New energy automobile battery pack charging and discharging state prediction and control method and system

Abstract

This invention provides a method and system for predicting and controlling the charge and discharge state of new energy vehicle battery packs, relating to the field of charge and discharge technology. The method includes acquiring real-time electrical parameters and historical operating data of the battery pack, extracting dual-time-scale feature vectors, constructing a degradation baseline trajectory, calculating the state deviation, determining safety boundary constraints, and generating charge and discharge control commands based on a multi-objective optimization algorithm when the state deviation exceeds a threshold, thereby executing power regulation. This method can effectively predict the battery pack state, optimize charge and discharge strategies, and improve battery life and safety.

Description

Technical Field

[0001] This invention relates to the field of charge and discharge technology, and in particular to a method and system for predicting and controlling the charge and discharge state of new energy vehicle battery packs. Background Technology

[0002] With the rapid development of the new energy vehicle industry, the performance and lifespan management of power batteries, as core components of new energy vehicles, have a decisive impact on the safety and user experience of the entire vehicle. During use, power battery packs are subject to degradation phenomena such as capacity decay and increased internal resistance due to various factors, affecting the vehicle's range and lifespan. Currently, the industry generally uses battery management systems to monitor and control the charging and discharging status of battery packs. By monitoring parameters such as battery voltage, current, and temperature in real time, the health status and usable capacity of the battery pack are assessed, and charging and discharging strategies are formulated accordingly.

[0003] Existing technologies mostly employ feature analysis methods based on a single time scale, which makes it difficult to simultaneously consider both the instantaneous operating state and long-term degradation trend of the battery. This results in an incomplete assessment of the current actual state of the battery pack, which can easily lead to overcharging, over-discharging, or premature entry into protection mode, affecting the battery pack's efficiency and lifespan.

[0004] Traditional battery management systems lack dynamic benchmarks for the degradation process of battery packs, often using fixed thresholds or simple linear models for state assessment. This fails to adapt to changes in the nonlinear degradation characteristics of battery packs under different operating conditions, resulting in insufficient state prediction accuracy, especially with large errors under complex operating conditions.

[0005] Existing battery management strategies mostly adopt a passive response mechanism, which only triggers protection measures when battery parameters reach their limits. They lack proactive control capabilities and cannot make forward-looking adjustments based on the deviation between the battery's real-time status and its historical degradation trajectory. This makes it difficult to achieve a balance between optimal lifespan and performance while ensuring battery safety. Summary of the Invention

[0006] This invention provides a method and system for predicting and controlling the charge and discharge state of new energy vehicle battery packs, which can solve the problems in the prior art.

[0007] A first aspect of this invention provides a method for predicting and controlling the state of charge and discharge of a new energy vehicle battery pack, comprising:

[0008] The real-time electrical parameters and historical operating data of the battery pack are acquired. The voltage response curve in the real-time electrical parameters and the capacity decay sequence in the historical operating data are coupled and analyzed to extract a dual-time-scale feature vector that simultaneously represents the instantaneous polarization state and the cumulative degradation degree. Based on the historical operating data, a degradation baseline trajectory reflecting the evolution of the battery pack capacity with the number of cycles is constructed.

[0009] Based on the cumulative degradation component in the dual-timescale feature vector, the deviation between the position of the dual-timescale feature vector in the state space and the degradation reference trajectory is calculated to obtain the state deviation; based on the instantaneous polarization state component in the dual-timescale feature vector, the safety boundary constraint corresponding to the current operating point of the battery pack is determined.

[0010] When the state deviation exceeds a preset deviation threshold, a charge / discharge control command is generated using a multi-objective optimization algorithm based on the deviation direction of the state deviation and the constraint margin of the safety boundary constraint. This command enables the battery pack to return to the degradation baseline trajectory without triggering the safety boundary constraint. Power regulation is then performed on the charge / discharge circuit of the battery pack based on the charge / discharge control command.

[0011] The voltage response curve in the real-time electrical parameters is coupled and analyzed with the capacity decay sequence in the historical operating data to extract a dual-timescale feature vector that simultaneously characterizes the instantaneous polarization state and the cumulative degradation degree, including:

[0012] A time-series analysis was performed on the amplitude variation trajectory of the electrochemical polarization component in the voltage response curve over multiple historical charge-discharge cycles to obtain the polarization evolution curve of the electrochemical polarization component as a function of the number of cycles.

[0013] A time-series analysis was performed on the numerical change of the internal resistance growth rate in the capacity decay sequence during the same historical charge-discharge cycles to obtain the impedance evolution curve of the internal resistance growth rate as a function of the number of cycles.

[0014] Align the polarization evolution curve and the impedance evolution curve on the time axis, and calculate the cross-correlation coefficient between the rate of change of the polarization evolution curve and the rate of change of the impedance evolution curve;

[0015] Based on the cross-correlation coefficient, the influence weight of the electrochemical polarization component on the internal resistance growth rate is determined, and a state correlation matrix is ​​constructed with the electrochemical polarization component as the row index, the internal resistance growth rate as the column index, and the influence weight as the matrix element.

[0016] The dual-time-scale feature vector is generated by weighting and fusing the multi-level voltage decay components and the degradation rate parameters based on the state correlation matrix.

[0017] The deviation between the position of the dual-timescale feature vector in the state space and the degenerate baseline trajectory is calculated to obtain the state deviation, which includes:

[0018] Based on the coordinate values ​​of the dual-time-scale feature vector in the state space, the current state point corresponding to the dual-time-scale feature vector is determined.

[0019] Based on the cumulative degradation degree component in the dual time-scale feature vector, the reference state point corresponding to the current cycle number is located on the degradation reference trajectory, and the forward neighborhood interval and backward neighborhood interval of the reference state point are extracted on the degradation reference trajectory.

[0020] The shortest distance from the current state point to the forward neighborhood interval is taken as the forward deviation distance, and the shortest distance from the current state point to the backward neighborhood interval is taken as the backward deviation distance.

[0021] The smaller of the forward deviation distance and the backward deviation distance is marked with a sign, and the deviation distance with the sign is taken as the state deviation degree. The magnitude of the state deviation degree represents the deviation range, and the positive or negative sign represents the deviation type.

[0022] Based on the deviation direction of the state deviation and the constraint margin of the safety boundary constraint, a multi-objective optimization algorithm is used to generate charge / discharge control commands that enable the battery pack to return to the degradation baseline trajectory without triggering the safety boundary constraint. These commands include:

[0023] Based on the deviation direction of the state deviation, determine the target regression speed required for the battery pack to return to the degradation baseline trajectory;

[0024] Based on the constraint margin of the safety boundary constraint, calculate the maximum allowable control intensity of the battery pack while maintaining a safety margin no lower than a preset safety threshold.

[0025] A constrained optimization problem is constructed with the target regression speed as the performance index and the maximum control intensity as the upper limit of the constraint. The constrained optimization problem is solved by the multi-objective optimization algorithm. When the constrained optimization problem has a feasible solution, the optimal regression speed that satisfies the constraint conditions is obtained. When the constrained optimization problem does not have a feasible solution, the target regression speed is iteratively decayed until the constrained optimization problem has a feasible solution, and the suboptimal regression speed under the constraint conditions is obtained.

[0026] The optimal or suboptimal regression speed is decomposed into a voltage regulation speed component and a current regulation speed component. A first dynamic adjustment command for the charging cutoff voltage is calculated based on the voltage regulation speed component, and a second dynamic adjustment command for the discharge current limiting threshold is calculated based on the current regulation speed component. The first dynamic adjustment command and the second dynamic adjustment command are combined to generate the charge and discharge regulation command.

[0027] The constrained optimization problem is solved using the multi-objective optimization algorithm. When a feasible solution exists for the constrained optimization problem, the optimal regression rate that satisfies the constraints is obtained. When no feasible solution exists for the constrained optimization problem, the objective regression rate is iteratively decayed until a feasible solution appears for the constrained optimization problem. The suboptimal regression rate under the constraints is obtained by:

[0028] Based on the target regression rate, the maximum control intensity, and the safety boundary constraints, and combined with the multi-objective optimization algorithm, a constrained optimization problem is constructed, which includes decision variables, objective functions, and constraints.

[0029] The multi-objective optimization algorithm is used to explore the feasible region of the constrained optimization problem and search for the range of values ​​of the decision variables that simultaneously satisfy all constraints. When the range of values ​​of the decision variables is not empty, it is determined that the constrained optimization problem has a feasible solution. Within the range of values ​​of the decision variables, the gradient descent method is used to search for the optimal charging current adjustment range and the optimal discharging current adjustment range that minimize the objective function. Based on the optimal charging current adjustment range and the optimal discharging current adjustment range, the optimal regression speed of the battery pack to regress the degradation baseline trajectory is calculated.

[0030] When the range of values ​​for the decision variables is empty, it is determined that there is no feasible solution to the constrained optimization problem. The target regression velocity is multiplied by a decay factor less than one to obtain the decayed target regression velocity. This process is repeated iteratively until the range of values ​​for the decision variables in the updated constrained optimization problem becomes non-empty. Within the range of values ​​for the decision variables at this time, the gradient descent method is used to search for the suboptimal charging current adjustment range and the suboptimal discharging current adjustment range that minimize the updated objective function, thus obtaining the suboptimal regression velocity.

[0031] The decision variables of the constrained optimization problem are the charging current adjustment magnitude and the discharging current adjustment magnitude. The objective function of the constrained optimization problem is to minimize the time required for the battery pack state point to move to the degradation reference trajectory along the deviation direction. The constraints of the constrained optimization problem include the control intensity constraint that the combined control intensity of the charging current adjustment magnitude and the discharging current adjustment magnitude does not exceed the maximum control intensity, the safety boundary constraint that the battery pack does not touch the safety boundary constraint during the control process, and the non-negativity constraint that both the charging current adjustment magnitude and the discharging current adjustment magnitude are non-negative values.

[0032] The optimal or suboptimal regression rate is decomposed into a voltage-controlled rate component and a current-controlled rate component. A first dynamic adjustment command for the charging cutoff voltage is calculated based on the voltage-controlled rate component. A second dynamic adjustment command for the discharge current-limiting threshold is calculated based on the current-controlled rate component.

[0033] Construct the Jacobian matrix of charge and discharge parameters-regression rate of the battery pack at the current state point. The Jacobian matrix includes the partial derivative elements of the charging cut-off voltage with respect to the regression rate and the partial derivative elements of the discharge current limiting threshold with respect to the regression rate. The partial derivative elements are obtained by numerically differentiating the relationship between the changes in charge and discharge parameters of the battery pack and the changes in regression rate at multiple historical state points.

[0034] Calculate the singular value decomposition of the Jacobian matrix to obtain a first singular vector representing the dominant direction of voltage regulation and a second singular vector representing the dominant direction of current regulation. Project the optimal regression velocity or the suboptimal regression velocity onto the orthogonal subspace spanned by the first singular vector and the second singular vector to obtain the voltage regulation velocity component along the first singular vector and the current regulation velocity component along the second singular vector.

[0035] The voltage regulation speed component is multiplied by the reciprocal of the predetermined voltage regulation effectiveness coefficient. The charging cut-off voltage increment required to realize the voltage regulation speed component is obtained by inverse operation of the partial derivative elements of the charging cut-off voltage with respect to the regression speed in the Jacobian matrix. The charging cut-off voltage increment is then superimposed on the current charging cut-off voltage setting value to generate the first dynamic adjustment command of the charging cut-off voltage.

[0036] Multiply the current regulation speed component by the reciprocal of the predetermined current regulation effectiveness coefficient, and obtain the discharge current limiting threshold increment required to realize the current regulation speed component by performing the inverse operation of the partial derivative elements of the discharge current limiting threshold with respect to the regression speed in the Jacobian matrix. Add the discharge current limiting threshold increment to the current discharge current limiting threshold setting value to generate the second dynamic adjustment command of the discharge current limiting threshold.

[0037] The method further includes:

[0038] The current state of charge and temperature of the battery pack are obtained. Based on the state of charge and temperature, the voltage regulation effectiveness coefficient and the current regulation effectiveness coefficient are calculated. The voltage regulation effectiveness coefficient represents the effectiveness of the charging cut-off voltage adjustment on the regression trajectory control in the current state. The current regulation effectiveness coefficient represents the effectiveness of the discharge current limiting threshold adjustment on the regression trajectory control in the current state.

[0039] A second aspect of the present invention provides a predictive and control system for the charge and discharge state of a new energy vehicle battery pack, comprising:

[0040] The analysis module is used to acquire the real-time electrical parameters and historical operating data of the battery pack, couple and analyze the voltage response curve in the real-time electrical parameters and the capacity decay sequence in the historical operating data, and extract a dual-time-scale feature vector that simultaneously represents the instantaneous polarization state and the cumulative degradation degree; and construct a degradation baseline trajectory reflecting the battery pack capacity evolution with the number of cycles based on the historical operating data.

[0041] The determination module is used to calculate the deviation between the position of the dual-time-scale feature vector in the state space and the degradation reference trajectory based on the cumulative degradation degree component in the dual-time-scale feature vector to obtain the state deviation degree; and to determine the safety boundary constraint corresponding to the current operating point of the battery pack based on the instantaneous polarization state component in the dual-time-scale feature vector.

[0042] The execution module is used to generate a charge / discharge control command that enables the battery pack to return to the degradation baseline trajectory without triggering the safety boundary constraint, based on the deviation direction of the state deviation and the constraint margin of the safety boundary constraint, using a multi-objective optimization algorithm when the state deviation exceeds a preset deviation threshold; and to perform power adjustment on the charge / discharge circuit of the battery pack based on the charge / discharge control command.

[0043] A third aspect of the present invention provides an electronic device, comprising:

[0044] processor;

[0045] Memory used to store processor-executable instructions;

[0046] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0047] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0048] The beneficial effects of this application are as follows:

[0049] By coupling and analyzing the voltage response curve and the capacity decay sequence, dual-timescale feature vectors are extracted, achieving a comprehensive characterization of the instantaneous polarization state and cumulative degradation degree of the battery pack, overcoming the limitation of existing technologies that only focus on characteristics of a single timescale.

[0050] By constructing a degradation baseline trajectory based on historical operating data and calculating the state deviation, the degree of deviation from the battery pack's health state can be accurately assessed, improving the accuracy of battery pack state monitoring. Innovatively, instantaneous polarization state components are used to determine safety boundary constraints, making the safety constraints more dynamic and precise, effectively avoiding the problem of overly conservative or risky safety boundaries in traditional methods.

[0051] When the deviation from the state exceeds the threshold, this method can generate charge and discharge control commands through a multi-objective optimization algorithm based on the margin between the deviation direction and the safety boundary constraints. This enables the battery pack to regress to the degradation baseline trajectory while ensuring that the safety boundary constraints are not triggered, thereby improving the battery pack's service life and operational safety.

[0052] This method organically combines real-time status monitoring of the battery pack with long-term degradation trajectory control to form a closed-loop control system. It can actively intervene in the working state of the battery pack, significantly extend the battery pack's service life, reduce safety risks, and improve the overall performance and reliability of new energy vehicles. Attached Figure Description

[0053] Figure 1 This is a flowchart illustrating the method for predicting and controlling the charge and discharge state of a new energy vehicle battery pack according to an embodiment of the present invention.

[0054] Figure 2 This is a flowchart of electrochemical impedance spectroscopy time-series analysis and dual-timescale feature extraction in an embodiment of the present invention. Detailed Implementation

[0055] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0056] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0057] Figure 1 This is a flowchart illustrating the method for predicting and controlling the charge and discharge state of a new energy vehicle battery pack according to an embodiment of the present invention. Figure 1 As shown, the method includes:

[0058] The real-time electrical parameters and historical operating data of the battery pack are acquired. The voltage response curve in the real-time electrical parameters and the capacity decay sequence in the historical operating data are coupled and analyzed to extract a dual-time-scale feature vector that simultaneously represents the instantaneous polarization state and the cumulative degradation degree. Based on the historical operating data, a degradation baseline trajectory reflecting the evolution of the battery pack capacity with the number of cycles is constructed.

[0059] Based on the cumulative degradation component in the dual-timescale feature vector, the deviation between the position of the dual-timescale feature vector in the state space and the degradation reference trajectory is calculated to obtain the state deviation; based on the instantaneous polarization state component in the dual-timescale feature vector, the safety boundary constraint corresponding to the current operating point of the battery pack is determined.

[0060] When the state deviation exceeds a preset deviation threshold, a charge / discharge control command is generated using a multi-objective optimization algorithm based on the deviation direction of the state deviation and the constraint margin of the safety boundary constraint. This command enables the battery pack to return to the degradation baseline trajectory without triggering the safety boundary constraint. Power regulation is then performed on the charge / discharge circuit of the battery pack based on the charge / discharge control command.

[0061] In one optional implementation, the voltage response curve in the real-time electrical parameters is coupled and analyzed with the capacity decay sequence in the historical operating data to extract a dual-timescale feature vector that simultaneously characterizes the instantaneous polarization state and the cumulative degradation degree, including:

[0062] A time-series analysis was performed on the amplitude variation trajectory of the electrochemical polarization component in the voltage response curve over multiple historical charge-discharge cycles to obtain the polarization evolution curve of the electrochemical polarization component as a function of the number of cycles.

[0063] A time-series analysis was performed on the numerical change of the internal resistance growth rate in the capacity decay sequence during the same historical charge-discharge cycles to obtain the impedance evolution curve of the internal resistance growth rate as a function of the number of cycles.

[0064] Align the polarization evolution curve and the impedance evolution curve on the time axis, and calculate the cross-correlation coefficient between the rate of change of the polarization evolution curve and the rate of change of the impedance evolution curve;

[0065] Based on the cross-correlation coefficient, the influence weight of the electrochemical polarization component on the internal resistance growth rate is determined, and a state correlation matrix is ​​constructed with the electrochemical polarization component as the row index, the internal resistance growth rate as the column index, and the influence weight as the matrix element.

[0066] The dual-time-scale feature vector is generated by weighting and fusing the multi-level voltage decay components and the degradation rate parameters based on the state correlation matrix.

[0067] like Figure 2 As shown, the method includes:

[0068] By coupling and analyzing real-time electrical parameters and historical operating data of the battery, dual-time-scale feature vectors are extracted to simultaneously characterize the instantaneous polarization state and cumulative degradation degree of the battery, providing a key basis for accurately estimating the remaining battery life.

[0069] The system acquires real-time electrical parameters and historical operating data of the battery. Real-time electrical parameters mainly include the voltage response curve, reflecting the instantaneous voltage changes during charging and discharging. Historical operating data mainly includes the capacity decay sequence, recording the gradual reduction of battery capacity over long-term use. These two types of data represent the battery's performance characteristics at different time scales.

[0070] Time-series analysis was performed on the electrochemical polarization component in the voltage response curve. This component is a crucial part of the voltage response curve and can typically be separated from the total voltage response using an equivalent circuit model. Specifically, a second-order RC equivalent circuit model was used to identify the parameters of the original voltage curve, separating the voltage drop caused by ohmic impedance, the voltage drop caused by electrochemical polarization, and the voltage drop caused by concentration polarization. For multiple historical charge-discharge cycles, the maximum amplitude of the electrochemical polarization component in each cycle was extracted, forming a sequence of data varying with the number of cycles, i.e., the polarization evolution curve. For example, the maximum value of the electrochemical polarization component in the first cycle was 35 mV, in the second it was 37 mV, and so on, up to the Nth cycle.

[0071] A time-series analysis of the internal resistance growth rate in the capacity decay sequence is performed. Based on historical capacity data, the capacity difference between two adjacent cycles is calculated, and divided by the cycle interval to obtain the capacity decay rate. The internal resistance growth rate and the capacity decay rate show a positive correlation and can be derived from the capacity decay rate through a linear transformation. Specifically, for each cycle point, the ratio of the current internal resistance value to the initial internal resistance value is calculated, and then the derivative of this ratio with respect to the number of cycles is calculated; this is the internal resistance growth rate. These rate values ​​are arranged according to the number of cycles to form an impedance evolution curve.

[0072] Align the polarization evolution curve and the impedance evolution curve on the time axis. Since the sampling points of the two curves are not completely consistent, interpolation is required to ensure that there are corresponding data values ​​at the same cycle points. Cubic spline interpolation is used to map the two curves onto a unified time axis, so that they have the same number of data points and corresponding cycle counts.

[0073] The cross-correlation coefficient between the rate of change of the polarization evolution curve and the rate of change of the impedance evolution curve is calculated. The first difference between the two curves is then calculated to obtain the rate of change series. The Pearson correlation coefficient is then used to calculate the degree of cross-correlation between the two rate of change series. The cross-correlation coefficient ranges from -1 to 1, with a larger absolute value indicating a stronger correlation. For example, a cross-correlation coefficient of 0.85 indicates a strong positive correlation between the change in the electrochemical polarization component and the change in the rate of internal resistance growth.

[0074] The influence weight of electrochemical polarization components on the rate of internal resistance growth is determined based on the cross-correlation coefficient. The absolute value of the cross-correlation coefficient is directly used as the initial influence weight, and then adjusted in combination with the characteristics of the battery chemical type. For example, for lithium iron phosphate batteries, the influence weight can be multiplied by an adjustment factor of 0.9; for ternary lithium batteries, it can be multiplied by an adjustment factor of 1.1 to reflect the differences in polarization characteristics between different types of batteries.

[0075] A state correlation matrix is ​​constructed, with electrochemical polarization components as row indices, internal resistance growth rates as column indices, and influence weights as matrix element values. The matrix dimension depends on the discretization precision; for example, dividing both polarization components and internal resistance rates into 10 levels results in a 10×10 state correlation matrix. Each element in the matrix represents the correlation strength between a specific polarization component value and a specific internal resistance growth rate, reflecting the battery's performance characteristics under different state combinations.

[0076] Based on the state correlation matrix, a weighted fusion of multi-level voltage decay components and degradation rate parameters is performed to generate a dual-timescale feature vector. Specifically, the currently measured electrochemical polarization component value and internal resistance growth rate value are located in the state correlation matrix, and their corresponding influence weights are extracted. The multi-level voltage decay components (including ohmic polarization, electrochemical polarization, and concentration polarization) and degradation rate parameters (including capacity decay rate and internal resistance growth rate) are then weighted and combined according to these weights. The generated feature vector typically has a dimension of 5 to 10, and simultaneously contains polarization information reflecting the battery's immediate response characteristics and impedance information reflecting its long-term degradation trend, achieving a dual-timescale feature representation.

[0077] In practical applications, this dual-timescale feature vector can effectively capture key features of battery performance degradation, providing more comprehensive and accurate input information for subsequent remaining life prediction. It is particularly suitable for battery system health management scenarios with complex and ever-changing operating conditions.

[0078] In one optional implementation, the deviation between the position of the dual-timescale feature vector in the state space and the degraded reference trajectory is calculated to obtain the state deviation, including:

[0079] Based on the coordinate values ​​of the dual-time-scale feature vector in the state space, the current state point corresponding to the dual-time-scale feature vector is determined.

[0080] Based on the cumulative degradation degree component in the dual time-scale feature vector, the reference state point corresponding to the current cycle number is located on the degradation reference trajectory, and the forward neighborhood interval and backward neighborhood interval of the reference state point are extracted on the degradation reference trajectory.

[0081] The shortest distance from the current state point to the forward neighborhood interval is taken as the forward deviation distance, and the shortest distance from the current state point to the backward neighborhood interval is taken as the backward deviation distance.

[0082] The smaller of the forward deviation distance and the backward deviation distance is marked with a sign, and the deviation distance with the sign is taken as the state deviation degree. The magnitude of the state deviation degree represents the deviation range, and the positive or negative sign represents the deviation type.

[0083] The current state point is determined based on the coordinates of the dual-timescale feature vector in the state space. The dual-timescale feature vector includes a fast-timescale component and a slow-timescale component, where the slow-timescale component represents the cumulative degradation degree. In one embodiment, assuming the dual-timescale feature vector is V=(S, F1, F2), where S represents the cumulative degradation degree component, and F1 and F2 represent two fast-timescale state feature components, then the coordinates of the current state point are (S, F1, F2).

[0084] Based on the cumulative degradation component in the dual-timescale feature vector, the baseline state point corresponding to the current cycle number is located on the degradation baseline trajectory. For example, if the current cycle number is n and the cumulative degradation component is S, then the baseline state point BP(S) corresponding to the cumulative degradation degree S is found on the degradation baseline trajectory. The degradation baseline trajectory can be understood as the degradation path formed by the device under ideal operating conditions as usage time increases, and this trajectory can be obtained by fitting historical normal operating data.

[0085] After determining the reference state point on the degradation reference trajectory, the forward and backward neighborhood intervals of the reference state point are extracted. In one embodiment, let the coordinates of the reference state point BP(S) on the degradation reference trajectory be (S, BF1, BF2). The forward neighborhood interval is selected as the trajectory segment extending forward from the cumulative degradation degree S within a range of δS, that is, the trajectory segment with a degradation degree range of [S-δS, S]. The backward neighborhood interval is selected as the trajectory segment extending backward from the cumulative degradation degree S within a range of δS, that is, the trajectory segment with a degradation degree range of [S, S+δS]. The neighborhood range δS can be adjusted according to the specific application scenario and accuracy requirements, for example, it can be set to 1% or 5% of the number of cycles.

[0086] The shortest distance from the current state point to the forward neighborhood interval is calculated as the forward deviation distance DF, and the shortest distance from the current state point to the backward neighborhood interval is calculated as the backward deviation distance DB. The shortest distance can be calculated using the Euclidean distance from the point to the curve segment. Specifically, for each point in the forward neighborhood interval, its Euclidean distance to the current state point is calculated, and the minimum value is taken as the forward deviation distance DF; similarly, for each point in the backward neighborhood interval, its Euclidean distance to the current state point is calculated, and the minimum value is taken as the backward deviation distance DB.

[0087] In one embodiment, assuming the current state point coordinates are (S, F1, F2) and the coordinates of a point in the forward neighborhood interval are (Sf, BF1f, BF2f), then the Euclidean distance between the two points is:

[0088] ;

[0089] The smaller of the forward and backward deviation distances is marked with a sign, and the marked deviation distance is taken as the state deviation degree. The marking rules are as follows: if the forward deviation distance is less than the backward deviation distance, the state deviation degree is negative, indicating that the equipment state degrades ahead of the reference trajectory; if the backward deviation distance is less than the forward deviation distance, the state deviation degree is positive, indicating that the equipment state degrades behind the reference trajectory; if the two are equal, a value of zero can be taken, indicating that the equipment state is exactly on the reference trajectory.

[0090] The state deviation calculated above includes both deviation magnitude and deviation type information: the numerical value represents the deviation magnitude, and the positive or negative sign represents the deviation type. In practical applications, the sign and numerical value of the state deviation can be used to determine whether the device is in an abnormal state. For example, when the absolute value of the state deviation exceeds a preset threshold, the device can be determined to be in an abnormal state; a positive value indicates that the device is in a lagging state of degradation, while a negative value indicates that the device is in a premature state of degradation.

[0091] In aero-engine health monitoring applications, a dual-timescale feature vector can include the engine's cumulative flight hours as a cumulative degradation component, and exhaust temperature and fuel flow rate as fast-timescale state feature components. By calculating the state deviation, abnormal changes in engine performance can be detected in a timely manner, providing a basis for maintenance decisions.

[0092] In one optional implementation, based on the deviation direction of the state deviation and the constraint margin of the safety boundary constraint, a charge / discharge control command is generated using a multi-objective optimization algorithm to enable the battery pack to return to the degradation baseline trajectory without triggering the safety boundary constraint. This includes:

[0093] Based on the deviation direction of the state deviation, determine the target regression speed required for the battery pack to return to the degradation baseline trajectory;

[0094] Based on the constraint margin of the safety boundary constraint, calculate the maximum allowable control intensity of the battery pack while maintaining a safety margin no lower than a preset safety threshold.

[0095] A constrained optimization problem is constructed with the target regression speed as the performance index and the maximum control intensity as the upper limit of the constraint. The constrained optimization problem is solved by the multi-objective optimization algorithm. When the constrained optimization problem has a feasible solution, the optimal regression speed that satisfies the constraint conditions is obtained. When the constrained optimization problem does not have a feasible solution, the target regression speed is iteratively decayed until the constrained optimization problem has a feasible solution, and the suboptimal regression speed under the constraint conditions is obtained.

[0096] The optimal or suboptimal regression speed is decomposed into a voltage regulation speed component and a current regulation speed component. A first dynamic adjustment command for the charging cutoff voltage is calculated based on the voltage regulation speed component, and a second dynamic adjustment command for the discharge current limiting threshold is calculated based on the current regulation speed component. The first dynamic adjustment command and the second dynamic adjustment command are combined to generate the charge and discharge regulation command.

[0097] The battery pack degradation baseline trajectory was established through historical operational data acquisition and processing. The data acquisition module continuously recorded state parameters of the battery pack, such as voltage, current, temperature, and capacity retention, at 10-second sampling intervals. The data was stored in a time-series database, retaining complete operational records for nearly two years. During the data preprocessing stage, outlier detection and filtering were performed on the acquired raw data. Data points with voltage exceeding ±5% of the rated value or current fluctuations greater than 100 amperes per second were marked as outliers and removed. Capacity degradation trend analysis employed a moving average algorithm, calculating the average capacity retention rate over a 30-day window to form a smooth degradation curve as the degradation baseline trajectory. This trajectory, with capacity retention rate on the ordinate and operating time on the abscissa, was fitted with a continuous curve using cubic spline interpolation, with interpolation accuracy controlled within 0.1%.

[0098] The deviation is calculated based on the difference between the current battery pack state and the corresponding time point on the degradation baseline trajectory. The real-time monitoring module acquires the current capacity retention rate of the battery pack every minute and uses linear interpolation to find the baseline value at the corresponding time point on the degradation baseline trajectory. The deviation is calculated as the difference between the current value and the baseline value. A positive value indicates that the battery pack state is better than the expected trajectory, while a negative value indicates that it is worse than the expected trajectory. The direction of deviation is determined with zero as the dividing point; an absolute value greater than 0.5% is considered a significant deviation requiring adjustment. Deviation data is stored in floating-point format with precision to three decimal places. Historical deviation records are used for trend analysis and optimization of adjustment strategies.

[0099] The safety boundary constraints cover multiple dimensions, including voltage, current, and temperature. The voltage safety boundary is set between 85% and 115% of the rated voltage; the current safety boundary is between -120% and +120% of the rated current; and the temperature safety boundary is between -10°C and 60°C. The constraint margin is calculated as the ratio of the distance from the current state value to the safety boundary to the safety boundary range. In the voltage constraint margin calculation, when the current voltage is 4.1 volts and the safety upper limit is 4.2 volts, the constraint margin is 0.1 divided by 0.3, which equals 0.333. The preset safety threshold is set to 0.2, meaning a safety warning is triggered when the constraint margin falls below 20%. Control commands must ensure that the adjusted constraint margin does not fall below this threshold.

[0100] The target regression rate is determined based on a comprehensive analysis of the deviation direction and magnitude. The regression rate calculation employs a proportional control strategy: the regression rate equals the deviation multiplied by a proportional coefficient. The default value for the proportional coefficient is 0.1, which can be adjusted within the range of 0.05 to 0.3 depending on the battery pack characteristics. When the deviation is positive 2%, the target regression rate is negative 0.2% per day, indicating that the degradation rate needs to be appropriately accelerated to bring the battery pack back to the baseline trajectory. The absolute value of the regression rate is limited to within 0.5% per day to avoid excessive adjustment that could damage the battery pack. The regression rate value is expressed as a percentage per day, and floating-point operations are used to maintain calculation accuracy.

[0101] The maximum control strength is calculated based on the minimum value of the constraint margins in each dimension. The voltage control strength is limited to the current voltage constraint margin multiplied by a safety factor of 0.8, and the current control strength is limited to the current current constraint margin multiplied by a safety factor of 0.8. The most stringent constraint among the multiple dimensions is taken as the maximum control strength to ensure that no dimension violates the safety boundary. The control strength value ranges from 0 to 1, representing the percentage of available controllability. When the voltage constraint margin is 0.3 and the current constraint margin is 0.5, the maximum control strength is the smaller value, 0.3 multiplied by 0.8, which equals 0.24.

[0102] The constrained optimization problem is constructed with the target regression velocity as the optimization objective and the maximum control strength as the upper limit of the constraint. The optimization variable is the actual regression velocity, and the objective function is the square of the difference between the actual regression velocity and the target regression velocity. The constraints include that the absolute value of the actual regression velocity does not exceed the maximum control strength, and that the state parameters after control do not violate the safety boundary. The multi-objective optimization algorithm adopts the particle swarm optimization method, with the number of particles set to 50, the upper limit of the number of iterations set to 200, the initial value of the inertia weight to be 0.9, the cognitive learning factor to be 2.0, and the social learning factor to be 2.0. The algorithm convergence criteria are that the change of the optimal solution is less than 0.001 in 10 consecutive iterations or the upper limit of iterations is reached.

[0103] Feasibility determination is achieved through constraint verification. In each iteration, the algorithm checks whether the current solution satisfies all constraints; solutions that meet the conditions are recorded as candidate feasible solutions. If a feasible solution exists after the iteration, the solution with the smallest objective function value is selected as the optimal regression rate. When constraints are too strict, resulting in no feasible solution, a target regression rate decay procedure is initiated. The decay strategy involves multiplying the target regression rate by a decay factor of 0.9 and then reconstructing the optimization problem for solving. The decay process is executed a maximum of 10 times, with the optimization algorithm rerun after each decay until a feasible solution is found or the maximum number of decay iterations is reached. While the obtained suboptimal regression rate deviates from the original objective, it ensures executability under safe constraints.

[0104] The regression velocity decomposition employs a vector decomposition method, decomposing the optimal or near-optimal regression velocity into voltage-regulating velocity components and current-regulating velocity components. The decomposition ratio is determined based on the weights of the influence of voltage and current on capacity decay, with the voltage component weight set to 0.6 and the current component weight set to 0.4. The voltage-regulating velocity component equals the regression velocity multiplied by the voltage weight, and the current-regulating velocity component equals the regression velocity multiplied by the current weight. The decomposition result maintains that the vector sum of the two components equals the original regression velocity, ensuring consistency in the regulation effect.

[0105] The calculation of the first dynamic adjustment command is based on the relationship between the voltage regulation rate component and the current charging cutoff voltage. The charging cutoff voltage adjustment is equal to the voltage regulation rate component multiplied by the voltage regulation sensitivity coefficient. The sensitivity coefficient is obtained through experimental calibration, with a typical value of 0.01 volts voltage adjustment per percentage of regression rate. When the voltage regulation rate component is -0.12% per day, the charging cutoff voltage adjustment is -0.0012 volts, indicating that the charging cutoff voltage needs to be reduced. The new charging cutoff voltage is equal to the current value plus the adjustment amount, and the adjusted voltage value must be within the safety boundary range. The first dynamic adjustment command includes fields such as the target charging cutoff voltage, effective time, and duration, and is transmitted to the battery management unit in the form of a data packet.

[0106] The second dynamic adjustment command is calculated based on the relationship between the current regulation rate component and the current discharge current limit threshold. The adjustment amount of the discharge current limit threshold is equal to the current regulation rate component multiplied by the current regulation sensitivity coefficient. A typical sensitivity coefficient value is 1 ampere current limit adjustment per percentage of the regression rate. When the current regulation rate component is -0.08% per day, the discharge current limit threshold adjustment amount is -0.8 amperes, indicating that the discharge current limit needs to be reduced. The new discharge current limit threshold is equal to the current value plus the adjustment amount, ensuring that the adjusted current limit value is within the equipment's rated range. The format of the second dynamic adjustment command is consistent with the first command, including key information such as the target discharge current limit threshold, effective time, and duration.

[0107] The combination of charge / discharge control commands is implemented through a command encapsulation module. The command encapsulation uses a unified data structure, including fields such as command type, target parameters, numerical range, timestamp, and checksum. The first dynamic adjustment command is marked as voltage type, and the second dynamic adjustment command is marked as current type. The two commands are combined in an array within a single data packet. A sequence number is added to the data packet for command tracking, and the priority field is set to high priority to ensure timely execution. The command activation time is uniformly set to 30 seconds after generation, allowing the battery management unit processing time. The combined charge / discharge control commands are sent to the battery management unit via the CAN bus, using a standard frame format, a data length of 8 bytes, and a communication rate of 500 kilobits per second.

[0108] In one optional implementation, the constrained optimization problem is solved using the multi-objective optimization algorithm. When a feasible solution exists for the constrained optimization problem, the optimal regression rate that satisfies the constraints is obtained. When no feasible solution exists for the constrained optimization problem, the objective regression rate is iteratively decayed until a feasible solution appears for the constrained optimization problem. Obtaining the suboptimal regression rate under the constraints includes:

[0109] Based on the target regression rate, the maximum control intensity, and the safety boundary constraints, and combined with the multi-objective optimization algorithm, a constrained optimization problem is constructed, which includes decision variables, objective functions, and constraints.

[0110] The multi-objective optimization algorithm is used to explore the feasible region of the constrained optimization problem and search for the range of values ​​of the decision variables that simultaneously satisfy all constraints. When the range of values ​​of the decision variables is not empty, it is determined that the constrained optimization problem has a feasible solution. Within the range of values ​​of the decision variables, the gradient descent method is used to search for the optimal charging current adjustment range and the optimal discharging current adjustment range that minimize the objective function. Based on the optimal charging current adjustment range and the optimal discharging current adjustment range, the optimal regression speed of the battery pack to regress the degradation baseline trajectory is calculated.

[0111] When the range of values ​​for the decision variables is empty, it is determined that there is no feasible solution to the constrained optimization problem. The target regression velocity is multiplied by a decay factor less than one to obtain the decayed target regression velocity. This process is repeated iteratively until the range of values ​​for the decision variables in the updated constrained optimization problem becomes non-empty. Within the range of values ​​for the decision variables at this time, the gradient descent method is used to search for the suboptimal charging current adjustment range and the suboptimal discharging current adjustment range that minimize the updated objective function, thus obtaining the suboptimal regression velocity.

[0112] The constrained optimization problem is constructed based on a multi-objective optimization algorithm framework. The decision variables are defined as two independent variables: the charging current adjustment amplitude and the discharging current adjustment amplitude. The range of the charging current adjustment amplitude is set from -50 amps to +50 amps, and the range of the discharging current adjustment amplitude is set from -40 amps to +40 amps, with numerical precision maintained to two decimal places. The decision variables are stored in a two-dimensional vector, with the charging current adjustment amplitude as the first component and the discharging current adjustment amplitude as the second component. The vector length is fixed at 2, and the data type is double-precision floating-point. The objective function is constructed as the weighted sum of squares of the difference between the actual regression rate and the target regression rate, with weight coefficients of 0.6 for the charging influence and 0.4 for the discharging influence. The actual regression rate is calculated by multiplying the charging current adjustment amplitude by the charging rate sensitivity coefficient and the discharging current adjustment amplitude by the discharging rate sensitivity coefficient. The sensitivity coefficients are set to correspond to 0.02% of the regression rate per amp and 0.015% of the regression rate per amp, respectively.

[0113] The constraints encompass the maximum regulation intensity limit, safety boundary constraints, and physical limitations on the current adjustment range. The maximum regulation intensity constraint is defined as the sum of the absolute values ​​of the charging current adjustment range and the discharging current adjustment range not exceeding the maximum regulation intensity value. Safety boundary constraints include the conditions that the adjusted charging current does not exceed the charging safety upper limit, the discharging current does not exceed the discharging safety upper limit, and the voltage remains within the safe range. The charging safety upper limit is set at 110% of the rated charging current, the discharging safety upper limit is set at 120% of the rated discharging current, and the voltage safety range is 90% to 110% of the rated voltage. The constraints are expressed as a system of inequalities, the number of which is determined by the safety boundary dimension, typically containing 6 to 8 constraint inequalities. The degree of constraint violation is calculated using the constraint function value; a constraint function value greater than zero indicates a constraint violation, while a value less than or equal to zero indicates that the constraint is satisfied.

[0114] Feasibility region detection employs a strategy combining grid search and boundary contraction. The decision variable space is divided into a 100x80 grid, with 100 sampling points for the charging current adjustment amplitude direction and 80 sampling points for the discharging current adjustment amplitude direction. The grid search process traverses all sampling points, checking if each point simultaneously satisfies all constraints. Points that meet the conditions are marked as feasible and recorded in the feasible point set. The feasible point set is stored using a dynamic array with an initial capacity of 1000, automatically expanding to 1.5 times the current capacity when storage space is insufficient. The boundary contraction strategy determines the feasible region boundary by identifying the minimum bounding rectangle of feasible points. The coordinates of the lower left and upper right corners of the bounding rectangle define the effective range of the feasible region. The computational complexity of feasible region detection is the number of grid points multiplied by the number of constraints, with the time cost of a single detection controlled within 100 milliseconds.

[0115] The determination of the range of decision variables is based on the size of the feasible point set. If the feasible point set is empty, the range of decision variables is considered empty, and the corresponding constraint optimization problem has no feasible solution. If the feasible point set is not empty, the range of decision variables is considered non-empty, and the corresponding constraint optimization problem has a feasible solution. The numerical representation of the range uses the coordinates of the minimum bounding rectangle, where the width and height of the rectangle represent the adjustable space of the two decision variables, respectively. The result of the feasibility determination is stored as a Boolean variable: True indicates the existence of a feasible solution, and False indicates the absence of a feasible solution. The computational overhead of the determination process mainly comes from feasible region probing. Parallel computing optimizes the execution efficiency of the grid search, with the number of threads set to 75% of the processor cores.

[0116] The gradient descent method is implemented based on the optimization of the objective function within the feasible region. Gradient calculation employs numerical differentiation with a differentiation step size of 0.01, and the dimension of the gradient vector is consistent with the dimension of the decision variables. The initial learning rate is set to 0.1, and an adaptive adjustment strategy is used. If the objective function value fails to improve after five consecutive iterations, the learning rate is multiplied by a decay factor of 0.8. The iteration starting point can be the center of the feasible region or a randomly selected feasible point; the selection strategy can be adjusted through configuration parameters. The iteration termination conditions include one of three cases: the magnitude of the gradient vector is less than 0.001, the change in the objective function value is less than 0.0001, or the maximum number of iterations (500) is reached. During gradient descent, it is crucial to ensure that the iteration points remain within the feasible region. Projection operations are used to pull out-of-bounds points back to the feasible region boundary.

[0117] The optimal charging current adjustment range and the optimal discharging current adjustment range are obtained through the convergence result of gradient descent. The decision variable value at the convergence point is the optimal adjustment range. The optimal charging current adjustment range corresponds to the first component of the decision variable vector, and the optimal discharging current adjustment range corresponds to the second component. The validity verification of the adjustment range includes constraint checks and objective function value calculation, ensuring that the optimal solution satisfies all constraints and the objective function value reaches a local minimum. The calculation of the optimal regression rate is based on the result of multiplying the optimal charging current adjustment range by the charging rate sensitivity coefficient and the optimal discharging current adjustment range by the discharging rate sensitivity coefficient. During the calculation, numerical precision is maintained to four decimal places to avoid accumulated errors affecting the accuracy of the regression rate.

[0118] The decay of the target regression rate employs an iterative decay strategy to handle cases where no feasible solution is found. The decay factor is set to 0.9. In each iteration, the current target regression rate is multiplied by the decay factor to obtain the decayed target regression rate. The decayed target regression rate is used to reconstruct the constrained optimization problem, updating the target regression rate parameters in the objective function while keeping the constraints unchanged. The maximum number of decay iterations is limited to 10 to avoid excessive decay that could lead to an excessively small regression rate, rendering the control meaningless. After each decay, the feasible region detection and feasible solution determination processes are repeated to check whether the updated constrained optimization problem produces a feasible solution. The decay iteration process records the target regression rate value and the existence of a feasible solution for each iteration, used for debugging and parameter tuning.

[0119] The solution process for the updated constrained optimization problem remains consistent with that of the initial problem, including feasible region exploration, gradient descent optimization, and optimal solution verification. The updated objective function uses the decayed objective regression rate as a reference value, while the form and weighting coefficients of the objective function remain unchanged. The suboptimal charging current adjustment magnitude and the suboptimal discharging current adjustment magnitude are obtained by searching within the updated feasible region using gradient descent. The quality of the suboptimal solution is evaluated by the objective function value and the degree of constraint violation; although the suboptimal solution deviates from the original objective, it ensures executability under the constraints. The calculation method for the suboptimal regression rate is consistent with that for the optimal regression rate, derived by weighted summation of the suboptimal charging current adjustment magnitude and the suboptimal discharging current adjustment magnitude using sensitivity coefficients.

[0120] In one optional implementation, the decision variables of the constrained optimization problem are the charging current adjustment magnitude and the discharging current adjustment magnitude. The objective function of the constrained optimization problem is to minimize the time required for the battery pack state point to move to the degradation reference trajectory along the deviation direction. The constraints of the constrained optimization problem include a control intensity constraint that the combined control intensity of the charging current adjustment magnitude and the discharging current adjustment magnitude does not exceed the maximum control intensity, a safety boundary constraint that the battery pack does not touch the safety boundary constraint during the control process, and a non-negativity constraint that both the charging current adjustment magnitude and the discharging current adjustment magnitude are non-negative values.

[0121] By acquiring the current state point information and degradation baseline trajectory information of the battery pack, the battery pack state point typically includes parameters such as state of charge (SOC) and state of health (SOH), which can be collected in real time by the battery management system (BMS). The degradation baseline trajectory refers to the pre-set ideal operating state trajectory of the battery, which is usually derived based on battery life models and usage scenarios.

[0122] After acquiring the status information, it is determined whether the current state point of the battery pack deviates from the degradation reference trajectory. If it does, the deviation direction between the state point and the degradation reference trajectory is calculated. The deviation direction can be represented as a vector pointing from the current state point to the nearest point on the degradation reference trajectory. This vector can be determined by calculating the distance from the current state point to each point on the degradation reference trajectory and finding the trajectory point corresponding to the minimum distance.

[0123] Once the deviation direction is determined, the maximum control intensity is set based on the battery pack's safety constraints and operating conditions. The maximum control intensity represents the maximum allowable range of adjustment for the charging and discharging currents, ensuring that the control process will not damage the battery. This parameter can be dynamically adjusted according to the battery type, ambient temperature, and the battery's current state.

[0124] A constrained optimization problem model is established, with the decision variables being the adjustment magnitudes of the charging current and discharging current. The objective function is defined as the minimum time required for the battery pack state point to move along the deviation direction to the degradation reference trajectory. This time minimization function can be expressed as the distance from the battery pack state point to the degradation reference trajectory divided by the state point movement rate, where the movement rate is related to the adjustment magnitudes of the charging current and discharging current.

[0125] The constraints include three aspects: First, the combined control intensity of the charging current adjustment amplitude and the discharging current adjustment amplitude shall not exceed the control intensity constraint of the maximum control intensity, that is, the sum of the squares of the two adjustment amplitudes shall not exceed the square of the maximum control intensity; Second, the constraint that the battery pack does not touch the safety boundary during the control process, ensuring that the battery state point is always kept within the safe operating range; Third, the non-negativity constraint that both the charging current adjustment amplitude and the discharging current adjustment amplitude are non-negative values.

[0126] When solving this constrained optimization problem, the Lagrange multiplier method or KKT conditions can be used for analysis. The solution process first involves constructing a Lagrange function, which includes the objective function and a weighted sum of all constraints. Then, by solving the system of equations where the partial derivatives of the Lagrange function with respect to the decision variables are zero, and combining this with complementary relaxation conditions, the optimal solution is obtained.

[0127] The calculated charging current adjustment range and discharging current adjustment range can be used to adjust the battery charging and discharging control strategy in real time. Specifically, the planned charging current is increased by the charging current adjustment range, and the planned discharging current is decreased by the discharging current adjustment range, thereby achieving precise control of the battery state.

[0128] In practical applications, a closed-loop control system can be designed to achieve continuous optimization. This system periodically collects battery state information, determines the deviation from the degradation baseline trajectory, solves the constrained optimization problem, and adjusts the charging and discharging current. The control cycle can be set according to the battery response characteristics and computing power, typically ranging from a few seconds to a few minutes.

[0129] To improve solution efficiency, numerical optimization algorithms such as gradient descent, interior-point methods, or sequential quadratic programming can be used. For embedded systems with limited computing resources, optimal adjustment strategies for different scenarios can be pre-calculated to form lookup tables for real-time control.

[0130] In electric vehicle applications, this method can significantly extend battery life. For example, after an electric vehicle has been driving at high speeds for an extended period, the battery's state point deviates from its optimal degradation trajectory due to increased temperature. In this case, by calculating the optimal charge and discharge current adjustment range, the on-board battery management system can appropriately reduce the discharge power or increase the cooling intensity, allowing the battery state point to gradually return to the ideal trajectory while ensuring that normal vehicle operation is not affected.

[0131] In photovoltaic energy storage systems, this method is also applicable to optimizing battery operating status. When changes in illumination conditions or fluctuations in electricity load are predicted, the system can calculate the charging and discharging current adjustment strategy in advance, so that the battery always works near the optimal degradation trajectory, which not only meets the energy dispatch requirements but also extends the battery life.

[0132] By using the above-mentioned constraint optimization method, precise control of the battery pack state can be achieved, minimizing the battery state correction time, improving battery management efficiency, and extending battery life while ensuring safety.

[0133] In one optional implementation, the optimal regression rate or the suboptimal regression rate is decomposed into a voltage-controlled rate component and a current-controlled rate component. A first dynamic adjustment command for the charging cutoff voltage is calculated based on the voltage-controlled rate component, and a second dynamic adjustment command for the discharge current-limiting threshold is calculated based on the current-controlled rate component.

[0134] Construct the Jacobian matrix of charge and discharge parameters-regression rate of the battery pack at the current state point. The Jacobian matrix includes the partial derivative elements of the charging cut-off voltage with respect to the regression rate and the partial derivative elements of the discharge current limiting threshold with respect to the regression rate. The partial derivative elements are obtained by numerically differentiating the relationship between the changes in charge and discharge parameters of the battery pack and the changes in regression rate at multiple historical state points.

[0135] Calculate the singular value decomposition of the Jacobian matrix to obtain a first singular vector representing the dominant direction of voltage regulation and a second singular vector representing the dominant direction of current regulation. Project the optimal regression velocity or the suboptimal regression velocity onto the orthogonal subspace spanned by the first singular vector and the second singular vector to obtain the voltage regulation velocity component along the first singular vector and the current regulation velocity component along the second singular vector.

[0136] The voltage regulation speed component is multiplied by the reciprocal of the predetermined voltage regulation effectiveness coefficient. The charging cut-off voltage increment required to realize the voltage regulation speed component is obtained by inverse operation of the partial derivative elements of the charging cut-off voltage with respect to the regression speed in the Jacobian matrix. The charging cut-off voltage increment is then superimposed on the current charging cut-off voltage setting value to generate the first dynamic adjustment command of the charging cut-off voltage.

[0137] Multiply the current regulation speed component by the reciprocal of the predetermined current regulation effectiveness coefficient, and obtain the discharge current limiting threshold increment required to realize the current regulation speed component by performing the inverse operation of the partial derivative elements of the discharge current limiting threshold with respect to the regression speed in the Jacobian matrix. Add the discharge current limiting threshold increment to the current discharge current limiting threshold setting value to generate the second dynamic adjustment command of the discharge current limiting threshold.

[0138] The construction of the charge / discharge parameter-regression rate Jacobian matrix is ​​based on statistical analysis of historical battery pack operating data. The data acquisition module extracts nearly 6 months of charge / discharge parameter change records and corresponding regression rate change records from the battery management database. The data recording interval is 24 hours, and the amount of data extracted at one time is approximately 180 historical state points. The selection criteria for historical state points include valid data points with a temperature range of 15 to 45 degrees Celsius, a battery pack capacity retention rate of 70% to 95%, and a charge / discharge parameter change amplitude greater than 1%. In the data preprocessing stage, outlier detection and smoothing filtering are performed on the raw data. Outliers are removed using a rule of 3 times the standard deviation, and the moving average window length is set to 7 days to ensure the temporal continuity and statistical validity of the data.

[0139] The numerical differential calculation employs the central difference method to obtain partial derivative elements. In calculating the partial derivative of the charging cut-off voltage with respect to the regression velocity, adjacent historical state points with a charging cut-off voltage change of 0.01 volts are selected, and the partial derivative value is obtained by dividing the regression velocity change by the charging cut-off voltage change. Similarly, in calculating the partial derivative of the discharge current limiting threshold with respect to the regression velocity, adjacent historical state points with a discharge current limiting threshold change of 1 ampere are selected, and the partial derivative value is obtained by dividing the regression velocity change by the discharge current limiting threshold change. The precision of the differential calculation is controlled to five decimal places, and multi-point averaging is used to reduce the impact of numerical errors on the accuracy of the partial derivatives. The Jacobian matrix has a dimension of 2 x 1. The first row contains the partial derivative of the charging cut-off voltage with respect to the regression velocity, and the second row contains the partial derivative of the discharge current limiting threshold with respect to the regression velocity. The matrix elements are of double-precision floating-point data type.

[0140] The Jacobian matrix of the current state point is estimated from historical data using an interpolation method. The state point matching algorithm searches for the five most similar neighboring state points in the historical state point set based on feature vectors such as the battery pack's current capacity retention rate, average temperature, and charge / discharge cycle count. Similarity calculation uses Euclidean distance as the metric, with distance weights of 0.5 for capacity retention rate, 0.3 for temperature, and 0.2 for cycle count. Interpolation uses an inverse distance weighting method, where the Jacobian matrix elements of neighboring state points are weighted by the inverse of their distances to obtain the Jacobian matrix of the current state point. The validity of the interpolation results is verified through residual analysis, with the interpolation error controlled within 5%. If the error exceeds this range, data re-collection and model retraining are triggered.

[0141] The singular value decomposition (SVD) is implemented using the classic SVD algorithm to decompose the Jacobian matrix. Since the Jacobian matrix is ​​a 2x1 matrix, the SVD result contains a singular value and its corresponding left and right singular vectors. The first singular vector corresponds to the left singular vector of the largest singular value, representing the dominant direction of voltage regulation's influence on the regression rate. This vector has a 2D dimension, and numerical normalization ensures its magnitude is 1. The second singular vector is obtained through orthogonal complement space construction, ensuring it is orthogonal to the first singular vector and spans a complete 2D space. The numerical stability of the SVD is guaranteed by a condition number check. When the condition number exceeds 1000, matrix regularization is triggered, with the regularization parameter set to 0.001.

[0142] Orthogonal subspace projection calculations decompose the optimal or suboptimal regression velocity into voltage-controlled velocity components and current-controlled velocity components. The projection operation is implemented using inner product calculations: the voltage-controlled velocity component equals the inner product of the regression velocity vector and the first singular vector multiplied by the first singular vector; the current-controlled velocity component equals the inner product of the regression velocity vector and the second singular vector multiplied by the second singular vector. During the projection calculation, numerical precision is maintained to four decimal places to avoid the accumulation of rounding errors affecting component accuracy. Component orthogonality is verified by calculating the inner product of the two component vectors; orthogonality is considered satisfied when the absolute value of the inner product is less than 0.0001. The integrity of the projection results is checked by calculating the difference between the component vector sum and the original regression velocity vector; the projection integrity is considered satisfied when the magnitude of the difference vector is less than 0.001.

[0143] The effectiveness coefficient of voltage regulation is determined based on statistical analysis of the impact of voltage regulation on the regression speed of the battery pack. The effectiveness coefficient is obtained through regression analysis, using the change in charging cut-off voltage as the independent variable and the change in regression speed as the dependent variable. A linear regression model is fitted to obtain the regression coefficient, which serves as the effectiveness coefficient. Coefficient calibration employs the least squares method, using historical voltage regulation data from the past three months for fitting, with a sample size of no less than 50 valid data points. The typical value range of the voltage regulation effectiveness coefficient is 0.8 to 1.2, with a default value set at 1.0. The coefficient update cycle is 30 days, and a sliding window method is used to maintain the timeliness of the coefficient. Anomaly detection is achieved through confidence interval determination; when the coefficient exceeds the range of 0.5 to 2.0, an anomaly warning and coefficient recalibration are triggered.

[0144] The calculation of the charging cutoff voltage increment is based on the inverse operation of the voltage regulation speed component and the partial derivative elements in the Jacobian matrix. In the inverse operation, the charging cutoff voltage increment equals the voltage regulation speed component divided by the voltage regulation effectiveness coefficient, and then divided by the partial derivative element of the charging cutoff voltage with respect to the regression speed. The numerical stability of the division operation is ensured by detecting the zero value of the denominator. When the absolute value of the partial derivative element is less than 0.0001, the default value of 0.01 is used to avoid numerical overflow. The rationality verification of the increment calculation result includes numerical range checking and physical meaning verification. When the absolute value of the increment exceeds 0.5 volts, a limiting process is triggered, with the limiting range set to -0.5 volts to +0.5 volts. The increment precision is maintained to three decimal places, and values ​​exceeding the precision are handled by rounding.

[0145] The current charging cutoff voltage setting is obtained through the battery management unit status query interface. The interface call is synchronous, with a timeout of 100 milliseconds. A retry strategy is implemented upon call failure, with a maximum of 3 retries. The validity of the setting value is verified through both numerical range and timestamp checks. A data refresh is triggered when the setting value exceeds the 3.0 to 4.5 volt range or when the timestamp exceeds 5 minutes. The incremental addition of the charging cutoff voltage uses floating-point addition, with the result maintained to three decimal places. The added charging cutoff voltage must meet safety boundary constraints; if it exceeds the safety range, boundary restrictions are applied to ensure the adjusted voltage remains within the device's allowable range.

[0146] The generation of the first dynamic adjustment command adopts a structured data format, including key fields such as command type, target voltage value, effective time, duration, and priority. The command type is identified as charging cut-off voltage adjustment, the target voltage value is the superimposed charging cut-off voltage, the effective time is set to 30 seconds after command generation, the duration is set to 24 hours, and the priority is set to high priority. The command data packet is encapsulated in JSON format, with field lengths limited to 12 characters for the target voltage value, 19 characters for the effective time, and 10 characters for the duration. Data packet integrity is guaranteed by CRC checksum, with a checksum length of 4 bytes; regeneration is triggered if checksum fails. Command transmission uses the TCP protocol, with the target address being the battery management unit IP address, port number 8080, and a transmission timeout set to 200 milliseconds.

[0147] The determination of the current regulation effectiveness coefficient employs the same statistical analysis method as the voltage regulation effectiveness coefficient. In regression analysis with the change in the discharge current limiting threshold as the independent variable and the change in the regression rate as the dependent variable, the typical range of the current regulation effectiveness coefficient is 0.7 to 1.3, with a default value set to 1.0. The coefficient calibration sample requires a discharge current limiting threshold change greater than 2 amperes and a sample size of no less than 40 valid data points. The coefficient update uses a recursive least squares algorithm, with a forgetting factor set to 0.95 to maintain sensitivity to recent data. The goodness of fit of the coefficient calibration is evaluated using the coefficient of determination. When the coefficient of determination falls below 0.7, model retraining is triggered, and the training data window is extended to 6 months.

[0148] The calculation of the discharge current limiting threshold increment is based on the inverse operation of the current regulation speed component and the partial derivative elements in the Jacobian matrix. In the inverse operation, the discharge current limiting threshold increment equals the current regulation speed component divided by the current regulation effectiveness coefficient, and then divided by the partial derivative element of the discharge current limiting threshold with respect to the regression speed. The numerical stability handling of the division operation is consistent with that of the voltage increment calculation; when the absolute value of the partial derivative element is less than 0.001, the default value of 0.1 is used instead. The limiting range for increment rationality verification is set to -20 amperes to +20 amperes; if it exceeds the range, truncation is performed. The increment accuracy is maintained to one decimal place to meet the engineering accuracy requirements of current control.

[0149] The current discharge current limiting threshold setting is obtained using the same interface as the charging cutoff voltage setting. Verification of the setting's validity includes checking the numerical range of 50 to 500 amperes and the timestamp's validity. The incremental summation of the discharge current limiting threshold must meet the device's rated power constraint; if the constraint is exceeded, power limit adjustment is performed. The precision control and safety boundary checks for the summation calculation are consistent with those for the charging cutoff voltage processing, ensuring that the adjusted current limiting threshold remains within a safe and controllable range.

[0150] The second dynamic adjustment command uses the same data structure and transmission protocol as the first dynamic adjustment command. The command type is identified as discharge current limiting threshold adjustment, and the target threshold is the superimposed discharge current limiting threshold. Other field configurations remain consistent. The coordination of the two dynamic adjustment commands is ensured through timestamp synchronization, with the effective time error controlled within 1 second to avoid abnormal battery pack operation caused by improper adjustment timing.

[0151] In one optional implementation, the method further includes:

[0152] The current state of charge and temperature of the battery pack are obtained. Based on the state of charge and temperature, the voltage regulation effectiveness coefficient and the current regulation effectiveness coefficient are calculated. The voltage regulation effectiveness coefficient represents the effectiveness of the charging cut-off voltage adjustment on the regression trajectory control in the current state. The current regulation effectiveness coefficient represents the effectiveness of the discharge current limiting threshold adjustment on the regression trajectory control in the current state.

[0153] Obtaining the current state of charge (SOC) and temperature status of a battery pack can be achieved by collecting parameters such as voltage, current, and temperature. Battery packs are typically equipped with a sensor network, including voltage acquisition units, current sensors, and temperature sensors, to monitor various parameters in real time. SOC represents the percentage of remaining charge in the battery, and can be calculated by the battery management system (BMS) using voltage-based, current-based, or combined estimation methods. Temperature status includes the temperature distribution across different areas of the battery pack; for large battery packs, multiple temperature points need to be monitored to obtain comprehensive temperature status information.

[0154] When calculating the voltage regulation effectiveness coefficient based on state of charge (SOC) and state of temperature (TRT), a voltage response model of the battery pack under different SOC and temperature conditions is established. This model can be expressed as a sensitivity function of voltage to changes in SOC. In the low SOC region (e.g., below 20%) or high SOC region (e.g., above 80%), the battery voltage is typically highly sensitive to changes in SOC, and adjusting the charging cut-off voltage has a significant impact on controlling the charging process. However, in the intermediate SOC region (e.g., 20%-80%), the voltage curve is relatively flat, and the effect of adjusting the charging cut-off voltage is relatively limited. Simultaneously, temperature also significantly affects the battery's voltage characteristics, especially in low-temperature environments, where the battery's internal resistance increases, voltage fluctuations intensify, and the effectiveness of voltage regulation decreases.

[0155] The voltage regulation effectiveness coefficient can be calculated as follows: First, determine the SOC sensitivity coefficient, which takes a larger value in the high and low SOC regions and a smaller value in the intermediate SOC region. Then, determine the temperature influence coefficient, which takes a larger value within the suitable temperature range (e.g., 15℃-35℃) and a smaller value under extremely low or high temperature conditions. Finally, weightedly combine the SOC sensitivity coefficient and the temperature influence coefficient to obtain the voltage regulation effectiveness coefficient. For example, when the battery SOC is 85% and the temperature is 25℃, the voltage regulation effectiveness coefficient is close to 0.9 (out of 1), indicating that adjusting the charging cut-off voltage can effectively control the charging process. However, when the SOC is 50% and the temperature is -10℃, the coefficient is only 0.3, indicating that the voltage regulation effect is limited.

[0156] The calculation of the current regulation effectiveness coefficient is similar to that of the voltage regulation effectiveness coefficient, but it focuses more on evaluating the impact of current limiting on battery behavior during discharge. In high-temperature environments (e.g., above 40°C), limiting the discharge current is significantly effective in protecting the battery from thermal runaway, resulting in a high current regulation effectiveness coefficient. However, in low-temperature environments (e.g., below 0°C), the battery's internal resistance is already high, naturally limiting its discharge capacity. Therefore, further limiting the discharge current has less marginal benefit, leading to a relatively low current regulation effectiveness coefficient. Furthermore, the state of charge (SOC) also affects the effectiveness of current regulation. In low SOC regions (e.g., below 20%), limiting the discharge current effectively prevents over-discharge and protects the battery; while in high SOC regions (e.g., above 80%), limiting the discharge current has a relatively smaller effect on battery protection.

[0157] The current regulation effectiveness coefficient can be calculated by assessing the impact of current on the internal temperature rise of the battery. First, a model is established to show the relationship between battery temperature rise and discharge current, reflecting the rate of temperature change of the battery under different currents. Then, considering the current state of charge (SOC) and ambient temperature, the effectiveness of limiting the discharge current in controlling temperature rise is evaluated. For example, when the battery is at a low SOC and the ambient temperature is high, the polarization effect during discharge is more pronounced. In this case, limiting the discharge current can effectively reduce internal heat generation, resulting in a higher current regulation effectiveness coefficient.

[0158] In practical applications, lookup tables or response surfaces for voltage regulation effectiveness coefficients and current regulation effectiveness coefficients can be established, allowing direct lookup of the corresponding coefficient values ​​based on real-time monitored SOC and temperature. Alternatively, machine learning methods can be used to train models based on historical operating data, enabling adaptive calculation of the coefficients.

[0159] After obtaining the voltage regulation effectiveness coefficient and the current regulation effectiveness coefficient, the control strategy for the charging cut-off voltage and the discharging current limiting threshold can be dynamically adjusted based on the relative magnitude of these two coefficients. When the voltage regulation effectiveness coefficient is significantly higher than the current regulation effectiveness coefficient, the battery state should be controlled to return to the expected trajectory by adjusting the charging cut-off voltage first; conversely, the discharging current limiting threshold should be adjusted first. Through this adaptive control strategy, the most effective control method can be selected according to the real-time state of the battery pack, improving the control effect of the battery management system and the service life of the battery pack.

[0160] The voltage regulation effectiveness coefficient and the current regulation effectiveness coefficient can also be used for smooth transitions in control strategies. For example, when switching from primarily relying on voltage regulation to primarily relying on current regulation, the changing trends of the two coefficients can be used to achieve a gradual switch in control mode, avoiding system instability caused by abrupt changes in control strategy.

[0161] A second aspect of the present invention provides a predictive and control system for the charge and discharge state of a new energy vehicle battery pack, comprising:

[0162] The analysis module is used to acquire the real-time electrical parameters and historical operating data of the battery pack, couple and analyze the voltage response curve in the real-time electrical parameters and the capacity decay sequence in the historical operating data, and extract a dual-time-scale feature vector that simultaneously represents the instantaneous polarization state and the cumulative degradation degree; and construct a degradation baseline trajectory reflecting the battery pack capacity evolution with the number of cycles based on the historical operating data.

[0163] The determination module is used to calculate the deviation between the position of the dual-time-scale feature vector in the state space and the degradation reference trajectory based on the cumulative degradation degree component in the dual-time-scale feature vector to obtain the state deviation degree; and to determine the safety boundary constraint corresponding to the current operating point of the battery pack based on the instantaneous polarization state component in the dual-time-scale feature vector.

[0164] The execution module is used to generate a charge / discharge control command that enables the battery pack to return to the degradation baseline trajectory without triggering the safety boundary constraint, based on the deviation direction of the state deviation and the constraint margin of the safety boundary constraint, using a multi-objective optimization algorithm when the state deviation exceeds a preset deviation threshold; and to perform power adjustment on the charge / discharge circuit of the battery pack based on the charge / discharge control command.

[0165] A third aspect of the present invention provides an electronic device, comprising:

[0166] processor;

[0167] Memory used to store processor-executable instructions;

[0168] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0169] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0170] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0171] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for predicting and controlling the state of charge and discharge of battery packs in new energy vehicles, characterized in that, include: The real-time electrical parameters and historical operating data of the battery pack are acquired. The voltage response curve in the real-time electrical parameters and the capacity decay sequence in the historical operating data are coupled and analyzed to extract a dual-time-scale feature vector that simultaneously represents the instantaneous polarization state and the cumulative degradation degree. Based on the historical operating data, a degradation baseline trajectory reflecting the evolution of the battery pack capacity with the number of cycles is constructed. Based on the cumulative degradation component in the dual-timescale feature vector, the deviation between the position of the dual-timescale feature vector in the state space and the degradation reference trajectory is calculated to obtain the state deviation; based on the instantaneous polarization state component in the dual-timescale feature vector, the safety boundary constraint corresponding to the current operating point of the battery pack is determined. When the state deviation exceeds a preset deviation threshold, based on the deviation direction of the state deviation and the constraint margin of the safety boundary constraint, a multi-objective optimization algorithm is used to generate a charge / discharge control command that returns the battery pack to the degradation baseline trajectory without triggering the safety boundary constraint, including: Based on the deviation direction of the state deviation, determine the target regression speed required for the battery pack to return to the degradation baseline trajectory; Based on the constraint margin of the safety boundary constraint, calculate the maximum allowable control intensity of the battery pack while maintaining a safety margin no lower than a preset safety threshold. A constrained optimization problem is constructed with the target regression speed as the performance index and the maximum control intensity as the upper limit of the constraint. The constrained optimization problem is solved by the multi-objective optimization algorithm. When the constrained optimization problem has a feasible solution, the optimal regression speed that satisfies the constraint conditions is obtained. When the constrained optimization problem does not have a feasible solution, the target regression speed is iteratively decayed until the constrained optimization problem has a feasible solution, and the suboptimal regression speed under the constraint conditions is obtained. The optimal regression speed or the suboptimal regression speed is decomposed into a voltage regulation speed component and a current regulation speed component. A first dynamic adjustment command for the charging cut-off voltage is calculated based on the voltage regulation speed component. A second dynamic adjustment command for the discharge current limiting threshold is calculated based on the current regulation speed component. The first dynamic adjustment command and the second dynamic adjustment command are combined to generate the charge and discharge regulation command. The power of the battery pack's charging and discharging circuit is adjusted based on the charging and discharging control command.

2. The method according to claim 1, characterized in that, The voltage response curve in the real-time electrical parameters is coupled and analyzed with the capacity decay sequence in the historical operating data to extract a dual-timescale feature vector that simultaneously characterizes the instantaneous polarization state and the cumulative degradation degree, including: A time-series analysis was performed on the amplitude variation trajectory of the electrochemical polarization component in the voltage response curve over multiple historical charge-discharge cycles to obtain the polarization evolution curve of the electrochemical polarization component as a function of the number of cycles. A time-series analysis was performed on the numerical change of the internal resistance growth rate in the capacity decay sequence during the same historical charge-discharge cycles to obtain the impedance evolution curve of the internal resistance growth rate as a function of the number of cycles. Align the polarization evolution curve and the impedance evolution curve on the time axis, and calculate the cross-correlation coefficient between the rate of change of the polarization evolution curve and the rate of change of the impedance evolution curve; Based on the cross-correlation coefficient, the influence weight of the electrochemical polarization component on the internal resistance growth rate is determined, and a state correlation matrix is ​​constructed with the electrochemical polarization component as the row index, the internal resistance growth rate as the column index, and the influence weight as the matrix element. The dual-time-scale feature vector is generated by weighting and fusing the multi-level voltage decay components and degradation rate parameters based on the state correlation matrix.

3. The method according to claim 1, characterized in that, The deviation between the position of the dual-timescale feature vector in the state space and the degenerate baseline trajectory is calculated to obtain the state deviation, which includes: Based on the coordinates of the dual-time-scale feature vector in the state space, determine the current state point corresponding to the dual-time-scale feature vector; Based on the cumulative degradation degree component in the dual time-scale feature vector, the reference state point corresponding to the current cycle number is located on the degradation reference trajectory, and the forward neighborhood interval and backward neighborhood interval of the reference state point are extracted on the degradation reference trajectory. The shortest distance from the current state point to the forward neighborhood interval is taken as the forward deviation distance, and the shortest distance from the current state point to the backward neighborhood interval is taken as the backward deviation distance. The smaller of the forward deviation distance and the backward deviation distance is marked with a sign, and the deviation distance with the sign is taken as the state deviation degree. The magnitude of the state deviation degree represents the deviation range, and the positive or negative sign represents the deviation type.

4. The method according to claim 1, characterized in that, The constrained optimization problem is solved using the multi-objective optimization algorithm. When a feasible solution exists for the constrained optimization problem, the optimal regression rate that satisfies the constraints is obtained. When no feasible solution exists for the constrained optimization problem, the objective regression rate is iteratively decayed until a feasible solution appears for the constrained optimization problem. The suboptimal regression rate under the constraints is obtained by: Based on the target regression rate, the maximum control intensity, and the safety boundary constraints, and combined with the multi-objective optimization algorithm, a constrained optimization problem is constructed, which includes decision variables, objective functions, and constraints. The multi-objective optimization algorithm is used to explore the feasible region of the constrained optimization problem and search for the range of values ​​of the decision variables that simultaneously satisfy all constraints. When the range of values ​​of the decision variables is not empty, it is determined that the constrained optimization problem has a feasible solution. Within the range of values ​​of the decision variables, the gradient descent method is used to search for the optimal charging current adjustment range and the optimal discharging current adjustment range that minimize the objective function. Based on the optimal charging current adjustment range and the optimal discharging current adjustment range, the optimal regression speed of the battery pack to regress the degradation baseline trajectory is calculated. When the range of values ​​for the decision variables is empty, it is determined that there is no feasible solution to the constrained optimization problem. The target regression velocity is multiplied by a decay factor less than one to obtain the decayed target regression velocity. This process is repeated iteratively until the range of values ​​for the decision variables in the updated constrained optimization problem becomes non-empty. Within the range of values ​​for the decision variables at this time, the gradient descent method is used to search for the suboptimal charging current adjustment range and the suboptimal discharging current adjustment range that minimize the updated objective function, thus obtaining the suboptimal regression velocity.

5. The method according to claim 4, characterized in that, The decision variables of the constrained optimization problem are the charging current adjustment magnitude and the discharging current adjustment magnitude. The objective function of the constrained optimization problem is to minimize the time required for the battery pack state point to move to the degradation reference trajectory along the deviation direction. The constraints of the constrained optimization problem include the control intensity constraint that the combined control intensity of the charging current adjustment magnitude and the discharging current adjustment magnitude does not exceed the maximum control intensity, the safety boundary constraint that the battery pack does not touch the safety boundary constraint during the control process, and the non-negativity constraint that both the charging current adjustment magnitude and the discharging current adjustment magnitude are non-negative values.

6. The method according to claim 1, characterized in that, The optimal or suboptimal regression rate is decomposed into a voltage-controlled rate component and a current-controlled rate component. A first dynamic adjustment command for the charging cutoff voltage is calculated based on the voltage-controlled rate component. A second dynamic adjustment command for the discharge current-limiting threshold is calculated based on the current-controlled rate component. Construct the Jacobian matrix of charge and discharge parameters-regression rate of the battery pack at the current state point. The Jacobian matrix includes the partial derivative elements of the charging cut-off voltage with respect to the regression rate and the partial derivative elements of the discharge current limiting threshold with respect to the regression rate. The partial derivative elements are obtained by numerically differentiating the relationship between the changes in charge and discharge parameters of the battery pack and the changes in regression rate at multiple historical state points. Calculate the singular value decomposition of the Jacobian matrix to obtain a first singular vector representing the dominant direction of voltage regulation and a second singular vector representing the dominant direction of current regulation. Project the optimal regression velocity or the suboptimal regression velocity onto the orthogonal subspace spanned by the first singular vector and the second singular vector to obtain the voltage regulation velocity component along the first singular vector and the current regulation velocity component along the second singular vector. The voltage regulation speed component is multiplied by the reciprocal of the predetermined voltage regulation effectiveness coefficient. The charging cut-off voltage increment required to realize the voltage regulation speed component is obtained by inverse operation of the partial derivative elements of the charging cut-off voltage with respect to the regression speed in the Jacobian matrix. The charging cut-off voltage increment is then superimposed on the current charging cut-off voltage setting value to generate the first dynamic adjustment command of the charging cut-off voltage. Multiply the current regulation speed component by the reciprocal of the predetermined current regulation effectiveness coefficient, and obtain the discharge current limiting threshold increment required to realize the current regulation speed component by performing the inverse operation of the partial derivative elements of the discharge current limiting threshold with respect to the regression speed in the Jacobian matrix. Add the discharge current limiting threshold increment to the current discharge current limiting threshold setting value to generate the second dynamic adjustment command of the discharge current limiting threshold.

7. The method according to claim 6, characterized in that, The method further includes: The current state of charge and temperature of the battery pack are obtained. Based on the state of charge and temperature, the voltage regulation effectiveness coefficient and the current regulation effectiveness coefficient are calculated. The voltage regulation effectiveness coefficient represents the effectiveness of the charging cut-off voltage adjustment on the regression trajectory control in the current state. The current regulation effectiveness coefficient represents the effectiveness of the discharge current limiting threshold adjustment on the regression trajectory control in the current state.

8. A new energy vehicle battery pack charge / discharge state prediction and control system, used to implement the method as described in any one of claims 1-7, characterized in that, include: The analysis module is used to acquire the real-time electrical parameters and historical operating data of the battery pack, couple and analyze the voltage response curve in the real-time electrical parameters and the capacity decay sequence in the historical operating data, and extract a dual-time-scale feature vector that simultaneously represents the instantaneous polarization state and the cumulative degradation degree; and construct a degradation baseline trajectory reflecting the battery pack capacity evolution with the number of cycles based on the historical operating data. The determination module is used to calculate the deviation between the position of the dual-time-scale feature vector in the state space and the degradation reference trajectory based on the cumulative degradation degree component in the dual-time-scale feature vector to obtain the state deviation degree; and to determine the safety boundary constraint corresponding to the current operating point of the battery pack based on the instantaneous polarization state component in the dual-time-scale feature vector. The execution module is used to generate a charge / discharge control command that enables the battery pack to return to the degradation baseline trajectory without triggering the safety boundary constraint, based on the deviation direction of the state deviation and the constraint margin of the safety boundary constraint, using a multi-objective optimization algorithm when the state deviation exceeds a preset deviation threshold; and to perform power adjustment on the charge / discharge circuit of the battery pack based on the charge / discharge control command.

9. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 7.

Images