Battery state-of-charge estimation method capable of resisting fault of voltage sensor

Abstract

The invention belongs to the technical field of battery management systems, and particularly relates to a battery state-of-charge estimation method for resisting a voltage sensor fault, which comprises the following steps: S1, data acquisition and system initialization; s2, state prediction; s3, measurement updating and online compensation are carried out; and S4, recursive circulation is carried out. According to the method, the direct-current bias quantity generated by a voltage sensor due to zero drift is widened into a state variable of a battery model, and the state variable, ohm internal resistance, polarization internal resistance, polarization voltage and charge state of a battery jointly form an expansion state vector; and performing real-time recursive estimation on the extended state vector by adopting an unscented Kalman filtering algorithm, and simultaneously obtaining an estimated value of the offset of the sensor and an estimated value of the SOC of the battery subjected to deviation compensation. Through combination of state extension and unscented Kalman filtering, fault bias of the voltage sensor can be identified and compensated online, non-Gaussian noise interference is effectively suppressed, and accuracy and robustness of battery SOC estimation in a complex industrial environment are remarkably improved.

Description

Technical Field

[0001] This invention belongs to the field of battery management system technology, and relates to lithium-ion battery state of charge estimation, specifically a battery state of charge estimation method that is resistant to voltage sensor failure. Background Technology

[0002] With the acceleration of the global energy transition, the booming development of electric vehicles, large-scale energy storage power stations, and smart microgrids has placed extremely high demands on the reliability and safety of lithium-ion battery management systems. As the core unit ensuring safe, efficient, and long-life operation, the performance of the battery management system directly affects the overall system's operational efficiency and safety baseline. Among the key state parameters of the battery management system, accurate estimation of the state of charge (SOC) is fundamental to achieving energy balance, preventing overcharging and over-discharging, and improving user experience. Currently, state estimation methods based on equivalent circuit models and Kalman filter algorithms have become the mainstream technical solution in industry and academia. However, in the complex industrial operating environment, voltage sensors face severe challenges to measurement reliability due to long-term operation under harsh conditions such as electromagnetic interference and temperature fluctuations. Voltage sensors may experience soft faults such as zero-point drift, resulting in a slowly changing or fixed DC bias in the measurement data. This deviation introduced by the fault couples with complex non-Gaussian background noise, severely undermining the ideal assumption that the measurement noise, upon which traditional state estimation algorithms rely, is Gaussian white noise. This leads to systematic biases in the SOC estimate and can even cause misjudgments by the battery management system, creating serious safety hazards. Therefore, developing a SOC estimation method that maintains high robustness and accuracy even under harsh operating conditions with sensor failure has become a core challenge that urgently needs to be addressed in the field of battery management technology.

[0003] In current research, methods for estimating the state of charge (SOC) of lithium-ion batteries have become increasingly diverse and refined. For example, Wang et al. (Wang Ruoqi, Wang Xiaojia, Yang Qi, et al. Joint estimation of lithium battery SOC / SOH based on dual adaptive unscented Kalman filter algorithm [J]. Mechanical Design and Manufacturing, 2023(1):1-4.) proposed a joint estimation method based on dual adaptive extended Kalman filter. This algorithm significantly improves the estimation accuracy under dynamic operating conditions by updating the noise statistical characteristics in real time. Meanwhile, Zhao et al. (Zhao Yani. Battery SOC estimation based on strong tracking Kalman filter [J]. Journal of Shenyang University of Technology, 2018, 40(2):192-197) developed an SOC estimation strategy based on strong tracking unscented Kalman filter. By introducing a fading factor to weaken the influence of old data, it effectively improves the tracking ability and robustness of the model during sudden changes.

[0004] However, most of these filtering methods focus on optimizing the algorithm's adaptability to model uncertainties and noise, while generally neglecting a key practical engineering problem: sensors may experience soft faults such as zero-point drift due to long-term operation or harsh environments. Such faults introduce an unknown, non-Gaussian DC bias into the measurement data. When the algorithm fails to model this specific fault mode, it not only fails to identify voltage sensor anomalies, but its estimation results (such as SOC) will also have significant errors due to systematic deviations in the measurements, thus posing a hidden danger to the safe management of the battery system. This limitation indicates that existing estimation techniques still need breakthroughs in the deep integration of fault diagnosis and state estimation, and there is an urgent need to develop a new algorithmic framework capable of online identification and compensation for sensor faults to ensure the estimation reliability of the Battery Management System (BMS) under various potential faults. Summary of the Invention

[0005] To address the shortcomings of existing technologies, the present invention aims to propose a battery state-of-charge estimation method to mitigate voltage sensor faults. This method, based on an unscented Kalman filter framework, expands the DC bias caused by zero-point drift of the voltage sensor into a system state variable, constructing an extended state-space model to achieve online identification of sensor faults and dynamic compensation for measurement errors. Furthermore, this method does not rely on the ideal assumption that measurement noise follows a Gaussian distribution, effectively handling common non-Gaussian noise interference in real-world industrial environments.

[0006] This invention is achieved using the following technical solution:

[0007] A method for estimating the state of charge of a battery to withstand voltage sensor failure includes the following steps:

[0008] Step S1: Data Acquisition and System Initialization

[0009] Collect operating current, terminal voltage, and time series data of lithium-ion batteries; initialize the extended state vector and error covariance matrix of the unscented Kalman filter;

[0010] Step S2, State Prediction

[0011] Based on a first-order RC equivalent circuit model with a constant-current pseudo-steady-state resistor, the state equation and observation equation of the battery system are established. The observation equation is characterized as follows: the battery terminal voltage is equal to the sum of the open-circuit voltage, ohmic voltage drop, polarization voltage, and the DC bias of the voltage sensor to be estimated.

[0012] Based on the posterior estimate of the extended state vector at the previous moment and the current current, a nonlinear transformation is performed through the state equation to predict the prior estimate of the extended state vector at the current moment and the prior value of its error covariance matrix.

[0013] Step S3: Measurement Update and Online Compensation

[0014] Substituting the prior estimate of the extended state vector into the observation equation yields the predicted terminal voltage. Subsequently, the Kalman gain is calculated and fused with actual voltage measurements to correct the prior state, resulting in the posterior estimate of the extended state vector at the current moment, which includes high-precision... The estimated value is the same as the estimated value of the DC bias of the voltage sensor;

[0015] Step S4, Recursive Loop

[0016] Using the posterior estimate of the current time step as the initial condition for the next time step, return to steps S2 and S3 to begin a new prediction-update loop, thus realizing the battery... Online, real-time, and collaborative estimation of DC bias of voltage sensors.

[0017] The technical solution provided by this invention has the following advantages compared with the prior art:

[0018] First, this invention, through a state extension method, introduces voltage sensor fault parameters (DC bias) as state variables into the estimation model for the first time, enabling the algorithm to identify the degree of fault online and proactively "sensor" and "compensate" for the fault. Through the collaborative estimation mechanism of state extension and UKF, accurate estimation of the voltage sensor bias is directly used to compensate the observation equation, eliminating systematic error sources at their root. This invention not only improves the estimation of the battery's internal state (such as...) This improves the estimation accuracy and enhances the ability to detect and tolerate abnormal operating states of voltage sensors, thus fundamentally improving fault tolerance.

[0019] Secondly, this invention employs unscented Kalman filtering to address non-Gaussian noise environments. Compared to traditional extended Kalman filtering based on the Gaussian assumption, unscented Kalman filtering accurately processes the probability distribution of nonlinear system states through unscented transformation. By combining state extension with unscented Kalman filtering, it can identify and compensate for fault biases in voltage sensors online, effectively suppressing non-Gaussian noise interference and significantly improving battery performance in complex industrial environments. The accuracy and robustness of the estimation.

[0020] Third, this invention utilizes a recursive "prediction-update" mechanism to achieve online, real-time collaborative estimation of the battery's state of charge and the DC bias of the voltage sensor until the entire charging and discharging process is complete. This closed-loop process ensures that even in the event of a zero-point drift fault in the voltage sensor, the system can continuously provide accurate values. information.

[0021] This invention is reasonably designed. By introducing a bias compensation mechanism into the observation model, it significantly improves the accuracy of state of charge estimation and system robustness of lithium-ion batteries under complex non-Gaussian noise environments such as soft faults in voltage sensors, and has great practical application value. Attached Figure Description

[0022] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 A schematic diagram showing the current and voltage under the World Light Vehicle Test Procedure (WLTP) conditions.

[0025] Figure 2 A schematic diagram showing the current and voltage under dynamic stress test (DST) conditions.

[0026] Figure 3 This graph shows a comparison between the original voltage and the measured voltage under dynamic stress test (DST) conditions.

[0027] Figure 4 This is a comparison chart showing the original voltage and the measured voltage under the World Light Vehicle Test Procedure (WLTP) conditions.

[0028] Figure 5 This graph shows a comparison of the results of the traditional UKF method and the method described in this paper under dynamic stress test (DST) conditions.

[0029] Figure 6 This chart compares the results of the traditional UKF method and the method described in this paper under the World Light Vehicles Test Procedure (WLTP) conditions. Detailed Implementation

[0030] To better understand the above-mentioned objectives, features, and advantages of the present invention, the solutions of the present invention will be further described below. It should be noted that, unless otherwise specified, the embodiments of the present invention and the features thereof can be combined with each other.

[0031] Many specific details are set forth in the following description in order to provide a full understanding of the invention, but the invention may also be practiced in other ways different from those described herein; obviously, the embodiments in the specification are only some embodiments of the invention, and not all embodiments.

[0032] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0033] This invention provides a method for estimating the state of charge of a battery to resist voltage sensor failures. Essentially, it is also a high-precision battery method capable of online identification and compensation for DC bias faults in voltage sensors. The estimation method is a closed-loop process executed cyclically. The core idea is to extend the state to estimate the DC bias of the voltage sensor. ) as a state variable, and the ohmic internal resistance of the battery ( ), polarization resistance ( ), polarization voltage ( ), state of charge ( Together, these elements form the extended state vector. An unscented Kalman filter (UKF) is used to estimate the state of the extended nonlinear system, obtaining high-precision results synchronously and in real-time within a unified "prediction-update" loop. The estimated values ​​and the estimated DC bias of the voltage sensor are used to eliminate the impact of voltage sensor faults. The estimated impact.

[0034] The specific steps are as follows:

[0035] Step S1: Data Acquisition and System Initialization

[0036] S1.1 Data Acquisition: Using a high-precision data acquisition card or the built-in sensors of the battery management system, the working current and terminal voltage of the lithium-ion battery during the charging and discharging process are collected synchronously at a fixed sampling frequency, and the timestamp of each data point is recorded to calculate the sampling time interval.

[0037] S1.2, Model and Parameter Initialization:

[0038] Analyzing the structure of a lithium battery, a constant-current pseudo-steady-state resistor is introduced into the basic first-order RC equivalent circuit. ) and total polarization resistance ( ), establish the equivalent circuit of a lithium battery with a constant current pseudo-steady-state resistor; analyze the typical architecture of the lithium battery management system and the fault modes of the voltage sensor, and introduce the DC bias of the voltage sensor into the state vector of the basic first-order RC equivalent circuit model (). () as a state variable.

[0039] As is easily understood, a typical battery management system (BMS) uses voltage and current sensors to collect data on the battery's operating state. In an ideal laboratory environment, voltage sensors can be considered linear and accurate measuring devices. However, in complex industrial applications, the performance of internal components in voltage sensors may degrade due to long-term operation, temperature fluctuations, and electromagnetic interference, resulting in zero-point and gain drift. Zero-point drift manifests as a non-zero output when the input is zero; its effect is equivalent to superimposing a fixed or slowly varying DC bias of the voltage sensor onto the actual voltage signal. The DC bias of this voltage sensor directly contaminates the measured value, causing systematic errors in the battery state estimation based on this measurement. Failure to effectively identify and compensate for this DC bias will severely affect the accuracy of battery state-of-charge estimation, thereby threatening the safe operation of the battery system.

[0040] The battery model used in this invention is a first-order RC equivalent circuit model with a constant current pseudo-steady-state resistor.

[0041] Initialize the extended state vector x and its error covariance matrix P of the unscented Kalman filter (UKF). The extended state vector includes the battery's ohmic internal resistance, polarization internal resistance, polarization voltage, state of charge, and the DC bias of the voltage sensor to be estimated; the extended state vector is defined as: ;

[0042] in, The internal resistance of the battery is ohmic. This represents the polarization internal resistance of the battery. This is the polarization voltage of the battery; This refers to the battery's state of charge. This represents the DC bias of the voltage sensor to be estimated.

[0043] Step S2: State Prediction (Time Update)

[0044] After introducing the DC bias of the voltage sensor, the system's observation equation is corrected to: That is, the observation equation is represented as: battery terminal voltage It equals the sum of open-circuit voltage, ohmic voltage drop, polarization voltage, and the DC bias of the voltage sensor to be estimated.

[0045] Within this framework, the unscented Kalman filter algorithm performs a "prediction-update" loop in each estimation cycle. In the update step, the algorithm updates the prior estimate of the DC bias of the voltage sensor obtained from the state prediction stage (…). Substituting these values ​​into the observation equation, the predicted terminal voltage is calculated. The core of this operation lies in pre-correcting the model's predicted value based on the algorithm's "cognition" of the fault (i.e., the predicted value of the voltage sensor's DC bias), thereby optimizing the comparison process with the measured value. Subsequently, the algorithm inversely optimizes and adjusts the measured voltage by calculating the residual (information) between the measured voltage and the corrected predicted voltage. and The algorithm estimates all state variables, including those mentioned above. Through this closed-loop feedback mechanism, the algorithm can gradually... The estimated value converges to the true bias, thus effectively eliminating its influence on the voltage sensor's fault level while simultaneously estimating the degree of fault. The impact of the estimated value enables high-precision state perception.

[0046] The equivalent circuit model established in this embodiment of the invention is a first-order RC equivalent circuit model with a constant current pseudo-steady-state resistor, utilizing the battery's... Information, namely, the introduction of ohmic internal resistance based on the first-order RC equivalent circuit model. and polarization resistance Sum of: Total polarization resistance Once the battery reaches a constant current pseudo-steady state, this... equal When the battery has not reached a constant current pseudo-steady state and is charged and discharged with current I, this Will to (SOC,I) converges.

[0047] Therefore, based on the collected experimental data, the physical relationship of the discrete time of the ternary lithium-ion battery, which is based on a first-order RC equivalent circuit model with a constant current pseudo-steady-state resistor, is described by the following equation:

[0048] ;

[0049] in, Let be the current flowing through the constant-current pseudo-steady-state resistor at time k-1. Let K be the current at time k-1. Let K be the current at time k. Let be the current flowing through the constant-current pseudo-steady-state resistor at time k; Let be the polarization voltage at time k. The polarization voltage at time k-1 Let be the terminal voltage at time k. The DC bias of the voltage sensor at time k; Let be the polarization resistance at time k. The polarization resistance at time k-1 Let be the Ohmic internal resistance at time k. The total polarization resistance at time k is... Let k be the state of charge at time k. The state of charge at time k-1, For time intervals, Let k be the time constant at time k-1. Open circuit voltage, This refers to the battery capacity.

[0050] From the above mathematical expression, it can be seen that when the battery is charged and discharged with a constant current I to a constant current pseudo-steady state, the polarization impedance... (in A polarizing capacitor will degenerate into a polarizing resistor. and with Combined, they form the total polarization resistance. As the system enters a constant-current pseudo-steady state, it will continuously converge, and its final convergence value is the constant-current pseudo-steady-state resistance. Therefore, when the battery is under dynamic discharge conditions, the state equation and observation equation of the adaptive estimation algorithm are:

[0051] ;

[0052] in, Let k be the time constant. Let be the Ohmic internal resistance at time k-1. This represents the DC bias of the voltage sensor at time k-1. State noise, To observe noise.

[0053] S2.1 Prior State Prediction: The posterior estimate of the extended state vector based on the previous time step (k-1). and the current collected at the current time (time k). By performing a nonlinear transformation on the system's state equations, the prior estimate of the extended state vector at the current time (time k) is predicted. .

[0054] S2.2, Prior Covariance Prediction: Simultaneously, predict the prior value of the error covariance matrix at the current time. .

[0055] Step S3: Measurement Update and Online Compensation

[0056] Substituting the prior estimate of the extended state vector into the observation equation yields the predicted terminal voltage. Specifically, the DC bias of the voltage sensor in the prior state is used to compensate for the voltage measurement, or directly used to calculate the predicted terminal voltage. Subsequently, the Kalman gain is calculated and fused with the actual voltage measurement to correct the prior state, resulting in a posterior estimate of the extended state vector at the current moment, which includes high-precision... The estimated value is the same as the estimated value of the DC bias of the voltage sensor.

[0057] S3.1, Terminal Voltage Prediction and Online Compensation: The prior estimate of the predicted extended state vector... (This already includes a priori estimates of the DC bias of the voltage sensor) Substitute the values ​​into the corrected observation equation to calculate the predicted terminal voltage. .

[0058] The revised observation equation: .

[0059] in, for The prior prediction value; for The prior prediction value; for The prior prediction value.

[0060] The key to this step is that the observation equation explicitly includes... Therefore, the calculated voltage prediction is naturally pre-compensated based on the current optimal prediction of the DC bias of the voltage sensor, which makes subsequent comparisons with the measured values ​​more accurate.

[0061] S3.2 Calculate the innovation sequence: Calculate the actual voltage measurement value Compared with the compensated predicted value The difference between them is called the innovation sequence: .

[0062] Since the predicted values ​​have already compensated for the estimated DC bias of the voltage sensor, this innovation sequence more accurately reflects the battery model itself (e.g., ...). The error is due to the estimation error of the internal resistance, rather than the error caused by the voltage sensor failure.

[0063] S3.3 Calculate the Kalman gain and update the state:

[0064] (1) Calculate the covariance of the innovation and the cross-covariance between the state and the observation to obtain the optimal Kalman gain. .

[0065] (2) Using Kalman gain, the innovation information is optimally integrated into the prior state estimate to correct the prior state and obtain the posterior estimate of the extended state vector at the current time (time k). and posterior error covariance matrix .

[0066] (3) Output: At this moment It includes a high-precision SOC estimate for the current time (time k) and the DC bias of the voltage sensor. The optimal estimate of ).

[0067] As is easily understood, the measurement update step is the core of the unscented Kalman filter. By fusing actual sensor measurements, it corrects the model's predictions, thus obtaining a more accurate state estimate. This is a crucial step in achieving online fault identification and compensation for voltage sensors. First, the prior state vector obtained in the prediction step is transformed again using an unscented method to generate a set of Sigma points. These points are then propagated through the observation equation to obtain the predicted Sigma point set of the observed quantities. A weighted sum is then used to obtain the predicted terminal voltage. This process crucially incorporates an online compensation mechanism because the observation equation explicitly includes... Therefore, the calculated voltage prediction value It already naturally includes the best predicted value based on the current DC bias of the voltage sensor. Compensation was then performed. Subsequently, the actual voltage measurement value was calculated. Compared with the predicted value The difference between them, i.e., the innovation sequence This innovation sequence has eliminated the influence of predicted sensor bias, thus more accurately reflecting the error in battery state estimation. Next, the algorithm calculates the covariance matrix of the innovation, the cross-covariance matrix of the state and observations, and finally obtains the optimal Kalman gain. Using this gain, the algorithm optimally integrates the innovation information into the prior estimate, completing the state update and updating the error covariance matrix. At this point, the algorithm outputs the optimal posterior estimate of the current state, which includes the battery state of charge. The precise positioning also includes the DC bias of the voltage sensor ( The optimal estimate of ).

[0068] Step S4, Recursive Loop

[0069] S4.1, the posterior estimate obtained at the current time (time k) and posterior error covariance matrix This serves as the initial condition for the filtering process at the next time step (k+1).

[0070] S4.2, Return to steps S2 (state prediction) and S3, then collect the current and voltage data at time k+1, and start a new "prediction-update" loop to realize the battery... Real-time, collaborative estimation of sensor bias.

[0071] This invention achieves control over the battery's state of charge (SOC) through a recursive closed-loop mechanism consisting of the above four steps. The system performs online, real-time coordinated estimation of the DC bias of the voltage sensor, a process that continues throughout the entire battery charge-discharge cycle. This ensures that the system can consistently provide accurate voltage readings even in the event of a zero-point drift fault in the voltage sensor. information.

[0072] Figure 1 The WLTP (World Light Vehicle Test Procedure) operating condition curves are displayed, with the core characteristic being the rapid and significant fluctuations in current. This accurately simulates the transient characteristics of vehicle acceleration (high-rate discharge) and regenerative braking (charging). The global voltage decay trend visually reflects the battery's state of charge (SOC). The process of decreasing as the test progresses.

[0073] Figure 2 The DST (Dynamic Stress Test) test was demonstrated, designed to reproduce battery behavior under complex application environments. Its testing strategy includes a multi-stage discharge phase starting from a full charge, with the current gradually increasing to correspond to varying load demands. In addition, intermittent charging phases were included to evaluate battery performance under simulated real-world fast-charging conditions.

[0074] Figure 3 The voltage conditions under DST (Dynamic Stress Testing) conditions are shown, and the observed voltage with noise is compared with the actual voltage. This reflects the noise environment faced by the voltage signal during dynamic stress testing.

[0075] Figure 4 This diagram illustrates the voltage conditions under WLTP (World Light Vehicle Test Procedure) conditions, specifically comparing the observed voltage with the actual voltage affected by noise interference. The diagram visually demonstrates the noise component contained in the voltage signal during actual measurement.

[0076] Figure 5 Demonstrates performance under DST (Dynamic Stress Testing) conditions based on Figure 3 The voltage data shown are analyzed using both traditional algorithms and the algorithm proposed in this invention. The results of the (State of Charge) estimation are compared. This figure clearly shows that, given the voltage data, the algorithm proposed in this invention can estimate the SOC more accurately.

[0077] Figure 6 Demonstrated performance under WLTP (World Light Vehicles Test Procedure) conditions, based on Figure 4 The voltage data shown are analyzed using both traditional algorithms and the algorithm proposed in this invention. The results of the (State of Charge) estimation are compared. This figure clearly shows that, given the voltage data, the algorithm proposed in this invention can estimate the SOC more accurately.

[0078] It can be seen that, Figure 1 and Figure 2 These represent the current and voltage conditions under two different test conditions. Figure 3 and Figure 4 The figures show the comparison between the measured voltage (which contains non-Gaussian noise and the effects of DC bias) and the actual voltage under two different test conditions, indicating that the test data is far from ideal. Figure 5 and Figure 6 The final output of the entire algorithm under two test conditions is the actual result. The line is black. The purpose of this invention is to make the estimated line as close as possible to the black line. It can be seen that the line (green) of the algorithm proposed in this invention is closer to the black line than the traditional UKF (red), which shows the superiority of the algorithm proposed in this invention.

[0079] The following table compares the SOC results:

[0080]

[0081] The table above shows the mean absolute error (MAE) and root mean square error (RMSE) based on the SOC estimation accuracy. The traditional UKF algorithm has MAE and RMSE exceeding 1% under both discharge conditions, while the algorithm proposed in this invention has MAE and RMSE not exceeding 1% under both conditions.

[0082] In summary, the method of this invention addresses the problem of inaccurate measurement data caused by decreased sensor reliability in complex industrial applications (this problem mainly manifests as non-Gaussian noise and DC bias caused by zero-point drift, which seriously affects the estimation accuracy of battery state of charge). By constructing an equivalent circuit model of a lithium-ion battery, the DC bias caused by zero-point drift of the voltage sensor is extended into a state variable of the battery model, forming an extended state vector together with the battery's ohmic internal resistance, polarization internal resistance, polarization voltage, and state of charge. An unscented Kalman filter algorithm is used to recursively estimate the extended state vector in real time, simultaneously obtaining the estimated value of the voltage sensor's DC bias and the battery's state of charge after bias compensation. Estimated value.

[0083] The "prediction-update" closed-loop design of the method in this invention enables it to adaptively adjust the estimation results as the battery's operating state evolves, achieving true online tracking and compensation. It also enables simple online real-time applications. The entire method uses a recursive algorithm, with controllable computational load and no need to store historical data, making it very suitable for online real-time operation in embedded systems with limited computing resources (such as automotive-grade BMS).

[0084] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the present invention. Although detailed descriptions have been provided with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments, and they should all be covered within the protection scope of the claims.

Claims

1. A method for estimating the state of charge of a battery to withstand voltage sensor failure, characterized in that: Includes the following steps: Step S1: Data Acquisition and System Initialization Collect operating current, terminal voltage, and time series data of lithium-ion batteries; initialize the extended state vector and error covariance matrix of the unscented Kalman filter; Step S2, State Prediction Based on a first-order RC equivalent circuit model with a constant-current pseudo-steady-state resistor, the state equation and observation equation of the battery system are established. The observation equation is characterized as follows: the battery terminal voltage is equal to the sum of the open-circuit voltage, ohmic voltage drop, polarization voltage, and the DC bias of the voltage sensor to be estimated. Based on the posterior estimate of the extended state vector at the previous moment and the current current, a nonlinear transformation is performed through the state equation to predict the prior estimate of the extended state vector at the current moment and the prior value of its error covariance matrix. Step S3: Measurement Update and Online Compensation Substituting the prior estimate of the extended state vector into the observation equation, the predicted value of the terminal voltage is obtained. Subsequently, the Kalman gain is calculated and fused with actual voltage measurements to correct the prior state, yielding a posterior estimate of the extended state vector at the current moment, including high-precision... The estimated value is the same as the estimated value of the DC bias of the voltage sensor; Step S4, Recursive Loop Using the posterior estimate of the current time step as the initial condition for the next time step, return to steps S2 and S3 to begin a new prediction-update loop, thus realizing the battery... Online, real-time, and collaborative estimation of DC bias of voltage sensors.

2. The battery state-of-charge estimation method for resisting voltage sensor failure according to claim 1, characterized in that: Step S1 is as follows: S1.1 Data Acquisition: At a fixed sampling frequency, synchronously acquire the operating current and terminal voltage of the lithium-ion battery during the charging and discharging process, and record the timestamp of each data point to calculate the sampling time interval; S1.2, Model and Parameter Initialization: The battery model used was determined to be a first-order RC equivalent circuit model with a constant current pseudo-steady-state resistor; Initialize the extended state vector x and its error covariance matrix P of the unscented Kalman filter; the extended state vector includes the battery's ohmic internal resistance, polarization internal resistance, polarization voltage, state of charge, and the DC bias of the voltage sensor to be estimated. The extended state vector is defined as: ; in, The internal resistance of the battery is ohmic. This represents the polarization internal resistance of the battery. This is the polarization voltage of the battery; This refers to the battery's state of charge. This represents the DC bias of the voltage sensor to be estimated.

3. The battery state-of-charge estimation method for resisting voltage sensor failure according to claim 2, characterized in that: Step S2 is as follows: Based on the first-order RC equivalent circuit model with a constant-current pseudo-steady-state resistor, its discrete-time physical relationship is described by the following equation: ； in, Let be the current flowing through the constant-current pseudo-steady-state resistor at time k-1. Let K be the current at time k-1. Let K be the current at time k. Let be the current flowing through the constant-current pseudo-steady-state resistor at time k; Let be the polarization voltage at time k. The polarization voltage at time k-1 Let be the terminal voltage at time k. The DC bias of the voltage sensor at time k; Let be the polarization resistance at time k. The polarization resistance at time k-1 Let be the Ohmic internal resistance at time k. The total polarization resistance at time k is... Let k be the state of charge at time k. The state of charge at time k-1, For time intervals, Let k be the time constant at time k-1. Open circuit voltage, Battery capacity; S2.1 Prior State Prediction: Based on the posterior estimate of the extended state vector at the previous time k-1. and the current collected at the current moment By performing a nonlinear transformation on the system's state equations, the prior estimate of the extended state vector at the current time k is predicted. ; S2.2, Prior Covariance Prediction: Simultaneously, predict the prior value of the error covariance matrix at the current time k. .

4. The battery state-of-charge estimation method for resisting voltage sensor failure according to claim 3, characterized in that: Step S3 is as follows: S3.1, Terminal Voltage Prediction and Online Compensation: The prior estimate of the predicted extended state vector... Substituting into the modified observation equation, where the prior estimate of the extended state vector is... The prior estimate of the DC bias of the voltage sensor to be estimated is already included. Calculate the predicted value of the terminal voltage. ; The revised observation equation: ; in, for The prior prediction value; for The prior prediction value; for The prior prediction value; S3.2 Calculate the innovation sequence: Calculate the actual voltage measurement value Compared with the compensated predicted value The difference between them is called the innovation sequence: ; S3.3 Calculate the Kalman gain and update the state: The covariance matrix of the innovation sequence and the cross-covariance matrix of the state and observations are calculated to obtain the optimal Kalman gain. ; Using Kalman gain, the innovation sequence is optimally fused into the prior state estimate, and the prior state is corrected to obtain the posterior estimate of the extended state vector at the current time k. and posterior error covariance matrix ; Output: The posterior estimate at this moment. Includes high precision of the current time k Estimated value and DC bias of voltage sensor The optimal estimate.

5. The battery state-of-charge estimation method for resisting voltage sensor failure according to claim 4, characterized in that: Step S4 is as follows: S4.1, take the posterior estimate obtained at the current time k. and posterior error covariance matrix As the initial condition for the next time step k+1; S4.2 Return to steps S2 and S3, then collect the current and voltage data at time k+1 and start a new prediction-update loop.

Images