SOC estimation method and system for high-capacity-difference battery pack

Abstract

The invention relates to the technical field of battery energy storage, in particular to an SOC estimation method and system for a high-capacity-difference battery pack, and the method comprises the steps: building a fractional-order equivalent circuit model, extracting a fast-change state vector and a slow-change state vector, and carrying out the parameter correction based on the slow-change vector, so as to build an SOC estimation model. According to the method, the SOC is estimated in real time by utilizing high sampling frequency and a fast-changing state vector is updated, and meanwhile, a slow-changing vector is updated by combining low sampling frequency with a variational Bayesian method, so that online correction of parameters is realized, the problem of model mismatching caused by battery aging, temperature change and inconsistency is solved, the precision, the adaptive capability and the real-time performance of SOC estimation under complex working conditions are improved, and the reliability of SOC estimation is improved. The problems of low SOC estimation precision and poor adaptability caused by inaccurate battery models and state estimation errors under complex working conditions are solved.

Description

Technical Field

[0001] This application relates to the technical field of battery energy storage, and in particular to a method and system for estimating the SOC of a high-capacity differential battery pack. Background Technology

[0002] With the rapid development of the electric vehicle and energy storage industries, the scale and complexity of battery systems continue to increase. The capacity, internal resistance, and other parameters of individual battery cells inevitably show significant differences due to variations in production batches, usage history, and operating environments, resulting in battery packs with large capacity discrepancies. Against this backdrop, the accurate estimation of the state of charge (SOC), a core task of the battery management system (BMS), has become a crucial link in ensuring system safety, improving energy utilization efficiency, and extending service life. As a key state parameter reflecting the remaining battery capacity, the accuracy of SOC estimation directly affects the system's operation control strategy, fault warning, and lifespan assessment.

[0003] In the field of battery state estimation, traditional solutions mainly rely on state observer methods based on equivalent circuit models (ECMs), such as Kalman filtering and its variants. These methods typically employ fixed-parameter models (such as first- or second-order RC models), provide a short-term benchmark through ampere-hour integration, and combine model voltage prediction for closed-loop correction. However, existing technologies have significant drawbacks: firstly, the fixed-parameter models they employ struggle to accurately describe the complex electrochemical behavior of lithium-ion batteries under dynamic operating conditions, over a wide temperature range, and throughout their entire lifecycle aging process, leading to model mismatch; secondly, when faced with strong inconsistencies within the battery pack caused by high capacity differences, traditional methods treat the entire battery pack as a homogeneous entity for estimation, lacking the ability to model and compensate for individual differences, resulting in a sharp decline in estimation accuracy in practical applications.

[0004] Therefore, existing technical solutions urgently need to address the following core issues: First, how to overcome the bottleneck of insufficient model accuracy and achieve high-precision SOC estimation in complex scenarios with dynamic operating conditions and high capacity differences within the battery pack; Second, how to enhance the adaptive capability of the estimation algorithm to time-varying characteristics of battery parameters (such as temperature fluctuations and aging degradation), thereby improving the robustness and reliability of the entire battery management system in real-world environments. Summary of the Invention

[0005] In view of this, to address the problems of low SOC estimation accuracy and poor adaptability under complex operating conditions due to inaccurate battery models and state estimation errors, this application provides an SOC estimation method and system for high-capacity differential battery packs, employing the following technical solution: A method for estimating the state of charge (SOC) of a high-capacity differential battery pack includes: Construct a fractional-order equivalent circuit model and obtain the associated fast-changing state vector and slow-changing state vector; Based on the slowly varying state vector, the parameters of the fractional-order equivalent circuit model are corrected to obtain the SOC estimation model. Real-time operating data of the battery pack is collected according to the preset first sampling frequency, and the real-time operating data and fast-change state vector are calculated by the SOC estimation model to obtain the real-time SOC estimate and update the fast-change state vector. Historical operating data of the battery pack is acquired based on a preset second sampling frequency, and the updated slow-change state vector is obtained by estimation using variational Bayes. Based on the updated slow-changing state vector and the updated fast-changing state vector, jump to the step of correcting the parameters of the fractional-order equivalent circuit model based on the slow-changing state vector to obtain the SOC estimation model. When the loop termination condition is met, output the real-time SOC estimate.

[0006] By adopting the above technical solution, the variational Bayesian method is used to accurately identify and correct model parameters online with low computational cost, realizing coordinated and frequency-division estimation of time-varying parameters and dynamic states inside the battery pack. This overcomes the model mismatch problem caused by battery aging, temperature changes, and inconsistencies. At the same time, the fractional-order model more accurately describes the dynamic characteristics of the battery, and combined with a dual-time-scale update mechanism, it ensures both the rapid response capability of real-time SOC estimation and the accuracy and adaptability of model parameters, ultimately improving the SOC estimation accuracy and robustness of high-capacity differential battery packs under complex operating conditions.

[0007] Optionally, the fractional-order equivalent circuit model is modified based on the slowly varying state vectors to obtain the SOC estimation model, including: The slowly changing state vector includes the battery's internal ohmic resistance and the battery's actual capacity. Replace the old parameter values ​​in the fractional-order equivalent circuit model with the battery's internal resistance and actual capacity, and use the fractional-order equivalent circuit model after parameter replacement as the SOC estimation model.

[0008] By adopting the above technical solution, the SOC estimation model is dynamically corrected by updating the key slowly varying parameters that have the greatest impact on the accuracy of SOC estimation online. This enables the model to continuously and accurately represent the current real working characteristics of the battery. The changes in ohmic internal resistance and actual capacity caused by aging, temperature fluctuations and individual cell differences during battery operation are fed back into the SOC estimation model in real time and accurately. This solves the problem of accumulated estimation errors caused by the inability of traditional fixed parameter models to adapt to time-varying parameters.

[0009] Optionally, the SOC estimation model is used to calculate the real-time operating data and the rapidly changing state vector to obtain a real-time SOC estimate and update the rapidly changing state vector, including: The rapidly changing state vector is the rapidly changing state vector of the previous time step; Based on the rapidly changing state vector of the previous time step, the rapidly changing state vector of the current time step is predicted by the SOC estimation model to obtain the prior estimate of the rapidly changing state vector. Based on the prior estimates, the model prediction of the battery terminal voltage is calculated; Obtain the measured value of the battery terminal voltage at the current moment, and calculate the residual between the model prediction value and the measured value; Based on residuals and real-time operational data, fuzzy inference and decision-making are performed using a fuzzy logic controller to obtain adjustment factors. The gain matrix of the Kalman filter algorithm in the SOC estimation model is dynamically adjusted based on the adjustment factor. Using the adjusted gain matrix, residuals, and prior estimates, the posterior estimate of the fast-changing state vector at the current time step is updated. The battery's state of charge (SOC) value is extracted from the posterior estimate and used as the real-time SOC estimate. The posterior estimate is then used as the updated fast-change state vector at the current time.

[0010] By adopting the above technical solution and introducing an adaptive adjustment mechanism based on a fuzzy logic controller, the strong nonlinearity and model uncertainty of the battery system under dynamic operating conditions can be addressed. By calculating the voltage prediction residual in real time and intelligently and dynamically adjusting the Kalman filter gain using fuzzy inference, the algorithm maintains optimal estimation performance when the system is stable, and quickly suppresses filter divergence when operating conditions change abruptly or the model exhibits large errors, thus enhancing robustness to complex external stimuli. Ultimately, stable and accurate tracking of rapidly changing states such as SOC and polarization voltage is achieved, ensuring the accuracy and reliability of real-time state estimation for high-capacity differential battery packs throughout their entire lifecycle and under various operating environments.

[0011] Optionally, the updated slowly varying state vector can be obtained by estimation using variational Bayes, including: Based on historical operational data, the variational Bayesian inference method is used to approximate the posterior probability distribution of the slowly varying state vector and obtain the variational distribution. Based on the variational distribution, the optimized posterior distribution of the slowly varying state vector is obtained by iteratively optimizing the variational distribution parameters by maximizing the lower bound of evidence. The mean of the optimized posterior distribution is extracted and used as the updated slowly varying state vector.

[0012] By adopting the above technical solution, the complex problem of solving the posterior distribution is transformed into a tractable optimization problem using variational Bayesian inference, reducing the computational complexity of large-scale battery pack parameter identification. By iteratively optimizing the variational distribution parameters to approximate the true posterior distribution, noise and uncertainty in the observed data are addressed, enabling robust estimation of slowly varying parameters such as battery ohmic internal resistance and actual capacity. This not only provides point estimates of the parameters but also preserves their probability distribution characteristics. Quantifying the estimation of uncertainties enhances the system's adaptability to slowly time-varying factors such as battery aging and temperature changes.

[0013] Optionally, the gain matrix of the Kalman filter algorithm in the SOC estimation model is dynamically adjusted based on an adjustment factor, including: The observation noise covariance matrix is ​​scaled according to the adjustment factor to obtain the scaled observation noise covariance matrix. The gain matrix of the Kalman filter algorithm in the SOC estimation model is calculated and updated based on the scaled observation noise covariance matrix.

[0014] By adopting the above technical solution, the confidence level of observation noise is adjusted in real time through the adjustment factor output by the fuzzy logic controller: when the system experiences large disturbances or model mismatch, the observation noise covariance is increased to reduce the confidence level of the current measurement and prevent filter divergence; when the system is running smoothly, the observation noise covariance is decreased to improve the utilization rate of measurement information and ensure estimation accuracy. By balancing the relationship between the system model and real-time measurements, the robustness and stability of the Kalman filter algorithm under complex conditions are increased.

[0015] Optionally, the loop termination condition includes at least one of the following: The rate of change of the real-time SOC estimate is less than the first preset threshold; The number of iterations has reached the preset maximum number of iterations; The battery pack's charge-discharge cycle has ended.

[0016] By adopting the above technical solution and employing a multi-condition judgment strategy, the computational efficiency of the SOC estimation value is ensured when it converges and stabilizes, avoiding unnecessary waste of computational resources. Furthermore, by setting a maximum number of iterations, the algorithm is prevented from falling into an infinite loop under abnormal conditions, ensuring the real-time requirements of the system. At the same time, it can promptly exit the current loop when the battery's operating state undergoes a fundamental change, providing a new estimation starting point for the next charge and discharge process. This balances the relationship between estimation accuracy, computational efficiency, and system reliability, ensuring the stable operation of the battery management system under various operating conditions.

[0017] A system for estimating the state of charge (SOC) of a high-capacity differential battery pack includes: The model building and parameter correction module is used to build a fractional equivalent circuit model and obtain the associated fast-changing state vector and slow-changing state vector, and to correct the parameters of the fractional equivalent circuit model based on the slow-changing state vector to obtain the SOC estimation model. The high-frequency estimation module is used to collect real-time operating data of the battery pack according to the preset first sampling frequency, and calculate the real-time operating data and fast-change state vector through the SOC estimation model to obtain the real-time SOC estimation value and update the fast-change state vector. The low-frequency update module is used to acquire the historical operating data of the battery pack based on a preset second sampling frequency, and to estimate the updated slow-change state vector through variational Bayes. The loop control module is used to control the jump execution of the model construction and parameter correction module based on the updated slow-changing state vector and the updated fast-changing state vector; The termination judgment and output module is used to determine whether the loop termination condition is met, and outputs the real-time SOC estimate when it is met.

[0018] By adopting the above technical solution, a precise battery model foundation is established through the model building and parameter correction module. A high-frequency estimation module enables rapid SOC tracking, a low-frequency update module facilitates slow parameter adaptation, a loop control module achieves closed-loop optimization between the model and the state, and a termination judgment and output module ensures the reliability and efficiency of the estimation process. By decomposing the complex SOC estimation problem into multiple specialized subtasks for parallel processing, the theoretical advantages of fractional-order models and variational Bayesian algorithms are fully utilized. Furthermore, modular design enables the rational allocation of computational resources, ultimately improving the state estimation accuracy, system robustness, and engineering practicality of high-capacity differential battery packs throughout their entire lifecycle and under various complex operating conditions.

[0019] Optional, the high-frequency estimation module includes: The prediction unit is used to predict the fast-changing state vector at the current time based on the fast-changing state vector at the previous time step using the SOC estimation model, and obtain the prior estimate of the fast-changing state vector. The voltage calculation unit is used to calculate the model prediction value of the battery terminal voltage based on the prior estimate; The residual calculation unit is used to obtain the measured value of the battery terminal voltage at the current moment and calculate the residual between the model prediction value and the measured value. The fuzzy inference unit is used to perform fuzzy inference and decision-making based on residuals and real-time operating data, according to the fuzzy logic controller, to obtain the adjustment factor. The gain adjustment unit is used to dynamically adjust the gain matrix of the Kalman filter algorithm in the SOC estimation model based on the adjustment factor. The state update unit is used to update the posterior estimate of the fast-changing state vector at the current time using the adjusted gain matrix, residual, and prior estimate. The SOC extraction unit is used to extract the battery's state of charge value from the posterior estimate as the real-time SOC estimate.

[0020] By adopting the above technical solution, the forward prediction of state and output is achieved through the prediction unit and voltage calculation unit. The residual calculation unit captures the deviation between the model and the actual situation in real time. Then, the fuzzy inference unit intelligently evaluates the uncertainty of the system and generates adjustment commands. Finally, the gain adjustment unit and the state update unit work together to complete the adaptive optimization of the filtering algorithm. This refined functional division not only maintains the framework advantages of the extended Kalman filter algorithm, but also enhances the algorithm's adaptability to dynamic operating conditions by introducing fuzzy logic control. This improves the numerical stability and convergence speed of the SOC estimation process under complex scenarios such as sudden current changes and temperature fluctuations.

[0021] A computer-readable storage medium having a computer program stored thereon, which, when executed, implements a method for estimating the state of charge (SOC) of a high-capacity differential battery pack.

[0022] A computer program product comprising a computer program stored on a non-transitory computer-readable storage medium, the computer program including program instructions that cause a computer to execute any of the SOC estimation methods for high-capacity differential battery packs.

[0023] In summary, this application includes at least one of the following beneficial technical effects: A fractional-order equivalent circuit model is constructed, and the associated fast-changing state vector and slow-changing state vector are obtained. Based on the slow-changing state vector, the parameters of the fractional-order equivalent circuit model are corrected to obtain the SOC estimation model. Real-time operating data of the battery pack is collected according to a preset first sampling frequency, and the real-time operating data and fast-changing state vector are calculated using the SOC estimation model to obtain the real-time SOC estimate and update the fast-changing state vector. Historical operating data of the battery pack is acquired based on a preset second sampling frequency, and estimation is performed using variational Bayes to obtain the updated slow-changing state vector. Based on the updated slow-changing state vector and the updated fast-changing state vector, the process jumps to the step of correcting the parameters of the fractional-order equivalent circuit model based on the slow-changing state vector to obtain the SOC estimation model. When the loop termination condition is met, the real-time SOC estimate is output. By using frequency-division co-estimation and fractional-order accurate modeling, the model mismatch problem caused by battery aging, temperature changes, and inconsistencies is overcome. Meanwhile, relying on the dual time scale update strategy, while ensuring the accuracy of parameter identification, it achieves rapid response and strong adaptability in real-time SOC estimation, solving the problems of low SOC estimation accuracy and poor adaptability caused by inaccurate battery models and state estimation errors under complex working conditions. Attached Figure Description

[0024] Figure 1 This is a flowchart of the SOC estimation method for a high-capacity differential battery pack provided in Embodiment 1 of this application; Figure 2 This is a flowchart of another method for estimating the SOC of a high-capacity differential battery pack provided in Embodiment 2 of this application; Figure 3 This is a structural block diagram of a high-capacity differential battery pack SOC estimation system provided in Embodiment 3 of this application. Detailed Implementation Example 1

[0025] Please refer to Figure 1 A method for estimating the state of charge (SOC) of a high-capacity, low-density battery pack, comprising: Step 101: Construct a fractional-order equivalent circuit model and obtain the associated fast-changing state vector and slow-changing state vector.

[0026] In this embodiment 1, the fractional-order equivalent circuit model refers to the circuit network structure obtained by modeling the internal electrochemical process of a lithium-ion battery based on fractional-order calculus theory, which is used to describe the dynamic response characteristics of the battery in a non-equilibrium state.

[0027] It should be noted that the fractional-order equivalent circuit model is composed of components such as fractional-order capacitors, ohmic internal resistance, polarization resistance, and diffusion impedance, which can more accurately characterize the relaxation process and frequency domain response behavior of lithium-ion batteries.

[0028] A rapidly changing state vector is a vector composed of observations representing instantaneous changes in external terminal voltage, current, and other parameters of the battery, reflecting the rapid dynamic process during charging and discharging. The rapidly changing state vector is the same as the rapidly changing state vector at the previous moment.

[0029] Slowly changing state vectors refer to vectors composed of state variables that characterize the slowly changing ion concentration distribution and solid-phase diffusion processes inside a battery, reflecting the long-term dynamic evolution characteristics of the battery.

[0030] In this embodiment 1, high-precision collaborative estimation of battery state can be achieved by jointly solving the fast-changing and slow-changing state vectors. A fractional-order differential operator is introduced based on the battery electrochemical mechanism to describe the non-integer-order dynamic behavior of polarization and diffusion processes; the model structure and parameters are determined by fitting impedance spectrum data; and the model is identified and verified using the least squares algorithm and frequency domain transformation method. The fast-changing state vector is obtained by real-time acquisition of battery terminal voltage and current data and processing it through Kalman filtering; the slow-changing state vector is obtained by iterative estimation based on voltage response and model parameters using a fractional-order state observer. Finally, by coupling the fast-changing and slow-changing state vectors, a full-state description system that can be used for predicting battery health and remaining life is constructed.

[0031] Step 102: Based on the slowly varying state vector, the parameters of the fractional-order equivalent circuit model are corrected to obtain the SOC estimation model.

[0032] Parameter correction refers to the calculation process of identifying and adjusting parameters such as fractional order, ohmic internal resistance, and polarization resistance in a fractional-order equivalent circuit model by using the slowly changing dynamic characteristics of the battery's internal structure (such as lithium-ion concentration distribution and active material decay) represented by the slowly changing state vector. The purpose is to enable the model parameters to track the actual aging state and operating characteristics of the battery in real time, thereby improving the model's adaptability throughout its entire life cycle.

[0033] The SOC estimation model refers to a fractional-order equivalent circuit model that, after parameter correction, can accurately reflect the current dynamic characteristics of the battery. When embedded with a state estimation algorithm, this model can achieve high-precision real-time estimation of the state of charge.

[0034] In this embodiment 1, an adaptive adjustment law is established with the slowly varying state vector as input and the model parameters as output. The parameter update rule is designed based on Lyapunov stability theory to ensure the convergence and robustness of the correction process.

[0035] Step 103: Collect real-time operating data of the battery pack according to the preset first sampling frequency, and calculate the real-time operating data and fast-changing state vector through the SOC estimation model to obtain the real-time SOC estimate and update the fast-changing state vector.

[0036] Preferably, the rapidly changing state vector is the rapidly changing state vector of the previous time step.

[0037] The first sampling frequency refers to the data acquisition time interval set to accurately capture the dynamic characteristics of the battery. In this embodiment, it is set to 1Hz-10Hz to meet the requirements for accurate measurement of rapidly changing signals such as battery terminal voltage and operating current.

[0038] Real-time operating data includes measurements of the battery pack's total voltage, total current, and temperature. These data are collected and preprocessed in real time by the sensor unit of the battery management system (BMS). The preprocessing process is a standard choice and will not be described in detail here.

[0039] Real-time SOC estimate refers to the estimated percentage of battery state of charge at the current moment, calculated based on the latest collected operational data and model.

[0040] Fast-changing state vector update refers to recursively correcting the state variables that characterize the external dynamic response of the battery using the observation data at the current moment.

[0041] In this embodiment 1, the BMS data acquisition module synchronously acquires the total voltage, total current, and temperature data of the battery pack at a first sampling frequency, and performs filtering and outlier removal on the raw data. The preprocessed real-time operating data (voltage, current) and the fast-changing state vector of the previous moment are used as inputs to the SOC estimation model. The fractional extended Kalman filter algorithm is used to recursively calculate the real-time SOC estimate at the current moment through two steps: state prediction and measurement update, and at the same time, the optimal estimate of the fast-changing state vector is updated. The calculated real-time SOC estimate and the updated fast-changing state vector are output and stored to maintain the accuracy of the state estimation.

[0042] Step 104: Obtain historical operating data of the battery pack based on the preset second sampling frequency, and estimate it using variational Bayes to obtain the updated slow-varying state vector.

[0043] The second sampling frequency refers to a lower data acquisition frequency set to capture the long-term degradation characteristics of the battery. In this embodiment, it is set to 0.001Hz-0.1Hz (i.e., the sampling interval is 10 seconds to 10 minutes) to obtain long-term historical operating data that can reflect the slow aging process of the battery.

[0044] Historical operating data includes historical battery pack voltage, historical current, historical temperature, and historical SOC sequence, which are extracted from the battery management system's storage unit. In this embodiment, the historical operating data is preferably the operating data acquired in the previous moment / previous cycle.

[0045] The updated slow-change state vector refers to the state estimate that has been processed by the variational Bayes algorithm and can more accurately characterize the slow-change processes inside the battery (such as loss of active lithium ions, increase of internal resistance, etc.).

[0046] In this embodiment 1, historical operation datasets within a specific time window (such as the past week or month) are extracted from the BMS historical database at a second sampling frequency. Then, a joint probabilistic graphical model containing the prior distribution of the slowly varying state vector and the system observation model is constructed. Next, the variational Bayes algorithm is used to iteratively optimize the parameters of the approximate posterior distribution by maximizing the lower bound of evidence (ELBO), thereby obtaining the optimal estimate of the slowly varying state vector. Finally, the updated slowly varying state vector is output and fed back to the parameter correction stage in step 102 to achieve adaptive adjustment of the model parameters, thereby completing the tracking and modeling of the long-term aging state of the battery.

[0047] Step 105: Based on the updated slow-changing state vector and the updated fast-changing state vector, jump to the step of correcting the parameters of the fractional-order equivalent circuit model based on the slow-changing state vector to obtain the SOC estimation model.

[0048] Step 106: When the loop termination condition is met, output the real-time SOC estimate.

[0049] Loop termination condition refers to the criterion for determining the end of the state estimation loop. Loop termination conditions include, but are not limited to, the following: The rate of change of the real-time SOC estimate is less than the first preset threshold; the number of iterations reaches the preset maximum number of iterations; the charge-discharge cycle of the battery pack ends.

[0050] In this embodiment 1, the updated slow-changing state vector obtained in step 104 and the updated fast-changing state vector in step 103 are input together into the parameter correction module, and the process of step 102 is re-executed. That is, based on the new slow-changing state vector, the key parameters (such as fractional order, polarization resistance, etc.) of the fractional-order equivalent circuit model are adjusted online to generate a new generation of SOC estimation model, thereby realizing closed-loop adaptive updating of model parameters. This process is executed cyclically until any loop termination condition in step 106 is met. When a system power-down signal is detected, or the SOC estimation value is found to fluctuate less than a preset threshold (such as 0.5%) for several consecutive cycles, or the loop counter reaches a preset upper limit, the loop is immediately terminated, and the latest calculated real-time SOC estimation value is output as the final result.

[0051] The implementation principle of this application embodiment is as follows: A fractional-order equivalent circuit model is constructed, and the associated fast-changing state vector and slow-changing state vector are obtained. Based on the slow-changing state vector, the parameters of the fractional-order equivalent circuit model are corrected to obtain the SOC estimation model. Real-time operating data of the battery pack is collected according to a preset first sampling frequency, and the real-time operating data and fast-changing state vector are calculated using the SOC estimation model to obtain the real-time SOC estimate and update the fast-changing state vector. Historical operating data of the battery pack is acquired based on a preset second sampling frequency, and estimation is performed using variational Bayes to obtain the updated slow-changing state vector. Based on the updated slow-changing state vector and the updated fast-changing state vector, the process jumps to the step of correcting the parameters of the fractional-order equivalent circuit model based on the slow-changing state vector to obtain the SOC estimation model. When the loop termination condition is met, the real-time SOC estimate is output. By using frequency-division co-estimation and fractional-order accurate modeling, the model mismatch problem caused by battery aging, temperature changes, and inconsistencies is overcome. Meanwhile, relying on the dual time scale update strategy, while ensuring the accuracy of parameter identification, it achieves rapid response and strong adaptability in real-time SOC estimation, solving the problems of low SOC estimation accuracy and poor adaptability caused by inaccurate battery models and state estimation errors under complex working conditions. Example 2

[0052] Please refer to Figure 2 A method for estimating the state of charge (SOC) of a high-capacity, low-density battery pack, comprising: Step 201: Construct a fractional-order equivalent circuit model and obtain the associated fast-changing state vector and slow-changing state vector.

[0053] In this embodiment 2, step 201 is similar to step 101 in embodiment 1, and will not be described again here.

[0054] Step 202: Replace the old parameter values ​​in the fractional-order equivalent circuit model with the battery's internal resistance and actual capacity, and use the fractional-order equivalent circuit model after parameter replacement as the SOC estimation model.

[0055] The slowly changing state vector includes the battery's internal ohmic resistance and its actual capacity.

[0056] The old parameter values ​​refer to the parameter values ​​that were originally obtained through offline identification or estimation in the previous cycle in the fractional equivalent circuit model. These parameters may deviate due to battery aging.

[0057] The internal ohmic resistance of a battery refers to the resistance parameter that characterizes the ohmic loss of ion conduction inside the battery, and can be identified online through a mixed pulse power characteristic test.

[0058] The actual battery capacity refers to the maximum usable capacity of the battery at present, reflecting its health status. It is obtained using the following formula: ; Where t1 and t2 are the start and end points of the time interval selected for capacity calculation, respectively. Let be the battery operating current at time t. This is a dynamic efficiency compensation factor. This is the corrected change in SOC. These are the weighting coefficients for the hysteresis voltage gradient. This is the hysteresis voltage gradient term. Wherein, The value range is 0.01 to 0.05.

[0059] Preferably, the dynamic efficiency compensation factor The calculation formula is: ; in, I represents the current temperature of the battery, and I represents the operating current. For reference temperature, , These are the fitting parameters.

[0060] It should be noted that the fitting parameters were calibrated through battery cycle testing, and respectively characterize the degree of influence of current rate and temperature deviation on efficiency.

[0061] Lag voltage gradient term The calculation formula is: ; in, The terminal voltage when the charge reaches level Q. This is the terminal voltage when the charge is discharged to Q. This represents the discharge charge for the current cycle.

[0062] Finally, the SOC dynamic calibration calculation formula is as follows: ; in, The SOC value is calculated using the ampere-hour integration method. The SOC value is estimated using the open-circuit voltage method. is the time-varying Kalman gain coefficient.

[0063] In this embodiment 2, the latest estimated battery ohmic resistance and actual battery capacity are extracted from the slowly changing state vector. Then, the original ohmic resistance parameter in the fractional equivalent circuit model is replaced with the latest estimated battery ohmic resistance. The reference capacity value used for SOC calculation in the SOC estimation model is replaced with the actual battery capacity value. Finally, the model after parameter replacement is verified. By comparing the predicted output voltage value with the actual measured value, the effectiveness of parameter updates is ensured, thereby forming an SOC estimation model that can adapt to the battery aging state and realize high-precision online SOC estimation based on the current actual characteristics of the battery.

[0064] Step 203: Collect real-time operating data of the battery pack according to the preset first sampling frequency, and calculate the real-time operating data and the fast-changing state vector through the SOC estimation model to obtain the real-time SOC estimate and update the fast-changing state vector.

[0065] Preferably, step 203 includes the following sub-steps: S31. Based on the rapidly changing state vector of the previous time step, the rapidly changing state vector of the current time step is predicted using the SOC estimation model to obtain the prior estimate of the rapidly changing state vector.

[0066] In this second embodiment, the predicted value of the fast-changing state vector at the current moment is calculated by using the fast-changing state vector that has been updated at the previous moment and combining it with the state equation describing the dynamic characteristics of the system in the SOC estimation model. This predicted value is called the prior estimate.

[0067] S32. Based on the prior estimate, calculate the model prediction value of the battery terminal voltage.

[0068] In this second embodiment, based on the prior estimate of the fast-changing state vector obtained in the previous step, the theoretical predicted value of the battery terminal voltage is calculated through the observation equation describing the relationship between output and state in the SOC estimation model.

[0069] S33. Obtain the measured value of the battery terminal voltage at the current moment, and calculate the residual between the model prediction value and the measured value.

[0070] In this embodiment 2, the actual measured value of the battery terminal voltage is collected in real time by a high-precision sensor, and the measured value is compared with the model prediction value obtained in the previous step. The difference between the two is the residual.

[0071] S34. Based on the residuals and real-time operating data, fuzzy reasoning and decision-making are performed according to the fuzzy logic controller to obtain the adjustment factor.

[0072] In this second embodiment, the calculated residuals and real-time current change rate, among other operating data, are input into the fuzzy logic controller. The controller performs inference decisions based on a preset fuzzy rule base and outputs an adjustment factor to dynamically adjust the parameters of the estimation algorithm.

[0073] S35. Based on the adjustment factor, dynamically adjust the gain matrix of the Kalman filter algorithm in the SOC estimation model.

[0074] S351. Scale the preset observation noise covariance matrix according to the adjustment factor to obtain the scaled observation noise covariance matrix.

[0075] In this embodiment 2, the adjustment factor output by the fuzzy controller is used to adaptively scale the pre-set observation noise covariance matrix to adapt to changes in noise characteristics under different operating conditions.

[0076] S352. Calculate and update the gain matrix of the Kalman filter algorithm in the SOC estimation model based on the scaled observation noise covariance matrix.

[0077] In this second embodiment, the gain matrix of the Kalman filter algorithm is recalculated using the scaled observation noise covariance matrix, thereby enabling online adjustment of the filter parameters.

[0078] S36. Using the adjusted gain matrix, residuals, and prior estimates, update the posterior estimate of the fast-changing state vector at the current time step.

[0079] In this embodiment 2, the adjusted gain matrix is ​​multiplied by the residual to obtain the state correction amount, and then this correction amount is used to correct the prior estimate to finally obtain a more accurate posterior estimate.

[0080] S37. Extract the battery's state of charge value from the posterior estimate and use it as the real-time SOC estimate. Then, use the posterior estimate as the updated fast-changing state vector at the current moment.

[0081] In this embodiment 2, the charged state component is parsed from the updated fast-changing state vector and output as the real-time SOC estimation result at the current moment. At the same time, the complete posterior estimated state vector is saved as the initial value for the state prediction at the next moment.

[0082] In this second embodiment, by introducing a fuzzy logic controller to adaptively adjust the Kalman filter parameters, the estimation bias caused by model mismatch and changes in noise statistical characteristics under dynamic conditions in traditional methods is overcome. Through high-frequency iterative updates, the system can quickly respond to complex operating conditions such as sudden load changes and temperature variations, improving the robustness and accuracy of SOC estimation.

[0083] Step 204: Based on historical operating data, use variational Bayesian inference to approximate the posterior probability distribution of the slowly varying state vector and obtain the variational distribution.

[0084] Preferably, the historical operating data includes historical voltage, historical current, historical temperature, and historical SOC sequence.

[0085] Variational Bayesian inference refers to a computational framework that approximates the true posterior distribution by finding a simple and decomposable probability distribution (i.e., variational distribution).

[0086] The posterior probability distribution of a slowly varying state vector refers to the probability distribution of the slowly varying states inside a battery under given observation data.

[0087] In this second embodiment, the variational distribution approximates the true posterior distribution, effectively solving the problem of excessively high posterior computational complexity in traditional methods, significantly reducing computational resource consumption, and meeting the real-time requirements of the battery management system. Secondly, the variational Bayesian framework can effectively handle the state estimation problem of nonlinear non-Gaussian systems, overcoming the limitations of traditional Kalman filter-type methods in battery aging modeling, and improving the estimation accuracy of slowly changing states (such as ohmic internal resistance and actual capacity). Finally, by outputting the estimation results in the form of a probability distribution, not only are point estimates of the state provided, but also a measure of the estimation uncertainty, providing richer decision-making information for the battery management system and enhancing the robustness and reliability of the system.

[0088] Step 205: Based on the variational distribution, the variational distribution parameters are iteratively optimized by maximizing the lower bound of evidence to obtain the optimized posterior distribution of the slowly varying state vector.

[0089] The lower bound of evidence refers to the lower bound of the KL divergence between the variational distribution and the true posterior distribution. Maximizing the lower bound of evidence is equivalent to minimizing the difference between the variational distribution and the true posterior distribution.

[0090] The optimized posterior distribution refers to the variational distribution that best approximates the true posterior distribution after iterative optimization.

[0091] In this embodiment 2, the lower bound of evidence is calculated as the objective function for optimization. Then, optimization algorithms such as stochastic gradient ascent or natural gradient method are used to iteratively update the parameters of the variational distribution. In each iteration, the lower bound of evidence is calculated and the variational parameters are updated by gradient. When the lower bound of evidence converges or the maximum number of iterations is reached, the optimization stops, and the optimized posterior distribution of the slowly varying state vector is obtained. This can meet the real-time performance requirements of the battery management system and enhance the adaptability and robustness of the system in different aging stages.

[0092] Step 206: Based on the updated slow-changing state vector and the updated fast-changing state vector, jump to the step of correcting the parameters of the fractional-order equivalent circuit model based on the slow-changing state vector to obtain the SOC estimation model.

[0093] In this embodiment 2, the mean of the optimized posterior distribution obtained in step 205 is used as the updated slow-varying state vector. Combined with the fast-varying state vector updated in step 203, the parameter correction process in step 202 is re-executed. That is, based on the new slow-varying state vector (containing the latest ohmic internal resistance and actual capacity estimates), the key parameters of the fractional-order equivalent circuit model are adjusted online to generate a new generation of SOC estimation model, realizing adaptive updating of model parameters.

[0094] Step 207: When the loop termination condition is met, output the real-time SOC estimate.

[0095] Preferably, the loop termination condition includes, but is not limited to, at least one of the following: The rate of change of the real-time SOC estimate is less than the first preset threshold; The number of iterations has reached the preset maximum number of iterations; The battery pack's charge-discharge cycle has ended.

[0096] In this embodiment 2, when the system detects that the rate of change of the real-time SOC estimate is less than a first preset threshold (e.g., 0.1% / s), it indicates that the SOC has stabilized and meets the accuracy requirements. The system automatically terminates the loop and outputs the current estimate. When the number of iterations reaches the preset maximum number of iterations (e.g., 100 times), the system forcibly terminates the calculation to avoid getting stuck in an infinite loop and ensure the real-time performance of the algorithm. When the battery management system detects that the charge / discharge cycle has ended (e.g., the current remains zero for more than a set time), the system also terminates the estimation process. This multi-condition coordinated termination mechanism significantly improves the robustness and computational efficiency of the system while ensuring estimation accuracy.

[0097] The implementation principle of this application embodiment is as follows: Construct a fractional-order equivalent circuit model and obtain the associated fast-changing and slow-changing state vectors; replace the corresponding old parameter values ​​in the fractional-order equivalent circuit model with the battery's ohmic internal resistance and actual battery capacity, and use the fractional-order equivalent circuit model after parameter replacement as the SOC estimation model; collect real-time operating data of the battery pack according to a preset first sampling frequency, and calculate the real-time SOC estimate and update the fast-changing state vector using the SOC estimation model; based on historical operating data, use variational Bayesian inference to approximate the posterior probability distribution of the slow-changing state vector to obtain the variational distribution; based on the variational distribution, iteratively optimize the variational distribution parameters by maximizing the lower bound of evidence to obtain the optimized posterior distribution of the slow-changing state vector; based on the updated slow-changing and fast-changing state vectors, jump to the step of replacing the corresponding old parameter values ​​in the fractional-order equivalent circuit model with the battery's ohmic internal resistance and actual battery capacity; when the loop termination condition is met, output the real-time SOC estimate. By calculating the actual battery capacity and employing a variational Bayesian probabilistic inference method, the estimation accuracy of slowly changing state vectors is significantly improved. Simultaneously, by combining a fuzzy logic adaptive adjustment Kalman filter algorithm, the impact of changes in noise statistical characteristics under dynamic operating conditions is overcome, achieving synergistic optimization of model parameters and state vectors. While ensuring computational efficiency, this significantly improves the adaptability and accuracy of SOC estimation to battery aging and complex operating conditions, resolving the long-term SOC estimation bias problem caused by fixed model parameters and inaccurate noise assumptions in traditional methods.

[0098] Example 3: Reference Figure 3 A system for estimating the state of charge (SOC) of a high-capacity differential battery pack, comprising: The model building and parameter correction module 301 is used to build a fractional-order equivalent circuit model and obtain the associated fast-changing state vector and slow-changing state vector, and to correct the parameters of the fractional-order equivalent circuit model based on the slow-changing state vector to obtain the SOC estimation model.

[0099] The high-frequency estimation module 302 is used to collect real-time operating data of the battery pack according to a preset first sampling frequency, and calculate the real-time operating data and fast-changing state vector through the SOC estimation model to obtain the real-time SOC estimation value and update the fast-changing state vector.

[0100] Preferably, the high-frequency estimation module 302 includes the following units: The prediction unit is used to predict the fast-changing state vector at the current time based on the fast-changing state vector at the previous time step using the SOC estimation model, and obtain the prior estimate of the fast-changing state vector. The voltage calculation unit is used to calculate the model prediction value of the battery terminal voltage based on the prior estimate; The residual calculation unit is used to obtain the measured value of the battery terminal voltage at the current moment and calculate the residual between the model prediction value and the measured value. The fuzzy inference unit is used to perform fuzzy inference and decision-making based on residuals and real-time operating data, according to the fuzzy logic controller, to obtain the adjustment factor. The gain adjustment unit is used to dynamically adjust the gain matrix of the Kalman filter algorithm in the SOC estimation model based on the adjustment factor. The state update unit is used to update the posterior estimate of the fast-changing state vector at the current time using the adjusted gain matrix, residual, and prior estimate. The SOC extraction unit is used to extract the battery's state of charge value from the posterior estimate as the real-time SOC estimate.

[0101] The low-frequency update module 303 is used to acquire historical operating data of the battery pack based on a preset second sampling frequency, and to estimate the updated slow-varying state vector through variational Bayes.

[0102] The loop control module 304 is used to control the operation of the jump execution model construction and parameter correction module based on the updated slow-changing state vector and the updated fast-changing state vector.

[0103] The termination judgment and output module 305 is used to determine whether the loop termination condition is met, and outputs the real-time SOC estimate when the condition is met.

[0104] The above describes a system corresponding to a SOC estimation method for high-capacity differential battery packs. Its implementation principle is similar to that of a SOC estimation method for high-capacity differential battery packs, and will not be elaborated here.

[0105] An electronic device according to an embodiment of the present invention includes: a memory and a processor, wherein the memory stores a computer program; when the computer program is executed by the processor, the processor performs a SOC estimation method for a high-capacity differential battery pack as described in any of the above embodiments.

[0106] The memory can be an electronic memory such as flash memory, EEPROM (Electrically Erasable Programmable Read-Only Memory), EPROM, hard disk, or ROM. The memory has storage space for program code used to perform any of the method steps described above. For example, the storage space for program code may include individual program codes for implementing the various steps in the methods described above. This program code can be read from or written to one or more computer program products. These computer program products include program code carriers such as hard disks, compact discs (CDs), memory cards, or floppy disks. The program code may be compressed, for example, in a suitable form. When run by a computing processing device, this code causes the computing processing device to perform the various steps in the methods described above.

[0107] This invention provides a computer-readable storage medium storing a computer program thereon, which, when executed, implements a SOC estimation method for a high-capacity differential battery pack as described in any embodiment of this invention.

[0108] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0109] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection between apparatuses or units through some interfaces, and may be electrical, mechanical, or other forms.

[0110] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0111] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0112] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0113] The embodiments described in this specific implementation are preferred embodiments of this application and are not intended to limit the scope of protection of this application. Identical components are represented by the same reference numerals. Therefore, all equivalent changes made to the structure, shape, and principle of this application should be covered within the scope of protection of this application.

Claims

1. A method for estimating the State of Charge (SOC) of a high-capacity differential battery pack, characterized in that, include: Construct a fractional-order equivalent circuit model and obtain the associated fast-changing state vector and slow-changing state vector; Based on the slowly varying state vector, the parameters of the fractional-order equivalent circuit model are corrected to obtain the SOC estimation model. Real-time operating data of the battery pack is collected according to a preset first sampling frequency, and the real-time operating data and the fast-change state vector are calculated using the SOC estimation model to obtain a real-time SOC estimate and update the fast-change state vector. Historical operating data of the battery pack is acquired based on a preset second sampling frequency, and the updated slow-change state vector is obtained by estimation using variational Bayes. Based on the updated slow-changing state vector and the updated fast-changing state vector, jump to the step of correcting the parameters of the fractional-order equivalent circuit model based on the slow-changing state vector to obtain the SOC estimation model. When the loop termination condition is met, the real-time SOC estimate is output.

2. The SOC estimation method for high-capacity differential battery packs according to claim 1, characterized in that, The step of correcting the parameters of the fractional-order equivalent circuit model based on the slowly varying state vector to obtain the SOC estimation model includes: The slowly changing state vector includes the battery's internal ohmic resistance and the battery's actual capacity. Replace the old parameter values ​​in the fractional-order equivalent circuit model with the battery's internal resistance and actual capacity, and use the fractional-order equivalent circuit model after parameter replacement as the SOC estimation model.

3. The SOC estimation method for high-capacity differential battery packs according to claim 1, characterized in that, The step of calculating the real-time SOC estimate and updating the fast-changing state vector using the SOC estimation model includes: The rapidly changing state vector is the rapidly changing state vector of the previous time step; Based on the rapidly changing state vector of the previous time step, the rapidly changing state vector of the current time step is predicted using the SOC estimation model to obtain a prior estimate of the rapidly changing state vector. Based on the prior estimate, the model prediction value of the battery terminal voltage is calculated; Obtain the measured value of the battery terminal voltage at the current moment, and calculate the residual between the model prediction value and the measured value; Based on the residuals and real-time operating data, fuzzy inference and decision-making are performed according to the fuzzy logic controller to obtain the adjustment factor; Based on the adjustment factor, the gain matrix of the Kalman filter algorithm in the SOC estimation model is dynamically adjusted; Using the adjusted gain matrix, the residual, and the prior estimate, the posterior estimate of the fast-changing state vector at the current time is updated. The battery's state of charge (SOC) value is extracted from the posterior estimate and used as the real-time SOC estimate. The posterior estimate is then used as the updated fast-change state vector at the current moment.

4. The SOC estimation method for high-capacity differential battery packs according to claim 2, characterized in that, The estimation using variational Bayes to obtain the updated slowly varying state vector includes: Based on the historical operating data, the variational Bayesian inference method is used to approximate the posterior probability distribution of the slowly changing state vector, and the variational distribution is obtained. Based on the variational distribution, the optimized posterior distribution of the slowly changing state vector is obtained by iteratively optimizing the variational distribution parameters by maximizing the lower bound of evidence. The mean of the optimized posterior distribution is extracted and used as the updated slowly varying state vector.

5. The SOC estimation method for high-capacity differential battery packs according to claim 3, characterized in that, The step of dynamically adjusting the gain matrix of the Kalman filter algorithm in the SOC estimation model based on the adjustment factor includes: The preset observation noise covariance matrix is ​​scaled according to the adjustment factor to obtain the scaled observation noise covariance matrix. The gain matrix of the Kalman filter algorithm in the SOC estimation model is calculated and updated based on the scaled observation noise covariance matrix.

6. The SOC estimation method for high-capacity differential battery packs according to claim 1, characterized in that, The loop termination condition includes at least one of the following: The rate of change of the real-time SOC estimate is less than a first preset threshold; The number of iterations has reached the preset maximum number of iterations; The charge-discharge cycle of the battery pack has ended.

7. A SOC estimation system for high-capacity differential battery packs, characterized in that, include: The model building and parameter correction module is used to build a fractional-order equivalent circuit model and obtain the associated fast-changing state vector and slow-changing state vector, and to correct the parameters of the fractional-order equivalent circuit model based on the slow-changing state vector to obtain the SOC estimation model. The high-frequency estimation module is used to collect real-time operating data of the battery pack according to a preset first sampling frequency, and calculate the real-time operating data and the fast-change state vector through the SOC estimation model to obtain the real-time SOC estimation value and update the fast-change state vector. The low-frequency update module is used to acquire the historical operating data of the battery pack based on a preset second sampling frequency, and to estimate the updated slow-change state vector through variational Bayes. The loop control module is used to control the jump to execute the operation of the model construction and parameter correction module based on the updated slow-changing state vector and the updated fast-changing state vector; The termination judgment and output module is used to determine whether the loop termination condition is met, and outputs the real-time SOC estimate when the condition is met.

8. The SOC estimation system for high-capacity differential battery packs according to claim 7, characterized in that, The high-frequency estimation module includes: The prediction unit is used to predict the fast-changing state vector at the current time based on the fast-changing state vector at the previous time step, through the SOC estimation model, to obtain a prior estimate of the fast-changing state vector. A voltage calculation unit is used to calculate the model prediction value of the battery terminal voltage based on the prior estimate; The residual calculation unit is used to obtain the measured value of the battery terminal voltage at the current moment and calculate the residual between the model prediction value and the measured value. The fuzzy inference unit is used to perform fuzzy inference and decision-making based on the residuals and real-time operating data, according to the fuzzy logic controller, to obtain the adjustment factor. A gain adjustment unit is used to dynamically adjust the gain matrix of the Kalman filter algorithm in the SOC estimation model based on the adjustment factor. The state update unit is used to update the posterior estimate of the fast-changing state vector at the current time using the adjusted gain matrix, the residual, and the prior estimate. The SOC extraction unit is used to extract the state of charge (SOC) value of the battery from the posterior estimate as the real-time SOC estimate.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed, it implements the SOC estimation method for high-capacity differential battery packs as described in any one of claims 1-6.

10. A computer program product, characterized in that, The computer program product includes a computer program stored on a non-transitory computer-readable storage medium, the computer program including program instructions that cause the computer to execute the SOC estimation method for high-capacity differential battery packs as described in any one of claims 1-6.

Images