Battery equivalent circuit model parameter identification method and system based on particle swarm optimization

Abstract

The invention belongs to the technical field of battery management, and relates to a battery equivalent circuit model parameter identification method and system based on particle swarm optimization, and the method comprises the steps: obtaining the dynamic test data of a battery; an OCV-SOC relation is obtained; calculating a real-time SOC sequence and performing interpolation to obtain a search OCV sequence; defining a second-order RC equivalent circuit model; constructing an objective function based on an error between prediction of the OCV sequence and search of the OCV sequence; and searching an optimal model parameter by adopting a particle swarm optimization algorithm and taking minimization of the target function as a target. By optimizing the matching degree of predicting the OCV and searching the OCV instead of directly fitting the terminal voltage, the sensitivity and accuracy of parameter identification on the dynamic characteristics of the battery are improved, and efficient and accurate identification on the parameters of the second-order RC model is realized in combination with the global optimization capacity and robustness design of PSO (Particle Swarm Optimization).

Description

Technical Field

[0001] The present invention relates to the technical field of battery management, and in particular to a battery equivalent circuit model parameter identification method and system based on particle swarm optimization. Background Art

[0002] The rapid development of electric vehicles, portable electronic devices, and large-scale energy storage systems is driving increasing demand for high-performance, high-safety rechargeable batteries, particularly lithium-ion batteries. Battery management systems (BMS) are central to ensuring safe, reliable, and efficient battery operation. Key functions include accurate estimation of state of charge (SOC) and state of health (SOH), as well as charge and discharge control, thermal management, and balancing. The implementation of these functions relies heavily on accurate battery models.

[0003] Equivalent circuit models (ECMs) are widely used in battery management systems (BMSs) due to their simple structure, clear physical meaning, and moderate computational complexity. Common ECMs include the Rint model, the first-order RC (Thevenin) model, the second-order RC model, and higher-order RC models. The second-order RC model, by adding a parallel RC network, can more accurately simulate the dynamic response of batteries at different time scales (such as electrochemical polarization and concentration polarization). This model strikes a good balance between accuracy and complexity, and has therefore attracted considerable attention. A second-order RC model typically consists of an ohmic internal resistor (R0), two parallel RC networks (R1 / C1 and R2 / C2), and a voltage-controlled voltage source (representing the open-circuit voltage (OCV)). The accuracy of the model depends largely on the accuracy of its parameters (R0, R1, C1, R2, and C2). However, these parameters are not constant; they vary nonlinearly with factors such as the battery's state of charge (SOC), state of OH (SOH), temperature, and charge / discharge current. Therefore, quickly and accurately identifying these model parameters is a key challenge in battery modeling and BMS development.

[0004] Existing parameter identification methods are mainly divided into time domain and frequency domain methods. Although frequency domain methods (such as electrochemical impedance spectroscopy (EIS)) can provide rich information, they generally require specialized testing equipment and are difficult to apply online. Time domain methods directly utilize time series data such as voltage, current, and temperature of the battery under actual operation or specific test conditions for identification, making them more suitable for BMS applications.

[0005] Among the time domain identification methods, the commonly used ones are:

[0006] (1) Based on least squares (LS) and its variants: such as recursive least squares (RLS). This type of method is computationally simple, but may not be effective when dealing with nonlinear systems and colored noise, and is prone to falling into local optimality. (2) Based on Kalman filter (KF) and its extensions: such as extended Kalman filter (EKF) and unscented Kalman filter (UKF). This type of method can simultaneously estimate the battery state and model parameters and is suitable for online applications. However, its performance depends on the prior knowledge of the statistical characteristics of the noise, and for strongly nonlinear systems, EKF may have accuracy and convergence problems. (3) Intelligent optimization algorithms: such as genetic algorithm (GA), particle swarm optimization (PSO), simulated annealing (SA), etc. This type of algorithm is based on swarm intelligence or probabilistic search, has global optimization capabilities, is not prone to falling into local optimality, and can effectively deal with nonlinear and multi-model optimization problems. (4) Particle swarm optimization (PSO) algorithm imitates the foraging behavior of bird flocks and finds the optimal solution through collaboration and information sharing among particles in the group. It has the advantages of simple concept, fewer parameters, fast convergence speed and easy implementation, and has been applied to battery model parameter identification.

[0007] However, the existing PSO-based battery parameter identification method still has some room for improvement in application:

[0008] (1) Construction of the objective function: Many methods directly use the error between the terminal voltage output by the model simulation and the measured terminal voltage (such as the root mean square error RMSE) as the objective function. Although intuitive, in some cases, the terminal voltage error may be affected by the OCV estimation error and the model structure error. Directly optimizing the terminal voltage error may not accurately reflect the matching degree of the model dynamic parameters (R / C parameters). (2) Processing of the initial state: Model simulation requires setting the initial voltage state of the RC network. Inappropriate initial state setting may affect the accuracy of the initial identification. (3) Optimization efficiency and robustness: The performance of the PSO algorithm is affected by its parameters (such as inertia weight, learning factor) and particle population size and number of iterations. How to set appropriate parameters and boundary conditions, and how to deal with numerical problems that may arise during the simulation process (such as instability caused by too small RC time constant) are crucial to ensuring the accuracy and robustness of the optimization results. (4) Combination with the OCV-SOC relationship: OCV is a key component of ECM, and an accurate OCV-SOC relationship is the basis for parameter identification. How to effectively utilize measured data and the known OCV-SOC relationship to guide the parameter optimization process is an issue worthy of in-depth study.

[0009] Therefore, it is urgent to propose an improved battery equivalent circuit model parameter identification method that can identify the parameters of the second-order RC model more accurately, robustly and efficiently. Summary of the Invention

[0010] The purpose of the present invention is to provide a battery equivalent circuit model parameter identification method and system based on particle swarm optimization to solve the above technical problems.

[0011] To achieve the above object, the present invention adopts the following technical solutions:

[0012] A battery equivalent circuit model parameter identification method based on particle swarm optimization includes the following steps: S1: obtaining dynamic test data of the battery, the dynamic test data including a time series, a measured terminal voltage series, a measured current series, and a measured capacity series;

[0013] S2: Obtaining a predetermined open circuit voltage-state of charge relationship of the battery;

[0014] S3: Based on the measured capacity sequence and the reference capacity of the battery, a real-time state of charge sequence is calculated;

[0015] S4: Based on the real-time state of charge sequence and the open circuit voltage-state of charge relationship, obtain an open circuit voltage sequence through interpolation calculation;

[0016] S5: defining a battery equivalent circuit model including parameters to be identified, wherein the parameters to be identified include at least an ohmic internal resistance R0 and a resistance R and a capacitance C of at least one RC parallel network;

[0017] S6: constructing an objective function based on an error metric between a predicted open-circuit voltage sequence and the searched open-circuit voltage sequence; wherein the predicted open-circuit voltage sequence is calculated by transforming the voltage equation of the equivalent circuit model and combining the measured terminal voltage sequence, the measured current sequence, and the voltage sequence of the RC parallel network iteratively calculated based on the current candidate parameters to be identified and the measured current sequence;

[0018] S7: Initialize the particle swarm optimization algorithm, including setting the particle population, parameter search boundary, and algorithm parameters;

[0019] S8: executing the particle swarm optimization algorithm, with the goal of minimizing the objective function, and iteratively updating the candidate parameter combinations to be identified represented by each particle in the particle swarm; in each iteration, for each particle, calculating the objective function value as its fitness based on the candidate parameters it represents;

[0020] S9: extracting, from the execution result of the particle swarm optimization algorithm, an optimal combination of parameters to be identified that minimizes or approaches the minimum of the objective function.

[0021] Further preferably, the equivalent circuit model is a second-order RC model, and the parameters to be identified include the resistance R1 and the capacitance C1 of the first RC network and the resistance R2 and the capacitance C2 of the second RC network.

[0022] Further preferably, the calculation method for predicting the open circuit voltage sequence includes: first, according to the discrete time state equation of the equivalent circuit model, using the current candidate parameters to be identified {R1, C1, R2, C2, ...} and the measured current sequence, iteratively calculating the RC network voltage sequence {V C1 (k),V C2 (k),…}; Then, for each time point k, the predicted open circuit voltage value is calculated based on the calculation relationship equivalent to the transformation of the model voltage equation.

[0023] Further preferably, before iteratively calculating the RC network voltage sequence, the RC network voltage at an initial time k=0 is set.

[0024] Further preferably, the error metric is a root mean square error between the predicted open circuit voltage sequence and the searched open circuit voltage sequence.

[0025] Further preferably, the step S7 includes applying a scaling process to the parameters to be identified whose value ranges have large differences in magnitude.

[0026] Further preferably, step S8 includes performing a parameter validity constraint check on each particle.

[0027] Further preferably, step S8 includes a mechanism for handling calculation anomalies. If a calculation error occurs in the process of calculating the objective function value for a particle, a penalty value is assigned to its fitness.

[0028] A battery equivalent circuit model parameter identification system includes: a data interface module for obtaining dynamic test data of the battery, the dynamic test data including time, voltage, current and capacity sequences;

[0029] a data processing module configured to calculate and generate an open circuit voltage sequence based on the capacity sequence and a pre-stored open circuit voltage-state of charge relationship;

[0030] A model operation module is configured to calculate and generate a predicted open circuit voltage sequence based on input candidate model parameters and the dynamic test data;

[0031] an objective function evaluation module configured to calculate an error metric between the predicted open circuit voltage sequence and the searched open circuit voltage sequence;

[0032] a parameter optimization module configured to execute a particle swarm optimization algorithm, with the goal of minimizing the error metric, and iteratively update the candidate model parameters until an optimal model parameter combination is found;

[0033] The result output module is used to output the optimal model parameter combination.

[0034] Beneficial effects of the present invention:

[0035] Improved parameter identification accuracy: By constructing an objective function based on the error between "predicting OCV" and "finding OCV", the optimization goal is directly focused on the impact of the model's dynamic characteristics (determined by the R / C parameters) on the OCV prediction ability, effectively decoupling the direct interference of OCV fluctuations or the uncertainty of its relationship with SOC on dynamic parameter identification, so that the identified R0, R1, C1, R2, and C2 can more accurately reflect the internal impedance and polarization characteristics of the battery.

[0036] Enhanced algorithm robustness: Using PSO as the optimization engine and leveraging its global search capabilities effectively avoids the local optimality often encountered with traditional gradient descent or least squares methods, increasing the probability of finding the globally optimal parameter solution. Furthermore, by setting reasonable parameter boundaries, addressing parameter scaling issues, adding RC time constant constraints, and implementing exception handling mechanisms, the stability of the optimization process and its adaptability to diverse operating conditions and data are enhanced.

[0037] The method is clear and easy to implement: The steps are well-defined and the logic is clear. It can be easily implemented using existing programming languages (such as Python) and optimization libraries (such as pyswarms). Data preparation, model calculation, objective function definition, and PSO optimization processes are all relatively standardized.

[0038] High Application Value: Accurately identifying second-order RC model parameters is essential for achieving high-precision BMS functions (such as SOC / SOH estimation, lifespan prediction, and safety warnings). The method of this invention can be applied to multiple aspects, including battery design, production quality control, and BMS algorithm development and verification, and has significant engineering application value.

[0039] Consideration of practical application details: The method of the present invention includes practical links such as data preprocessing, RC initial state estimation, and parameter scaling, making the method closer to actual application needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 is a flow chart of a battery equivalent circuit model parameter identification method of the present invention;

[0041] Figure 2 It is the structure diagram of the second-order RC equivalent circuit model;

[0042] Figure 3 It is a schematic diagram of the key signal relationships involved in the constructed objective function;

[0043] Figure 4 It is a schematic diagram of the particle swarm optimization process;

[0044] Figure 5 It is a functional block diagram of the battery equivalent circuit model parameter identification system of the present invention. DETAILED DESCRIPTION

[0045] The technical solutions in the embodiments of the present invention will be described clearly and completely below with reference to the accompanying drawings in the embodiments of the present invention.

[0046] Embodiment of the battery equivalent circuit model parameter identification method based on particle swarm optimization of the present invention:

[0047] like Figure 1-Figure 5 As shown in FIG, the battery equivalent circuit model parameter identification method based on particle swarm optimization includes the following steps:

[0048] S1: Acquire dynamic test data of the battery, where the dynamic test data includes a time series, a measured terminal voltage series, a measured current series, and a measured capacity series.

[0049] Dynamic test data refers to the test data of the battery under dynamic working conditions. The dynamic working conditions can be standard test cycles (such as HPPC, DST, UDDS) or actual operating data.

[0050] Preprocess the acquired data, for example:

[0051] (1) Unify the data format, such as reading CSV files and renaming column labels.

[0052] (2) Processing the current direction, for example, uniformly stipulating that the discharge current is positive (or negative) and performing corresponding conversion.

[0053] (3) (Optional) Data cleaning, such as removing outliers, data smoothing and filtering, etc.

[0054] (4) Determine the data sampling time interval dt.

[0055] S2: Obtaining a predetermined open circuit voltage-state of charge relationship of the battery.

[0056] Load pre-measured or calibrated battery OCV-SOC (open circuit voltage-state of charge) data. This data is typically stored in a lookup table or function, containing a series of SOC values and their corresponding OCV values. Also retrieve the battery's rated capacity or the total available capacity (capacityOCV) under the current test conditions.

[0057] S3: Based on the measured capacity sequence and the reference capacity of the battery, a real-time state of charge sequence is calculated.

[0058] Using the acquired real-time capacity sequence and capacityOCV, the real-time state-of-charge (SOC) sequence testSOC is calculated: testSOC[k] = capacity[k] / capacityOCV. The calculated testSOC is bounded to ensure that its value is within the range [0, 1] (or [0%, 100%]).

[0059] S4: Based on the real-time state of charge sequence and the open circuit voltage-state of charge relationship, an open circuit voltage sequence is obtained through interpolation calculation.

[0060] By using a table lookup and interpolation (linear interpolation np.interp), the "lookup OCV" sequence testOCV corresponding to the testSOC sequence is found from the loaded OCV-SOC relationship data. This sequence represents the expected open circuit voltage value obtained from the standard OCV-SOC curve based on the real-time SOC.

[0061] S5: Define a battery equivalent circuit model including parameters to be identified, where the parameters to be identified include at least an ohmic internal resistance R0 and a resistance R and a capacitance C of at least one RC parallel network.

[0062] Figure 2 As shown, the equivalent circuit model is a second-order RC model, and the parameters to be identified include the ohmic internal resistance R0, the resistor R1 and capacitor C1 of the first RC network, the resistor R2 and capacitor C2 of the second RC network, and a voltage-controlled voltage source representing the open circuit voltage OCV.

[0063] S6: Construct an objective function based on an error metric between a predicted open-circuit voltage sequence and the searched open-circuit voltage sequence; wherein the predicted open-circuit voltage sequence is transformed according to the voltage equation of the equivalent circuit model, and is calculated in combination with the measured terminal voltage sequence, the measured current sequence, and the voltage sequence of the RC parallel network obtained by iterative calculation based on the current candidate parameters to be identified and the measured current sequence.

[0064] The calculation method for predicting the open circuit voltage sequence includes: first, according to the discrete time state equation of the equivalent circuit model, using the current candidate parameters to be identified {R1, C1, R2, C2, ...} and the measured current sequence, iteratively calculating the RC network voltage sequence {V C1 (k),V C2 (k),…}; Then, for each time point k, the predicted open circuit voltage value is calculated based on the calculation relationship equivalent to the transformation of the model voltage equation.

[0065] V T (k)=OCV(k)-I(k)*R0-V C1(k)-V C2 (K);

[0066] The state equation for the voltage in the RC network describes its dynamic response:

[0067]

[0068] Discretize the continuous-time state equation to obtain the iterative update formula of the RC network voltage. For example, using the exact discretization method, we can obtain:

[0069]

[0070] Where k is the discrete time step index and dt is the sampling time interval.

[0071] The terminal voltage model equation V T (k)=OCV(k)-I(k)*R0-V C1 (k)-V C2 (k) is transformed to obtain the expression of OCV:

[0072] OCV(k)=V T (k)+I(k)*R0+V C1 (k)+V C2 (k) (assuming I is positive for outflow)

[0073] Define the "predicted OCV" sequence predictOCV. For a set of candidate model parameters {R0, R1, C1, R2, C2}, and the measured terminal voltage sequence obtained in step 1 and the measured current sequence I meas :

[0074] Determine the initial RC voltage: Set the RC network voltage at the initial time (k=0) V C1 (0),V C2 (0). For example, the measured terminal voltage at the initial moment can be The initial open circuit voltage testOCV(0) and the measured current I meas (0) and the candidate parameter R0 to calculate the initial total overpotential, which is then distributed to the two RC networks according to the ratio of the candidate parameters R1 and R2.

[0075] Iterative calculation of RC voltage sequence: Using the above discretized state equation, from k = 1 to N (data sequence length), according to V C1 (k-1),V C2 (k-1),I meas (k) and candidate parameters {R1, C1, R2, C2}, calculate the complete RC network voltage sequence VC1 ,V C2 .

[0076] Calculate and predict the OCV sequence: Use the transformed OCV expression combined with the measured voltage Measured current I meas , candidate parameter R0 and calculated RC voltage sequence V C1 ,V C2 , calculate the "predicted OCV" sequence predictOCV:

[0077]

[0078] Define the objective function J: Calculate the error metric between the "predicted OCV" sequence predictOCV and the "searched OCV" sequence testOCV obtained above, such as the root mean square error (RMSE), as the objective function J to be optimized:

[0079]

[0080] The goal of optimization is to find a set of parameters {R0, R1, C1, R2, C2} that minimizes the value of the objective function J.

[0081] S7: Initialize the particle swarm optimization algorithm, including setting the particle population, parameter search boundary, and PSO algorithm parameters, as follows:

[0082] Set the number of particles;

[0083] Determine the dimension of the optimization variables (5 for the second-order RC model);

[0084] Set a reasonable search range (upper and lower bounds) for each parameter to be identified (R0, R1, R2, C1, C2);

[0085] Handling parameter scale differences: For parameters with significant differences in the order of magnitude of their value ranges (such as resistor R and capacitor C), scale transformation can be introduced. The scaled parameter values are operated in the optimization search space, and the scale is reversed before being substituted into the model to calculate the objective function to restore their physical units. For example, optimizing Use C1=C1 in calculation scaled *scaleFactor;

[0086] Configure the control parameters of the PSO algorithm, such as inertia weight, learning factor, and neighborhood topology parameters (if using local optimal PSO);

[0087] The position (representing the initial parameter combination) and velocity of the initial particle swarm are randomly generated within the set boundaries.

[0088] S8: Execute the particle swarm optimization algorithm, with the goal of minimizing the objective function, and iteratively update the candidate parameter combinations to be identified represented by each particle in the particle swarm; in each iteration, for each particle, calculate the objective function value as its fitness based on the candidate parameters it represents. The details are as follows:

[0089] For each particle in the particle swarm, its current position represents a set of candidate model parameters.

[0090] Fitness evaluation: For each particle, use the candidate parameters it represents and execute the complete calculation process described in step 6 (including RC voltage iteration and predicted OCV calculation) to finally obtain the value of the objective function J. This J value is used as the fitness (or cost) of the particle.

[0091] Implement constraint checks: When evaluating fitness, you can include checks on the validity of parameter combinations. For example, verify that the RC time constants (e.g., R1*C1, R2*C2) calculated from candidate parameters meet physical or numerical stability requirements (e.g., greater than the sampling interval dt). Parameter combinations that do not meet the constraints are assigned a significant penalty fitness value, causing them to be naturally eliminated from the optimization process.

[0092] Handling computational exceptions: When executing model calculations to evaluate fitness, if a numerical calculation error (such as division by zero, overflow, etc.) occurs, the exception should be caught and a penalty fitness value should be assigned to the corresponding particle.

[0093] Update individual optimal records: Each particle tracks and records its own best position (i.e., parameter combination) and its corresponding optimal fitness value in history.

[0094] Update the optimal record of the group: According to the topological structure (global or local) of the PSO algorithm, determine the best position found among all the current particles (or particles in the neighborhood) and update the optimal record of the group.

[0095] Update particle state: Based on the particle's current speed, individual best record, group best record, and the PSO algorithm's control parameters, update each particle's speed and position to guide it to a more optimal search area. Ensure that the updated position remains within the preset boundaries.

[0096] Repeat the above iterative process of fitness evaluation, optimal record update, and particle state update until the termination condition is met (such as reaching the maximum number of iterations, stagnation of fitness value improvement, etc.).

[0097] S9: extracting, from the execution result of the particle swarm optimization algorithm, an optimal combination of parameters to be identified that minimizes or approaches the minimum of the objective function.

[0098] After the iteration is completed, the parameter combination represented by the recorded optimal position of the group (note the reverse scaling process) is used as the final identified model parameters {R0, R1, C1, R2, C2}.

[0099] S10: Result output and verification

[0100] Output the final optimized parameter set {R0, R1, C1, R2, C2} obtained by identification. The optimized parameters can be saved to a storage medium.

[0101] Verification: For example, using the identified optimal parameters, calculate a simulated terminal voltage series through forward simulation of the model under the same measured current input. Compare this simulated voltage series with the measured terminal voltage series (for example, calculate the RMSE between the two or draw a comparison curve) to evaluate the accuracy of parameter identification and the model fit.

[0102] The battery equivalent circuit model parameter identification system of the present invention includes:

[0103] The data interface module is used to obtain dynamic test data of the battery, which includes time, voltage, current and capacity sequences;

[0104] a data processing module configured to calculate and generate an open circuit voltage sequence based on the capacity sequence and a pre-stored open circuit voltage-state of charge relationship;

[0105] The model operation module is configured to iteratively calculate the RC network voltage sequence based on the input candidate model parameters and the dynamic test data, according to the preset equivalent circuit model structure and discrete state equation, and further combine the measured voltage, current and candidate parameters to calculate and generate a predicted OCV sequence.

[0106] The objective function evaluation module is configured to receive the predicted OCV sequence generated by the model operation module and the search OCV sequence generated by the data processing module, calculate the error metric between the two, and use it as the objective function output.

[0107] a parameter optimization module configured to: execute a particle swarm optimization algorithm, receive the error metric value output by the objective function evaluation module as fitness feedback, and iteratively adjust and update the candidate model parameters with the goal of minimizing the error metric value until an optimal parameter set that minimizes the objective function is found;

[0108] The result output module is used to output or store the optimal model parameter set determined by the parameter optimization module.

[0109] Compared with the prior art, the present invention has the following beneficial effects:

[0110] Improved parameter identification accuracy: By constructing an objective function based on the error between "predicting OCV" and "finding OCV", the optimization goal is directly focused on the impact of the model's dynamic characteristics (determined by the R / C parameters) on the OCV prediction ability, effectively decoupling the direct interference of OCV fluctuations or the uncertainty of its relationship with SOC on dynamic parameter identification, so that the identified R0, R1, C1, R2, and C2 can more accurately reflect the internal impedance and polarization characteristics of the battery.

[0111] Enhanced algorithm robustness: Using PSO as the optimization engine and leveraging its global search capabilities effectively avoids the local optimality often encountered with traditional gradient descent or least squares methods, increasing the probability of finding the globally optimal parameter solution. Furthermore, by setting reasonable parameter boundaries, addressing parameter scaling issues, adding RC time constant constraints, and implementing exception handling mechanisms, the stability of the optimization process and its adaptability to diverse operating conditions and data are enhanced.

[0112] The method is clear and easy to implement: The steps are well-defined and the logic is clear. It can be easily implemented using existing programming languages (such as Python) and optimization libraries (such as pyswarms). Data preparation, model calculation, objective function definition, and PSO optimization processes are all relatively standardized.

[0113] High Application Value: Accurately identifying second-order RC model parameters is essential for achieving high-precision BMS functions (such as SOC / SOH estimation, lifespan prediction, and safety warnings). The method of this invention can be applied to multiple aspects, including battery design, production quality control, and BMS algorithm development and verification, and has significant engineering application value.

[0114] Consideration of practical application details: The method of the present invention includes practical links such as data preprocessing, RC initial state estimation, and parameter scaling, making the method closer to actual application needs.

[0115] The present invention is not limited to the above-mentioned optimal implementation mode. Anyone can derive other forms of products under the inspiration of the present invention. However, no matter what changes are made in the shape or structure, any technical solution that is the same or similar to that of the present application falls within the scope of protection of the present invention.

Claims

1. A battery equivalent circuit model parameter identification method based on particle swarm optimization, characterized by: The method comprises the following steps: S1: obtaining dynamic test data of the battery, wherein the dynamic test data comprises a time series, a measured terminal voltage series, a measured current series, and a measured capacity series; S2: Obtaining a predetermined open circuit voltage-state of charge relationship of the battery; S3: Based on the measured capacity sequence and the reference capacity of the battery, a real-time state of charge sequence is calculated; S4: Based on the real-time state of charge sequence and the open circuit voltage-state of charge relationship, an open circuit voltage sequence is obtained by interpolation calculation; S5: defining a battery equivalent circuit model including parameters to be identified, wherein the parameters to be identified include at least an ohmic internal resistance R0 and a resistance R and a capacitance C of at least one RC parallel network; S6: constructing an objective function based on an error metric between a predicted open-circuit voltage sequence and the searched open-circuit voltage sequence; wherein the predicted open-circuit voltage sequence is calculated by transforming the voltage equation of the equivalent circuit model and combining the measured terminal voltage sequence, the measured current sequence, and the voltage sequence of the RC parallel network iteratively calculated based on the current candidate parameters to be identified and the measured current sequence; S7: Initialize the particle swarm optimization algorithm, including setting the particle population, parameter search boundary, and algorithm parameters; S8: executing the particle swarm optimization algorithm, with the goal of minimizing the objective function, and iteratively updating the candidate parameter combinations to be identified represented by each particle in the particle swarm; in each iteration, for each particle, calculating the objective function value as its fitness based on the candidate parameters it represents; S9: extracting, from the execution result of the particle swarm optimization algorithm, the optimal combination of parameters to be identified that makes the objective function reach a minimum or close to a minimum.

2. The battery equivalent circuit model parameter identification method based on particle swarm optimization according to claim 1, characterized in that: The equivalent circuit model is a second-order RC model, and the parameters to be identified include the resistor R1 and the capacitor C1 of the first RC network and the resistor R2 and the capacitor C2 of the second RC network.

3. The battery equivalent circuit model parameter identification method based on particle swarm optimization according to claim 1 or 2, characterized in that: The calculation method for predicting the open circuit voltage sequence includes: first, according to the discrete time state equation of the equivalent circuit model, using the current candidate parameters to be identified {R1, C1, R2, C2,...} and the measured current sequence, iteratively calculate the RC network voltage sequence at all times {V C1 (k),V C2 (k),…}; Then, for each time point k, the predicted open circuit voltage value is calculated based on the calculation relationship equivalent to the transformation of the model voltage equation.

4. The battery equivalent circuit model parameter identification method based on particle swarm optimization according to claim 3, characterized in that: Before iteratively calculating the RC network voltage sequence, the RC network voltage at the initial time k=0 is set.

5. The battery equivalent circuit model parameter identification method based on particle swarm optimization according to claim 1, characterized in that: The error metric is a root mean square error between the predicted open circuit voltage sequence and the searched open circuit voltage sequence.

6. The battery equivalent circuit model parameter identification method based on particle swarm optimization according to claim 1, characterized in that: The step S7 includes applying a scaling process to the parameters to be identified whose value ranges have large differences in magnitude.

7. The battery equivalent circuit model parameter identification method based on particle swarm optimization according to claim 1, characterized in that: Step S8 includes performing parameter validity constraint checks on each particle.

8. The battery equivalent circuit model parameter identification method based on particle swarm optimization according to claim 1, characterized in that: Step S8 includes a mechanism for handling calculation anomalies. If a calculation error occurs in the process of calculating the objective function value for a particle, a penalty value is assigned to its fitness.

9. Battery equivalent circuit model parameter identification system, characterized in that: include: The data interface module is used to obtain dynamic test data of the battery, which includes time, voltage, current and capacity sequences; a data processing module configured to calculate and generate an open circuit voltage sequence based on the capacity sequence and a pre-stored open circuit voltage-state of charge relationship; A model operation module is configured to calculate and generate a predicted open circuit voltage sequence based on input candidate model parameters and the dynamic test data; an objective function evaluation module configured to calculate an error metric between the predicted open circuit voltage sequence and the searched open circuit voltage sequence; a parameter optimization module configured to execute a particle swarm optimization algorithm, with the goal of minimizing the error metric, and iteratively update the candidate model parameters until an optimal model parameter combination is found; The result output module is used to output the optimal model parameter combination.

Images