Health management method and system for large-scale energy storage power station

Abstract

The application provides a large-scale energy storage power station health management method and system, relates to the technical field of energy storage power station health management, and comprises the following steps: constructing a charging curve corresponding to cumulative charging capacity and terminal voltage, and calculating differential capacity entropy based on the charging curve; performing time-frequency analysis on total current and terminal voltage, and extracting impedance phase dispersion; determining a thermal consistency index based on the dispersion degree between temperature change rates of each battery monomer; constructing a health feature matrix, performing trend filtering on each health feature sequence in the health feature matrix, and obtaining a corresponding degradation trend value; standardizing and performing principal component analysis on the degradation trend matrix, constructing an instant health index of a current effective segment, judging the instant health index, and automatically outputting early warning information. Through multi-dimensional feature fusion, the detection sensitivity of battery cluster inconsistency degradation and potential thermal runaway precursors is significantly improved.

Description

Technical Field

[0001] This invention relates to the field of energy storage power station health management technology, specifically to a method and system for large-scale energy storage power station health management. Background Technology

[0002] With the large-scale grid connection of renewable energy sources such as wind power and photovoltaics, electrochemical energy storage power stations, as a key resource for the flexible regulation of the power system, are becoming increasingly important for their safe and stable operation. Energy storage power stations typically consist of hundreds or thousands of battery cells connected in series and parallel. Operating under complex conditions such as frequent charging and discharging and high current surges, battery performance deteriorates rapidly, easily leading to safety hazards such as abnormal heating, internal short circuits, and thermal runaway. According to incomplete statistics, more than 100 fire accidents involving energy storage power stations have occurred globally in the past decade, making the safety of battery energy storage systems the primary bottleneck restricting their large-scale application. In actual operation, due to differences in manufacturing, uneven temperature distribution, and charging and discharging strategies, the aging and performance degradation of individual cells within a battery cluster often vary. This inconsistency worsens over time, increasing the risk of overcharging, over-discharging, and thermal runaway. The patent document with publication number CN116581804A proposes a method for real-time monitoring and evaluation of the state of charge, power loss, and remaining cycle count of an energy storage subsystem using a SOC assessment model, a charge / discharge loss model, and a lifetime assessment model within a full lifecycle model. However, this method's health assessment mainly relies on macroscopic electrical parameters such as SOC and cycle count, lacking direct perception of the degradation of internal electrochemical characteristics of the battery and making it difficult to capture subtle changes in the early stages of battery aging. The thermal behavior analysis of this method is limited to the power loss calculation level and does not involve a quantitative assessment of the temperature consistency between individual cells within the battery cluster, resulting in limited ability to identify thermal runaway risks. The evaluation results of this method are mainly based on SOC equalization control and lifetime prediction, focusing on operational regulation rather than early fault warning. It lacks a mechanism to integrate multi-dimensional health characteristics into a unified quantitative indicator and trigger proactive alarms accordingly, making it difficult to issue effective warnings in the early stages of battery cluster consistency degradation. Therefore, there is an urgent need for a health management method and system for large-scale energy storage power stations.

[0003] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0004] The purpose of this invention is to provide a health management method and system for large-scale energy storage power stations to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: The health management methods for large-scale energy storage power stations include the following specific steps: S1: Collect the total current, terminal voltage and temperature of each battery cell of the battery cluster at a preset sampling frequency, and identify the constant current charging state based on the total current, and determine the continuous charging period that meets the preset stability conditions as the effective segment. S2: For the effective segment, determine the cumulative charging capacity at each sampling time based on the total current and terminal voltage, construct a charging curve corresponding to the cumulative charging capacity and terminal voltage, and calculate the differential capacity entropy based on the charging curve; at the same time, perform time-frequency analysis on the total current and terminal voltage to extract the impedance phase dispersion. S3: Calculate the corresponding temperature change rate based on the temperature of each battery cell, and determine the thermal consistency index based on the dispersion between the temperature change rates of each battery cell; use the differential capacity entropy, impedance phase dispersion and thermal consistency index as health features, and construct a health feature matrix according to the time sequence of the effective segments; S4: Perform trend filtering on each health feature sequence in the health feature matrix to obtain the corresponding degradation trend value and construct a degradation trend matrix; perform standardization and principal component analysis on the degradation trend matrix, fuse the real-time health index based on the contribution of each principal component, and output the corresponding health warning information according to the comparison result between the real-time health index and the preset health threshold.

[0006] Furthermore, the identification logic for valid segments is as follows: the total current of the battery cluster is collected at fixed sampling times. Within a preset sliding window, if the deviation of the total current from the preset constant current charging setting value is always within the preset allowable deviation range, and the duration of this deviation exceeds the preset minimum duration, then the state is determined as a constant current charging state, and the continuous charging period corresponding to the constant current charging state is marked as a valid segment.

[0007] Furthermore, a charging curve corresponding to the cumulative charging capacity and terminal voltage is constructed, specifically including: The total current of each sampling point within the current valid segment is integrated using the rectangular method to obtain the cumulative charging capacity corresponding to each sampling point. Using the cumulative charging capacity and terminal voltage of each of two adjacent sampling points, the adjacent sampling points are connected sequentially with straight line segments in the plane formed by the cumulative charging capacity and terminal voltage through linear interpolation to form the charging curve.

[0008] Furthermore, the logic for calculating the differential capacity entropy based on the charging curve is as follows: A preset voltage range is selected from the charging curve, and equally spaced voltage points are generated within this preset voltage range. The cumulative charging capacity corresponding to each voltage point is calculated by reverse calculation of the charging curve. The differential cumulative charging capacity within each adjacent voltage point range is calculated, and all differential cumulative charging capacities are normalized to obtain a probability distribution. The information entropy is calculated based on the probability distribution and used as the differential capacity entropy.

[0009] Furthermore, the impedance phase dispersion within the current valid segment is extracted, specifically including: Apply sliding time windows to the total current sequence and terminal voltage sequence within the current effective segment and perform fast Fourier transform to obtain the current spectrum and voltage spectrum corresponding to each frequency point within each time window, and then calculate the impedance phase corresponding to each frequency point. The variance of all impedance phases relative to their overall average value is calculated for all time windows and all frequency points, and this variance is used as the impedance phase dispersion.

[0010] Furthermore, the thermal consistency index of the current valid fragment is generated, specifically including: For the temperature of each battery cell at each sampling point within the current valid segment, the corresponding temperature change rate is calculated using first-order backward difference. For each sampling point, calculate the sum of squares of the temperature differences between all adjacent units, take the average and then take the square root to obtain the discrete value of the temperature change rate at that sampling point. The thermal consistency index is obtained by averaging the discrete values ​​of the temperature change rate of all sampling points within the current valid segment.

[0011] Furthermore, The trend filter uses Savitzky. The Golay filter is implemented, and the method for constructing the degradation trend matrix is ​​as follows: Extract each column of the health feature matrix as a health feature sequence; The half-window length and polynomial order of the preset filter are used. With each effective segment position as the center position, polynomial least squares fitting is performed on the health feature sequence values ​​within the window. The value of the fitted polynomial at the effective segment corresponding to the center position is taken as the degradation trend value of the effective segment. The degradation trend matrix is ​​formed by combining the degradation trend values ​​corresponding to the differential capacity entropy, impedance phase dispersion, and thermal uniformity index column by column.

[0012] Furthermore, the method for generating the real-time health index is as follows: The degradation trend matrix is ​​standardized column by column, and each standardized row is used as a degradation trend feature vector. Calculate the covariance matrix of the standardized degradation trend matrix, and perform eigenvalue decomposition on the covariance matrix, taking the eigenvector corresponding to the largest eigenvalue as the principal component direction; The degradation trend feature vector of the current effective fragment is projected onto the direction of the principal component to obtain the instant health index; When the real-time health index of a preset number of consecutive valid segments exceeds the preset health threshold, an early warning message is output.

[0013] The present invention also provides a large-scale energy storage power station health management system, which is used to execute the above-described large-scale energy storage power station health management method, including: The data acquisition module is used to collect the total current, terminal voltage and temperature of each battery cell of the battery cluster at a preset sampling frequency, and to identify the constant current charging state based on the total current, and to determine the continuous charging period that meets the preset stability conditions as the effective segment. The parameter calculation module is used to determine the cumulative charging capacity at each sampling time based on the total current and terminal voltage for the effective segment, construct a charging curve corresponding to the cumulative charging capacity and terminal voltage, and calculate the differential capacity entropy based on the charging curve; at the same time, it performs time-frequency analysis on the total current and terminal voltage to extract the impedance phase dispersion. The matrix construction module is used to calculate the corresponding temperature change rate based on the temperature of each battery cell, and to determine the thermal consistency index based on the dispersion between the temperature change rates of each battery cell; using the differential capacity entropy, impedance phase dispersion and thermal consistency index as health features, a health feature matrix is ​​constructed according to the time order of the effective segments. The health assessment module is used to perform trend filtering on each health feature sequence in the health feature matrix to obtain the corresponding degradation trend value and construct the degradation trend matrix; to perform standardization and principal component analysis on the degradation trend matrix, to obtain an instant health index based on the contribution of each principal component, and to output corresponding health warning information based on the comparison result of the instant health index and the preset health threshold.

[0014] Compared with the prior art, the beneficial effects of the present invention are: This solution identifies valid segments based on constant current charging status, ensuring consistency of operating conditions for subsequent feature extraction and avoiding interference from invalid data such as charge / discharge switching and current fluctuations. Differential capacity entropy, impedance phase dispersion, and thermal consistency index are extracted as health features from three dimensions: electrochemistry, impedance spectroscopy, and thermal behavior. Compared to existing health assessment methods that rely on only a single parameter, this approach more comprehensively characterizes the degradation information of internal electrode phase transitions, electrochemical process consistency, and thermal diffusion uniformity, improving the accuracy and reliability of health assessment. A real-time health index is obtained by fusing the health feature sequence with trend filtering and principal component analysis, effectively filtering out operational noise and extracting long-term degradation trends in battery performance, making the health status assessment results more robust. An early warning is issued when the real-time health index of multiple consecutive valid segments exceeds a preset threshold, allowing for early identification of battery cluster consistency degradation and potential thermal runaway risks. This provides a scientific basis for energy storage power station operation and maintenance decisions, helping to reduce unplanned downtime and extend battery life. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the overall method flow of the present invention; Figure 2 This is a graph showing the relationship between the proportion of differential charging capacity to total differential charging capacity in this invention and the cumulative charging capacity. Figure 3 This is a schematic diagram of the overall system modules of the present invention. Detailed Implementation

[0016] 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.

[0017] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0018] Example: Please see Figures 1-2 The present invention provides a technical solution: The health management methods for large-scale energy storage power stations include the following specific steps: S1: Collect the total current, terminal voltage and temperature of each battery cell of the battery cluster at a preset sampling frequency, and identify the constant current charging state based on the total current, and determine the continuous charging period that meets the preset stability conditions as the effective segment. In this embodiment, the identification logic of the valid segment is as follows: the total current of the battery cluster is collected at fixed sampling times. Within a preset sliding window, if the deviation of the total current from the preset constant current charging setting value is always within the preset allowable deviation range, and the duration of the deviation exceeds the preset minimum duration, then the state is determined as a constant current charging state, and the continuous charging period corresponding to the constant current charging state is marked as a valid segment.

[0019] In the above process, under constant current charging conditions, the charging current remains constant, and the change in terminal voltage is mainly determined by the internal electrochemical state of the battery. At this time, the correspondence between the accumulated charging capacity and the terminal voltage can truly reflect the capacity increment characteristics of the battery. If the charging curve is constructed and the relevant characteristics are calculated under current fluctuation or non-constant current conditions, the change in terminal voltage will be affected by both current changes and changes in the internal state of the battery. The extracted feature values ​​cannot accurately reflect the degradation information of the battery itself. Therefore, by identifying the constant current charging state and extracting effective segments, a unified operating condition basis is established for the accurate calculation of subsequent multi-dimensional health characteristics, ensuring the reliability and repeatability of the entire health assessment process. The preset sampling frequency needs to comprehensively consider the hardware acquisition capabilities of the battery management system (BMS) in the energy storage power station and the data resolution requirements for subsequent feature extraction. In modular battery energy storage systems, the optimization of the sampling frequency needs to minimize the hardware and computational processing requirements of the BMS while ensuring the accuracy of the state of charge estimation. For the calculation of differential capacity entropy and impedance phase dispersion, a sampling frequency of no less than 1Hz is usually required to capture the dynamic changes in terminal voltage. In actual engineering, the current and voltage sampling frequencies of the BMS are mostly between 1Hz and 10Hz. The preset constant current charging setting value is directly taken from the current command value set in the charging strategy of the energy storage power station. It is determined by the charging and discharging management strategy of the power station. This value varies under different cell types and power station scales. For example, the constant current charging current of a single lithium-ion battery can be programmed in the range of 100mA to several amperes. The preset allowable deviation range is usually set based on the combined effect of the charger current ripple and the measurement error of the current sensor, and is generally between ±1% and ±5% of the constant current setting value. The preset minimum duration is generally set based on the minimum data length requirements of the battery's electrochemical characteristics and the typical charging time of the energy storage station. The value range is usually from tens of seconds to several minutes. The preset sliding window length can be determined by multiplying the BMS sampling frequency by the number of sampling points to be covered. When the sampling frequency is 1Hz, the preset sliding window length can usually be a value between 30 seconds and 300 seconds; when the sampling frequency is 10Hz, the preset sliding window length can be a value between 10 seconds and 60 seconds, so as to ensure that the window contains enough sampling points for reliable judgment of the deviation persistence state.

[0020] S2: For the effective segment, determine the cumulative charging capacity at each sampling time based on the total current and terminal voltage, construct a charging curve corresponding to the cumulative charging capacity and terminal voltage, and calculate the differential capacity entropy based on the charging curve; at the same time, perform time-frequency analysis on the total current and terminal voltage to extract the impedance phase dispersion. In this embodiment, the charging curve corresponding to the cumulative charging capacity and the terminal voltage is constructed, specifically including: The total current of each sampling point within the current valid segment is integrated using the rectangular method to obtain the cumulative charging capacity corresponding to each sampling point. Using the cumulative charging capacity and terminal voltage of each of two adjacent sampling points, the adjacent sampling points are connected sequentially with straight line segments in the plane formed by the cumulative charging capacity and terminal voltage through linear interpolation to form the charging curve.

[0021] Construct the charging curve corresponding to the cumulative charging capacity and terminal voltage using the following formula: in, Indicates the first valid segment of the current segment. The cumulative charging capacity corresponding to each sampling point; Indicates the first valid segment of the current segment. The terminal voltage corresponding to each sampling point; This indicates the total number of sampling points for the currently valid segment; Indicates the first valid segment of the current segment. The cumulative charging capacity corresponding to each sampling point; Indicates an indicator function, when The value is 1 if the condition is met, and 0 otherwise. Indicates cumulative charging capacity; The first valid fragment of the current segment The terminal voltage corresponding to each sampling point; Indicates the index of the sampling point in the current valid segment; Indicates the cumulative charging capacity is The charging curve under the given conditions.

[0022] The technical effect achieved by the above process is that by using the rectangular method to numerically integrate the discrete total current sample values, the current within each sampling interval is approximated as a constant value and multiplied by the sampling time interval. This realizes the conversion from discrete current data to continuous cumulative charging capacity, transforming the original acquired current sequence into a cumulative charging capacity sequence that can be directly used for subsequent curve construction, thus providing a data foundation for establishing the charging curve. Simultaneously, by using linear interpolation to sequentially connect the discrete cumulative charging capacity and terminal voltage data points with straight line segments, a continuous and usable charging curve is formed. This ensures that the corresponding terminal voltage value can be obtained at any cumulative charging capacity value, thereby meeting the requirement for continuous values ​​of cumulative charging capacity derived from equally spaced voltage points in subsequent differential capacity entropy calculations. The specific implementation of the rectangular integration method is as follows: Within the current valid segment, for each sampling point, the time interval from the previous sampling time to the current sampling time is considered as the width of the rectangle, and the total current value measured within this time interval is considered as the height of the rectangle. Multiplying the two yields the area of ​​the rectangle, which is the cumulative charging capacity increment within this time interval. Subsequently, starting from the beginning of the valid segment, the areas of all rectangles are accumulated sequentially in time to obtain the cumulative charging capacity value corresponding to each sampling point. The rectangular integration method is used because the total current data collected by the battery management system (BMS) of the energy storage power station is a discrete sampled value, not an analytical function over continuous time, and cannot be directly integrated analytically. Within each sampling interval, the total current can be approximated as constant. Therefore, multiplying this constant current value by the sampling time interval to approximate the actual charging capacity increment within that interval is a numerical integration method that meets engineering accuracy requirements and has the least computational cost. Linear interpolation is used to connect adjacent sampling points sequentially with straight line segments to form a charging curve. This is because the actual sampled cumulative charging capacity and terminal voltage are discrete data points, while the subsequent calculation of differential capacity entropy requires generating equally spaced voltage points within a continuous voltage range and then calculating the corresponding cumulative charging capacity. This necessitates that the charging curve must have continuous values ​​between each sampling point. Linear interpolation, as the simplest interpolation method, can obtain a continuous curve that meets engineering accuracy requirements with extremely low computational cost, providing a reliable curve basis for the subsequent calculation of differential capacity entropy.

[0023] The logic for calculating the differential capacity entropy based on the charging curve is as follows: A preset voltage range is extracted from the charging curve, and equally spaced voltage points are generated within this preset voltage range. The cumulative charging capacity corresponding to each voltage point is calculated by reverse calculation of the charging curve. The differential cumulative charging capacity within each adjacent voltage point range is calculated, and all differential cumulative charging capacities are normalized to obtain a probability distribution. The information entropy is calculated based on the probability distribution and used as the differential capacity entropy.

[0024] The formula used to further calculate the differential capacity entropy of the current effective segment based on the probability distribution is: in, in, This represents the differential capacity entropy of the currently valid segment; Indicates the first The normalized difference capacity corresponding to each interval; Indicates the first The cumulative charging capacity corresponding to the right end of each voltage range; Indicates the voltage range index; Indicates the first The cumulative charging capacity corresponding to the left end of each voltage range; This represents the total number of equally spaced voltage points generated within a preset voltage range; Indicates the first The proportion of the differential charging capacity corresponding to each voltage range to the total differential charging capacity within the entire preset voltage range.

[0025] In the above process, the dependent variable The differential capacity entropy, representing the current effective segment, characterizes the uniformity of capacity increment distribution during constant current charging. Its magnitude directly reflects the consistency and orderliness of the phase transition process of electrochemical active materials within the battery. A larger value indicates that the differential cumulative charging capacity is more evenly distributed across voltage ranges, and the probability distribution tends to be flat. This suggests that the electrochemical phase transition process inside the battery is dispersed over a wide voltage range without obvious local concentration characteristics. When... When the value is small, it means that the differential cumulative charging capacity is concentrated in a few specific voltage ranges, and the probability distribution is peak-shaped. This indicates that there is a concentrated phase change or side reaction behavior in the battery near a specific voltage point. This feature is closely related to aging mechanisms such as structural degradation of electrode active materials and obstruction of lithium ion insertion and extraction paths. Independent variable Indicates the first The proportion of the differential cumulative charging capacity corresponding to each voltage range to the total differential cumulative charging capacity within the entire preset voltage range is the normalized probability distribution value. The value of directly determines Size: When each voltage range When the distribution is relatively uniform, Larger; when When concentrated in a few intervals, The capacity is relatively small. From a physical perspective, differential cumulative charging capacity essentially reflects the amount of charge added to the battery per unit voltage change, i.e., the capacity increment. It is directly related to the phase transition process inside the battery, the structural transformation of the active materials, and the lithium-ion insertion / extraction behavior. As the battery ages, the electrode active materials gradually lose their lithium insertion ability, and some phase transition peaks weaken or disappear, leading to a change in the probability distribution of the differential cumulative charging capacity. The uniformity changes accordingly, ultimately reflected in The numerical change of entropy value makes it a sensitive indicator of battery health. This formula defines differential capacity entropy in the form of information entropy. Its logic lies in the fact that information entropy can effectively quantify the uniformity of the probability distribution, without being limited by the specific form of the probability distribution. For a differential cumulative charging capacity distribution consisting of multiple voltage ranges, a higher entropy value indicates a more uniform distribution of capacity increments on the voltage scale, while a lower entropy value indicates that capacity increments are more concentrated in a specific voltage range. This characteristic makes... It can effectively capture the decay and drift of peak capacity increments during battery aging. Regarding its formal rationality, the formula... Define a method to ensure that the sum of the normalized differential capacities across all voltage ranges is 1, satisfying the basic mathematical constraints of the probability distribution, such that... The value of has a clear physical boundary and depends only on The relative distribution rather than the absolute value ensures that the entropy values ​​calculated for the same battery at different charging rates or different sampling frequencies are still comparable, thus enabling a stable characterization of the degradation trend of battery health.

[0026] In the above embodiments, 20 sets of data are provided, showing the ratio of the cumulative charging capacity corresponding to the right end of the voltage range to the total differential charging capacity, to reflect how the ratio changes with the cumulative charging capacity corresponding to the right end of the voltage range, as shown in Table 1: Table 1: Relationship between the proportion of differential charging capacity to total differential charging capacity and cumulative charging capacity. In Table 1 above, the preset voltage range is set to 3.0V to 4.2V, and the number of equally spaced voltage points is: Total differential charging capacity: The difference capacity of each interval divided by the total difference capacity of 100.0 gives the result. According to the instruction manual Figure 2As can be seen, the differential capacity ratio exhibits a typical unimodal distribution, peaking near the 10th to 11th interval, gradually decreasing at both ends. Furthermore, the differential capacity intervals are no longer equidistant, reflecting the non-uniform distribution of capacity increment with voltage during actual battery charging. The impedance phase dispersion within the current effective segment is extracted, specifically including: Apply sliding time windows to the total current sequence and terminal voltage sequence within the current effective segment and perform fast Fourier transform to obtain the current spectrum and voltage spectrum corresponding to each frequency point within each time window, and then calculate the impedance phase corresponding to each frequency point. The variance of all impedance phases relative to their overall average value is calculated for all time windows and all frequency points, and this variance is used as the impedance phase dispersion.

[0027] Calculate the impedance phase dispersion using the following formula: in, Indicates the impedance phase dispersion; Indicates all The arithmetic mean; Represented as the first The first time window, the first The impedance phase corresponding to each frequency point; Index for sliding time windows; This represents the frequency index corresponding to the Fast Fourier Transform; Indicates the total number of frequency points; Indicates the total number of time windows; Indicates the first The first time window, the first The voltage spectrum corresponding to each frequency point; Indicates the first The first time window, the first The current spectrum corresponding to each frequency point; This indicates taking the imaginary part of a complex number; This indicates taking the real part of a complex number.

[0028] In the above process, the dependent variable This represents the impedance phase dispersion within the current effective segment, used to characterize the degree of impedance phase fluctuation of the battery cluster at different time windows and frequency points during charging. Its magnitude directly reflects the stability of the consistency of the electrochemical processes inside the battery. A large value indicates a significant difference in impedance phase between different time windows and frequency points, suggesting that the dynamic characteristics of the electrochemical reaction inside the battery exhibit significant time-varying and frequency-dependent variations. This may be related to the instability of the electrode interface state, localized degradation of the solid electrolyte interface film, or differential changes in the lithium-ion transport pathway. When the value is small, it means that the impedance phase remains highly consistent across different time windows and different frequency points, indicating that the electrochemical process inside the battery has good stability and consistency, and the battery is in good health. Influence The core independent variable is the impedance phase at each frequency point. It is the overall arithmetic mean across all time windows and all frequency points. The degree of deviation determines Size. The voltage-current spectrum is determined by the arctangent function of the ratio of the imaginary to the real part of the voltage spectrum to the current spectrum. Essentially, it reflects the degree of phase lag between the voltage response and the current excitation at a specific frequency. From a physical perspective, when a battery is in a healthy state, the kinetic parameters of its internal electrochemical processes, such as charge transfer resistance, diffusion coefficient, and SEI film impedance, remain consistent across different time windows. Therefore, the impedance phase at each frequency is relatively concentrated. The resistance is relatively low; however, as the battery ages, degradation effects such as loss of electrode active materials, electrolyte decomposition, and lithium salt consumption increase charge transfer resistance and hinder lithium-ion diffusion. Different frequencies exhibit varying sensitivities to these degradation effects, leading to differential changes in the phase response of diffusion impedance in the low-frequency range and charge transfer impedance in the mid-to-high-frequency range. This, in turn, exacerbates the dispersion of impedance phase across different time windows and frequency points. Increase. Therefore By quantifying the degree of fluctuation in impedance phase, the degree of uniform degradation of internal electrochemical parameters of the battery can be indirectly reflected. The phase calculation stage employs a four-quadrant arctangent function, whose output range covers... It can automatically determine the quadrant where the impedance phase lies based on the signs of the real and imaginary parts of the complex ratio, avoiding the limitation of ordinary arctangent functions that can only output... The resulting phase information folding or loss problem is addressed by ensuring that the impedance phase calculated at different frequencies and under different battery conditions has complete angular information and correct physical meaning. This is achieved by first calculating the overall average impedance phase across all time windows and frequencies, and then calculating the mean of the squared deviations of each phase value from this average. It can comprehensively reflect the fluctuation characteristics of impedance phase in both the time domain (different time windows) and the frequency domain (different frequency points). Regarding the rationality of its form, the factors in this formula... make sure The average of all deviations, rather than their sum, makes... The value of is not directly affected by the total number of time windows and the total number of frequency points, ensuring that the dispersion calculated under different data lengths or different fast Fourier transform parameter settings is comparable, thus enabling it to serve as a consistent evaluation index for battery health status. In practical engineering applications, the sliding time window can be determined by performing a short-time Fourier transform with a time window length of 300 seconds. Within this window length, frequency components up to 300Hz can be included. A new time window is generated by sliding with a step size of one second. Each Fourier transform can obtain multiple frequency components containing real and imaginary parts.

[0029] S3: Calculate the corresponding temperature change rate based on the temperature of each battery cell, and determine the thermal consistency index based on the dispersion between the temperature change rates of each battery cell; use the differential capacity entropy, impedance phase dispersion and thermal consistency index as health features, and construct a health feature matrix according to the time sequence of the effective segments; In this embodiment, generating the thermal consistency index of the current valid fragment specifically includes: For the temperature of each battery cell at each sampling point within the current valid segment, the corresponding temperature change rate is calculated using first-order backward difference. For each sampling point, calculate the sum of squares of the temperature differences between all adjacent units, take the average and then take the square root to obtain the discrete value of the temperature change rate at that sampling point. The thermal consistency index is obtained by averaging the discrete values ​​of the temperature change rate of all sampling points within the current valid segment.

[0030] Calculate the thermal uniformity index of the battery cluster using the following formula: in, This indicates the thermal uniformity index of the battery cluster; Indicates the index of the sampling point in the current valid segment; This indicates the total number of sampling points in the current valid segment; Indicates the total number of monomers; Indicates a single-entity index; Indicates the first segment within the currently valid segment. The sampling point of the first sampling point The temperature of each monomer; Indicates the first segment within the currently valid segment. The sampling point of the first sampling point The temperature of each monomer.

[0031] In the above process, the dependent variable The thermal uniformity index represents the overall level of the dispersion of temperature change rates among individual cells within the battery cluster, and its value directly reflects the quality of thermal behavior uniformity within the battery cluster. A large value indicates a significant difference in the temperature change rate between individual cells, suggesting a marked non-uniformity in the heating rate and heat dissipation conditions of different cells within the battery cluster. This may be caused by factors such as differences in internal resistance between battery cells, inconsistent connection impedance, blocked heat dissipation channels, or localized hotspot effects. When... When the value is small, it means that the temperature change rate between each cell is highly consistent, indicating that the thermal behavior of each cell in the battery cluster has good synchronicity and uniformity, and the overall thermal management state of the battery cluster is relatively ideal. Influence The core independent variable is the temperature difference between adjacent cells at each sampling time. The discrete value obtained by taking the square root of the sum of squares directly determines the... The value of this discrete value is [value]. The larger the discrete value, the [value]. The larger the current, the smaller the temperature change rate; conversely, the smaller the current, the smaller the temperature change rate. From a physical mechanism perspective, the temperature change rate of a single battery cell during charging and discharging is directly related to the Joule heat generated per unit time and the heat of reaction absorbed or released. The magnitude of Joule heat depends on the product of the square of the current and the battery's internal resistance, while the intensity of the heat of reaction is closely related to the electrochemical reaction rate and entropy change. When there are differences in the internal resistance, state of charge, or aging degree of individual cells within a battery cluster, under the same charging and discharging current excitation, the heat generation rate of each cell will be different, leading to a discrete temperature change rate. Consequently, the performance difference between individual cells increases. As battery aging intensifies, the performance differentiation between individual cells becomes increasingly significant. It shows a gradual upward trend, so the index can effectively reflect the degradation process of the consistency of thermal behavior of each cell in the battery cluster. This formula defines the thermal consistency index using the mathematical form of root mean square error (RMSE). Its logic lies in the fact that RMSE can effectively quantify the dispersion of a set of data relative to an ideal consistency state. For each sampling point, the discrete value of the temperature change rate at that sampling point is obtained by calculating the sum of squares of the temperature differences between all adjacent cells, taking the square root of the average, and then obtaining the discrete value. This operation first eliminates the problem of positive and negative values ​​canceling each other out by squaring the differences between adjacent cells, and then restores the dimensions by taking the square root, so that the discrete value has the same physical unit as the temperature change rate. Subsequently, the discrete values ​​of all sampling points are averaged to obtain the thermal consistency index of the entire current valid segment. This averaging operation can effectively suppress the random deviations caused by instantaneous temperature fluctuations, making the thermal consistency index more consistent with the temperature change rate. It can stably characterize the overall consistency level of the thermal behavior of battery clusters. Regarding formal rationality, the denominator in the formula... The unbiased estimation setting ensures that discrete value calculations are unaffected by the total number of individuals, and the averaging process between sampling points ensures... The value of has a clear physical boundary, which facilitates horizontal comparison between different valid segments.

[0032] S4: Perform trend filtering on each health feature sequence in the health feature matrix to obtain the corresponding degradation trend value and construct a degradation trend matrix; perform standardization and principal component analysis on the degradation trend matrix, fuse the real-time health index based on the contribution of each principal component, and output the corresponding health warning information according to the comparison result between the real-time health index and the preset health threshold.

[0033] The trend filter uses Savitzky. The Golay filter is implemented, and the method for constructing the degradation trend matrix is ​​as follows: Extract each column of the health feature matrix as a health feature sequence; The half-window length and polynomial order of the preset filter are used. With each effective segment position as the center position, polynomial least squares fitting is performed on the health feature sequence values ​​within the window. The value of the fitted polynomial at the effective segment corresponding to the center position is taken as the degradation trend value of the effective segment. The degradation trend values ​​corresponding to the differential capacity entropy, impedance phase dispersion, and thermal uniformity index are combined column-wise to form the degradation trend matrix. The preset half-window length of the Savitzky-Golay filter is... The order of the polynomial is , and All are preset positive integers, and full... ; In each valid segment At that point, take the continuous path centered on it. A valid fragment, using one A polynomial of order 1 is used to perform least-squares fitting on the feature sequence values ​​at each point within the window, and the polynomial is then taken at the center point. The fitted value at point is used as the smoothing result, and smoothing is performed according to the following formula: in, Indicates the effective segment offset from the center point The convolution coefficients at each position, Indicates the first The degradation trend value of each valid segment; Indicates the first The feature sequence of each effective segment; Indicates the length of half a window; This represents the position index of the valid segment offset from the center point; Indicates the index of a valid fragment; Convolution coefficients The rules for determining it are as follows: in, express The matrix whose first... line, number Column elements are ,Right now ,in , ; Representation matrix transpose; Indicates the polynomial order index; Indicates the total order of the polynomial; For length is The unit vector, only in the first... One position is 1, and the rest are 0; Let represent the row vector of convolution coefficients, satisfying the normalization condition. ; when When the range exceeds [1,N], the sequence is completed using a symmetric mirror extension method, where N represents the total number of valid segments.

[0034] In the above process, the number of effective segment points covered by the filter width window is typically between 5 and 15, corresponding to a half-window length. The value is 2 to 7, while ensuring that Greater than the order of the polynomial To ensure that the least squares fitting problem has a solution; polynomial order Typically set to a value between 2 and 4, but if more complex trends need to be captured, the value should be adjusted while still meeting the window width requirement. Greater than Under the premise of [the above], it can be appropriately increased; The specific implementation of least squares fitting is as follows: at each effective segment position t, take the continuous segment centered at that position. Given a sequence of health feature values, construct a P-order polynomial. ,in The polynomial represents the offset of each position within the window relative to the center position at each offset. The value at that point should be as close as possible to the original health characteristic sequence value. The goal of least squares fitting is to determine the polynomial. coefficients This minimizes the sum of squared residuals between the predicted and original values ​​of the polynomial at each position within the window, i.e. Obtain the minimum value; after obtaining the polynomial coefficients, adjust the offset corresponding to the center position. Substituting into the polynomial, the result is the fitted value of the polynomial at the center point, which is used as the smoothed degradation trend value. The determination rule originates from the core derivation logic of the Savitzky-Golay filter, which solves for the closed-form of the convolution coefficients through least-squares fitting. Specifically, it involves calculating the offsets within the window. Powers form a matrix Its elements are Then the polynomial fitting problem is transformed into a linear least squares problem, and the coefficient vector corresponding to the least squares solution can be obtained from... We obtain, among which This is a column vector composed of the feature sequence values ​​at each position within the window. Since only the fitted value at the center point (i.e., ...) is ultimately needed... The value of the polynomial at the point where = 0 is the constant term in the polynomial representation. ,and Through unit vector Extracted from the coefficient vector, i.e. .Will The whole thing is viewed as a vector of convolution coefficients. The smoothing result can then be further expressed as Convolution coefficients Dot product with the feature sequence values ​​within the window; The specific completion method of symmetric mirror continuation is: when a certain position Out of range of valid indexes for the sequence When the missing position is considered as having the nearest sequence boundary as its axis of symmetry, the healthy feature sequence value at the corresponding position inside the boundary is taken as the completion value; that is, when When, the padding value is That is, by index As a symmetrical point, map the missing position on the left to the corresponding position on the right; when When, the padding value is That is, by index Using the symmetrical point as an example, map the missing position on the right to the corresponding position on the left.

[0035] The method for generating the real-time health index is as follows: The degradation trend matrix is ​​standardized column by column, and each standardized row is used as a degradation trend feature vector. Calculate the covariance matrix of the standardized degradation trend matrix, and perform eigenvalue decomposition on the covariance matrix, taking the eigenvector corresponding to the largest eigenvalue as the principal component direction; In the above process, the covariance matrix is ​​calculated as follows: Let the standardized degradation trend matrix be... Its dimensions are ,in This represents the total number of valid segments. Each column corresponds to a health characteristic, namely, the standardized degradation trend sequence of differential capacity entropy, impedance phase dispersion, and thermal uniformity index. Since each column has been standardized column-by-column, the covariance matrix... Can be directly from The calculated dimension is The diagonal elements represent the variance of each health trait, while the off-diagonal elements represent the covariance among different health traits, reflecting the synergistic relationship between the degradation trends of various health traits. Eigenvalue decomposition is then performed on this covariance matrix to solve for... Three eigenvalues ​​were obtained. and its corresponding eigenvectors The largest eigenvalue Corresponding feature vector This is the direction of the first principal component; this principal component direction represents the direction with the largest variance in the standardized degradation trend matrix, that is, the dominant trend direction in which the three health characteristics change most significantly in synergy during battery degradation. The degradation trend feature vector of the current effective fragment is projected onto the direction of the principal component to obtain the instant health index; When the real-time health index of a preset number of consecutive valid segments exceeds the preset health threshold, an early warning message is output.

[0036] The real-time health index of the current valid segment is calculated using the following formula: in, This represents the eigenvector corresponding to the largest eigenvalue obtained by eigenvalue decomposition of the covariance matrix; This indicates the real-time health index of the currently valid segment; This represents the degradation trend feature vector after standardization of the current valid fragment; In the above process, the dependent variable The instantaneous health index represents the current effective segment and is used to characterize the overall health status of the battery cluster at the current moment. Its value directly reflects the degree of comprehensive degradation of the battery cluster in three dimensions: electrochemical characteristics, impedance characteristics, and thermal behavior. When the value is small, it means that the projection value of the degradation trend feature vector of the battery cluster onto the principal component direction is low, indicating that the differential capacity entropy, impedance phase dispersion, and thermal consistency index of the current battery cluster are all at a low level, and the battery cluster is in good health; when When the value is large, it means that the battery cluster shows a significant degradation trend in all three health characteristic dimensions, indicating that the loss of electrode active materials inside the battery cluster is aggravated, the consistency of electrochemical processes is worsened, the uniformity of thermal behavior is deteriorated, and the battery cluster as a whole is in a high health risk state. Influence The core independent variable is the standardized degradation trend feature vector. It includes standardized degradation trend values ​​of three health characteristics: differential capacity entropy, impedance phase dispersion, and thermal uniformity index. Depend on In the principal component direction The magnitude of the projection on the surface is determined by the physical mechanism. An increase or decrease in differential capacity entropy reflects a change in the phase transition behavior of the electrode active material; an increase in impedance phase dispersion reflects a deterioration in the uniformity of the electrochemical process; and an increase in the thermal uniformity index reflects a deterioration in the uniformity of thermal distribution within the battery cluster. The greater the degradation trend values ​​in these three dimensions, the more... The larger the projection value along the principal component direction, The larger the value, the more likely it is to indicate the direction of the principal component. The eigenvalues ​​of the covariance matrix of the standardized degradation trend matrix are obtained, representing the main coordinated change directions of the three health characteristic sequences in a statistical sense. It can comprehensively reflect the degree of simultaneous degradation of the three health characteristic dimensions, thereby effectively characterizing the overall health risk level of the battery cluster; The formula uses linear projection to reduce the dimensionality of a multidimensional degradation trend feature vector into a single real-time health index. Its logic lies in the fact that principal component analysis can extract the main direction of change in multivariate data, projecting sample points in the high-dimensional feature space onto this direction, thereby achieving dimensionality compression with minimal information loss. Regarding formal rationality, the formula... The eigenvector corresponding to the largest eigenvalue of the covariance matrix represents the direction of the greatest variation in the three healthy feature sequences, i.e., the dominant trend direction of the co-evolution of the three features during battery cluster degradation. The standardized degradation trend eigenvector... Project to Above, making In a statistical sense, it maximizes the preservation of the differences in the original three-dimensional features during the degradation process, making the index the most sensitive to changes in the health status of battery clusters.

[0037] When the real-time health index of a number of consecutive valid segments exceeds the preset health index threshold, it indicates that the current battery cluster is in a dangerous state and an alarm is issued.

[0038] In the above process, the value of the consecutive preset number is usually set according to the total number of valid segments. The total number of valid segments reflects the data accumulation of the battery under constant current charging conditions. For scenarios where the total number of valid segments is sufficient, the consecutive preset number can be set to an integer between 3 and 5. When the total number of valid segments is limited, this value can be reduced accordingly. The specific method for determining the preset health index threshold is as follows: During the initial operation of the energy storage power station or after the battery clusters have undergone comprehensive maintenance and confirmed to be in a healthy state, collect real-time health index calculation values ​​for several valid segments to establish a baseline dataset under healthy conditions. Perform statistical analysis on this baseline data and calculate its mean. and standard deviation The preset health index threshold can be set to ,in This is the multiplier, which can be selected according to the desired confidence level. It is usually a value between 3 and 5. For example, a value of 3 corresponds to a confidence level of 99.7%.

[0039] Please see Figure 3 The present invention also provides a large-scale energy storage power station health management system, which is used to execute the above-described large-scale energy storage power station health management method, including: The data acquisition module is used to collect the total current, terminal voltage and temperature of each battery cell of the battery cluster at a preset sampling frequency, and to identify the constant current charging state based on the total current, and to determine the continuous charging period that meets the preset stability conditions as the effective segment. The parameter calculation module is used to determine the cumulative charging capacity at each sampling time based on the total current and terminal voltage for the effective segment, construct a charging curve corresponding to the cumulative charging capacity and terminal voltage, and calculate the differential capacity entropy based on the charging curve; at the same time, it performs time-frequency analysis on the total current and terminal voltage to extract the impedance phase dispersion. The matrix construction module is used to calculate the corresponding temperature change rate based on the temperature of each battery cell, and to determine the thermal consistency index based on the dispersion between the temperature change rates of each battery cell; using the differential capacity entropy, impedance phase dispersion and thermal consistency index as health features, a health feature matrix is ​​constructed according to the time order of the effective segments. The health assessment module is used to perform trend filtering on each health feature sequence in the health feature matrix to obtain the corresponding degradation trend value and construct the degradation trend matrix; to perform standardization and principal component analysis on the degradation trend matrix, to obtain an instant health index based on the contribution of each principal component, and to output corresponding health warning information based on the comparison result of the instant health index and the preset health threshold.

[0040] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution.

[0041] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0042] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A health management method for large-scale energy storage power stations, characterized in that, The specific steps include: S1: Collect the total current, terminal voltage and temperature of each battery cell of the battery cluster at a preset sampling frequency, and identify the constant current charging state based on the total current, and determine the continuous charging period that meets the preset stability conditions as the effective segment. S2: For the effective segment, determine the cumulative charging capacity at each sampling time based on the total current and terminal voltage, construct a charging curve corresponding to the cumulative charging capacity and terminal voltage, and calculate the differential capacity entropy based on the charging curve; at the same time, perform time-frequency analysis on the total current and terminal voltage to extract the impedance phase dispersion. S3: Calculate the corresponding temperature change rate based on the temperature of each battery cell, and determine the thermal consistency index based on the dispersion between the temperature change rates of each battery cell; use the differential capacity entropy, impedance phase dispersion and thermal consistency index as health features, and construct a health feature matrix according to the time sequence of the effective segments; S4: Perform trend filtering on each health feature sequence in the health feature matrix to obtain the corresponding degradation trend value and construct a degradation trend matrix; perform standardization and principal component analysis on the degradation trend matrix, fuse the real-time health index based on the contribution of each principal component, and output the corresponding health warning information according to the comparison result between the real-time health index and the preset health threshold.

2. The health management method for large-scale energy storage power stations according to claim 1, characterized in that, The identification logic for the effective segment is as follows: the total current of the battery cluster is collected at fixed sampling times. Within a preset sliding window, if the deviation of the total current from the preset constant current charging setting value is always within the preset allowable deviation range, and the duration of this deviation exceeds the preset minimum duration, then the state is determined as a constant current charging state, and the continuous charging period corresponding to the constant current charging state is marked as an effective segment.

3. The health management method for large-scale energy storage power stations according to claim 1, characterized in that, Constructing a charging curve corresponding to the cumulative charging capacity and terminal voltage, specifically including: The total current of each sampling point within the current valid segment is integrated using the rectangular method to obtain the cumulative charging capacity corresponding to each sampling point. Using the cumulative charging capacity and terminal voltage of each of two adjacent sampling points, the adjacent sampling points are connected sequentially with straight line segments in the plane formed by the cumulative charging capacity and terminal voltage through linear interpolation to form the charging curve.

4. The health management method for large-scale energy storage power stations according to claim 1, characterized in that, The logic for calculating the differential capacity entropy based on the charging curve is as follows: A preset voltage range is selected from the charging curve, and equally spaced voltage points are generated within the preset voltage range. The cumulative charging capacity corresponding to each voltage point is then calculated by reverse calculation using the charging curve. Calculate the differential cumulative charging capacity within each adjacent voltage point interval, and normalize all differential cumulative charging capacities to obtain a probability distribution; calculate the information entropy based on the probability distribution, and use it as the differential capacity entropy.

5. The health management method for large-scale energy storage power stations according to claim 1, characterized in that, Extracting the impedance phase dispersion within the current valid segment, specifically including: Apply sliding time windows to the total current sequence and terminal voltage sequence within the current effective segment and perform fast Fourier transform to obtain the current spectrum and voltage spectrum corresponding to each frequency point within each time window, and then calculate the impedance phase corresponding to each frequency point. The variance of all impedance phases relative to their overall average value is calculated for all time windows and all frequency points, and this variance is used as the impedance phase dispersion.

6. The health management method for large-scale energy storage power stations according to claim 1, characterized in that, The thermal consistency index of the currently valid fragment is generated, specifically including: For the temperature of each battery cell at each sampling point within the current valid segment, the corresponding temperature change rate is calculated using first-order backward difference. For each sampling point, calculate the sum of squares of the temperature differences between all adjacent units, take the average and then take the square root to obtain the discrete value of the temperature change rate at that sampling point. The thermal consistency index is obtained by averaging the discrete values ​​of the temperature change rate of all sampling points within the current valid segment.

7. The health management method for large-scale energy storage power stations according to claim 1, characterized in that, The trend filter uses Savitzky. The Golay filter is implemented, and the method for constructing the degradation trend matrix is ​​as follows: Extract each column of the health feature matrix as a health feature sequence; The half-window length and polynomial order of the preset filter are used. With each effective segment position as the center position, polynomial least squares fitting is performed on the health feature sequence values ​​within the window. The value of the fitted polynomial at the effective segment corresponding to the center position is taken as the degradation trend value of the effective segment. The degradation trend matrix is ​​formed by combining the degradation trend values ​​corresponding to the differential capacity entropy, impedance phase dispersion, and thermal uniformity index column by column.

8. The health management method for large-scale energy storage power stations according to claim 7, characterized in that, The method for generating the real-time health index is as follows: The degradation trend matrix is ​​standardized column by column, and each row after standardization is used as a degradation trend feature vector. Calculate the covariance matrix of the standardized degradation trend matrix, and perform eigenvalue decomposition on the covariance matrix, taking the eigenvector corresponding to the largest eigenvalue as the principal component direction; The degradation trend feature vector of the current effective fragment is projected onto the direction of the principal component to obtain the instant health index; When the real-time health index of a preset number of consecutive valid segments exceeds the preset health threshold, an early warning message is output.

9. A health management system for large-scale energy storage power stations, characterized in that: The large-scale energy storage power station health management system is used to execute the large-scale energy storage power station health management method according to any one of claims 1-8, including: The data acquisition module is used to collect the total current, terminal voltage and temperature of each battery cell of the battery cluster at a preset sampling frequency, and to identify the constant current charging state based on the total current, and to determine the continuous charging period that meets the preset stability conditions as the effective segment. The parameter calculation module is used to determine the cumulative charging capacity at each sampling time based on the total current and terminal voltage for the effective segment, construct a charging curve corresponding to the cumulative charging capacity and terminal voltage, and calculate the differential capacity entropy based on the charging curve; at the same time, it performs time-frequency analysis on the total current and terminal voltage to extract the impedance phase dispersion. The matrix construction module is used to calculate the corresponding temperature change rate based on the temperature of each battery cell, and to determine the thermal consistency index based on the dispersion between the temperature change rates of each battery cell; using the differential capacity entropy, impedance phase dispersion and thermal consistency index as health features, a health feature matrix is ​​constructed according to the time order of the effective segments. The health assessment module is used to perform trend filtering on each health feature sequence in the health feature matrix to obtain the corresponding degradation trend value and construct the degradation trend matrix; to perform standardization and principal component analysis on the degradation trend matrix, to obtain an instant health index based on the contribution of each principal component, and to output corresponding health warning information based on the comparison result of the instant health index and the preset health threshold.

Images