SOC-SOH joint estimation method and system based on lithium battery expansion force

Abstract

The application provides a SOC-SOH joint estimation method based on lithium battery expansion force, comprising the following steps: acquiring voltage, current and expansion force data of a lithium battery cell in real time; calculating expansion force change and expansion force first derivative; inputting voltage, current, expansion force change, expansion force first derivative and SOH estimation value of the last charge-discharge cycle as multi-modal data into a first mapping function to obtain an SOC estimation value in real time; and after the current charge-discharge cycle ends, locating a target inflection point of expansion force change in the charge-discharge cycle, extracting an inflection point key characteristic value related to SOH, inputting the inflection point key characteristic value into a second mapping function to obtain an SOH estimation value of the current charge-discharge cycle. The application further discloses a SOC-SOH joint estimation system based on lithium battery expansion force. The application can effectively improve the real-time performance and accuracy of SOC and SOH estimation, and is helpful to improve the safety and use efficiency of lithium batteries.

Description

Technical Field

[0001] This invention belongs to the field of lithium-ion battery technology, specifically relating to a joint SOC-SOH estimation technology based on the expansion force of lithium batteries. Background Technology

[0002] As a core component of new energy vehicles, the performance of the battery system directly affects vehicle performance, driving range, and user safety; therefore, ensuring the safety of the battery system is of paramount importance. The Battery Management System (BMS) is the core of ensuring the safety of the battery system in new energy vehicles. The BMS determines the state of the entire battery system by detecting the state of each individual cell (i.e., battery cell) in the battery pack, and implements corresponding control adjustments and strategies based on the state parameters. This enables the management of the charging and discharging of the battery system and each individual cell, ensuring the safe and stable operation of the battery system, while also providing charging and discharging protection for the lithium battery and ensuring that the battery provides optimal energy output to the vehicle.

[0003] State of Charge (SOC) and State of Health (SOH) are two crucial parameters in a battery management system (BMS). SOC represents the ratio of the current remaining capacity to the total usable capacity of the battery when fully charged in its current state; simply put, it's the percentage of remaining capacity. SOH measures the degree of battery degradation due to use and aging, typically expressed as the ratio of the battery's current maximum usable capacity to its factory rated capacity; simply put, it's the percentage of battery health. A complete BMS should have functions such as battery state estimation and detection, overcharge / discharge protection, and battery equalization. The implementation of these functions largely relies on accurate estimation of SOC and SOH. Therefore, improving the accuracy and real-time performance of battery state (SOC and SOH) estimation is of great significance.

[0004] Currently, most Battery Management Systems (BMS) estimate battery state by combining voltage, current, and other signals with corresponding algorithm models. While these methods are simple and low-cost, they fail to reflect the internal operating mechanisms of the battery, resulting in errors and lag in the battery state estimation. Furthermore, as... Figure 1 As shown, electrical signals such as voltage and current have limited sensitivity to changes in battery state, are prone to long plateau regions, and are greatly affected by external environmental interference, making it difficult to accurately reflect the true state of the battery.

[0005] Existing methods for estimating SOC and SOH mainly include open-circuit voltage method, model-based method, and data-driven method. However, these methods all have certain limitations: the open-circuit voltage method requires a long settling time, which is difficult to meet the needs of real-time applications; model-based methods are often complex in structure and have high requirements for parameter accuracy; and data-driven methods rely on a large amount of high-quality data and have insufficient generalization ability when operating conditions change. Summary of the Invention

[0006] To address the aforementioned issues of insufficient accuracy and high delay in estimating battery state (SOC and SOH), this invention proposes a joint SOC-SOH estimation method and system based on lithium battery expansion force.

[0007] The technical solution of the present invention includes the following aspects:

[0008] The first aspect of this invention discloses a joint SOC-SOH estimation method based on lithium battery expansion force, comprising:

[0009] Real-time acquisition of voltage, current, and expansion force data of lithium battery cells;

[0010] Calculate the change in battery expansion force and the first derivative of expansion force during the current charge-discharge cycle;

[0011] The voltage, current, expansion force change, first derivative of expansion force, and SOH estimate of the previous charge-discharge cycle are used as multimodal data and input into a predefined first mapping function to obtain SOC estimate in real time.

[0012] Furthermore, after the current charge-discharge cycle ends, the target inflection point of the expansion force change in the charge-discharge cycle is located based on the expansion force change. Based on the target inflection point, key inflection point feature values ​​related to SOH are extracted. The key inflection point feature values ​​are input into a predefined second mapping function to obtain the SOH estimate of the current charge-discharge cycle.

[0013] As an optional approach, the target inflection point of the expansion force change during the charge-discharge cycle is located based on the expansion force change, specifically including:

[0014] Calculate the second derivative of the change in battery expansion force during the current charge-discharge cycle;

[0015] Points where the second derivative is zero and the sign changes are defined as candidate inflection points;

[0016] The candidate inflection points are screened to obtain the target inflection points corresponding to the electrode phase transition behavior.

[0017] As an alternative approach, candidate inflection points are screened to obtain target inflection points corresponding to the electrode phase transition behavior, specifically including:

[0018] Remove pseudo-inflection points from the candidate inflection points whose slope is less than the curvature slope threshold;

[0019] Calculate the difference in expansion force between several sampling points before and after the remaining candidate inflection points and the candidate inflection points respectively, and retain the candidate inflection points whose difference in expansion force exceeds the difference threshold.

[0020] Determine whether the remaining candidate inflection points are within their corresponding SOC intervals; if so, they are identified as target inflection points.

[0021] As an optional approach, key inflection point feature values ​​related to SOH are extracted based on the target inflection point, specifically including:

[0022] Calculate the inflection point characteristic values ​​between each pair of inflection points among all target inflection points; the characteristic values ​​include one or more of the following: change in expansion force at the inflection point, inflection point time, time interval between inflection points, relative change in expansion force between inflection points, curve slope, local curvature, and area characteristics.

[0023] Correlation analysis was used to select key inflection point features that were highly correlated with SOH from the inflection point feature values.

[0024] As an optional approach, correlation analysis is used to select key inflection point features that are highly correlated with SOH from the inflection point feature values, specifically including:

[0025] The inflection point feature value is paired with the corresponding SOH calibration value of the cycle;

[0026] Calculate the Pearson correlation coefficient between the inflection point eigenvalue and the SOH calibration value;

[0027] If the absolute value of the Pearson correlation coefficient is greater than a set threshold, and the trend of the inflection point characteristic value is basically consistent with the SOH in more than a set number of charge-discharge cycles, then the inflection point characteristic value is considered to be highly correlated with the SOH and is used as a key attenuation factor.

[0028] As an alternative, the first mapping function is obtained by using the multimodal data as input and the measured SOC value as output, through machine learning training or regression analysis; the second mapping function is obtained by using the inflection point key feature value as input and the measured SOH value as output, through machine learning training or regression analysis.

[0029] As an optional approach, the first mapping function can be any one of nonlinear regression, support vector regression, random forest, or neural network; the second mapping function can be any one of linear model, nonlinear regression model, or neural network model.

[0030] A second aspect of the present invention discloses a joint SOC-SOH estimation system based on lithium battery expansion force, comprising:

[0031] The signal acquisition module is used to acquire real-time data on the voltage, current, and expansion force of the lithium battery cell.

[0032] The signal processing module is used to calculate the change in battery expansion force and the first derivative of expansion force during the current charge-discharge cycle.

[0033] The main control module is configured to: calculate the SOC estimate in real time based on the voltage, current, expansion force change, first derivative of expansion force, and SOH estimate of the previous charge-discharge cycle using a predefined first mapping function; and, after the current charge-discharge cycle ends, locate the target inflection point of expansion force change in the charge-discharge cycle based on the expansion force change, extract key inflection point feature values ​​related to SOH based on the target inflection point, and calculate the SOH estimate of the current charge-discharge cycle using a predefined second mapping function.

[0034] A third aspect of the present invention discloses an electronic device, including a processor and a memory; the processor is configured to call a computer program stored in the memory to execute the SOC-SOH joint estimation method based on lithium battery expansion force as described in the first aspect of the present invention or any alternative embodiment thereof.

[0035] A fourth aspect of the present invention discloses a computer-readable storage medium for storing program code for executing the SOC-SOH joint estimation method based on lithium battery expansion force as described in the first aspect of the present invention or any alternative embodiment thereof.

[0036] The present invention has the following beneficial effects:

[0037] This invention leverages the high sensitivity and stability of expansion force signals by integrating them into a Battery Management System (BMS) for joint estimation of State of Charge (SOC) and State of Health (SOH). This overcomes the problems of low sensitivity and susceptibility to interference inherent in relying solely on voltage and current signals. Furthermore, because the expansion force signal can better pinpoint changes in battery state, it enables more accurate and efficient joint estimation of SOC and SOH, contributing to further improvements in the safety and efficiency of lithium batteries.

[0038] This invention overcomes the non-monotonicity of the expansion force signal within the SOC variation period by utilizing the first derivative signal of the expansion force, thus realizing a unique mapping relationship between the expansion force and SOC. By using the special inflection point in the expansion force curve, key characteristic values ​​strongly correlated with SOH are obtained, overcoming the problem that the characteristic values ​​of traditional voltage and current signals are not strongly correlated with SOH.

[0039] This invention uses multimodal signals based on expansion force to achieve joint modeling and dynamic compensation of SOC and SOH, overcoming the lag of passive compensation in traditional methods. It can estimate SOC in real time and SOH in each charge-discharge cycle, effectively improving the real-time performance and accuracy of SOC and SOH estimation, and helping to prevent the safety risks of battery failure in advance.

[0040] Furthermore, because the expansion force signal can stably reflect changes in battery state, this invention has good applicability and generalization ability under various operating conditions. Attached Figure Description

[0041] Figure 1 This is a schematic diagram illustrating the voltage signal changes during the charging and discharging process.

[0042] Figure 2 Flowchart for joint SOC-SOH estimation;

[0043] Figure 3 The curve showing the relationship between PV and PV;

[0044] Figure 4 This is a schematic diagram illustrating the changes in the expansion force signal during charging and discharging.

[0045] Figure 5 The graph shows the change in the expansion force of the battery during 1200 aging cycles.

[0046] Figure 6 The relationship between the degradation factor extracted at the inflection point and the battery's state of harmonic equilibrium (SOH) is given by: Figure 6 (a) shows the change in SOH during battery cycling. Figure 6 (b) to Figure 6 (g) in the figure represents the change of the key degradation factor extracted during battery cycling.

[0047] Figure 7 The relationship between expansion force and SOC during battery charging and discharging is given, where: Figure 7 In the figure, (a) represents the relationship between the change in expansion force and the state of charge (SOC) during the charging process of four discontinuous charge-discharge cycles of the battery. Figure 7 (b) in the figure represents the relationship between the change in expansion force and the state of charge (SOC) during the discharge process of four discontinuous charge-discharge cycles.

[0048] Figure 8 This is a schematic diagram of the SOH estimation results based on voltage characteristics;

[0049] Figure 9 This is a schematic diagram of the SOH estimation results based on the eigenvectors of the inflection point of the expansion force curve;

[0050] Figure 10This is a schematic diagram of the SOC estimation results for parallel experiment 1;

[0051] Figure 11 This is a schematic diagram of the SOC estimation results for parallel experiment 2;

[0052] Figure 12 This is a schematic diagram of the SOC estimation results for parallel experiment 3. Detailed Implementation

[0053] To better understand the present invention, the technical solution of the present invention will be further described below in conjunction with the accompanying drawings and specific embodiments.

[0054] Studies have shown that batteries experience changes in expansion pressure during charging and discharging, stemming from lithium plating and intercalation processes between the positive and negative electrodes. These changes indicate a correlation between expansion pressure and the battery's State of Charge (SOC) and State of Hypothesis (SOH). For SOC estimation, compared to the relatively flat curve of the relationship between open-circuit voltage (OCV) and SOC over a narrow voltage range, the curve of expansion pressure versus SOC shows more significant changes and higher resolution, allowing for better inference of SOC at different times based on expansion pressure. For SOH estimation, as lithium-ion batteries age, especially near the end of their lifespan, voltage, current, and temperature typically do not change significantly. However, internal physicochemical processes can trigger significant changes in expansion pressure. For example, the SEI film on the negative electrode gradually forms and thickens, and the formation of lithium dendrites leads to irreversible electrode expansion, resulting in significant pressure changes during charging, discharging, and operation. Furthermore, under unconventional operating conditions, processes such as dendrite formation, electrode decomposition, and electrolyte vaporization can also cause significant changes in internal pressure. These pressure changes can serve as early warning signals of battery failure. Therefore, incorporating pressure signals into the battery management system (BMS) is of great significance. By monitoring the changes in expansion pressure during charging and discharging in real time, the battery's state of charge (SOC) and state of equilibrium (SOH) can be assessed more accurately, and potential failure risks can be detected early under abnormal conditions, thereby improving the safety and reliability of battery operation. This monitoring method based on mechanical signals effectively complements traditional detection methods that rely solely on voltage and current signals, making the BMS more comprehensive in battery state assessment.

[0055] However, although the use of expansion force signals has many obvious advantages, its use is limited to some extent due to its non-monotonic relationship with SOC and its relatively obvious coupling relationship with SOH.

[0056] The expansion force curve of a lithium battery during charge and discharge contains rich characteristic information. The inflection points on the curve (i.e., the locations where the rate of change of expansion force is significant during battery charge and discharge) are closely related to the electrochemical processes inside the battery. Based on this, this invention overcomes the non-monotonicity problem of the expansion force signal within the SOC change cycle by utilizing the first derivative signal of the expansion force, achieving a unique mapping relationship between expansion force and SOC. Furthermore, by locating and analyzing the mathematical characteristics of these inflection points, the battery's SOH and SOC can be effectively estimated.

[0057] State of Charge (SOC) and State of Health (SOH) exhibit a significant coupling relationship during battery use. Estimating them independently often fails to capture this coupling effect, limiting estimation accuracy. For example, as the battery ages, capacity decline directly impacts the calculated SOC, causing it to deviate from the true value. Therefore, traditional methods struggle to simultaneously meet the requirements for accurate SOC and SOH estimation, and consequently, fail to guarantee the accuracy and reliability of battery state assessment, thus affecting the efficient and safe utilization of the battery.

[0058] Based on this, this invention proposes a joint SOC-SOH estimation method based on the expansion force of lithium batteries. The method calculates the SOH based on the expansion force data of the lithium battery (hereinafter referred to as "battery"), and then combines multi-modal data (expansion force, first derivative of expansion force, voltage, current, and SOH calibration data) to calculate the SOC. This method fuses mechanical signals with voltage, current, and other signals to achieve battery state estimation based on multi-sensor signal fusion.

[0059] like Figure 2 As shown in the figure, this invention discloses a joint SOC-SOH estimation method based on lithium battery expansion force, which mainly includes the following parts:

[0060] The first part is the signal acquisition section. This section mainly acquires data on the battery's expansion force, voltage, and current. The battery expansion force data can be measured using a pressure sensor. The type of pressure sensor is not limited to load cells, strain gauges, thin-film sensors, or fiber optic sensors; the arrangement of the pressure sensor includes internal implantation and external attachment. This invention does not impose any particular limitations on the above details. Although the type and arrangement of the sensor may affect the accuracy of the expansion force data to some extent, this is not within the scope of this patent.

[0061] In a specific application case, a 1*8 channel ionotropic sensor can be selected as the pressure sensor. Ionotropic sensors have advantages such as high sensitivity, high resolution, and strong anti-interference ability. The ionotropic sensor converts the mechanical signal into a capacitance signal, which can be converted into an analog voltage value by an operational amplifier and integrator circuit, and then the analog voltage value is converted into a digital voltage value by an AD converter, thereby establishing the relationship between the mechanical signal and the output voltage value.

[0062] Using a standard mechanical testing instrument, the PV (pressure-digital voltage) correlation relationship was tested, and the correlation curve between the two was obtained as follows: Figure 3 As shown, the output digital voltage signal has a good linear relationship with the pressure value, so an accurate pressure signal can be obtained based on the digital voltage signal collected from the sensor.

[0063] Battery voltage data can be obtained through a voltage detection chip. For example, the voltage signal across each cell in the battery can be measured in real time using a dedicated voltage detection chip (AFE) acquisition channel.

[0064] Battery current data can be obtained through current detection devices. For example, common current detection devices such as Hall effect current sensors or shunt resistors can be used to obtain current data during battery charging and discharging.

[0065] It should be noted that the parameters mentioned above are all collected from individual lithium-ion batteries, i.e., cells, and the related calculations are also performed on the SOH and SOC of individual cells. For simplicity, an individual cell can be simply referred to as a battery or a cell, and those skilled in the art should understand that this refers to the cells in the battery pack.

[0066] Furthermore, battery temperature data can be acquired. For example, thermistors, thermocouples, and semiconductor temperature sensors can be placed on the battery surface to obtain temperature information during battery operation. Temperature data can also be used as one of the multimodal signals to estimate the State of Charge (SOC).

[0067] The second part is the data processing section. This section mainly deals with the processing of expansion force data.

[0068] Battery state of charge (SOC) estimation can be calculated and monitored in real time throughout the entire charge and discharge process, providing relatively continuous and real-time assessment data. In contrast, battery state of health (SOH) assessment is typically performed very infrequently, perhaps every few dozen cycles or at specific time points. The relevant calculations can be completed by the battery management system (BMS), which pre-stores calculation models for SOC and SOH.

[0069] Understandably, a complete charge-discharge cycle of a lithium battery is defined based on the relationship between the cumulative discharge amount and the battery's rated capacity. That is, when the cumulative discharge amount reaches 100% of the current maximum usable capacity, one cycle is completed. In this embodiment of the invention, SOC=0 is defined as the starting point of a charge-discharge cycle. Therefore, a charge-discharge cycle includes the entire process from starting with the current remaining capacity of 0 (i.e., SOC=0), charging to 100% of the current maximum usable capacity (i.e., SOC=100%), and then discharging back to the current remaining capacity of 0. This process can be simply referred to as "one cycle".

[0070] In this embodiment of the invention, the expansion force signal of the battery is collected in real time, and the change in expansion force of the battery during each charge and discharge cycle is calculated based on this signal. Then, the first and second derivatives of the change in expansion force during the cycle are calculated.

[0071] definition The value of the expansion force at the current time, i.e., time t, and the change in expansion force. This refers to the change in expansion force at the current time, i.e., time t, relative to time SOC=0.

[0072] (1).

[0073] The first derivative of the expansion force can be approximated using the discrete derivative method. During the sampling of the expansion force signal, a pressure sensor can be used to calculate the approximate derivative at fixed time intervals. Record expansion force data And form a discrete sequence ,in The index representing the sampling point is a discrete integer. Indicates the first The expansion force values ​​measured at each sampling point; This represents the time interval between adjacent sampling points (also called the "sampling period"). Therefore, the first derivative of the expansion force can be expressed as:

[0074] (2),

[0075] The second derivative of the expansion force can be expressed as:

[0076] (3).

[0077] After the data processing is complete, the joint estimation of SOC-SOH can begin. The joint estimation of SOC-SOH first requires the estimation of SOH.

[0078] To facilitate understanding, this invention plots an expansion force variation curve based on sampled expansion force data, as shown below. Figure 4 As shown in the figure, the characteristics of each charge-discharge cycle can be intuitively seen from the expansion force change curve, and the inflection point of expansion force change can be found. Then, based on these inflection points, key attenuation factors can be further extracted.

[0079] The expansion force curve of a lithium battery during charging and discharging contains a wealth of characteristic information. Figure 4 The inflection points on the curves shown are closely related to the electrochemical processes inside the battery, mainly including two categories: valley points and peak points. By locating and analyzing the mathematical characteristics of these inflection points, the SOH and SOC of the battery can be effectively estimated.

[0080] In this invention, an inflection point can be specifically defined as the point on the expansion force change curve where the second derivative of the expansion force is zero and its sign changes. The inflection point corresponds to a significant position on the expansion force curve during battery charging and discharging, reflecting a critical characteristic point in the phase transition stage of the electrode material. Specifically, inflection points mainly include two types: peak points formed when the second derivative changes from positive to negative, such as... Figure 4 The blue inflection point; the valley point formed when the second derivative changes from negative to positive, such as... Figure 4 The red inflection points in the diagram reflect the material phase transitions that occur in the electrodes during lithium insertion and extraction, and are important characteristics of the battery's physical state.

[0081] Combination Figure 5 The diagram shows the changes in the expansion force curve of a battery during 1200 aging cycles. The horizontal axis represents the charge-discharge cycle time, and the vertical axis represents the change in expansion force. Different colors represent different cycle numbers. It can be seen that the expansion force curve and its inflection point change significantly during battery aging. Since the inflection point of expansion force is related to different phase transition degrees, as the number of cycles increases, the battery gradually ages, and the phase transition changes, thus changing the inflection point. With battery aging, the position and mechanical magnitude of the inflection point in the expansion force curve undergo quantifiable changes. Therefore, the mathematical characteristics of these inflection points can be used as characteristic parameters to characterize SOH (State of Health), thereby enabling SOH estimation based on changes in expansion force.

[0082] It is worth noting that in actual signal processing, all zero-crossing points that meet the conditions (i.e., from positive to negative and from negative to positive) are first identified by calculating the second derivative, and these points are used as candidate inflection points. Considering that there may be spurious feature points among the candidate inflection points caused by noise or small fluctuations, the slope of the candidate inflection points, the difference in expansion force between the candidate inflection points and the sampling points, and the SOC interval constraints can be further combined to filter from the candidate inflection points and obtain the stable inflection points corresponding to the electrode phase transition behavior, i.e., the target inflection points. By selecting the target inflection points, noise interference can be effectively eliminated, and the true mechanical response of the battery can be characterized more accurately.

[0083] The selection of target inflection points can be accomplished through the following steps: First, for the obtained candidate inflection points, eliminate pseudo-inflection points with small changes based on the slope threshold; then, calculate the difference in expansion force between the remaining candidate inflection points and several sampling points before and after them, and compare this difference with a pre-set difference threshold, retaining only candidate inflection points with significant expansion force changes (i.e., differences exceeding the threshold); finally, determine whether the candidate inflection point is within its corresponding SOC range, thus ultimately determining the target inflection point. This process ensures that the selected inflection point accurately reflects the significant physical changes inside the battery, that is, ensures that the selected target inflection point is stable and corresponds to the electrode phase transition behavior (including the lithium insertion and delithiation phase transition stages), possessing high reliability, thereby obtaining a target inflection point that reflects the true mechanical response of the battery. The slope threshold, difference threshold, and SOC range can be determined based on experiments and requirements.

[0084] The slope threshold can be determined based on the local slope at the candidate inflection point. To determine the slope threshold, a method based on statistical quantiles is used.

[0085] (4),

[0086] In the formula, The slope threshold, for The p-quantile is determined, with p ranging from 0.6 to 0.9, preferably 0.75.

[0087] For example, slope threshold It can be set to 0.005-0.015, with 0.01 being preferred.

[0088] The difference threshold can be determined based on the difference between the change in expansion force at each of the W sampling points before and after the candidate inflection point and at the inflection point. For example, W can be 5 to 30, preferably 10 to 20; the difference threshold can be set to 0.05 to 0.35, preferably 0.2.

[0089] Based on the target inflection points, the eigenvalues ​​between all pairs of inflection points (including adjacent and non-adjacent inflection points) are further calculated, and these inflection point eigenvalues ​​are used as candidate eigenvalues. These eigenvalues ​​typically include: the change in expansion force at the inflection point, the inflection point time, the time interval between inflection points, the relative change in expansion force between inflection points, the curve slope, local curvature, area characteristics, etc. Then, through correlation analysis with SOH, key degradation factors highly correlated with the battery aging process, i.e., key inflection point eigenvalues, are screened from multiple candidate eigenvalues.

[0090] It should be noted that, among these characteristic values, only the change in expansion force at the inflection point and the inflection point time can be directly obtained from the inflection point identification results. The remaining characteristic values ​​can be calculated from relevant data of the expansion force curve. Specifically:

[0091] Relative change in expansion force between inflection points: This represents the magnitude of the change in expansion force between two target inflection points, used to reflect the deformation difference of the electrode in adjacent phase transition stages.

[0092] Curve slope: This refers to the slope of the expansion force curve at the target inflection point, also known as the first derivative. Specifically, it can be obtained by linear fitting based on several sampling points before and after the target inflection point.

[0093] Local curvature: This refers to the local curvature of the expansion force curve at the target inflection point. It can be calculated using the first and second derivatives, and is defined as follows:

[0094] (5).

[0095] Area characteristic: The area defined by the expansion force curve between the inflection points of the two targets can be obtained by integrating the expansion force curve between the inflection points of the two targets.

[0096] (6).

[0097] Among them, the key degradation factor is related to the battery's state of health (SOH). This invention selects six key feature units with the highest correlation, such as... Figure 6 As shown, the value of the key degradation factor exhibits a significant positive correlation with the battery's SOH value during the battery degradation process. Figure 6 (a) shows the change in SOH during battery cycling. Figure 6 (b) to Figure 6 (g) in the diagram represents the changes in the extracted key degradation factors during battery cycling, demonstrating a strong correlation between the key degradation factors and SOH. Therefore, the selected key degradation factors can be used to estimate SOH, specifically through methods such as multiple regression and machine learning.

[0098] The selection of key attenuation factors can be based on the correlation between key attenuation factors and SOH (State of Health), calculated through statistical analysis. Specifically, candidate feature factors extracted from different cycles are first paired with the corresponding SOH calibration values, then the Pearson correlation coefficient is calculated, and correlation analysis is performed using the Pearson correlation coefficient. If the absolute value of the correlation coefficient is greater than a set threshold, and the changing trend of these inflection point feature values ​​in multiple charge-discharge cycles is basically consistent with SOH, then the inflection point feature is considered to have a significant correlation with SOH and can be used as a key attenuation factor. The set threshold can be selected based on experimental data, ranging from 0.80 to 0.95, with a preferred value of 0.95. In this way, the mapping strength between the expansion force inflection point feature and SOH can be quantitatively determined, providing a reliable input variable for the SOH estimation model.

[0099] The Pearson correlation coefficient is calculated as follows:

[0100] (7),

[0101] In the formula, For the first The characteristic values ​​corresponding to each charge-discharge cycle For the first SOH calibration value for each charge-discharge cycle and These are the average values ​​for all charge-discharge cycles during the battery's lifespan. This is the effect number for charge-discharge cycles. When When the value is close to 1, it indicates that the characteristic is strongly correlated with SOH.

[0102] Ultimately, the estimation of SOH relies only on the key inflection point features formed by these target inflection points, rather than all candidate inflection points or all inflection point features of the target inflection points, thus ensuring the robustness and accuracy of the estimation.

[0103] Key feature values ​​of inflection points formed by multiple target inflection points The resulting feature array is represented as follows:

[0104] (8).

[0105] Furthermore, define the SOH estimation function. The estimated SOH values ​​are as follows:

[0106] (9),

[0107] In the formula: The model parameter set; the estimation function It can be a mapping function for nonlinear regression, support vector regression, random forest, neural networks, etc.

[0108] It is worth noting that while the traditional ampere-hour integration method for estimating state of equilibrium (SOH) is simple and easy to implement, it suffers from severe error accumulation, relies on complete charge-discharge cycles, cannot capture complex aging mechanisms, and lacks accuracy and real-time performance. This invention utilizes the inflection point of expansion force change for SOH estimation. It can obtain the inflection point of expansion force change during each charge-discharge cycle, and based on the inflection point information, obtain the estimated SOH value for the current cycle.

[0109] Once the SOH is estimated, the combined SOC-SOH estimation can be performed.

[0110] During charging and discharging, lithium insertion / extraction reactions occur in the active materials of the electrodes, causing volume changes, which manifest as measurable changes in expansion force in the battery structure. This expansion force change exhibits a monotonic or approximately monotonic relationship with the state of charge (SOC) change; therefore, expansion force can be used as an effective indicator for SOC estimation. Since the expansion force change in each charge-discharge cycle is related to the degree of lithiation of the electrodes, the correlation between expansion force change and SOC is strong. Therefore, the expansion force change during each charge-discharge cycle can be used for SOC estimation. Specifically, regression methods, observer methods, and machine learning methods can be used for calculation.

[0111] like Figure 7 As shown, Figure 7 In the figure, (a) represents the relationship between the change in expansion force and the state of charge (SOC) during the charging process of four discontinuous charge-discharge cycles of the battery. Figure 7 (b) in the figure represents the relationship between the change in expansion force and SOC during the discharge process of four discontinuous charge-discharge cycles. It can be seen that as the battery ages, the mapping relationship between expansion force and SOC changes significantly, indicating that in the use of the expansion force signal, its relationship with SOC and its relationship with SOH have a natural coupling in the battery operating mechanism.

[0112] As the State of Harmony (SOH) decreases, the battery's capacity decay and internal resistance changes cause drift in signals such as voltage curves and expansion force response, thus reducing the accuracy of traditional SOC estimation based on signals such as voltage and current. To address this, this invention proposes using the SOH estimate calculated in the previous charge-discharge cycle as one of the input dimensions for SOC estimation in the current charge-discharge cycle. This allows for dynamic calibration and compensation of SOC during the calculation process, improving the accuracy and robustness of SOC estimation.

[0113] Considering the coupling relationship between SOC and SOH, SOH will affect SOC to some extent. Therefore, this invention introduces a function. SOH is used as one of the input dimensions in SOC estimation to obtain the SOC estimate in real time.

[0114] Specifically, the formula for calculating SOC calibration, using voltage, current, changes in expansion force, the first derivative of expansion force, and the estimated SOH value from the previous charge-discharge cycle, is as follows:

[0115] (10)

[0116] In the formula, t represents any sampling time during the charging and discharging process of the lithium battery; This represents the change in expansion force at time t; This represents the battery voltage at time t; This represents the battery current at time t; This represents the estimated SOH value of the previous charge-discharge cycle, used for dynamic compensation of SOC; The first derivative of the expansion force; This represents the set of model parameters, such as regression coefficients, network weights, or other system parameters. This represents the mapping function, which can be a linear model, a nonlinear regression model, or a neural network model, etc.

[0117] It is worth noting that, considering the strong coupling relationship between SOC and SOH, and that changes in SOH directly affect the calculation results of SOC, this invention no longer uses the post-calculation weighted compensation method, but instead uses SOH together with signals such as expansion force, voltage, and current as input dimensions for joint modeling.

[0118] State of Health (SOH) is directly used as a state characteristic in the construction of the State of Charge (SOC) estimation model. This allows the model to utilize both battery health and operational information during the estimation process, enabling joint modeling and coupled estimation of SOC and SOH. This allows the two to constrain and complement each other during the estimation process, thereby effectively improving the model's ability to represent battery states under complex operating conditions and enhancing the accuracy and robustness of the estimation results.

[0119] Based on the above description, the specific calculation process of this invention is as follows:

[0120] Step 1, calculate SOH: ;

[0121] Step 2, calculate SOC: .

[0122] It is important to note that the calculation of SOH and SOC is divided into real-time and periodic phases:

[0123] Real-time phase (at each sampling time t):

[0124] Get ,calculate ;

[0125] and Once the data is input into the model, the compensated data will be output. .

[0126] Periodic phases (per lap):

[0127] After completing one charge-discharge cycle (i.e., a full charge-discharge cycle), the entire process is analyzed. curve;

[0128] Extracting inflection point features Estimate the current loop ;

[0129] Update the input dimensions for the next round of SOC estimation.

[0130] It should be noted that for brand new batteries that have just left the factory, the SOH value can be set to 100% by default during the SOC estimation process in the first charge-discharge cycle.

[0131] It is worth noting that traditional battery SOC estimation methods typically employ the ampere-hour integration method, which calculates a preliminary SOC value through current integration or voltage characteristics, and then corrects it based on an independently estimated SOH value. However, this method has significant drawbacks. First, since SOH estimation usually relies on data accumulation or capacity testing over a long timescale, SOC correction exhibits a significant lag. Second, traditional methods fail to fully utilize the coupling relationship between SOC and SOH, making it difficult to accurately reflect the instantaneous state of the battery, thus affecting the accuracy and real-time performance of SOC.

[0132] Furthermore, the SOC-SOH joint estimation method proposed in this invention, based on multimodal signal fusion, uses SOH along with signals such as expansion force, voltage, and current as input features, directly participating in the SOC modeling process. This avoids the drawback of traditional methods that require "calculating SOC first and then supplementing SOH." Expansion force, as a key indicator of battery state, has a stronger correlation with SOC and SOH and exhibits higher sensitivity to internal physicochemical changes in the battery. Compared to traditional current integration or voltage characteristic methods, changes in expansion force more accurately reflect real-time changes in the battery's state, making the estimation of SOC and SOH more accurate and real-time.

[0133] This invention detects changes in battery expansion force in real time, first locating the target inflection point from the expansion force curve, extracting key attenuation factors for SOH estimation, and then using SOH as one of the input dimensions for SOC estimation and dynamic compensation, achieving synchronous estimation and coupled modeling of SOC and SOH. Unlike traditional methods, the introduction of the expansion force signal allows the estimation of SOC and SOH to dynamically and in real time reflect changes in battery state, greatly improving the accuracy, real-time performance, and robustness of the estimation, thus providing strong support for accurate battery prediction and safe management.

[0134] In a specific embodiment 1, regression relationship is taken as an example.

[0135] The steps for calculating SOH are as follows:

[0136] Let the inflection point feature vector extracted in the current loop be... ,but:

[0137] (11),

[0138] In the formula, The polynomial regression function calculated for SOH. For feature index, and They represent the first The and the first Inflection point characteristic quantity, For constant terms, The coefficient of the linear term is the constant term. The coefficient of the interaction term. The coefficient of the quadratic term, This indicates that it can be extended to higher-order polynomials.

[0139] Preferably, the function A multinomial regression model is adopted, which includes not only a linear combination of inflection point features, but also quadratic and interaction terms, in order to better capture the nonlinear relationship between SOH and inflection point features.

[0140] The SOC calculation steps are as follows:

[0141] Using expansion force, voltage, current, and SOH as inputs, a polynomial regression model is established to calculate the preliminary SOC:

[0142] (12),

[0143] In the formula, The polynomial regression function calculated for SOC. where are the regression coefficients of each parameter, used to describe the nonlinear relationship between the multimodal input variables and SOC, i=1,2,……,5.

[0144] The matrix formed by the polynomial regression coefficients in the SOH and SOC calculation formulas can be determined using the least squares method. Specifically, the eigenvalues ​​can be calculated based on the terms, as will be understood by those skilled in the art. The matrix expression is as follows:

[0145] (13)

[0146] In the formula, A is the polynomial regression coefficient matrix. The feature matrix formed by the key eigenvalues ​​of the inflection point. The target vector is the measured value of SOH.

[0147] in, It can be represented as follows:

[0148] (14)

[0149] Each row represents a sample, specifically a key feature value at an inflection point within a charge-discharge cycle, with a total of N charge-discharge cycles. Each column represents an input feature, with a total of M feature values ​​(e.g., using...). Figure 6 (The six key attenuation factors); the 1 in the first column is used to solve for the constant term. .

[0150] in, It can be represented as follows:

[0151] (15)

[0152] In the formula, This represents the number of training samples, i.e., the number of charge-discharge cycles.

[0153] In another specific embodiment 2, the estimation of SOC and SOH is modeled using a multilayer perceptron neural network.

[0154] The steps for estimating SOH are as follows:

[0155] Inflection point features extracted from the expansion force curve As input, construct the MLP model:

[0156] (16)

[0157] In the formula, This represents the multilayer perceptron neural network function used for SOH computation. This is the set of parameters for the neural network model, including the weights and biases of each layer.

[0158] The constructed MLP model has an input layer dimension of n (i.e., multimodal feature dimension), and can have 2-4 hidden layers (e.g., 3 layers), with 32-64 neurons per layer. The activation function can be either ReLU or Sigmoid, and the output layer has 1 neuron.

[0159] For example, using the ReLU function as the activation function, its expression is:

[0160] ,

[0161] In the formula, x is the weighted input of the neuron.

[0162] Using Sigmoid activation, the calculation form for each layer is as follows:

[0163] (17)

[0164] In the formula, For activation function, .

[0165] The loss function of the MLP model is:

[0166] (18)

[0167] In the formula, Let M be the loss function and M be the number of samples. No. The true SOH value of each sample. This is the corresponding estimated value.

[0168] The SOC estimation steps are as follows:

[0169] Constructing multimodal input vectors:

[0170] (19)

[0171] The input is fed into another multilayer perceptron neural network, and the estimated SOC is shown below:

[0172] (20)

[0173] In the formula, This represents the multilayer perceptron neural network function used for SOC computation. Represents a multimodal input vector. This represents the set of parameters for a neural network model, including the weights and biases of each layer.

[0174] The input dimension is 5 (expansion force, voltage, current, SOH, and the first derivative of the expansion force). There can be 4-6 hidden layers, with 64-128 neurons per layer. The activation function is ReLU. For example, three hidden layers can be set, with 64, 128, and 256 neurons per layer, respectively. The output layer consists of one neuron, activated using the Sigmoid function.

[0175] The loss function is:

[0176] (twenty one),

[0177] In the formula, Let N represent the loss function, and N represent the number of samples. Indicates the first The true SOC value of each sample Indicates the first The predicted SOC value for each sample.

[0178] During the training process, the Adam optimizer can be used to optimize the model, with the learning rate set to 0.01.

[0179] For example, in a specific experimental case, under the 1C constant current and constant voltage charging and discharging strategy, in a 25-degree Celsius operating environment, and with an external preload of 100 kgf applied to the battery cell, the pressure sensor is installed by external attachment, specifically placed between the battery cell and the clamp and positioned in the middle of the battery cell.

[0180] Voltage, current, and expansion force data are sampled simultaneously, with a sampling frequency of 0.5–2 Hz, preferably 1 Hz. For comparative analysis, data from 200–400 complete charge-discharge cycles of the battery can be used, preferably 300 cycles. It is important to note that the number of cycles should generally be evenly distributed throughout the battery's entire lifespan, for example, from an initial SOH of 100% to an SOH of 80%.

[0181] Among them, for SOH estimation, Figure 8 This is a comparison chart of the fitted curves of the estimated SOH based on voltage-related signals and the fitted curves of the measured SOH. Figure 9 This is a comparison chart of the fitting curves of the SOH estimated value obtained by the estimation method described in this invention and the fitting curves of the measured SOH value. Calculations show that the root mean square error (RMSE) of the SOH estimated value based on the voltage-related signal is 1.430%, and the mean absolute error (MAE) is 0.731%; the root mean square error (RMSE) of the SOH estimated value obtained by the estimation method described in this invention is 0.058%, and the mean absolute error (MAE) is 0.043%.

[0182] The root mean square error (RMSE) measures the dispersion of the prediction error, while the mean absolute error (MAE) characterizes the overall error magnitude. Smaller RMSE and MAE values ​​indicate higher model prediction accuracy. The definitions of RMSE and MAE are as follows:

[0183] (twenty two),

[0184] (twenty three),

[0185] In the formula, i is the sample size. This represents the true SOH (i.e., the measured SOH value). This is the predicted SOH (i.e., the predicted SOH value).

[0186] It is evident that the SOH estimate obtained by the present invention based on the characteristic vector of the inflection point of the expansion force curve has excellent performance. Compared with the traditional electrochemical characteristic method, the estimation error is reduced by at least one order of magnitude.

[0187] Furthermore, based on the SOC estimation method proposed in Embodiment 2 of this invention, a comparative analysis of SOC estimation is performed using different feature combinations. The settings for each combination are as follows:

[0188] Combination 1: Constructing a multi-mode input vector based on traditional electrical signal voltage V and current I And input into a multilayer perceptron neural network , .

[0189] Combination 2: Multimodal input vector Introducing an expansion force signal based on Combination 1 ,get .

[0190] Combination 3: Multimodal input vector Based on combination 2, the first derivative feature of expansion force is further added. ,get .

[0191] Combination 4: The multimodal input vector introduces aging state-related features based on Combination 2. ,get .

[0192] Combination 5: The multimodal input vector adopts the scheme described in this invention, that is, it integrates voltage, current, expansion force, and In addition to multimodal information such as SOH, we obtain .

[0193] The errors in calculating SOC were compared using three parallel experiments under different feature combinations. The comparison results are shown in Table 1 and... Figures 10 to 12 As shown:

[0194] Table 1:

[0195]

[0196] In Table 1, the definitions of Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) have been explained above. During the calculation, it is only necessary to... Replace it with the actual SOC obtained by the volumetric method (i.e., the measured SOC value). Replace with the model-predicted SOC (i.e., the predicted SOC value).

[0197] Taking Experiment 3 as an example, the SOC estimation error using the proposed method decreased significantly. The results show that when only electrical signals are used, the model error is significantly amplified during rate changes, with the MAE and RMSE of Combination 1 reaching 3.528% and 4.578%, respectively. With the expansion force, and aging status information With the gradual introduction of [the method], the SOC estimation error continued to decrease, with MAE decreasing to 1.234% and RMSE decreasing to 1.589% in combination 5. This indicates that the accuracy of the method of this invention is greatly improved compared with the traditional SOC estimation method.

[0198] Furthermore, this invention also discloses a SOC-SOH joint estimation system, which mainly includes a signal acquisition module, a signal processing module, and a main control module. The signal acquisition module is used to implement the data acquisition part of the SOC-SOH joint estimation method, acquiring battery expansion force, voltage, and current data; the signal processing module is used to receive various types of battery data and process the received expansion force data, i.e., to implement the data processing part of the SOC-SOH joint estimation method; the main control module is used to implement the joint estimation part of SOC-SOH, the details of which are not elaborated further.

[0199] Furthermore, the present invention also discloses an electronic device, which mainly includes a processor and a memory. The processor is mainly used to call and execute the computer program stored in the memory to implement the joint estimation method of SOC-SOH in the foregoing embodiments.

[0200] Furthermore, the present invention also provides a computer-readable storage medium for storing program code, which is used to execute the vehicle painted text detection and recognition method described in the foregoing embodiments. Optional computer-readable storage media may include: flash drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks, etc.

[0201] Finally, it should be noted that although the embodiments of the present invention have been described above in conjunction with the accompanying drawings, the present invention is not limited to the specific embodiments and application fields described above. The specific embodiments described above are merely illustrative and instructive, and not restrictive. Those skilled in the art, guided by this specification, can make many other forms without departing from the scope of protection of the claims of the present invention, and all of these are within the scope of protection of the present invention.

Claims

1. A joint estimation method for SOC-SOH based on the expansion force of lithium batteries, characterized in that, include: Real-time acquisition of voltage, current, and expansion force data of lithium battery cells; Calculate the change in battery expansion force and the first derivative of expansion force during the current charge-discharge cycle; The voltage, current, expansion force change, first derivative of expansion force, and SOH estimate of the previous charge-discharge cycle are used as multimodal data and input into a predefined first mapping function to obtain SOC estimate in real time. Furthermore, after the current charge-discharge cycle ends, the target inflection point of the expansion force change in the charge-discharge cycle is located based on the expansion force change. Based on the target inflection point, key inflection point feature values ​​related to SOH are extracted. The key inflection point feature values ​​are input into a predefined second mapping function to obtain the SOH estimate of the current charge-discharge cycle. The target inflection point is an inflection point that corresponds to the phase transition behavior of the electrode. The inflection point is defined as the point on the expansion force change curve during the charge-discharge cycle where the second derivative of the expansion force change is zero and the sign changes.

2. The SOC-SOH joint estimation method as described in claim 1, characterized in that, Based on the aforementioned change in expansion force, the target inflection point of expansion force change during the charge-discharge cycle is located, specifically including: Calculate the second derivative of the change in battery expansion force during the current charge-discharge cycle; Points where the second derivative is zero and the sign changes are defined as candidate inflection points; The candidate inflection points are screened to obtain the target inflection points corresponding to the electrode phase transition behavior.

3. The SOC-SOH joint estimation method as described in claim 2, characterized in that, Candidate inflection points are screened to obtain target inflection points corresponding to electrode phase transition behavior, specifically including: Remove pseudo-inflection points from the candidate inflection points whose slope is less than the curvature slope threshold; Calculate the difference in expansion force between several sampling points before and after the remaining candidate inflection points and the candidate inflection points respectively, and retain the candidate inflection points whose difference in expansion force exceeds the difference threshold. Determine whether the retained candidate inflection point is within its corresponding SOC interval; if so, determine it as the target inflection point.

4. The SOC-SOH joint estimation method as described in claim 1, characterized in that, Based on the target inflection point, key inflection point feature values ​​related to SOH are extracted, specifically including: Calculate the inflection point characteristic values ​​between each pair of inflection points among all target inflection points; the characteristic values ​​include one or more of the following: change in expansion force at the inflection point, inflection point time, time interval between inflection points, relative change in expansion force between inflection points, curve slope, local curvature, and area characteristics. Correlation analysis was used to select key inflection point features that were highly correlated with SOH from the inflection point feature values.

5. The SOC-SOH joint estimation method as described in claim 4, characterized in that, Through correlation analysis, key inflection point features highly correlated with SOH are selected from the inflection point feature values, specifically including: The inflection point feature value is paired with the corresponding SOH calibration value of the cycle; Calculate the Pearson correlation coefficient between the inflection point eigenvalue and the SOH calibration value; If the absolute value of the Pearson correlation coefficient is greater than a set threshold, and the trend of the inflection point characteristic value is basically consistent with the SOH in more than a set number of charge-discharge cycles, then the inflection point characteristic value is considered to be highly correlated with the SOH and is used as a key attenuation factor.

6. The SOC-SOH joint estimation method as described in claim 1, characterized in that, The first mapping function is obtained by taking the multimodal data as input and the measured SOC value as output, through machine learning training or regression analysis; the second mapping function is obtained by taking the inflection point key feature value as input and the measured SOH value as output, through machine learning training or regression analysis.

7. The SOC-SOH joint estimation method as described in claim 6, characterized in that, The first mapping function is any one of nonlinear regression, support vector regression, random forest, and neural network; the second mapping function is any one of linear model, nonlinear regression model, or neural network model.

8. A joint SOC-SOH estimation system based on lithium battery expansion force, characterized in that, include: The signal acquisition module is used to acquire real-time data on the voltage, current, and expansion force of the lithium battery cell. The signal processing module is used to calculate the change in battery expansion force and the first derivative of expansion force during the current charge-discharge cycle. The main control module is configured to: calculate the SOC estimate in real time based on the voltage, current, expansion force change, first derivative of expansion force, and SOH estimate of the previous charge-discharge cycle using a predefined first mapping function; and after the current charge-discharge cycle ends, locate the target inflection point of expansion force change in the charge-discharge cycle based on the expansion force change, extract key feature values ​​of the inflection point related to SOH based on the target inflection point, and calculate the SOH estimate of the current charge-discharge cycle using a predefined second mapping function. The target inflection point is an inflection point that corresponds to the phase transition behavior of the electrode. The inflection point is defined as the point on the expansion force change curve during the charge-discharge cycle where the second derivative of the expansion force change is zero and the sign changes.

9. An electronic device, characterized in that, It includes a processor and a memory; the processor is used to call a computer program stored in the memory to execute the SOC-SOH joint estimation method based on lithium battery expansion force as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium is used to store program code for executing the SOC-SOH joint estimation method based on lithium battery expansion force as described in any one of claims 1 to 7.

Images