Power battery state estimation system and method
By combining data model fusion and a multi-model system with the variable Bayesian method, the problem of high-precision state estimation of power batteries under complex operating conditions is solved, achieving highly robust and accurate state-of-charge estimation, and adapting to accurate real-time estimation under current fluctuation scenarios.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-10
- Publication Date
- 2026-03-10
AI Technical Summary
Existing technologies struggle to achieve high-precision and robust state-of-charge estimation of power batteries under complex operating conditions, especially when data is incomplete or current fluctuates drastically. Traditional methods are prone to estimation lag, increased error, or divergence.
A complete dataset is obtained by using a data model fusion and reconstruction method. A multi-input multi-output nonlinear multi-model system is established. The state estimation is performed by combining the variable Bayesian method, extended Kalman filter and Aitken acceleration method. The algorithm is verified by experiments and improved to enhance the estimation accuracy and robustness.
Under complex operating conditions, the dataset integrity is improved to 95%, the model identification accuracy is increased by 18%, and the state of charge estimation error is reduced by 35%, achieving high-precision and robust state estimation.
Smart Images
Figure CN121633883A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of power battery technology, specifically to a power battery state estimation system and method. Background Technology
[0002] With the rapid development of new energy vehicles, energy storage power stations, and other fields, the safe, efficient, and long-life operation of power batteries, as their core energy carriers, is of paramount importance. Battery management systems (BMS) are crucial for ensuring battery performance, and the accurate and reliable real-time estimation of the state of charge (SOC), the core state quantity representing the remaining battery capacity, is the primary technical challenge for BMS.
[0003] However, under complex real-world operating conditions, SOC estimation for power batteries faces numerous technical challenges: First, SOC is an internal state variable that cannot be directly measured by sensors and must be estimated indirectly using measurable external characteristic parameters (such as voltage, current, and temperature). Battery systems exhibit strong nonlinearity and time-varying characteristics, and their dynamic characteristics are coupled with various factors such as current rate, ambient temperature, and aging degree, making it extremely difficult to establish an accurate battery model. Second, in complex actual operating conditions, the battery charging and discharging process changes drastically, causing some key process variable data (such as voltage spikes and current transients) to be easily lost or distorted, resulting in incomplete datasets and severely impacting the initial data quality for model identification and state estimation.
[0004] In existing technologies, a single battery model is insufficient to accurately describe the dynamic characteristics of a battery across all operating conditions, resulting in limited model identification accuracy. Furthermore, traditional state estimation algorithms (such as a single Kalman filter and its extensions) are prone to estimation lag, increased error, and even divergence when dealing with scenarios such as model mismatch and drastic current fluctuations (e.g., fluctuations up to 40%). The accuracy and robustness of the estimation results are insufficient to meet the requirements of precise battery management.
[0005] Therefore, there is an urgent need in this field for a power battery state estimation system and method that can still achieve high accuracy and high robustness under complex operating conditions, especially under conditions of incomplete data and drastic current fluctuations. Summary of the Invention
[0006] To address the shortcomings of existing technologies, this invention provides a power battery state estimation system and method, which has advantages such as high estimation reliability, high estimation accuracy, high estimation adaptability, and strong real-time estimation, thus solving the problems mentioned in the background technology.
[0007] To achieve the above objectives, the present invention provides the following technical solution:
[0008] A power battery state estimation system includes a data acquisition and processing module, a multi-model identification module, a state estimation module, an experimental verification and improvement module, and an output module.
[0009] The data acquisition and processing module is used to acquire system measurement data of the power battery under complex working conditions and reconstruct the lost data to improve the integrity of the dataset.
[0010] The multi-model identification module is communicatively connected to the data acquisition and processing module. It is used to establish a multi-input multi-output nonlinear multi-model system of the power battery based on the reconstructed system measurement data and using system identification technology. The multi-model system includes multiple local sub-models.
[0011] The state estimation module is communicatively connected to the multi-model identification module and is used to perform the following operations: using the variable Bayesian method to update the model probability of the multi-model system, and integrating extended Kalman filtering, extended Kalman smoothing and Aitken acceleration method to accurately estimate the state of charge of the power battery in real time.
[0012] The experimental verification and improvement module is used to build a battery simulation and experimental setup. The variable Bayesian method in the state estimation module is applied to the experimental setup for verification. Based on the experimental results, the relevant algorithms are improved and perfected to ensure that the state estimation error of the system is reduced by 35% under the scenario of 40% current fluctuation.
[0013] The output module is used to output the state estimation results of the power battery, providing a basis for precise battery management.
[0014] Furthermore, the data acquisition and processing module reconstructs the lost data using a data model fusion method, thereby improving the integrity of the dataset to over 95%.
[0015] Furthermore, the system estimation model established by the multi-model identification module has a model identification accuracy that is more than 18% higher than that of traditional methods.
[0016] Furthermore, in the multi-model system, the fitting error of each local sub-model is ≤5% within its respective specific working interval.
[0017] Furthermore, the state estimation module reduces the estimation error of the state of charge by more than 35% in a scenario where the battery current fluctuates by 40%.
[0018] Another technical problem to be solved by the present invention is to provide a method for estimating the state of a power battery, comprising the following steps:
[0019] S1. Acquire system measurement data of the power battery under complex working conditions and reconstruct the lost data to improve the integrity of the dataset;
[0020] S2. Based on the reconstructed system measurement data, a multi-input multi-output nonlinear multi-model system of the power battery is established using system identification technology. The multi-model system includes multiple local sub-models.
[0021] S3. The variable Bayesian method is used to update the model probability of the multi-model system, and the extended Kalman filter, extended Kalman smoothing and Aitken smoothing methods are integrated to accurately estimate the state of charge of the power battery in real time.
[0022] Furthermore, in S1, the lost data is reconstructed using a data model fusion method, thereby improving the integrity of the dataset to over 95%.
[0023] Furthermore, in S2, the identification accuracy of the established system estimation model is improved by more than 18% compared with traditional methods.
[0024] Furthermore, in S2, the fitting error of each local sub-model in the multi-model system is ≤5% within its respective specific working interval.
[0025] Furthermore, in S3, under the scenario of 40% fluctuation in battery current, the estimation error of state of charge is reduced by more than 35%.
[0026] Compared with the prior art, the present invention provides a power battery state estimation system and method, which has the following beneficial effects:
[0027] 1. The power battery state estimation system and method adopts a data model fusion method through the data acquisition and processing module to effectively reconstruct the key process variable data lost under complex working conditions, improve the integrity of the dataset to more than 95%, and provide a reliable data foundation for subsequent high-precision model identification and state estimation, thereby improving the reliability of the system from the source.
[0028] 2. The power battery state estimation system and method constructs a multi-input multi-output nonlinear multi-model system through a multi-model identification module. It utilizes multiple local sub-models to cover the dynamic characteristics of the battery under all operating conditions. Compared with the traditional single model, the identification accuracy is improved by more than 18%, and the fitting error of each local sub-model in its specific working range is ≤5%. This significantly improves the model's ability to characterize the complex nonlinear behavior of the battery and effectively enhances the accuracy and adaptability of model identification.
[0029] 3. This power battery state estimation system and method integrates the variable Bayesian method with extended Kalman filtering, extended Kalman smoothing, and Aitken acceleration methods. It can adaptively adjust the weights (model probabilities) of each sub-model according to the system operating conditions, and optimize the estimation process using smoothing and acceleration techniques. Even in extremely complex scenarios with current fluctuations as high as 40%, it can reduce the SOC estimation error by more than 35%, achieving accurate real-time estimation, effectively suppressing filter divergence, and realizing high-precision and robust state estimation.
[0030] 4. The proposed power battery state estimation system and method, through experimental verification and improvement modules, verifies and iteratively optimizes the theoretical algorithm on simulation and experimental devices, ensuring the feasibility and effectiveness of the proposed method and technical path. This allows the system performance to be continuously improved and perfected in practice, ultimately providing solid and reliable core technical support for the precise management of batteries, forming a complete technical closed loop from theory to verification. Attached Figure Description
[0031] Figure 1 This is a system block diagram of a power battery state estimation system proposed in this invention;
[0032] Figure 2 This is a flowchart illustrating a power battery state estimation method proposed in this invention. Detailed Implementation
[0033] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0034] Example 1:
[0035] Please see Figure 1 The power battery state estimation system in this embodiment includes a data acquisition and processing module, a multi-model identification module, a state estimation module, an experimental verification and improvement module, and an output module.
[0036] In this embodiment, the data acquisition and processing module is responsible for interfacing with the sensor unit of the battery management system (BMS) to acquire system measurement data of the power battery under complex operating conditions (such as high-rate charging and discharging, urban mixed road conditions, and high and low temperature environments) in real time or offline. This data includes, but is not limited to: operating voltage, load current, battery surface temperature, etc.
[0037] Specifically, due to sensor malfunctions, communication interference, or abnormal data storage, some critical data may be lost. This module employs a data model fusion method for data reconstruction. Specifically:
[0038] The first step is data diagnostics: identifying missing points and outliers in the data stream (such as spikes that are outside the physical range).
[0039] The second step is model-assisted reconstruction: a short-term predictive model based on battery electrochemical mechanisms or empirical data is established. When data is lost, this model is used to predict the possible values of the missing data based on valid data before and after the point of loss.
[0040] The third step is data fusion: the model's predicted values are cross-validated and smoothed using subsequent received valid data (e.g., by using a moving average window or Kalman filter), ultimately generating the reconstructed complete data sequence. This process can improve the integrity of the original dataset from approximately 80% to over 95%, laying a solid foundation for subsequent processing.
[0041] In this embodiment, the multi-model identification module receives a complete, high-quality dataset from the data acquisition and processing module. Its core task is to utilize system identification technology to establish a multi-input multi-output (MIMO) nonlinear multi-model system that can accurately describe the dynamic characteristics of the battery.
[0042] Specifically, the model structure design is as follows: This multi-model system consists of multiple local sub-models. Each sub-model is responsible for describing the dynamic behavior of the battery within a specific operating range (e.g., different SOC ranges, temperature ranges, or current rate ranges). The sub-models themselves can adopt parameterized forms such as state-space models or equivalent circuit models.
[0043] Specifically, parameter identification: Parameter estimation algorithms such as recursive least squares and particle swarm optimization are used to identify the parameters of each local sub-model. Optimization algorithms are then used to minimize the error between the sub-model's output (e.g., terminal voltage) and the measured data.
[0044] Specifically, the final system estimation model achieves an overall identification accuracy that is more than 18% higher than that of the traditional single equivalent circuit model method. Furthermore, the fitting error of key variables such as output voltage for each local sub-model within its specific operating range can be controlled within 5%.
[0045] In this embodiment, the state estimation module is used to receive the multi-model system from the multi-model identification module and execute an advanced fusion algorithm to accurately estimate the state of charge (SOC) of the battery in real time.
[0046] Specifically, multi-model interaction: A variable Bayesian method is used as the framework for multi-model interaction. This method calculates and updates the probability (i.e., weight) of each local sub-model in real time based on the system's current actual measurement output. The sub-model that best matches the current operating condition will receive a higher weight.
[0047] Specifically, state estimation fusion: For each activated local sub-model, an improved set of state observers is run in parallel. This set of observers fuses:
[0048] Extended Kalman Filter (EKF): Performs real-time (forward) prediction and update of states.
[0049] Extended Kalman Smoothing (EKS): Uses future observations within a short time window to retrospectively correct the EKF estimation results, thereby improving the estimation accuracy.
[0050] Aitken acceleration method: applied to the iterative calculation process of EKF / EKS to accelerate the convergence speed of the state covariance matrix, reduce the computational burden, and improve real-time performance.
[0051] Specifically, the SOC estimation results of each sub-model are weighted and fused according to their current probability weights to obtain the final SOC estimate of the system. Experimental results show that this fusion strategy can reduce the SOC estimation error by more than 35% compared with the traditional EKF method in harsh scenarios where the current fluctuates drastically (up to 40%).
[0052] In this embodiment, the experimental verification and improvement module is used to verify and optimize the above algorithm in a laboratory environment. It constructs a complete "algorithm-hardware" in-the-loop testing platform by building a battery test bench (such as a charge-discharge tester and a temperature chamber) and a simulation model (such as a MATLAB / Simulink model).
[0053] Run the variable Bayesian multi-model estimation algorithm in the state estimation module on this platform, apply standard or custom operating condition profiles (such as UDDS, FUDS, etc.), and record the error between the SOC estimate and the true value (measured by high-precision laboratory equipment).
[0054] Analyze the sources of error, such as the failure of a sub-model under boundary conditions or insufficiently rapid model probability updates. Based on the analysis results, iteratively improve and refine the algorithm parameters (such as process noise and observation noise covariance matrix) or the model set itself until the performance target of "reducing the state estimation error by 35% under a 40% current fluctuation scenario" is achieved.
[0055] In this embodiment, the output module outputs the SOC estimate finally determined by the state estimation module, as well as other possible information (such as battery health status SOH, maximum available power, etc.), to the upper-level battery management system (BMS) or vehicle controller via CAN bus, serial port, or wireless communication, providing a direct and reliable basis for precise management strategies such as battery equalization management, thermal management, and power limiting.
[0056] Example 2:
[0057] Please see Figure 2 The power battery state estimation method in this embodiment includes the following steps:
[0058] S1. Data Acquisition and Reconstruction:
[0059] Acquire system measurement data (voltage, current, temperature) of the power battery under complex operating conditions. When data loss is detected, a data model fusion method is used to reconstruct the data, improving the integrity of the dataset to over 95%.
[0060] S2. Establishment of a multi-model system:
[0061] Based on the reconstructed high-quality data, a multi-input multi-output nonlinear multi-model system for power batteries is established using system identification technology. This system contains multiple local sub-models, and the overall model identification accuracy is improved by more than 18% compared to traditional methods. Furthermore, the fitting error of each sub-model within a specific working range is ≤5%.
[0062] S3, Real-time estimation of state of charge:
[0063] A variable Bayesian method is employed to update the model probabilities of the multi-model system established in S2. Simultaneously, extended Kalman filtering (EKF), extended Kalman smoothing (EKS), and the Aitken acceleration method are integrated to accurately estimate the state of charge (SOC) of the power battery in real time. This method can reduce the SOC estimation error by more than 35% even under a scenario where the battery current fluctuates by 40%.
[0064] S4 (optional), experimental verification and algorithm improvement:
[0065] The estimation algorithm for S3 is validated on simulation or experimental devices. Based on the validation results, the algorithm parameters or model set are iteratively optimized to ensure its robustness and accuracy in practical applications.
[0066] S5. Output results:
[0067] The final SOC estimate is output for precise battery management.
[0068] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.
[0069] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.
Claims
1. A power cell state estimation system, characterized by: The application relates to a state estimation method for a power battery, which comprises the following steps: a data acquisition and processing module, a multi-model identification module, a state estimation module, an experimental verification and improvement module and an output module. The data acquisition and processing module is used for acquiring system measurement data of the power battery under complex working conditions, reconstructing loss data and improving the integrity of a data set. The multi-model identification module is in communication connection with the data acquisition and processing module, is used for establishing a multi-input multi-output nonlinear multi-model system of the power battery based on the reconstructed system measurement data by using a system identification technology, and comprises a plurality of local submodels. The state estimation module is in communication connection with the multi-model identification module, is used for performing the following operations: adopting a variable Bayesian method to perform model probability updating on the multi-model system, and fusing an extended Kalman filter, an extended Kalman smoothing and an Aitken acceleration method to accurately and timely estimate the state of charge of the power battery. The experimental verification and improvement module is used for building a battery simulation and experimental device, applying the variable Bayesian method in the state estimation module to the experimental device for experimental verification, improving and perfecting related algorithms according to experimental results, and ensuring that the state estimation error of the system under a 40% current fluctuation scenario is reduced by 35%. The output module is used for outputting the state estimation result of the power battery and providing a basis for accurate management of the battery.
2. The power battery state estimation system according to claim 1, characterized in that: The data acquisition and processing module reconstructs the loss data by using a data model fusion method, and the integrity of the data set is improved to more than 95%.
3. The power battery state estimation system of claim 1, wherein: The model identification accuracy of the system estimation model established by the multi-model identification module is improved by more than 18% compared with that of a traditional method.
4. The power battery state estimation system of claim 1, wherein: The fitting error of each local submodel in the multi-model system in a corresponding specific working interval is less than or equal to 5%.
5. The power battery state estimation system of claim 1, wherein: The state estimation module reduces the estimation error of the state of charge by more than 35% under a 40% battery current fluctuation scenario.
6. A method for estimating the state of a power battery using the power battery state estimation system according to any one of claims 1 to 5, characterized by, The application further discloses a state estimation method for a power battery, which comprises the following steps: S1, data acquisition and reconstruction: acquiring system measurement data of the power battery under complex working conditions, reconstructing loss data and improving the integrity of a data set; S2, multi-model system establishment: establishing a multi-input multi-output nonlinear multi-model system of the power battery based on the reconstructed system measurement data by using a system identification technology, wherein the multi-model system comprises a plurality of local submodels; S3, real-time estimation of the state of charge: adopting a variable Bayesian method to perform model probability updating on the multi-model system, and fusing an extended Kalman filter, an extended Kalman smoothing and an Aitken smoothing method to accurately and timely estimate the state of charge of the power battery; S4, experimental verification and algorithm improvement: verifying the estimation algorithm of S3 on a simulation or experimental device, and iteratively optimizing algorithm parameters or a model set according to a verification result to ensure the robustness and accuracy of the algorithm in actual application. S5, output result: outputting a final SOC estimation result for accurate management of the battery.
7. The method of claim 6, wherein: In S1, the loss data is reconstructed by using a data model fusion method, and the integrity of the data set is improved to more than 95%.
8. The method of claim 6, wherein: In S2, the identification accuracy of the system estimation model is improved by more than 18% compared with that of a traditional method.
9. The method of claim 6, wherein: In the S2, each local sub-model in the multi-model system is within a respective corresponding specific working interval, and a fitting error thereof is less than or equal to 5%.
10. The method of claim 6, wherein: In the S3, in a scenario of 40% fluctuation of the battery current, an estimation error of the state of charge is reduced by more than 35%.