A two-dimensional weight-based multi-temperature ocv-soc curve cooperative correction method

Abstract

The application relates to the technical field of batteries and discloses a multi-temperature OCV-SOC curve cooperative correction method based on two-dimensional weights, which comprises the following steps: constructing an initial curve cluster; judging whether the initial curve cluster is corrected; if yes, calculating correction weights of all battery residual capacity points on each OCV-SOC curve in all multi-temperature OCV-SOC initial curve clusters; adjusting the correction weights by using a topological neighborhood aggregation effect to obtain weight coefficients after topological manifold space correction; calculating a correction amount; cooperatively correcting the initial curve cluster and performing shape-preserving smoothing to obtain an OCV-SOC curve after shape-preserving smoothing; and dynamically adjusting bandwidth parameters of a two-dimensional Gaussian weight function by using a variance of an end voltage prediction error and a battery aging degree. The application has improvements in the aspects of curve consistency, correction precision, stability, adaptability, engineering practicability and the like.

Description

Technical Field

[0001] This invention relates to the field of battery technology, and more specifically, to a method for collaborative correction of multi-temperature OCV–SOC curves based on two-dimensional weights. Background Technology

[0002] In a BMS (Battery Management System), SOC (State of Charge) is a core control parameter that directly affects battery charge and discharge control, safety protection, and lifespan. The SOC estimation method based on a first-order RC equivalent circuit model and extended Kalman filter (EKF) is currently the most widely used technique for SOC estimation in mass-produced BMS applications. Its estimation accuracy is highly dependent on the accuracy of the OCV-SOC curve.

[0003] The OCV–SOC relationship of chemical batteries exhibits a significant temperature dependence: the voltage levels of the OCV–SOC curves differ at different temperatures, but the curve shapes (concavity / convexity, inflection points, and slopes) are basically consistent, displaying similar characteristics within the same family. During the offline modeling phase, OCV–SOC curves are typically tested at multiple temperature ranges (e.g., -20℃, 0℃, 25℃, 40℃, 55℃) to form an initial OCV–SOC curve cluster, which serves as a benchmark for SOC estimation.

[0004] However, during actual battery operation, factors such as aging, cycle count, ambient temperature fluctuations, and changes in charge / discharge rate can cause the offline-constructed OCV-SOC curve to gradually deviate from the battery's true characteristics, leading to increased SOC estimation errors (especially in low-temperature and aging scenarios, where the error can rise from 1%~2% to over 5%). Therefore, online correction of the OCV-SOC curve is necessary to maintain the accuracy of SOC estimation.

[0005] Currently, online correction of multi-temperature OCV–SOC curves has become a core technical requirement for high-end BMS. The core challenge lies in how to achieve coordinated correction of multiple temperature curves while ensuring the physical self-consistency of the curves (no crossover, no jump), adapting to the characteristic changes throughout the battery's entire life cycle, and taking into account the low computing power requirements of embedded BMS.

[0006] The existing technical solution is a single-point online correction method for the single-temperature OCV–SOC curve, and its core implementation logic is as follows:

[0007] a) Construct a cluster of initial OCV-SOC curves for multiple temperatures offline. Each curve is stored independently and there is no linkage between them.

[0008] b) During online operation, the battery temperature, terminal voltage, and current are collected in real time, and the SOC is calculated.

[0009] c) When the static conditions are met (current approaches 0, SOC is stable), calculate the deviation between the measured OCV and the model OCV at the corresponding SOC point of the current temperature curve, and directly add the deviation to the SOC point of the current temperature curve to complete the single-point correction of the OCV–SOC curve.

[0010] d) After correction, without smoothing the curve, directly substitute the corrected curve into the first-order RC model for Kalman filter SOC estimation.

[0011] e) The correction parameters (such as the correction range) are fixed, and the effects of battery aging, temperature distance, and SOC distance on the correction effect are not considered, nor are other temperature curves linked for synchronous correction.

[0012] The typical application scenarios for this existing technology are BMS for new energy vehicles and BMS for energy storage systems. Its core purpose is to improve the accuracy of SOC estimation by compensating for deviations in offline curves through online correction. However, it does not solve the problem of coordinated correction of multiple temperature curves, and has obvious technical defects. The above method only achieves single-point correction and does not utilize the similarity of multiple temperature curves for joint correction, which cannot guarantee the consistency of the curve family. Therefore, it cannot meet the requirements for high-precision SOC estimation.

[0013] Specifically, based on the aforementioned existing technical solutions, and considering the physical characteristics of batteries and the actual operating scenarios of BMS, its disadvantages are as follows:

[0014] a) Existing technologies use single-temperature, single-point correction without linking other temperature curves. The OCV-SOC curves at different temperatures lack consistency constraints. After correcting the SOC point at a certain temperature, the OCV at other temperatures with the same SOC point is not adjusted synchronously, which may result in temperature curve intersections. For example, the OCV at low temperatures may be higher than the OCV at high temperatures, which violates the physical law that the battery OCV increases slightly with increasing temperature. When Kalman filtering is used for SOC estimation, temperature switching (such as a vehicle going from cold to hot) will cause SOC jumps, resulting in decreased estimation accuracy.

[0015] b) Existing technologies lack weight constraints, and corrections are directly superimposed. Abnormal measurements (such as sensor noise and transient interference) can lead to excessive single-point OCV corrections, resulting in spikes and abnormal inflection points on the curve. In Kalman filtering, the dOCV / dSOC (the derivative of OCV with respect to SOC) used for voltage prediction changes abruptly, increasing voltage prediction errors, causing Kalman gain to fluctuate, filtering to diverge, and deteriorating SOC estimation stability.

[0016] c) The existing technology has fixed correction parameters (correction range and magnitude), which cannot adapt to the aging characteristics of the battery (the deviation pattern of the battery OCV-SOC curve after aging) or to different operating conditions (low noise in static operating conditions and high noise in dynamic operating conditions). After long-term operation, the correction accuracy gradually decreases, and the SOC estimation error of the aged battery cannot be controlled within 1%, which cannot meet the accuracy requirements of mass production BMS.

[0017] d) The existing technology does not perform conformal smoothing after correction, resulting in local jumps in the corrected curve, dOCV / dSOC exceeding the reasonable range, increased terminal voltage prediction error of the first-order RC model, failure of the measurement update link of Kalman filter, accumulation of SOC estimation bias, and ultimately SOC drift.

[0018] e) Existing technologies lack a closed-loop feedback mechanism, and the correction effect cannot be linked to the SOC estimation accuracy. Even if the corrected curve has a deviation, the correction strategy cannot be adjusted in time. The correction process is in an open-loop state and cannot achieve long-term adaptive optimization.

[0019] The core flaw of the existing technology is that it does not utilize the similarity characteristics of the OCV-SOC curves at multiple temperatures, lacks a multi-temperature linkage correction mechanism, and the correction process has no weight constraints, no conformal processing, and no closed-loop feedback, resulting in poor curve consistency, low accuracy, and weak stability. It cannot adapt to the entire battery life cycle and complex operating conditions, and it is difficult to support the long-term high-precision requirements of first-order RC-EKF SOC estimation.

[0020] In general, the main problems with existing technologies include:

[0021] a) Individual correction of multi-temperature OCV–SOC curve clusters is prone to curve intersection, which violates physical properties.

[0022] b) Single-point correction has no weight constraints and is easily affected by measurement noise and other interferences, which can lead to curve distortion.

[0023] c) Fixed correction parameters cannot adapt to battery aging and changes in operating conditions.

[0024] d) The lack of curve conformal constraints leads to instability in the first derivative (dOCV / dSOC) of the OCV–SOC curve.

[0025] e) Without closed-loop feedback, it cannot achieve adaptive correction in the long term.

[0026] There are currently no effective solutions to the problems in the relevant technologies. Summary of the Invention

[0027] To address the problems in related technologies, this invention proposes a collaborative correction method for multi-temperature OCV–SOC curves based on two-dimensional weights, in order to overcome the aforementioned technical problems existing in the prior art.

[0028] Therefore, the specific technical solution adopted by the present invention is as follows:

[0029] A method for collaborative correction of multi-temperature OCV–SOC curves based on two-dimensional weights includes:

[0030] The open-circuit voltage and remaining charge of the battery at multiple temperatures are obtained to construct a multi-temperature OCV-SOC initial curve cluster. Based on the chaotic nonlinear evolution method, and comparing the Euclidean distance from the current trajectory point in the chaotic phase space to the center of the steady-state attractor with the adaptive conditions, it is determined whether to modify the multi-temperature OCV-SOC initial curve cluster. If so, a two-dimensional Gaussian weight function is constructed based on the temperature distance and the distance of the remaining battery charge, and the modification weights of all remaining battery charge points on each OCV-SOC curve in all multi-temperature OCV-SOC initial curve clusters are calculated. If not, the multi-temperature OCV-SOC initial curve clusters are not modified. The modification weights are adjusted using the topological neighborhood aggregation effect. The weighting coefficients after topological manifold space correction are obtained; the correction amount is calculated based on the measured open-circuit voltage and the model open-circuit voltage in the multi-temperature OCV-SOC initial curve cluster; based on the weighting coefficients and correction amount after topological manifold space correction, the multi-temperature OCV-SOC initial curve cluster is collaboratively corrected to obtain the corrected curve; the corrected curve is then smoothed in a conformal manner to obtain the shape-preserving smoothed OCV-SOC curve; the terminal voltage prediction error is calculated using the shape-preserving smoothed OCV-SOC curve and the first-order RC equivalent circuit model; based on the effective feedback enable condition, and using the variance of the terminal voltage prediction error and the degree of battery aging, the bandwidth parameter of the two-dimensional Gaussian weighting function is dynamically adjusted.

[0031] Furthermore, the open-circuit voltage and remaining capacity of the battery at multiple temperatures are obtained to construct a multi-temperature OCV-SOC initial curve cluster, including: selecting several temperature ranges based on the principle of covering the actual operating temperature range of the battery; and obtaining the remaining capacity and corresponding open-circuit voltage of the battery at each temperature range through a static method combined with current charge-discharge testing, thus forming a multi-temperature OCV-SOC initial curve cluster.

[0032] Furthermore, based on the chaotic nonlinear evolution method, and by comparing the Euclidean distance from the current trajectory point in the chaotic phase space to the center of the steady-state attractor with the adaptive conditions, it is determined whether to correct the initial curve cluster of the multi-temperature OCV-SOC. This includes: constructing a chaotic phase space based on the current, remaining battery charge, voltage, and temperature at a certain moment; determining the center of the steady-state attractor based on the mean vector of the steady-state trajectories in the most recent frames; calculating the Euclidean distance from the current trajectory point in the chaotic phase space to the center of the steady-state attractor and configuring the attractor radius; and determining whether to correct the initial curve cluster of the multi-temperature OCV-SOC based on the comparison results of the Euclidean distance from the current trajectory point to the center of the steady-state attractor and the attractor radius, as well as whether the number of dwell points meets the preset requirements.

[0033] Furthermore, based on temperature distance and remaining battery capacity distance, a two-dimensional Gaussian weighting function is constructed, and the correction weights for all remaining battery capacity points on each OCV-SOC curve in all multi-temperature OCV-SOC initial curve clusters are calculated. This includes: constructing a two-dimensional Gaussian weighting function that incorporates temperature distance and remaining battery capacity distance, based on the principle that the closer the temperature and the closer the remaining battery capacity, the greater the correction weight; and using the expression of the two-dimensional Gaussian function, calculating the correction weights for all remaining battery capacity points on each OCV-SOC curve in all multi-temperature OCV-SOC initial curve clusters.

[0034] Furthermore, the expression for the two-dimensional Gaussian function is:

[0035] ;

[0036] In the formula, The correction weights for the i-th temperature curve and the j-th battery remaining charge point; T i Ti is the calibration temperature of the i-th curve in the OCV–SOC curve cluster; T0 is the real-time battery temperature at the current correction trigger point; SOC j σ is the j-th SOC node in the curve cluster; SOC0 is the remaining battery power value at the current correction trigger point; T The bandwidth is weighted in the temperature direction; σ SOC The bandwidth weighted by the remaining battery power.

[0037] Furthermore, leveraging the topological neighborhood aggregation effect, the correction weights are adjusted to obtain the weight coefficients after correction via the topological manifold space. This includes: constructing a battery operating condition topological feature vector based on battery temperature, remaining battery capacity, battery open-circuit voltage, and current sampling; recording the normalized battery operating condition topological feature vector after each trigger condition is met, forming a set of operating condition points; based on the Euclidean distance calculation results, identifying the mutually adjacent points of the current battery operating condition within the set of operating condition points and configuring a first topological connectivity; simultaneously configuring a second topological connectivity between non-mutually adjacent points; calculating the quantized value of the similarity between the current operating condition and the operating condition points at the time of historical trigger correction in the topological manifold space based on the first or second topological connectivity and combined with the similarity decay coefficient; multiplying the quantized value of the similarity between the current operating condition and the operating condition points at the time of historical trigger correction with the correction weights to obtain the weight coefficients after correction via the topological manifold space.

[0038] Furthermore, based on the measured open-circuit voltage and the model open-circuit voltage in the multi-temperature OCV-SOC initial curve cluster, the correction amount is calculated, including: acquiring the measured open-circuit voltage at the current battery temperature and remaining charge, obtaining the model open-circuit voltage in the multi-temperature OCV-SOC initial curve cluster at the current battery temperature and remaining charge; and subtracting the measured open-circuit voltage from the model open-circuit voltage to obtain the correction amount.

[0039] Furthermore, based on the weight coefficients and correction amounts after topological manifold space correction, the initial curve cluster of multiple temperature OCV–SOC is collaboratively corrected, including: multiplying the weight coefficients after topological manifold space correction with the correction amount to obtain the weighted correction amount; and using the weighted correction amount to collaboratively correct the initial curve cluster of multiple temperature OCV–SOC.

[0040] Furthermore, using the conformally smoothed OCV–SOC curve and the first-order RC equivalent circuit model, the terminal voltage prediction error is calculated, including: substituting the conformally smoothed OCV–SOC curve into the first-order RC equivalent circuit model to calculate the terminal voltage prediction value; and subtracting the measured terminal voltage from the predicted terminal voltage value to obtain the terminal voltage prediction error.

[0041] Furthermore, based on the effective feedback enable conditions and utilizing the variance of the terminal voltage prediction error and the degree of battery aging, the bandwidth parameter of the two-dimensional Gaussian weighting function is dynamically adjusted, including: pre-configuring several effective feedback enable conditions, and when any effective feedback enable condition is met, incorporating the terminal voltage prediction error into a sliding window and calculating the variance of the terminal voltage prediction error; and dynamically adjusting the bandwidth parameter of the two-dimensional Gaussian weighting function based on the comparison results of the terminal voltage prediction error variance and the degree of battery aging with their respective preset values.

[0042] The beneficial effects of this invention are as follows:

[0043] 1. Multi-temperature curve linkage correction to ensure curve consistency and physical self-consistency: This invention uses a two-dimensional Gaussian weighting function to correct a certain temperature and a certain SOC point, and simultaneously corrects the neighborhood points of all temperature curves. The closer the temperature and the closer the SOC, the greater the correction amount, avoiding the intersection and jump of different temperature curves. This conforms to the physical law of battery OCV changing with temperature. When estimating SOC using Kalman filtering, temperature switching will not cause SOC jumps, and the estimation stability is significantly improved.

[0044] 2. Strong anti-interference capability and smooth curve without abnormalities: This invention introduces weight constraints and correction amount rationality verification. The correction weight of abnormal measurement values ​​(noise, sensor error) is reduced, and the correction is performed to preserve the shape and smooth the curve, avoiding sharp spikes and abnormal inflection points. The dOCV / dSOC is stable, the voltage prediction error of Kalman filter is reduced, the filter convergence is better, and the SOC estimation accuracy is improved.

[0045] 3. Adapts to the entire battery life cycle, ensuring long-term stable correction accuracy: The weight bandwidth of this invention is adaptively adjusted according to SOH, and narrow bandwidth correction is used for aged batteries to avoid parameter fluctuation interference. The correction strategy is adapted to the aging characteristics of the battery, and can maintain high correction accuracy even after long-term operation. This solves the problem of correction accuracy decay after aging in existing technologies and extends the stable period of BMS SOC estimation accuracy.

[0046] 4. Closed-loop adaptive, no manual intervention required: This invention designs a closed-loop feedback mechanism to feed back the terminal voltage prediction error to the correction stage, dynamically adjust the weight bandwidth, and the correction strategy can be automatically optimized according to the actual operating status of the battery without manual adjustment of correction parameters, adapting to complex operating conditions (static / dynamic, high and low temperatures) and reducing the operation and maintenance cost of BMS.

[0047] 5. Low computing power consumption and can be implemented in embedded mass production: This invention adopts the storage method of benchmark curve and deviation table, and the correction calculation adopts fixed-point operation and simplified interpolation to avoid complex floating-point operation. It is compatible with 8-bit / 16-bit low computing power BMS main control chip, without the need to increase hardware costs. It can be directly applied to mass production BMS and has extremely high engineering practicality.

[0048] 6. Strong compatibility, adaptable to the first-order RC-EKF SOC estimation framework: The modified OCV-SOC curve of this invention can be directly substituted into the first-order RC model, seamlessly connecting with the Kalman filter algorithm. No modification to the existing SOC estimation framework is required; only the modification module of this invention needs to be added to improve the SOC estimation accuracy. It is compatible with the algorithm architecture of existing mass-produced BMS, and the upgrade cost is low. Attached Figure Description

[0049] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0050] Figure 1 This is a flowchart of a multi-temperature OCV–SOC curve collaborative correction method based on two-dimensional weights according to an embodiment of the present invention;

[0051] Figure 2 This is a block diagram of a multi-temperature OCV–SOC curve collaborative correction system based on two-dimensional weights according to an embodiment of the present invention;

[0052] Figure 3 This is a flowchart according to an embodiment of the present invention;

[0053] Figure 4 This is a schematic diagram of the functional modules in the corresponding BMS according to an embodiment of the present invention;

[0054] Figure 5 This is a flowchart of constructing a two-dimensional Gaussian weight function and calculating the corrected weights. Detailed Implementation

[0055] To further illustrate the various embodiments, the present invention provides accompanying drawings, which are part of the disclosure of the present invention. These drawings are mainly used to illustrate the embodiments and can be used in conjunction with the relevant descriptions in the specification to explain the operating principles of the embodiments. With reference to these drawings, those skilled in the art should be able to understand other possible implementation methods and the advantages of the present invention. The components in the drawings are not drawn to scale, and similar component symbols are generally used to represent similar components.

[0056] According to an embodiment of the present invention, a collaborative correction method for multi-temperature OCV–SOC curves based on two-dimensional weights is provided.

[0057] The present invention will now be further described in conjunction with the accompanying drawings and specific embodiments, such as... Figure 1 As shown, according to an embodiment of the present invention, a method for collaborative correction of multi-temperature OCV–SOC curves based on two-dimensional weights is provided, comprising:

[0058] S1. Obtain the open-circuit voltage and remaining charge of the battery at multiple temperatures to construct a cluster of initial OCV–SOC curves at multiple temperatures.

[0059] In one embodiment, the open-circuit voltage and remaining charge of the battery at multiple temperatures are obtained to construct a multi-temperature OCV-SOC initial curve cluster, including: selecting several temperature ranges based on the principle of covering the actual operating temperature range of the battery; and obtaining the remaining charge and corresponding open-circuit voltage of the battery at each temperature range through a static method combined with current charge-discharge testing, thereby forming a multi-temperature OCV-SOC initial curve cluster.

[0060] S2. Based on the chaotic nonlinear evolution method, and by comparing the Euclidean distance from the current trajectory point in the chaotic phase space to the center of the steady-state attractor with the adaptive conditions, determine whether to correct the initial curve cluster of the multi-temperature OCV–SOC.

[0061] In one embodiment, based on the chaotic nonlinear evolution method, and by comparing the Euclidean distance from the current trajectory point in the chaotic phase space to the center of the steady-state attractor with the adaptive conditions, it is determined whether to correct the initial curve cluster of multi-temperature OCV-SOC. This includes: constructing a chaotic phase space based on the current, remaining battery charge, voltage, and temperature at a certain moment; determining the center of the steady-state attractor based on the mean vector of the steady-state trajectories in the most recent frames; calculating the Euclidean distance from the current trajectory point in the chaotic phase space to the center of the steady-state attractor and configuring the attractor radius; and determining whether to correct the initial curve cluster of multi-temperature OCV-SOC based on the comparison result of the Euclidean distance from the current trajectory point to the center of the steady-state attractor and the attractor radius, as well as whether the number of dwell points meets the preset requirements.

[0062] S3. If yes, then construct a two-dimensional Gaussian weighting function based on the temperature distance and the remaining battery charge distance, and calculate the correction weight of all remaining battery charge points on each OCV-SOC curve in all multi-temperature OCV-SOC initial curve clusters; if no, then do not correct the multi-temperature OCV-SOC initial curve clusters.

[0063] In one embodiment, a two-dimensional Gaussian weighting function is constructed based on temperature distance and remaining battery capacity distance, and the corrected weights for all remaining battery capacity points on each OCV-SOC curve in all multi-temperature OCV-SOC initial curve clusters are calculated, such as... Figure 5 As shown, the process includes: constructing a two-dimensional Gaussian weighting function that includes temperature distance and battery remaining capacity distance, based on the principle that the closer the temperature and the closer the battery remaining capacity, the greater the correction weight; and using the expression of the two-dimensional Gaussian function to calculate the correction weight of all battery remaining capacity points on each OCV-SOC curve in all multi-temperature OCV-SOC initial curve clusters.

[0064] In one embodiment, the expression for the two-dimensional Gaussian function is:

[0065] ;

[0066] In the formula, The corrected weight for the i-th temperature curve (i.e., the OCV-SOC curve in the multi-temperature OCV-SOC initial curve cluster); T is the corrected weight for the j-th battery remaining capacity point. i T0 is the calibration temperature of the i-th curve in the OCV–SOC curve cluster; T0 is the real-time battery temperature at the current correction trigger point; SOC j σ is the j-th SOC node in the curve cluster; SOC0 is the remaining battery power value at the current correction trigger point; T The bandwidth is weighted in the temperature direction; σ SOC The bandwidth weighted by the remaining battery power.

[0067] S4. Utilize the topological neighborhood aggregation effect to adjust the correction weights and obtain the weight coefficients after topological manifold space correction; calculate the correction amount based on the measured open-circuit voltage and the model open-circuit voltage in the multi-temperature OCV–SOC initial curve cluster.

[0068] In one embodiment, the topological neighborhood aggregation effect is used to adjust the correction weights to obtain the weight coefficients after correction by the topological manifold space. This includes: constructing a battery operating condition topological feature vector based on battery temperature, remaining battery capacity, battery open-circuit voltage, and current sampling; recording the normalized battery operating condition topological feature vector after each trigger condition is met to form an operating condition point set; based on the Euclidean distance calculation results, obtaining the mutually adjacent points of the current battery operating condition in the operating condition point set and configuring a first topological connectivity; simultaneously configuring a second topological connectivity between non-mutually adjacent points; calculating the quantized value of the similarity between the current operating condition and the operating condition points at the time of historical trigger correction in the topological manifold space according to the first or second topological connectivity and combined with the similarity decay coefficient; multiplying the quantized value of the similarity between the current operating condition and the operating condition points at the time of historical trigger correction with the correction weights to obtain the weight coefficients after correction by the topological manifold space.

[0069] In one embodiment, the correction amount is calculated based on the measured open-circuit voltage and the model open-circuit voltage in the multi-temperature OCV-SOC initial curve cluster, including: acquiring the measured open-circuit voltage at the current battery temperature and remaining charge, obtaining the model open-circuit voltage in the multi-temperature OCV-SOC initial curve cluster at the current battery temperature and remaining charge, and subtracting the measured open-circuit voltage from the model open-circuit voltage to obtain the correction amount.

[0070] S5. Based on the weight coefficients and correction amounts after correction in the topological manifold space, the initial OCV–SOC curve clusters at multiple temperatures are collaboratively corrected to obtain the corrected curves; the corrected curves are then smoothed in a shape-preserving manner to obtain the smoothed OCV–SOC curves.

[0071] In one embodiment, the multi-temperature OCV–SOC initial curve family is collaboratively corrected based on the weighting coefficients and correction amounts after topological manifold space correction, including:

[0072] Multiply the weight coefficients after topological manifold space correction by the correction amount to obtain the weighted correction amount;

[0073] We use weighted corrections to collaboratively correct the initial OCV–SOC curve clusters at multiple temperatures.

[0074] S6. Using the conformally smoothed OCV–SOC curve and the first-order RC equivalent circuit model, the terminal voltage prediction error is calculated. Based on the effective feedback enable condition, and using the variance of the terminal voltage prediction error and the degree of battery aging, the bandwidth parameter of the two-dimensional Gaussian weighting function is dynamically adjusted.

[0075] In one embodiment, the terminal voltage prediction error is calculated using the conformally smoothed OCV–SOC curve and a first-order RC equivalent circuit model, including: substituting the conformally smoothed OCV–SOC curve into the first-order RC equivalent circuit model to calculate the terminal voltage prediction value; and subtracting the measured terminal voltage from the predicted terminal voltage value to obtain the terminal voltage prediction error.

[0076] In one embodiment, based on effective feedback enable conditions and utilizing the variance of the terminal voltage prediction error and the degree of battery aging, the bandwidth parameter of the two-dimensional Gaussian weighting function is dynamically adjusted, including: pre-configuring several effective feedback enable conditions, and when any effective feedback enable condition is met, incorporating the terminal voltage prediction error into a sliding window and calculating the variance of the terminal voltage prediction error; and dynamically adjusting the bandwidth parameter of the two-dimensional Gaussian weighting function based on the comparison results of the terminal voltage prediction error variance and the degree of battery aging with their respective preset values.

[0077] like Figure 2 As shown, according to another embodiment of the present invention, a multi-temperature OCV–SOC curve collaborative correction system based on two-dimensional weights is also provided, comprising:

[0078] The curve cluster construction module is used to obtain the open-circuit voltage and remaining capacity of the battery at multiple temperatures in order to construct an initial curve cluster of OCV–SOC at multiple temperatures.

[0079] The weight acquisition module is used to determine whether to modify the multi-temperature OCV-SOC initial curve cluster based on the chaotic nonlinear evolution method and by comparing the Euclidean distance from the current trajectory point in the chaotic phase space to the center of the steady-state attractor with the adaptive conditions. If not, the multi-temperature OCV-SOC initial curve cluster is not modified; if so, a two-dimensional Gaussian weight function is constructed based on the temperature distance and the distance of the remaining battery charge, and the modification weight of all remaining battery charge points on each OCV-SOC curve in all multi-temperature OCV-SOC initial curve clusters is calculated.

[0080] The weight adjustment and correction amount acquisition module is used to adjust the correction weights by utilizing the topological neighborhood aggregation effect to obtain the weight coefficients after topological manifold space correction; the correction amount is calculated based on the measured open-circuit voltage and the model open-circuit voltage in the multi-temperature OCV–SOC initial curve cluster.

[0081] The curve correction module is used to collaboratively correct the initial OCV–SOC curve cluster of multiple temperatures based on the weight coefficients and correction amounts after correction by the topological manifold space, so as to obtain the corrected curves; and to perform shape-preserving smoothing on the corrected curves to obtain the shape-preserving smoothed OCV–SOC curves.

[0082] The closed-loop feedback module is used to calculate the terminal voltage prediction error using the conformally smoothed OCV–SOC curve and the first-order RC equivalent circuit model. Based on the effective feedback enable condition, and using the variance of the terminal voltage prediction error and the degree of battery aging, the bandwidth parameter of the two-dimensional Gaussian weighting function is dynamically adjusted.

[0083] To facilitate understanding of the above technical solutions of the present invention, the working principle of the present invention in actual process will be described in detail below.

[0084] In existing battery management systems (BMS) for chemical batteries, the OCV–SOC (Open Circuit Voltage – Remaining Capacity) curve is widely used to correct for SOC (Single-Pole Capacity). Since the open circuit voltage of a chemical battery is affected by battery temperature, multiple OCV–SOC curve families under different temperature conditions are typically used to correct for the battery SOC at different actual battery temperatures. Furthermore, the OCV–SOC curves at each battery temperature change with battery aging. To avoid increased SOC estimation errors due to changes in this curve family caused by prolonged battery use, during the productization stage, the curve family can be corrected online by measuring battery voltage, temperature, and current during long-term charge and discharge cycles, thereby minimizing the increase in SOC estimation errors caused by battery aging.

[0085] In actual operation, online correction can only correct a portion of points within the adjacent OCV-SOC curve cluster at (T0, SOC0) at a time. If there is no overlap, this typically involves four adjacent points on two adjacent curves. However, the curves in this cluster have certain relationships; for example, they can only overlap at most, but absolutely cannot intersect. Furthermore, a change in SOC at a certain voltage and temperature usually implies a change in SOC at other temperatures under that voltage. However, correcting only the voltage points of one or two adjacent curves at a time is too inefficient.

[0086] This invention designs a correction mechanism involving multi-temperature curve linkage, two-dimensional weight adaptation, shape-preserving smoothing, and closed-loop feedback to ensure the physical consistency and long-term accuracy of a multi-temperature OCV-SOC curve cluster, providing reliable model input for SOC estimation. Based on the similarity of multi-temperature curves, a two-dimensional weight linkage correction mechanism is constructed, combining shape-preserving smoothing and closed-loop feedback to achieve online adaptive correction of multi-temperature OCV-SOC curves. The overall process is divided into offline initialization, online triggering, weight calculation, synchronous correction and shape-preserving smoothing, and closed-loop feedback, adapting to the first-order RC-EKF SOC estimation framework. Figure 3 As shown, the specific technical solution is as follows.

[0087] I. Offline Initialization: Constructing a cluster of initial OCV–SOC curves at multiple temperatures.

[0088] II. Online Operation: The data acquisition module collects the battery's temperature (T0), SOC value (SOC0), and terminal voltage (V) in real time. t The data includes current (I) and aging degree (SOH); sliding window filtering is applied to the collected data.

[0089] 3. Meet the correction triggering conditions: Use the chaotic nonlinear evolution method for correction triggering.

[0090] IV. Calculate the correction amount: Collect the current data. Actual measurement at the location ; This is the measured open-circuit voltage at the current battery temperature and SOC. (Query the current model.) ; This is the model open-circuit voltage obtained from the multi-temperature OCV–SOC curve cluster at the current battery temperature and SOC. This value is the expected battery voltage value in the battery model. The correction amount is then calculated. .

[0091] V. Weight Calculation: Based on Temperature Distance Distance from SOC Construct a two-dimensional Gaussian weighting function; calculate the corrected weights for all SOC points on each OCV-SOC curve in all multi-temperature OCV-SOC initial curve clusters. Based on the topological neighborhood aggregation effect, the modified weights are adjusted twice to obtain the weight coefficients w after correction by the topological manifold space.

[0092] VI. Synchronous Correction: Based on Synchronously correct the corresponding SOC points of all temperature curves with w, and update the OCV–SOC curve cluster.

[0093] 7. Shape Preservation and Smoothing: Apply a moving average filter to each corrected OCV–SOC curve to constrain the rate of change of dOCV / dSOC and avoid curve anomalies.

[0094] 8. Closed-loop feedback: Substitute the corrected OCV–SOC curve into the first-order RC model for Kalman filter SOC estimation, calculate the terminal voltage prediction error e, and adjust the adaptive bandwidth (σ) of the weighting function based on e and SOH. T σ SOC ).

[0095] The functional module schematic diagram in the BMS corresponding to this invention is as follows: Figure 4 As shown, where:

[0096] Module M1: Offline Initialization. Used for offline testing and storage of multi-temperature OCV–SOC initial curve clusters, baseline curves, and deviation table ΔV(T, SOC), providing an initial baseline for online correction.

[0097] Module M2: Data Acquisition. Composed of a current sensor, voltage sensor, temperature sensor, and SOH estimation unit, it is used to acquire real-time operating data of the battery and output filtered I and V values. t T, SOC, and SOH data.

[0098] Module M3: Storage. Used to store the initial curve, the corrected curve, weight parameters, and historical correction data, providing support for subsequent corrections and retrospective analysis.

[0099] Module M4: Trigger Judgment. Used to determine whether the correction trigger conditions are met based on the collected data, and to control the start and stop of the correction process.

[0100] Module M5: Correction Calculation. Used to calculate the deviation between the measured OCV and the model OCV. The reasonableness of the correction amount is checked. If the value is greater than 0.01V, reduce the correction weight.

[0101] Module M6: Weight Calculation. Calculates corrected weights based on temperature distance and SOC distance. Furthermore, based on the topological neighborhood aggregation effect, the modified weights are adjusted a second time. The weight bandwidth σ is dynamically adjusted according to the SOH (Solar Optimization). T σ SOC .

[0102] Module M7: Synchronization Correction. Receives w and ΔV. corr It synchronously updates the corresponding SOC points of all temperature OCV–SOC curves and outputs the corrected OCV–SOC curve cluster.

[0103] Module M8: Conformal Smoothing. Employs a moving average filter to smooth the corrected curves, constrains the rate of change of dOCV / dSOC, and outputs a smoothed OCV–SOC curve cluster.

[0104] Module M9: Closed-loop feedback. It receives the conformal OCV–SOC curve set, substitutes it into a first-order RC model to calculate the predicted terminal voltage, and compares it with the measured V. t The terminal voltage prediction error e is obtained by comparison. The weighted bandwidth is adjusted according to e and SOH to achieve adaptive correction strategy.

[0105] Module M10: SOC estimation. It receives the corrected OCV–SOC curve set, combines it with the first-order RC model parameters, executes the EKF algorithm, and outputs a high-precision SOC estimation result.

[0106] Key aspects of this invention are described below:

[0107] I. Offline Initialization: Six temperature ranges (-20℃, 0℃, 10℃, 25℃, 40℃, and 55℃) were selected (covering the actual operating temperature range of the battery). A calibration principle of 0.02C ultra-low current stepped charge-discharge combined with long-term static placement was adopted: low-current charge-discharge weakens battery polarization voltage interference. After each 1% change in SOC, the battery was kept at a constant temperature for more than 4 hours to allow lithium-ion concentration polarization and interface polarization to completely dissipate, and the battery terminal voltage to converge to the true open-circuit voltage (OCV). Through static placement combined with low-current charge-discharge testing, OCV values ​​for SOCs of 0%~100% (in 1% steps) at each temperature were obtained, forming an initial OCV-SOC curve cluster. Using the 25℃ curve as the baseline curve, the voltage deviation between each temperature curve and the baseline curve at the same SOC point was calculated, forming a deviation table. ,in The OCV value is the reference curve at 25℃. This involves creating battery voltage models corresponding to the State of Charge (SOC) at different temperatures. The resulting curve set and deviation table are stored in the BMS's Flash memory as an initial reference for online correction.

[0108] 2. Trigger Condition Judgment: The existing conventional steady-state judgment only relies on a single current threshold for judgment, which has two major defects: ① When the current is zero for a short period but the internal polarization of the battery has not dissipated, the steady state is misjudged, the collected OCV is distorted, causing a deviation in the correction of the OCV curve; ② Under small charge-discharge dynamic conditions and sensor noise disturbances, frequent false trigger corrections will occupy the embedded computing power of BMS and reduce the stability of SOC estimation. The design purpose of the trigger condition is to ensure the accuracy of the measured OCV. Only when the battery is at rest and in a steady state, the terminal voltage is approximately equal to the OCV. Here, the nonlinear evolution theory of the chaotic system is used to realize the trigger condition, and the specific steps are as follows:

[0109] 1. Determine the chaotic phase space and state vector, and construct a four-dimensional phase space:

[0110] ;

[0111] Where: is the current at time k, is the SOC at time k, is the terminal voltage at time k, is the temperature at time k.

[0112] 2. Determine the steady-state attractor center (the current real steady-state point), take the mean vector of the latest N frames of steady-state trajectories, N is 8 points, and the point sampling step is 1s, which meets the actual conditions of embedded hardware computing power.

[0113] ;

[0114] Where: ; ; ; .

[0115] 3. Calculate the Euclidean distance :

[0116] The distance from the current trajectory point X(k) to the attractor center :

[0117] .

[0118] 4. Judge whether it falls into the attractor region:

[0119] Set the attractor radius R (adaptive threshold): if D(k)<R, it falls into the steady-state attractor; if M consecutive points all satisfy the condition, it is determined that the system enters a real steady state. M is generally 3~8 points to prevent noise jump.

[0120] 5. The adaptiveness of the attractor radius R and the consecutive M points is shown in Table 1. R and M are not fixed, and are adaptive according to SOH and temperature:

[0121] Table 1. Attractor radius R and number of residence points M

[0122] Advantages of the four-dimensional phase space steady-state determination scheme for chaos:

[0123] 1. The four-dimensional vector fusion of current, SOC, terminal voltage and temperature features can accurately distinguish pseudo steady-state conditions with apparent zero current and internal polarization not disappearing compared to single current threshold judgment. OCV acquisition is triggered only after the actual polarization dissipates, which greatly improves the accuracy of measured OCV and reduces OCV curve correction error from the source.

[0124] 2. Based on the mean value of chaotic attractor trajectory to determine steady state, it has the ability to filter and suppress random noise of current and voltage sensors, avoids frequent false triggering correction under small disturbances, reduces memory lifespan loss caused by frequent curve reading and writing, and reduces the computing power overhead of embedded main control.

[0125] 3. The center of the attractor is calculated by taking the average of the 1-second step trajectory of 8 frames. The computational load is extremely low, which is suitable for the computing power constraints of low-cost BMS chips and takes into account both judgment accuracy and hardware real-time performance.

[0126] 4. Relying on the Euclidean distance constraint of continuous multi-frame trajectory to achieve steady state, ineffective corrections are avoided during vehicle start-stop and idling small charging and discharging phases, ensuring long-term stability of online OCV correction.

[0127] III. Two-dimensional Gaussian weight calculation:

[0128] Based on the similarity characteristics of multi-temperature OCV–SOC curves, a two-dimensional Gaussian weighting function with temperature distance and SOC distance as the core is constructed to achieve collaborative correction of multiple temperature curves with local correction, global linkage, and adaptive constraints. The weight allocation follows the principle that the closer the temperature and SOC, the greater the correction weight, which ensures that adjacent curves are adjusted synchronously while avoiding excessive interference to distant points.

[0129] The core of the weighting function is that the closer the temperature and SOC are, the greater the correction weight. A two-dimensional Gaussian function is used to ensure the linkage correction of multiple temperature curves while avoiding interference from distant points. The specific expression is as follows:

[0130] ;

[0131] in, The correction weight for the i-th temperature curve and the j-th SOC point; T i Ti is the calibration temperature of the i-th curve in the OCV–SOC curve cluster; T0 is the real-time battery temperature at the current correction trigger point; SOC j σ is the j-th SOC node in the curve cluster; SOC0 is the SOC value of the current correction trigger point;T The bandwidth is weighted in the temperature direction; σ SOC The bandwidth is the weighted bandwidth in the SOC direction. The bandwidth parameter is not fixed and is also adaptively determined by SOH and temperature, where σ... SOC As mentioned earlier, the attractor radius R is the same, σ T As shown on the far right of Table 2:

[0132] Table 2 Bandwidth Parameters

[0133] To further improve the adaptability of weight allocation to the nonlinear characteristics of battery operating conditions, this invention introduces the concept of topological manifolds to supplement and optimize the traditional Euclidean distance weighting logic:

[0134] 1. Constructing a topological manifold space: Mapping battery temperature, SOC, OCV, and current acquisition data to a low-dimensional topological manifold space, and characterizing the intrinsic similarity of operating points through the manifold structure, replacing the single numerical distance judgment.

[0135] 2. Topological connectivity and similarity measurement: By calculating the topological connectivity between operating points, the approximation of different operating conditions is quantified, achieving a similarity judgment that is more in line with battery characteristics than the traditional Euclidean distance.

[0136] 3. Topological neighborhood aggregation effect correction: Based on the topological neighborhood aggregation effect, the correction weight is adjusted a second time. The stronger the neighborhood correlation of a work case in the topological manifold, the stronger the aggregation effect, and the corresponding correction weight increases adaptively. The greater the topological difference of the work cases, the weaker the aggregation effect, and the correction weight decreases synchronously. Details are as follows:

[0137] ;

[0138] Where w is the weight coefficient after correction for the topological manifold space. And S topo S is the quantified value of the similarity between the current working condition and the working condition points at the time of historical trigger corrections in the topological manifold space. topo The value ranges from 0 to 1, where 1 indicates that the two working points are completely adjacent in the topological manifold and have highly similar characteristics, and 0 indicates that they are completely disconnected and have extremely different characteristics. topo The calculation method is as follows:

[0139] Constructing the battery operating condition topology feature vector: Where T is the battery temperature (°C), SOC is the remaining battery charge (%), and OCV is the battery open-circuit voltage (V). Let be the rate of change of current. For Normalization was performed on each dimension to obtain To eliminate dimensional differences.

[0140] Pre-built reference point set: Records the battery operating conditions after each trigger condition is met, thus obtaining the operating condition point set. .

[0141] Construct a nearest neighbor graph: based on the current working condition Find the K nearest neighbors, with K set to 5, which is suitable for the computing power of embedded hardware. The Euclidean distance method is used to calculate the nearest neighbors.

[0142] Calculate topological connectivity: in the set of operating points Find a point and If the points are neighbors, the topological connectivity C=1; otherwise, C=0.

[0143] calculate : ,in, The similarity attenuation coefficient is adaptively adjusted based on SOH and temperature range, and is related to attractor radius R, number of residence points M, and temperature-direction weighted bandwidth σ. T SOC direction weighted bandwidth σ SOC The value selection strategy is uniform, as shown in Table 3:

[0144] Table 3 Similarity decay coefficient

[0145] IV. Synchronous Correction: For each SOC point of all temperature curves, correction is performed according to the following formula: This ensures that multiple temperature curves are synchronized and linked, avoiding curve intersections. Among these, These are the battery voltage values ​​corresponding to the temperature and SOC in the original OCV–SOC curve cluster. For the process The battery voltage values ​​corresponding to temperature and SOC in the weighted modified OCV–SOC curve cluster.

[0146] V. Conformity Preservation and Smoothing: A 3-5 point moving average filter is used to smooth each corrected curve, constraining the rate of change of dOCV / dSOC to ≤0.005V / % (configurable) to avoid sharp spikes and abrupt changes in the curve. For example, for the SOC = 48%~52%, the OCV values ​​of three adjacent points are averaged to eliminate local fluctuations and ensure the smoothness and physical consistency of the curve.

[0147] VI. Closed-loop feedback: Substitute the conformally smoothed OCV–SOC curve into the first-order RC equivalent circuit model to calculate the predicted terminal voltage. , and the measured terminal voltage The difference is used to obtain the terminal voltage prediction error e:

[0148] ;

[0149] The predicted terminal voltage value is calculated using the following formula:

[0150] ;

[0151] In the formula, The current temperature and open-circuit voltage at the current SOC are after two-dimensional weight correction and shape-preserving smoothing; I is the real-time battery current, positive for discharging and negative for charging; R0 is the battery's ohmic internal resistance. The first-order RC polarization voltage is calculated using the discrete iterative formula:

[0152] ;

[0153] In the formula, Δt is the sampling period. C p These are the polarization internal resistance and polarization capacitance, respectively. For the first Real-time battery current at the sampling time.

[0154] The principle of the first-order RC equivalent circuit model is that ohmic polarization and concentration polarization occur simultaneously during the charging and discharging process of a lithium-ion battery. The first-order RC equivalent circuit is a low-order electrochemical simplification model adapted to embedded BMS. The model consists of an ohmic internal resistance R0 connected in series with an RC polarization branch, where R0 represents the ohmic voltage drop corresponding to the instantaneous current, and the polarization branch... C p The model simulates the slow decay of concentration polarization voltage after battery charging and discharging; it balances computational complexity and simulation accuracy, and can accurately reproduce the dynamic terminal voltage variation of the battery across the entire temperature and SOC range.

[0155] To avoid prolonged periods of battery inactivity (i.e., when current is 0, temperature and SOC remain unchanged), Equal to This results in a naturally smaller variance in the prediction error e, which fails to accurately reflect the quality of the OCV–SOC curve correction.

[0156] This invention introduces an effective feedback enable condition, whereby the error e is included in the sliding window and the variance is calculated only when any of the following conditions are met: the absolute value of the real-time battery current abs(I) ≥ 0.02C; the temperature change ΔT within 5 minutes ≥ 1℃; the SOC change ΔSOC within 5 minutes ≥ 1%; or an OCV–SOC curve correction has just been completed. If the battery is in a static steady state (simultaneously satisfying the three conditions abs(I) < 0.02C, ΔT < 1℃, and ΔSOC < 1%), then error accumulation and variance updates are paused, and the current weighted bandwidth parameters remain unchanged.

[0157] For N within the effective feedback interval ’ A series of error samples e1~e N’ The variance is calculated as follows:

[0158] Calculate the mean error within the window: ; Calculate the variance of the error: Based on the error variance Var(e) and the battery aging degree SOH, the bandwidth parameter σ of the two-dimensional Gaussian weighting function is dynamically adjusted. T σ SOC :

[0159] 1. When SOH ≥ 80% (new battery / lightly aged): If Var(e) > 0.0002, it indicates a large curve deviation, and σ should be reduced. T σ SOC This improves the locality of the correction and suppresses anomalous interference; if Var(e) ≤ 0.0002, it indicates that the correction effect is good, and σ is maintained or moderately increased. T σ SOC Improve the coverage of corrections.

[0160] 2. When SOH < 80% (re-aging): Regardless of the error variance, a narrow bandwidth σ is used. T =3℃、σ SOC =3%, to avoid correction failure due to fluctuations in aging battery parameters.

[0161] Through the aforementioned closed-loop feedback mechanism, the correction strategy is adapted to the battery status and operating conditions in real time, ensuring the long-term accuracy and physical self-consistency of the OCV-SOC curve cluster.

[0162] This invention utilizes the family-like similarity property of multi-temperature OCV–SOC curves to construct a two-dimensional Gaussian weighting function for temperature and SOC, achieving synchronized correction of multiple temperature curves and solving the problem of inconsistency caused by single-temperature, single-point correction in existing technologies; the weighting bandwidth (σ) T σ SOC The correction algorithm adaptively adjusts the State of Harmony (SOH) based on battery aging, adapting to the characteristic changes throughout the battery's lifecycle and improving the long-term stability of the correction. After correction, conformal smoothing is added to constrain the rate of change of dOCV / dSOC, ensuring a smooth curve without jumps, and adapting to the voltage prediction logic of Kalman filtering. A closed-loop mechanism of correction, estimation, and feedback is designed to feed back the terminal voltage prediction error to the correction stage, dynamically adjusting the correction strategy to achieve adaptive optimization of correction accuracy. The correction algorithm is simple and efficient, employing fixed-point calculation and simplified interpolation, adapting to the low computing power requirements of embedded BMS without requiring additional hardware costs.

[0163] This invention includes: obtaining the measured open-circuit voltage OCV based on the current battery temperature T0 and the current SOC0. meas According to temperature distance |T i -T0|Distance from SOC|SOC j-SOC0| Constructs a two-dimensional Gaussian weighting function to calculate the corrected weights for all temperatures and SOC points in the OCV–SOC curve family. The corrected weights are then corrected using topological neighborhood aggregation. Finally, based on the adjusted corrected weights and amounts, the OCV–SOC curves at all temperatures are simultaneously corrected, achieving joint correction of multiple temperature curves. The bandwidth parameter σ of the two-dimensional Gaussian weighting function... T σ SOC The system adaptively adjusts the SOH (State of Health) based on the battery's aging level; a wide bandwidth is used when SOH ≥ 80%, and a narrow bandwidth is used when SOH < 80%.

[0164] The trigger setting has been revised to include a two-stage cascaded judgment logic. The first stage is a preliminary coarse screening threshold: the second stage, chaotic nonlinear evolution steady-state fine verification, can only proceed if the battery meets the following conditions: absolute current value ≤ preset current threshold, SOC fluctuation ≤ preset fluctuation threshold, and resting time ≥ preset resting time. OCV correction is only performed after both conditions are met. The preset current threshold is 0.02C; the preset SOC fluctuation threshold is 1%; and the preset resting time threshold is 300s.

[0165] Shape-preserving smoothing step: Perform moving average filtering on each corrected OCV–SOC curve to constrain the first derivative of OCV with respect to SOC, dOCV / dSOC, within a preset range to avoid sharp spikes or jumps in the curve.

[0166] Closed-loop feedback steps: Substitute the corrected OCV–SOC curve into the battery model to estimate SOC and obtain the terminal voltage prediction error e; based on the statistical characteristics of error e, adaptively adjust the bandwidth parameter of the weighting function.

[0167] In the offline phase, a deviation table ΔV(T, SOC) is constructed for each temperature curve relative to the reference curve. During online correction, the curves are updated based on this deviation table, reducing computational load. The correction amount ΔV is then calculated. corr Then a rationality check is performed, when ΔV corr When the voltage exceeds the preset range, the correction weight is reduced or the correction is rejected to suppress abnormal measurement interference.

[0168] This invention can be widely used in BMS systems for new energy vehicles, energy storage, and backup power supplies, and has high industrial value. Compared with existing technologies, it improves in terms of curve consistency, correction accuracy, stability, adaptability, and engineering practicality. It can effectively solve the technical pain points of existing online correction of multi-temperature OCV-SOC curves, and provide reliable model support for first-order RC-EKF SOC estimation, thus having high industrial application value.

[0169] This invention addresses the inconsistencies and overlaps in multiple temperature curves caused by single-temperature, single-point correction. It leverages the similarity of OCV-SOC curves at multiple temperatures to achieve coordinated correction across these curves, ensuring their physical consistency. A weighted constraint mechanism is introduced to suppress interference from abnormal measurements, preventing sharp spikes and abrupt changes in the curves and ensuring the smoothness and continuity of the OCV-SOC curves, thus meeting the dOCV / dSOC calculation requirements of Kalman filtering. Adaptive correction parameters are designed to accommodate battery aging and variations in operating conditions, achieving long-term stability of correction accuracy and controlling SOC estimation errors. Conformal smoothing processing and a closed-loop feedback mechanism are added to feed back SOC estimation errors and terminal voltage prediction errors to the correction stage, enabling adaptive optimization of the correction strategy. Considering the low computational requirements of embedded BMS, a simple and efficient correction algorithm is designed that can be directly embedded into the main control chip of mass-produced BMS without additional hardware costs. Overall, it provides high-precision and high-stability multi-temperature OCV-SOC curve inputs for first-order RC-EKF SOC estimation, improves the SOC estimation accuracy and long-term reliability of BMS, and adapts to the actual application needs of scenarios such as new energy vehicles and energy storage systems.

[0170] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for collaborative correction of multi-temperature OCV–SOC curves based on two-dimensional weights, characterized in that, include: Obtain the open-circuit voltage and remaining capacity of the battery at multiple temperatures to construct a multi-temperature OCV–SOC initial curve cluster; Based on the chaotic nonlinear evolution method, and by comparing the Euclidean distance from the current trajectory point in the chaotic phase space to the center of the steady-state attractor with the adaptive conditions, it is determined whether to correct the initial curve cluster of multi-temperature OCV–SOC; where the current trajectory point is the coordinate point in the chaotic phase space of the state vector composed of the current, remaining battery charge, terminal voltage and temperature at the current moment. If so, a two-dimensional Gaussian weighting function is constructed based on the temperature distance and the remaining battery capacity distance, and the correction weights for all remaining battery capacity points on each OCV-SOC curve in all multi-temperature OCV-SOC initial curve clusters are calculated; if not, the multi-temperature OCV-SOC initial curve clusters are not corrected. The expression for the two-dimensional Gaussian weighting function is: ； In the formula, The correction weights for the i-th temperature curve and the j-th battery remaining charge point; T i T0 is the calibration temperature of the i-th curve in the OCV–SOC curve cluster; T0 is the real-time battery temperature at the current correction trigger point; SOC j σ is the j-th SOC node in the curve cluster; SOC0 is the remaining battery power value at the current correction trigger point; T The bandwidth is weighted in the temperature direction; σ SOC Weighted bandwidth in the direction of remaining battery power; The correction weights are adjusted by utilizing the topological neighborhood aggregation effect to obtain the weight coefficients after correction by the topological manifold space; the correction amount is calculated based on the measured open-circuit voltage and the model open-circuit voltage in the multi-temperature OCV–SOC initial curve cluster. Based on the weight coefficients and correction amounts after correction in the topological manifold space, the initial OCV–SOC curve clusters at multiple temperatures are collaboratively corrected to obtain the corrected curves; the corrected curves are then smoothed in a shape-preserving manner to obtain the smoothed OCV–SOC curves. Specifically, the terminal voltage prediction error is calculated using the conformally smoothed OCV–SOC curve and the first-order RC equivalent circuit model. Based on the effective feedback enable condition, and using the variance of the terminal voltage prediction error and the degree of battery aging, the bandwidth parameter of the two-dimensional Gaussian weighting function is dynamically adjusted.

2. The method for collaborative correction of multi-temperature OCV–SOC curves based on two-dimensional weights according to claim 1, characterized in that, The process of acquiring the open-circuit voltage and remaining charge of the battery at multiple temperatures to construct a multi-temperature OCV–SOC initial curve cluster includes: Several temperature settings are selected based on the principle of covering the actual operating temperature range of the battery. By combining the static method with current charge-discharge testing, the remaining capacity and corresponding open-circuit voltage of the battery at each temperature level are obtained, forming a cluster of initial OCV-SOC curves for multiple temperatures.

3. The method for collaborative correction of multi-temperature OCV-SOC curves based on two-dimensional weights according to claim 1, characterized in that, The method, based on chaotic nonlinear evolution, uses the comparison of the Euclidean distance from the current trajectory point in the chaotic phase space to the center of the steady-state attractor with the adaptive conditions to determine whether to correct the initial curve cluster of multi-temperature OCV–SOC, including: Based on the current, remaining battery charge, voltage, and temperature at a certain moment, a chaotic phase space is constructed; the center of the steady-state attractor is determined based on the mean vector of the steady-state trajectories of the most recent frames. Calculate the Euclidean distance from the current trajectory point in the chaotic phase space to the center of the steady-state attractor, and configure the attractor radius; By comparing the Euclidean distance from the current trajectory point to the center of the steady-state attractor with the attractor radius, and by checking whether the number of dwell points meets the preset requirements, it is determined whether to correct the initial cluster of multi-temperature OCV–SOC curves.

4. The method for collaborative correction of multi-temperature OCV–SOC curves based on two-dimensional weights according to claim 1, characterized in that, The method involves constructing a two-dimensional Gaussian weighting function based on temperature distance and remaining battery capacity distance, and calculating the corrected weights for all remaining battery capacity points on each OCV-SOC curve in all multi-temperature OCV-SOC initial curve clusters, including: Based on the principle that the closer the temperature and the closer the remaining battery power, the greater the correction weight, a two-dimensional Gaussian weight function including temperature distance and remaining battery power distance is constructed. The corrected weights of all battery remaining charge points on each OCV-SOC curve in all multi-temperature OCV-SOC initial curve clusters are calculated using the expression of a two-dimensional Gaussian weighting function.

5. The method for collaborative correction of multi-temperature OCV-SOC curves based on two-dimensional weights according to claim 1, characterized in that, The method of adjusting the modified weights using the topological neighborhood aggregation effect to obtain the weight coefficients after topological manifold space correction includes: Based on battery temperature, remaining battery capacity, battery open-circuit voltage and current data, a topological feature vector of battery operating conditions is constructed. Record the normalized battery operating condition topology feature vector after each trigger condition is met, forming a set of operating condition points; Based on the Euclidean distance calculation results, the mutually adjacent points of the current battery condition are obtained in the set of operating points, and the first topological connectivity is configured; at the same time, the second topological connectivity is configured between non-mutually adjacent points. Based on the first or second topological connectivity and combined with the similarity decay coefficient, calculate the quantified value of the similarity between the current working condition and the working condition point at the time of historical trigger correction in the topological manifold space. Multiply the quantified value of the similarity between the current working condition and the working condition point at the time of historical correction with the correction weight to obtain the weight coefficient after correction by the topological manifold space.

6. The method for collaborative correction of multi-temperature OCV-SOC curves based on two-dimensional weights according to claim 1, characterized in that, The correction amount, calculated based on the measured open-circuit voltage and the model open-circuit voltage in the multi-temperature OCV–SOC initial curve cluster, includes: Collect the measured open-circuit voltage at the current battery temperature and remaining charge, and obtain the model open-circuit voltage in the multi-temperature OCV-SOC initial curve cluster at the current battery temperature and remaining charge. The difference between the measured open-circuit voltage and the model open-circuit voltage is used to obtain the correction amount.

7. The method for collaborative correction of multi-temperature OCV–SOC curves based on two-dimensional weights according to claim 1, characterized in that, The collaborative correction of the multi-temperature OCV–SOC initial curve cluster based on the weight coefficients and correction amounts after topological manifold space correction includes: Multiply the weight coefficients after topological manifold space correction by the correction amount to obtain the weighted correction amount; We use weighted corrections to collaboratively correct the initial OCV–SOC curve clusters at multiple temperatures.

8. The method for collaborative correction of multi-temperature OCV-SOC curves based on two-dimensional weights according to claim 1, characterized in that, The method of calculating the terminal voltage prediction error using the conformally smoothed OCV–SOC curve and the first-order RC equivalent circuit model includes: The conformally smoothed OCV–SOC curve is substituted into the first-order RC equivalent circuit model to calculate the predicted terminal voltage. The difference between the measured terminal voltage and the predicted terminal voltage is used to obtain the terminal voltage prediction error.

9. The method for collaborative correction of multi-temperature OCV–SOC curves based on two-dimensional weights according to claim 1, characterized in that, The method of dynamically adjusting the bandwidth parameter of the two-dimensional Gaussian weighting function based on effective feedback enable conditions and utilizing the variance of the terminal voltage prediction error and the degree of battery aging includes: Several effective feedback enable conditions are pre-configured, and when any effective feedback enable condition is met, the terminal voltage prediction error is included in the sliding window and the variance of the terminal voltage prediction error is calculated. Based on the comparison results of the terminal voltage prediction error variance and the battery aging degree with their respective preset values, the bandwidth parameter of the two-dimensional Gaussian weighting function is dynamically adjusted.

Images