Reconfigurable battery system open-circuit voltage online prediction method and device, reconfigurable battery system and medium

Abstract

The invention discloses an open-circuit voltage (OCV) online prediction method and device based on the switching process of a reconfigurable battery system, the reconfigurable battery system and a medium. By identifying a cut-off-short-time relaxation fragment naturally formed by a battery module during series / parallel / bypass topology switching, a port voltage sequence during the period is collected, and five types of features including a mean value, a standard deviation, a voltage drop amplitude, a rolling mean value and a rolling standard deviation are extracted. A pre-trained Gaussian process regression model (GPR) is utilized to predict the static OCV of the battery cell, and a confidence interval is output. According to the method, long-time standing is not needed, the OCV can be obtained on line in the operation process of the battery pack, the prediction error can be controlled within the range of 10 mV, and the method is suitable for active balance control, health state evaluation, parameter identification and the like of a reconfigurable battery system.

Description

Technical Field

[0001] This invention belongs to the field of battery management system (BMS) technology, and particularly relates to a method, device and reconfigurable battery system and medium for online prediction of open circuit voltage of reconfigurable battery system. It is applicable to scenarios that require real-time monitoring and management of battery status, such as energy storage systems, electric vehicles, and drones. Background Technology

[0002] Open-circuit voltage (OCV) is a key parameter characterizing the thermodynamic equilibrium state of a battery, widely used in battery state of charge (SoC) estimation, state of health (SoH) assessment, parameter identification, and active equalization control. Traditional OCV acquisition requires the battery to be left to stand for tens of minutes to several hours to allow the port voltage to approach thermodynamic equilibrium. However, in energy storage systems or electric vehicles, prolonged standing is not possible, making it difficult to obtain OCV in real time, thus affecting equalization, SoH assessment, and parameter identification.

[0003] Reconfigurable battery systems (RBS) have attracted widespread attention due to their ability to dynamically adjust the series and parallel topology of battery cells through a switching matrix. During topology switching, some cells experience a brief current interruption, forming a natural voltage relaxation window. Effectively utilizing these short relaxation segments holds promise for replacing the traditional resting process and enabling rapid online prediction of open voltage voltage (OCV). However, existing technologies still have the following shortcomings: ① Relaxation fragments are short and noisy, making it difficult for traditional modeling methods based on long-term static periods to reliably extract effective information; ② It lacks an automated relaxation segment recognition mechanism, relying on manual intervention or fixed time window interception, resulting in poor adaptability; ③ There is limited research on OCV prediction methods for reconfigurable battery systems, especially in scenarios with low data volume and high noise, where robust and efficient prediction models are lacking.

[0004] Chinese patent document CN120855588A discloses a "reconfigurable modular intelligent battery system and control method," which achieves dynamic adjustment of the battery cell connection topology through a modular design that integrates a switch network and a control unit within the battery module. This solves the problems of low balancing efficiency and poor fault tolerance in traditional battery packs. However, it still relies on traditional static or complex excitation injection, making it difficult to obtain OCV in real time during operation, thus limiting its application effectiveness in dynamic balancing and state assessment.

[0005] Therefore, a new method is needed that can automatically collect relaxation fragments, extract robust features, and use machine learning methods (such as GPR) to predict OCV during RBS operation. Summary of the Invention

[0006] This invention aims to overcome the dependence of traditional methods on resting time, as well as the shortcomings of existing methods utilizing relaxation segments, such as noise sensitivity, lack of automatic identification mechanism, and insufficient robustness of prediction models. It provides a method, device, and reconfigurable battery system and dielectric for online prediction of open-circuit voltage of a reconfigurable battery system. By automatically capturing and utilizing the short-time current interruption window naturally formed during system topology switching, relaxation voltage characteristics are extracted, and fast and high-precision online prediction of OCV is achieved through a pre-trained machine learning model.

[0007] The technical solution of the present invention is as follows: A method for online prediction of open-circuit voltage of a reconfigurable battery system, characterized by the following steps: S1: During the normal operation of the reconfigurable battery system, monitor and identify in real time the moment when at least one cell enters a natural disconnection state from the loaded state due to the switching matrix performing series, parallel or bypass topology switching operations. S2: Starting from the moment when the natural current interruption state begins, automatically capture the dynamic voltage decay sequence of the cell within a first preset time after the current interruption, as a short relaxation segment for open circuit voltage prediction, wherein the first preset time is less than the standard resting time required to obtain the thermodynamic equilibrium open circuit voltage. S3: Extract features from the dynamic voltage decay sequence of the captured short-time relaxation segment and construct a feature vector containing at least two different types of statistical features. The different types of statistical features are selected from at least: global statistical features representing the overall level, stability features representing fluctuation characteristics, and dynamic features representing local change trends. S4: Input the feature vector into a pre-trained Gaussian process regression prediction model, and the model outputs the predicted open-circuit voltage of the battery cell in the current state and its corresponding confidence interval.

[0008] Furthermore, in step S1, the identification of the natural interruption state is achieved based on real-time monitoring of the digital control signal of the switch matrix.

[0009] Furthermore, in step S2, the first preset duration is 1-10 minutes.

[0010] Furthermore, in step S3, the feature vector specifically includes at least three of the following five types of features: The average voltage value of the dynamic voltage decay sequence; The voltage standard deviation of the dynamic voltage decay sequence; The total voltage drop amplitude of the dynamic voltage decay sequence within the first preset time period; The rolling average of the dynamic voltage decay sequence calculated based on a sliding window; The rolling standard deviation of the dynamic voltage decay sequence calculated based on a sliding window.

[0011] Furthermore, in step S4, the Gaussian process regression prediction model is trained based on historical relaxation segment data containing different charge / discharge rates, ambient temperatures, and cell aging degrees. The confidence interval is used to quantify the uncertainty of the current prediction result and serves as an indicator for evaluating the quality of the short-term relaxation segment.

[0012] Furthermore, step S4 is followed by: S5: Using the predicted open-circuit voltage value and the confidence interval, perform at least one of the following battery management functions: - When the confidence interval meets the preset reliability threshold, active equalization control based on open-circuit voltage consistency is triggered; - Correct or calibrate the estimated state of charge of the battery cell; - Used as input parameters to update the battery health status assessment model.

[0013] Second, the present invention provides an online open-circuit voltage prediction device for implementing the above method, characterized in that it is integrated into the battery management system of a reconfigurable battery system, comprising: The switch status monitoring and segment recognition unit is used to monitor the switch matrix status in real time and identify natural interruption events caused by topology switching, triggering relaxation segment capture. The relaxation data acquisition and feature extraction unit is connected to the switch state monitoring and segment recognition unit, and is used to acquire the dynamic voltage decay sequence of the short-time relaxation segment and extract multi-dimensional statistical features to construct a feature vector. The Gaussian process regression prediction unit has a built-in pre-trained Gaussian process regression model and is connected to the relaxation data acquisition and feature extraction unit. It is used to predict the open-circuit voltage based on the feature vector and output the prediction result with a confidence interval.

[0014] Furthermore, the prediction results with confidence intervals output by the Gaussian process regression prediction unit are configured to be directly input to at least one of the equalization control module, state estimation module, and health management module of the battery management system.

[0015] Third, a reconfigurable battery system, characterized in that it includes a battery module, a switch matrix, a sampling circuit, a control module, and an online open-circuit voltage prediction device as described above.

[0016] Fourth, a computer-readable storage medium storing a computer program thereon, characterized in that, when the program is executed by a processor, it controls the processor to execute the above-described method.

[0017] Compared with the prior art, the technical effects of the present invention are as follows: By utilizing the short-term interruption window naturally formed during topology switching in reconfigurable battery systems (series, parallel, bypass, etc.), rapid prediction of open-circuit voltage (OCV) without the need for prolonged resting is achieved. Verified in actual operating environments, with typical relaxation segments lasting less than 5 minutes, the proposed feature extraction method stably captures relaxation trends, keeping the OCV error of the prediction model within 10 mV under different rate conditions, with deviations of less than 5 mV for most segments. Compared to traditional methods that rely on a 60-120 minute resting period to obtain approximate OCV, this invention reduces the time available for OCV acquisition by over 90%, while maintaining good robustness even under conditions of high noise and significant load fluctuations. Furthermore, the probabilistic prediction method employed in this invention outputs an uncertainty index including a confidence interval, which can not only be used to identify poor-quality relaxation segments but also provide decision-making support for equalization control, health assessment, and cell selection. In reconfigurable battery systems, introducing the online OCV prediction capability of this invention improves SoC estimation accuracy in the charge / discharge region, thereby increasing usable capacity and reducing the risk of overcharging and over-discharging. In summary, this invention significantly improves the state awareness, safety, and energy management efficiency of reconfigurable battery systems under actual operating conditions, and has outstanding engineering application value and technical advantages. Attached Figure Description

[0018] Figure 1 This is a schematic diagram illustrating the functional location of the method of the present invention in a reconfigurable battery system. Figure 2 The voltage-time curve of three parallel battery cells during reconfigurable operation. Figure 3 This is a schematic diagram of the three-switch reconfigurable battery cell used in this invention. Figure 4 This is a design diagram of the reconfigurable battery drive circuit used in this invention. Detailed Implementation The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and do not constitute a limitation thereof.

[0019] This invention monitors the state of the switching matrix of a reconfigurable battery system and automatically identifies the moment when a cell switches from a series, parallel, or bypass state to a current-disconnected state as the starting point of a short-time relaxation segment. Because the current in a cell rapidly drops to zero after current disconnection, its port voltage undergoes a natural relaxation process that decays over time. This invention utilizes this physical characteristic to automatically extract a relaxation voltage sequence of a certain duration after each topology switch, replacing the voltage evolution curve required for a traditional long resting process.

[0020] To improve the robustness and modelability of relaxation information, this invention extracts features from the extracted voltage sequence. Multiple statistical features capable of stably characterizing the relaxation trend are selected, including but not limited to the average voltage, voltage fluctuation level, voltage drop magnitude, and local statistics based on window functions. By extracting representative low-dimensional features, this invention effectively mitigates the impact of scene noise, sampling jitter, and rate changes, making short-term relaxation segments suitable for the generalization requirements of machine learning models.

[0021] After feature extraction, this invention employs a regression method based on probabilistic modeling to predict OCV. This method can simultaneously output predicted values ​​and uncertainties, thus reflecting the impact of relaxation segment quality and current operating conditions on the prediction results. The prediction model is trained based on a large amount of offline experimental data, including relaxation segment features under multiple rate, temperature, and cycle conditions, enabling the model to adapt to changes in cell behavior under different operating conditions. The OCV inferred by the model can approximate the true static OCV without requiring a resting period.

[0022] Through the above technical solution, this invention achieves the ability to automatically acquire open-circuit voltage during the normal operation of a reconfigurable battery system. Compared with the prior art, this invention has the following significant advantages: First, prediction does not require a long resting process and OCV can be acquired online, thereby achieving real-time balancing, cell consistency adjustment, and health status assessment during vehicle operation or energy storage system operation; Second, the relaxation window naturally formed by topology switching does not require additional excitation or increased hardware costs; Third, based on the combination of statistical characteristics and probabilistic models, the method has strong noise resistance and can be applied to cells with different rates, temperatures, and aging levels; Fourth, the prediction results include uncertainty, which can assist the BMS in making safety decisions and adaptive control.

[0023] The embodiments of the present invention are based on Figure 1The reconfigurable battery system (RBS) architecture is shown. This system comprises several battery cells, a switching matrix, a sampling circuit, and a control module. During system operation, each battery cell can be reconfigured via the switching matrix in different ways, such as series connection, parallel connection, or bypass. Because topology switching generates a short-term current interruption window, the cell port voltage exhibits a regular relaxation change after the current is interrupted. This invention utilizes this characteristic to achieve online prediction of the open-circuit voltage.

[0024] The control module first continuously monitors the digital input status of the switching matrix to identify when the battery cell switches from a loaded state to a disconnected state. When a change in the topology is detected, the control module records the moment and automatically extracts a subsequent relaxation voltage sequence as the analysis object. This sequence covers the short-time voltage change process of the battery cell from a loaded state to a near-resting state.

[0025] The system preprocesses the extracted relaxation voltage sequence and extracts several statistical features that characterize the voltage drop trend, fluctuation characteristics, and overall level. These features stably reflect relaxation behavior and are not significantly affected by sampling noise and rate fluctuations, making them suitable as model input.

[0026] After feature extraction, the control module inputs it into a pre-trained prediction model to obtain an estimate of the open-circuit voltage of the corresponding battery cell. This model, based on a probabilistic regression structure, can simultaneously output predicted values ​​and uncertainties, used to evaluate cell quality and prediction reliability. The prediction results can be periodically reported to the battery management system for subsequent equalization control, state estimation, and health assessment.

[0027] The embodiments of the present invention are characterized by simple implementation, low hardware dependence, and minimal interference with system operation, and are applicable to various topologies of reconfigurable battery systems. It should be noted that, without affecting the core idea of ​​the present invention, the method of extracting relaxation segments, the feature construction method, and the prediction model structure can all be replaced or adjusted according to actual needs, and all such modifications should be considered to fall within the protection scope of the present invention.

[0028] In summary, this invention proposes a rapid prediction method for open-circuit voltage (OCV) based on the short-time relaxation segment of a reconfigurable battery system. Through a simple structural solution that requires no additional sensors and no static setup to obtain OCV, it significantly improves the state estimation capability of the battery management system under dynamic operating conditions. This provides an important foundation for cell consistency maintenance, online balancing strategy optimization, and health assessment, demonstrating significant innovation and practical value.

[0029] Specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art can make various modifications or variations within the scope of the claims, which do not affect the essence of the present invention. Where there is no conflict, the above embodiments and features described therein can be combined with each other.

Claims

1. A method for online prediction of open-circuit voltage of a reconfigurable battery system, characterized in that, Includes the following steps: S1: During the normal operation of the reconfigurable battery system, monitor and identify in real time the moment when at least one cell enters a natural disconnection state from the loaded state due to the switching matrix performing series, parallel or bypass topology switching operations. S2: Starting from the moment when the natural current interruption begins, automatically capture the dynamic voltage decay sequence of the cell within a first preset time after the current interruption, as a short relaxation segment for open circuit voltage prediction, wherein the first preset time is less than the standard resting time required to obtain the thermodynamic equilibrium open circuit voltage. S3: Extract features from the dynamic voltage decay sequence of the captured short-time relaxation segment and construct a feature vector containing at least two different types of statistical features. The different types of statistical features are selected from at least: global statistical features representing the overall level, stability features representing fluctuation characteristics, and dynamic features representing local change trends. S4: Input the feature vector into a pre-trained Gaussian process regression prediction model, and the model outputs the predicted open-circuit voltage of the battery cell in the current state and its corresponding confidence interval.

2. The online prediction method for open-circuit voltage according to claim 1, characterized in that, In step S1, the identification of the natural interruption state is achieved based on the real-time monitoring of the digital control signal of the switch matrix.

3. The online prediction method for open-circuit voltage according to claim 1, characterized in that, In step S2, the first preset duration is 1-10 minutes.

4. The online prediction method for open-circuit voltage according to claim 1, characterized in that, In step S3, the feature vector specifically includes at least three of the following five types of features: The average voltage value of the dynamic voltage decay sequence; The voltage standard deviation of the dynamic voltage decay sequence; The total voltage drop amplitude of the dynamic voltage decay sequence within the first preset time period; The rolling average of the dynamic voltage decay sequence calculated based on a sliding window; The rolling standard deviation of the dynamic voltage decay sequence calculated based on a sliding window.

5. The online prediction method for open-circuit voltage according to claim 1, characterized in that, In step S4, the Gaussian process regression prediction model is trained based on historical relaxation segment data containing different charge / discharge rates, ambient temperatures, and cell aging degrees. The confidence interval is used to quantify the uncertainty of the current prediction result and serves as an indicator for evaluating the quality of the short-term relaxation segment.

6. The online prediction method for open-circuit voltage according to claim 1, characterized in that, Step S4 is followed by: S5: Using the predicted open-circuit voltage value and the confidence interval, perform at least one of the following battery management functions: - When the confidence interval meets the preset reliability threshold, active equalization control based on open-circuit voltage consistency is triggered; - Correct or calibrate the estimated state of charge of the battery cell; - Used as input parameters to update the battery health status assessment model.

7. An online open-circuit voltage prediction device for implementing the method of any one of claims 1 to 6, characterized in that, The battery management system integrated into the reconfigurable battery system includes: The switch status monitoring and segment recognition unit is used to monitor the switch matrix status in real time and identify natural interruption events caused by topology switching, triggering relaxation segment capture. The relaxation data acquisition and feature extraction unit is connected to the switch state monitoring and segment recognition unit, and is used to acquire the dynamic voltage decay sequence of the short-time relaxation segment and extract multi-dimensional statistical features to construct a feature vector. The Gaussian process regression prediction unit has a built-in pre-trained Gaussian process regression model and is connected to the relaxation data acquisition and feature extraction unit. It is used to predict the open-circuit voltage based on the feature vector and output the prediction result with a confidence interval.

8. The online open-circuit voltage prediction device according to claim 7, characterized in that, The prediction results with confidence intervals output by the Gaussian process regression prediction unit are configured to be directly input to at least one of the equalization control module, state estimation module, and health management module of the battery management system.

9. A reconfigurable battery system, characterized in that, It includes a battery module, a switch matrix, a sampling circuit, a control module, and an online open-circuit voltage prediction device as described in claim 7 or 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it controls the processor to perform the method of any one of claims 1 to 6.

Images