information loss caused by compressing multi-dimensional features into a single scalar.

[0044] Step S106: Using the distortion rate offset of the current discharge cycle as the capacity decay driving factor of the capacity degradation topology tensor in the current discharge cycle, calculate the remaining capacity of the energy storage power supply at the current moment based on the product relationship between the capacity decay driving factor and the rated capacity of the energy storage power supply.

[0045] In this step, the discharge current signal of the energy storage power source is continuously collected during the current discharge cycle, and the discharge current signal is integrated over time from the start of the discharge to the current time to obtain the cumulative discharge capacity. Calculate the matrix trace of the capacity degradation topology tensor, divide the matrix trace by the cumulative discharge capacity, and obtain the current unit capacity topology distortion rate. The matrix trace is the sum of the diagonal elements of the capacity degradation topology tensor, and the current unit capacity topology distortion rate characterizes the average degree of distortion of the geometric topology of the voltage response manifold during the release of a unit capacity by the energy storage power source. Obtain the initial topology distortion rate per unit charge of the energy storage power source under factory calibration conditions; Calculate the difference between the current topology distortion rate per unit of power and the initial topology distortion rate per unit of power, and divide the difference by the initial topology distortion rate per unit of power to obtain the distortion rate offset; Subtracting the distortion rate offset from the value 1 yields the capacity retention coefficient, and multiplying the capacity retention coefficient by the rated capacity of the energy storage power source yields the remaining capacity of the energy storage power source at the current moment.

[0046] In this embodiment, the topology distortion rate per unit charge is obtained by calculating the ratio of the matrix trace of the capacity degradation topology tensor to the cumulative discharged charge. This distortion rate offset is then obtained by comparing it with the initial value under factory calibration. Finally, the remaining capacity is directly calculated by multiplying the rated capacity by the capacity retention coefficient. The entire calculation process requires only an initial reference value from the factory. During the usage phase, it relies entirely on analytical calculations based on measurement data from the current discharge cycle. It eliminates the need to accumulate large amounts of historical discharge data, fit empirical mapping curves, or rely on any machine learning models or equivalent circuit parameter identification, thus possessing clear physical interpretability and engineering practicality.

[0047] Please see Figure 2 The diagram shows a structural block diagram of a capacity analysis system for an energy storage power source according to this application.

[0048] like Figure 2 As shown, the capacity analysis system 200 includes an acquisition module 210, an extraction module 220, a mapping module 230, a calculation module 240, an aggregation module 250, and an output module 260.

[0049] The acquisition module 210 is configured to acquire the terminal voltage time-series signal continuously sampled by the energy storage power supply during the current discharge cycle, and construct the original response manifold trajectory of the terminal voltage time-series signal in a two-dimensional coordinate system with the sampling time as the horizontal axis and the terminal voltage value as the vertical axis; the extraction module 220 is configured to perform Gaussian regularized convolution processing on the original response manifold trajectory to obtain a smoothed regularized response manifold trajectory, and extract the geometric invariant sequence of the regularized response manifold trajectory according to a preset resampling rule, wherein the geometric invariant sequence includes at least the geodesic curvature sequence and geodesic torsion sequence of the regularized response manifold trajectory at each sampling point; the mapping module 230 is configured to map the regularized response manifold trajectory to the curvature-torsion phase space with the geodesic curvature in the geodesic curvature sequence and the geodesic torsion in the geodesic torsion sequence as coordinate axes to obtain a phase space point cloud distribution map; the calculation module 240 is configured to perform homology analysis on the phase space point cloud distribution map and calculate different scales. The parameters are used to determine the change trajectory of the phase space point cloud distribution map, and the persistent feature point set of the phase space point cloud distribution map is extracted based on the change trajectory. The persistent feature point set contains multiple persistent feature points, and each persistent feature point corresponds to a generation scale and a decay scale of a topological feature. The aggregation module 250 is configured to calculate the persistence entropy value and lifetime value of each persistent feature point based on the generation scale and decay scale of each persistent feature point, construct a two-dimensional persistent contribution vector of each persistent feature point using the persistence entropy value and the lifetime value as dimensions, and perform weighted tensor aggregation on the two-dimensional persistent contribution vectors of all persistent feature points to obtain the capacity degradation topological tensor of the current discharge cycle. The output module 260 is configured to use the distortion rate offset of the current discharge cycle as the capacity decay driving factor of the capacity degradation topological tensor in the current discharge cycle, and calculate the remaining capacity of the energy storage power supply at the current moment based on the product relationship between the capacity decay driving factor and the rated capacity of the energy storage power supply.

[0050] In summary, the method of this application acquires the terminal voltage time-series signal and constructs the original response manifold trajectory; extracts the geodesic curvature and geodesic torsion of the smooth trajectory through Gaussian regularized convolution; maps it to the curvature-torsion phase space to obtain a point cloud distribution map; performs homology analysis to extract a persistent feature point set; calculates the persistent entropy and lifetime values ​​of each feature point and constructs a two-dimensional persistent contribution vector, and aggregates the weighted tensors to obtain the capacity degradation topology tensor; calculates the current topology distortion rate per unit of charge and compares it with the initial factory value to obtain the distortion rate offset, and corrects the rated capacity accordingly to obtain the remaining capacity; transforms capacity estimation into a manifold geometric topology analysis problem, which does not rely on historical data and empirical mapping and has operating condition adaptability.

[0051] It should be understood that Figure 2 The modules and references described in the document Figure 1The steps described in the text correspond to those in the method described above. Therefore, the operations, features, and corresponding technical effects described above also apply to the method described in the text. Figure 2 The various modules in the document will not be described in detail here.

[0052] In other embodiments, the present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein when the program instructions are executed by a processor, the processor performs the capacity analysis method for the energy storage power supply in any of the above method embodiments. In one embodiment, the computer-readable storage medium of the present invention stores computer-executable instructions, which are configured as follows: The terminal voltage timing signal obtained by continuous sampling of the energy storage power supply during the current discharge cycle is acquired. The original response manifold trajectory of the terminal voltage timing signal is constructed in a two-dimensional coordinate system with sampling time as the horizontal axis and terminal voltage value as the vertical axis. The original response manifold trajectory is subjected to Gaussian regularized convolution to obtain a smoothed regularized response manifold trajectory. The geometric invariant sequence of the regularized response manifold trajectory is extracted according to a preset resampling rule. The geometric invariant sequence includes at least the geodesic curvature sequence and geodesic torsion sequence of the regularized response manifold trajectory at each sampling point. Using the geodesic curvature in the geodesic curvature sequence and the geodesic torsion in the geodesic torsion sequence as coordinate axes, the regular response manifold trajectory is mapped to the curvature-torsion phase space to obtain the phase space point cloud distribution map; Homology analysis is performed on the phase space point cloud distribution map to calculate the change trajectory of the phase space point cloud distribution map under different scale parameters, and the persistent feature point set of the phase space point cloud distribution map is extracted based on the change trajectory. The persistent feature point set contains multiple persistent feature points, and each persistent feature point corresponds to the generation scale and the disappearance scale of a topological feature. The persistence entropy and lifetime values ​​of each persistent feature point are calculated based on the generation and extinction scales of each persistent feature point. A two-dimensional persistent contribution vector of each persistent feature point is constructed using the persistence entropy and lifetime values ​​as dimensions. Weighted tensor aggregation is performed on the two-dimensional persistent contribution vectors of all persistent feature points to obtain the capacity degradation topology tensor of the current discharge cycle. Using the distortion rate offset of the current discharge cycle as the capacity decay driving factor of the capacity degradation topology tensor in the current discharge cycle, the remaining capacity of the energy storage power source at the current moment is calculated based on the product relationship between the capacity decay driving factor and the rated capacity of the energy storage power source.

[0053] Computer-readable storage media may include a stored program area and a stored data area, wherein the stored program area may store an operating system and an application program required for at least one function; the stored data area may store data created based on the use of the energy storage power supply's capacity analysis system, etc. Furthermore, the computer-readable storage medium may include high-speed random access memory, and may also include memory, such as at least one disk storage device, flash memory device, or other non-volatile solid-state storage device. In some embodiments, the computer-readable storage medium may optionally include memory remotely disposed relative to a processor, which can be connected to the energy storage power supply's capacity analysis system via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0054] Figure 3 This is a schematic diagram of the structure of the electronic device provided in the embodiment of the present invention, such as... Figure 3 As shown, the device includes a processor 310 and a memory 320. The electronic device may also include an input device 330 and an output device 340. The processor 310, memory 320, input device 330, and output device 340 can be connected via a bus or other means. Figure 3 Taking a bus connection as an example, the memory 320 is the computer-readable storage medium described above. The processor 310 executes various server functions and data processing by running non-volatile software programs, instructions, and modules stored in the memory 320, thereby implementing the capacity analysis method for the energy storage power supply described in the above embodiment. The input device 330 can receive input digital or character information and generate key signal inputs related to user settings and function control of the energy storage power supply capacity analysis system. The output device 340 may include a display screen or other display device.

[0055] The aforementioned electronic device can execute the method provided in the embodiments of the present invention, and has the corresponding functional modules and beneficial effects for executing the method. Technical details not described in detail in this embodiment can be found in the method provided in the embodiments of the present invention.

[0056] In one implementation, the above-described electronic device is applied to a capacity analysis system for energy storage power supplies, serving as a client, and includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to: The terminal voltage timing signal obtained by continuous sampling of the energy storage power supply during the current discharge cycle is acquired. The original response manifold trajectory of the terminal voltage timing signal is constructed in a two-dimensional coordinate system with sampling time as the horizontal axis and terminal voltage value as the vertical axis. The original response manifold trajectory is subjected to Gaussian regularized convolution to obtain a smoothed regularized response manifold trajectory. The geometric invariant sequence of the regularized response manifold trajectory is extracted according to a preset resampling rule. The geometric invariant sequence includes at least the geodesic curvature sequence and geodesic torsion sequence of the regularized response manifold trajectory at each sampling point. Using the geodesic curvature in the geodesic curvature sequence and the geodesic torsion in the geodesic torsion sequence as coordinate axes, the regular response manifold trajectory is mapped to the curvature-torsion phase space to obtain the phase space point cloud distribution map; Homology analysis is performed on the phase space point cloud distribution map to calculate the change trajectory of the phase space point cloud distribution map under different scale parameters, and the persistent feature point set of the phase space point cloud distribution map is extracted based on the change trajectory. The persistent feature point set contains multiple persistent feature points, and each persistent feature point corresponds to the generation scale and the disappearance scale of a topological feature. The persistence entropy and lifetime values ​​of each persistent feature point are calculated based on the generation and extinction scales of each persistent feature point. A two-dimensional persistent contribution vector of each persistent feature point is constructed using the persistence entropy and lifetime values ​​as dimensions. Weighted tensor aggregation is performed on the two-dimensional persistent contribution vectors of all persistent feature points to obtain the capacity degradation topology tensor of the current discharge cycle. Using the distortion rate offset of the current discharge cycle as the capacity decay driving factor of the capacity degradation topology tensor in the current discharge cycle, the remaining capacity of the energy storage power source at the current moment is calculated based on the product relationship between the capacity decay driving factor and the rated capacity of the energy storage power source.

[0057] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of various embodiments or some parts of embodiments.

[0058] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A capacity analysis method for an energy storage power source, characterized in that, include: The terminal voltage timing signal obtained by continuous sampling of the energy storage power supply during the current discharge cycle is acquired. The original response manifold trajectory of the terminal voltage timing signal is constructed in a two-dimensional coordinate system with sampling time as the horizontal axis and terminal voltage value as the vertical axis. The original response manifold trajectory is subjected to Gaussian regularized convolution to obtain a smoothed regularized response manifold trajectory. The geometric invariant sequence of the regularized response manifold trajectory is extracted according to a preset resampling rule. The geometric invariant sequence includes at least the geodesic curvature sequence and geodesic torsion sequence of the regularized response manifold trajectory at each sampling point. Using the geodesic curvature in the geodesic curvature sequence and the geodesic torsion in the geodesic torsion sequence as coordinate axes, the regular response manifold trajectory is mapped to the curvature-torsion phase space to obtain the phase space point cloud distribution map; Homology analysis is performed on the phase space point cloud distribution map to calculate the change trajectory of the phase space point cloud distribution map under different scale parameters, and the persistent feature point set of the phase space point cloud distribution map is extracted based on the change trajectory. The persistent feature point set contains multiple persistent feature points, and each persistent feature point corresponds to the generation scale and the disappearance scale of a topological feature. The persistence entropy and lifetime values ​​of each persistent feature point are calculated based on the generation and extinction scales of each persistent feature point. A two-dimensional persistent contribution vector of each persistent feature point is constructed using the persistence entropy and lifetime values ​​as dimensions. Weighted tensor aggregation is performed on the two-dimensional persistent contribution vectors of all persistent feature points to obtain the capacity degradation topology tensor of the current discharge cycle. Using the distortion rate offset of the current discharge cycle as the capacity decay driving factor of the capacity degradation topology tensor in the current discharge cycle, the remaining capacity of the energy storage power source at the current moment is calculated based on the product relationship between the capacity decay driving factor and the rated capacity of the energy storage power source.

2. The capacity analysis method for an energy storage power source according to claim 1, characterized in that, The original response manifold trajectory of the terminal voltage time-series signal, constructed in a two-dimensional coordinate system with sampling time as the horizontal axis and terminal voltage value as the vertical axis, includes: The terminal voltage values ​​at each sampling time in the terminal voltage timing signal are arranged sequentially according to the sampling time. Each sampling time is used as the horizontal axis value, and the corresponding terminal voltage value at each sampling time is used as the vertical axis value to generate a discrete voltage sampling point column in the two-dimensional coordinate system. Interpolation fitting is performed on the discrete voltage sampling point sequence to generate a continuously differentiable two-dimensional curve, which is then used as the original response manifold trajectory.

3. The capacity analysis method for an energy storage power source according to claim 1, characterized in that, The step of extracting the geometric invariant sequence of the regular response manifold trajectory according to a preset resampling rule includes: Perform a first-order differential operation on the terminal voltage timing signal to obtain a voltage differential sequence. Detect all local extreme points on the voltage differential sequence and arrange the voltage values ​​corresponding to each local extreme point in descending order. Select the voltage values ​​corresponding to the first few local extreme points as feature voltage values. Centered on each characteristic voltage value, a voltage window interval is determined with a preset voltage window radius. The union of all voltage window intervals is taken as the set of characteristic voltage intervals. The voltage window radius is a preset percentage of the total voltage change amplitude of the terminal voltage timing signal. On the regular response manifold trajectory, all trajectory segments whose ordinate values ​​fall within the set of characteristic voltage intervals are determined. A first resampling density is set in each of the trajectory segments according to the equal arc length method. A second resampling density is set on the remaining trajectory segments on the regular response manifold trajectory other than the set of characteristic voltage intervals. The first resampling density is greater than the second resampling density. Multiple resampling positions are determined on the regular response manifold trajectory according to the first resampling density and the second resampling density, and the first-order directional derivative vector and the second-order directional derivative vector of the regular response manifold trajectory are calculated at each resampling position; At each resampling location, the curvature component of the resampling location is calculated based on the first-order directional derivative vector and the second-order directional derivative vector, and the curvature components are arranged in the order of the resampling locations on the regular response manifold trajectory to form the geodesic curvature sequence. At each resampling location, the torsion component of the resampling location is calculated based on the first-order directional derivative vector and the second-order directional derivative vector, and the torsion components are arranged in the order of the resampling locations on the regular response manifold trajectory to form the geodesic torsion sequence.

4. The capacity analysis method for an energy storage power source according to claim 1, characterized in that, The step of performing homology analysis on the phase space point cloud distribution map, calculating the change trajectory of the phase space point cloud distribution map under different scale parameters, and extracting the persistent feature point set of the phase space point cloud distribution map based on the change trajectory includes: Calculate the pairwise Euclidean distances between all point clouds in the phase space point cloud distribution map, sort all the pairwise Euclidean distances in ascending order of value to obtain an ordered distance sequence, remove the pairwise Euclidean distances that are greater than a preset distance threshold from the ordered distance sequence, and arrange the remaining pairwise Euclidean distances in ascending order to obtain a candidate scale parameter sequence. Each candidate scale parameter value in the candidate scale parameter sequence is taken out as the current distance threshold. Under each current distance threshold, any two point clouds in the phase space point cloud distribution map whose Euclidean distance is less than the current distance threshold are connected by a straight line segment to obtain the connection map under the current distance threshold. In the connection graph at the current distance threshold, all closed loops formed by the straight line segments are detected, and the number of closed loops is counted to obtain the trajectory curve of the number of closed loops as the distance threshold changes. The perimeter of the closed loop is equal to the sum of the lengths of the straight line segments that make up the closed loop. All local extreme points are detected on the closed loop number trajectory curve. The distance threshold corresponding to the starting point of the rising edge of the closed loop number trajectory curve where the number of closed loops first changes from zero to a positive integer is used as the generation scale of the corresponding closed loop. The distance threshold corresponding to the ending point of the falling edge of the closed loop number trajectory curve where the number of closed loops changes from a positive integer to zero is used as the extinction scale of the corresponding closed loop. The generation and extinction scales of each closed loop are grouped into pairs, and the pairs corresponding to all closed loops are collected into a set of pairs, which yields the set of persistent feature points.

5. The capacity analysis method for an energy storage power source according to claim 4, characterized in that, The step of calculating the persistence entropy and lifetime values ​​of each persistent feature point based on the generation and decay scales of each persistent feature point, and constructing a two-dimensional persistent contribution vector for each persistent feature point using the persistence entropy and lifetime values ​​as dimensions, includes: For each persistent feature point, the extinction scale of the persistent feature point is subtracted from the generation scale of the persistent feature point to obtain the lifetime value of the persistent feature point. The lifetime value represents the span of the corresponding closed loop that persists throughout the entire range of the distance threshold from the generation scale to the extinction scale. For each persistent feature point, the generation scale of the persistent feature point is divided by the extinction scale of the persistent feature point to obtain a scale ratio. The persistent entropy value of the persistent feature point is determined based on the scale ratio. The persistent entropy value characterizes the uniformity of the distribution of the area enclosed by the corresponding closed loop in the connectivity graph from generation to extinction as a function of a distance threshold. For each persistent feature point, the persistent entropy value of the persistent feature point is used as the horizontal axis value, and the lifetime value of the persistent feature point is used as the vertical axis value to form a two-dimensional persistent contribution vector of the persistent feature point on a two-dimensional plane.

6. The capacity analysis method for an energy storage power source according to claim 1, characterized in that, The weighted tensor aggregation of the two-dimensional persistent contribution vectors of all persistent feature points to obtain the capacity degradation topology tensor of the current discharge cycle includes: All persistent feature points are arranged in descending order of their lifetime values ​​to obtain a sorted persistent feature point sequence. The persistent feature points ranked from the first to the Mth in the persistent feature point sequence are assigned to the first weight level, and the persistent feature points ranked from the M+1th to the Nth are assigned to the second weight level, where M is a positive integer less than N and N is the total number of all persistent feature points. Configure a first weight coefficient for the first weight level and a second weight coefficient for the second weight level. The first weight coefficient is greater than the second weight coefficient, and the ratio of the first weight coefficient to the second weight coefficient is not less than 2. For each persistent feature point belonging to the first weight level, the two-dimensional persistent contribution vector of the persistent feature point is externally productted with the transpose of the two-dimensional persistent contribution vector to obtain the second-order contribution matrix of the persistent feature point. The second-order contribution matrix is ​​then multiplied by the first weight coefficient to obtain the first weighted contribution matrix of the persistent feature point. For each persistent feature point belonging to the second weight level, the two-dimensional persistent contribution vector of the persistent feature point is externally producted with the transpose of the two-dimensional persistent contribution vector to obtain the second-order contribution matrix of the persistent feature point. The second-order contribution matrix is ​​then multiplied by the second weight coefficient to obtain the second weighted contribution matrix of the persistent feature point. The first weighted contribution matrix of all persistent feature points and the second weighted contribution matrix of all persistent feature points are summed by matrix addition to obtain a 2×2 dimensional summation matrix. The summation matrix is ​​used as the capacity degradation topology tensor of the current discharge cycle.

7. The capacity analysis method for an energy storage power source according to claim 1, characterized in that, The step of using the distortion rate offset of the current discharge cycle as the capacity decay driving factor of the capacity degradation topology tensor within the current discharge cycle, and calculating the remaining capacity of the energy storage power source at the current moment based on the product relationship between the capacity decay driving factor and the rated capacity of the energy storage power source, includes: The discharge current signal of the energy storage power source is continuously collected during the current discharge cycle, and the discharge current signal is integrated over time from the start of the discharge to the current time to obtain the cumulative discharge capacity. Calculate the matrix trace of the capacity degradation topology tensor, divide the matrix trace by the cumulative discharge capacity, and obtain the current unit capacity topology distortion rate. The matrix trace is the sum of the diagonal elements of the capacity degradation topology tensor, and the current unit capacity topology distortion rate characterizes the average degree of distortion of the geometric topology of the voltage response manifold during the release of a unit capacity by the energy storage power source. Obtain the initial topology distortion rate per unit charge of the energy storage power source under factory calibration conditions; Calculate the difference between the current topology distortion rate per unit of power and the initial topology distortion rate per unit of power, and divide the difference by the initial topology distortion rate per unit of power to obtain the distortion rate offset; Subtracting the distortion rate offset from the value 1 yields the capacity retention coefficient, and multiplying the capacity retention coefficient by the rated capacity of the energy storage power source yields the remaining capacity of the energy storage power source at the current moment.

8. A capacity analysis system for an energy storage power source, characterized in that, include: The acquisition module is configured to acquire the terminal voltage timing signal continuously sampled by the energy storage power supply during the current discharge cycle, and construct the original response manifold trajectory of the terminal voltage timing signal in a two-dimensional coordinate system with the sampling time as the horizontal axis and the terminal voltage value as the vertical axis. The extraction module is configured to perform Gaussian regularized convolution processing on the original response manifold trajectory to obtain a smoothed regularized response manifold trajectory, and extract the geometric invariant sequence of the regularized response manifold trajectory according to a preset resampling rule. The geometric invariant sequence includes at least the geodesic curvature sequence and geodesic torsion sequence of the regularized response manifold trajectory at each sampling point. The mapping module is configured to map the regular response manifold trajectory to the curvature-torsion phase space using the geodesic curvature in the geodesic curvature sequence and the geodesic torsion in the geodesic torsion sequence as coordinate axes, thereby obtaining a phase space point cloud distribution map. The calculation module is configured to perform homology analysis on the phase space point cloud distribution map, calculate the change trajectory of the phase space point cloud distribution map under different scale parameters, and extract the persistent feature point set of the phase space point cloud distribution map based on the change trajectory. The persistent feature point set contains multiple persistent feature points, and each persistent feature point corresponds to the generation scale and the disappearance scale of a topological feature. The aggregation module is configured to calculate the persistence entropy and lifetime values ​​of each persistent feature point based on the generation and extinction scales of each persistent feature point, construct a two-dimensional persistent contribution vector of each persistent feature point using the persistence entropy and lifetime values ​​as dimensions, and perform weighted tensor aggregation on the two-dimensional persistent contribution vectors of all persistent feature points to obtain the capacity degradation topology tensor of the current discharge cycle. The output module is configured to use the distortion rate offset of the current discharge cycle as the capacity decay driving factor of the capacity degradation topology tensor in the current discharge cycle, and calculate the remaining capacity of the energy storage power supply at the current moment based on the product relationship between the capacity decay driving factor and the rated capacity of the energy storage power supply.

9. An electronic device, characterized in that, include: At least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method according to any one of claims 1 to 7.

Images