Battery remote monitoring method, device and equipment and storage medium

Abstract

The invention provides a battery remote monitoring method, device and equipment and a storage medium, and the method comprises the steps: carrying out the dimension reduction of a preset state space model of a target battery, and obtaining a state equation and a measurement equation after the dimension reduction; calculating a center and a shape matrix of a predicted ellipsoid according to the state equation, calculating a center and a shape matrix of an estimated ellipsoid according to the measurement equation and the measurement data, and respectively obtaining the predicted ellipsoid and the estimated ellipsoid; calculating an intersection-union ratio of the predicted ellipsoid and the estimated ellipsoid, and comparing the intersection-union ratio with a threshold value to obtain an anomaly detection result; and performing interval calculation on the battery state according to the estimated ellipsoid or the predicted ellipsoid in case of an abnormal detection result to obtain a confidence interval of the battery state. According to the method, the confidence interval of the battery state is provided through the ellipsoid set estimation method, data anomaly detection is carried out based on the intersection-to-union ratio of the predicted ellipsoid and the estimated ellipsoid, and the reliability and safety of remote monitoring of the battery are improved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method, apparatus, device, and storage medium for remote battery monitoring. Background Technology

[0002] With the rapid development of electric vehicles and energy storage systems, accurate monitoring of battery status by battery management systems has become increasingly important. Traditional remote battery monitoring methods typically use Kalman filtering and other point estimation algorithms to estimate the battery's state of charge and health. These methods can only provide a single estimate and cannot provide a range of reliability for the estimation results, making it difficult for users to judge the reliability of the estimation results and posing significant safety risks in practical applications.

[0003] Furthermore, during remote battery monitoring, measurement data needs to be transmitted to the monitoring center via a wireless network. However, existing technologies lack an effective data anomaly detection mechanism. When measurement data during transmission is affected by network attacks or sensor malfunctions, the monitoring system cannot detect abnormal data in time and continues to use contaminated data for state estimation, resulting in estimation results that deviate significantly from the true value, affecting the safety and reliability of battery management. Summary of the Invention

[0004] The main objective of this invention is to solve the technical problem that existing battery remote monitoring methods cannot provide a reliable range for state estimation and lack an effective data anomaly detection mechanism, resulting in poor reliability of the estimation results. This invention provides a method for remote battery monitoring, the method comprising: The pre-defined state-space model of the target battery is reduced in dimension to obtain the reduced state equation and measurement equation. The center and shape matrices of the predicted ellipsoid are calculated based on the reduced state equation, and the center and shape matrices of the estimated ellipsoid are calculated based on the reduced measurement equation and measurement data, thus obtaining the predicted ellipsoid and the estimated ellipsoid, respectively. The intersection-union ratio (IUU) of the predicted ellipsoid and the estimated ellipsoid is calculated, and the IUU is compared with a threshold to obtain the anomaly detection result. When the anomaly detection result does not indicate that an anomaly has been detected, the battery state is calculated in intervals based on the estimated ellipsoid. When the anomaly detection result indicates that an anomaly has been detected, the predicted ellipsoid is used to calculate in intervals to obtain the confidence interval of the battery state.

[0005] The present invention also provides a battery remote monitoring device, the battery remote monitoring device comprising: The dimension reduction module is used to reduce the dimension of the preset state space model of the target battery to obtain the reduced state equation and measurement equation. The ellipsoid estimation module is used to calculate the center and shape matrix of the predicted ellipsoid based on the reduced state equation, and to calculate the center and shape matrix of the estimated ellipsoid based on the reduced measurement equation and measurement data, so as to obtain the predicted ellipsoid and the estimated ellipsoid respectively. An anomaly detection module is used to calculate the intersection-union ratio (IUR) between the predicted ellipsoid and the estimated ellipsoid, and compare the IUR with a threshold to obtain an anomaly detection result. The interval calculation module is used to perform interval calculation on the battery state based on the estimated ellipsoid when the anomaly detection result does not indicate that an anomaly has been detected, and to perform interval calculation using the predicted ellipsoid when the anomaly detection result indicates that an anomaly has been detected, so as to obtain the confidence interval of the battery state.

[0006] The present invention also provides a battery remote monitoring device, comprising: a memory and at least one processor, wherein the memory stores instructions, and the memory and the at least one processor are interconnected via a line; the at least one processor invokes the instructions in the memory to cause the battery remote monitoring device to perform the steps of the battery remote monitoring method described above.

[0007] The present invention also provides a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the steps of the above-described battery remote monitoring method.

[0008] Beneficial effects: The aforementioned battery remote monitoring method, device, equipment, and storage medium reduce the computational complexity of ellipsoid set estimation by dimensionality reduction of the state space model, enabling confidence interval calculation to run in real time in resource-constrained battery management systems. By providing prediction results based on the physical model and estimation results based on fused measurement data through predicted and estimated ellipsoids respectively, and representing the uncertainty range of the state in ellipsoidal form, the system outputs a confidence interval for the battery state instead of a single estimated value, solving the problem that existing point estimation methods cannot provide a confidence range. Anomaly detection is performed by calculating the intersection-union ratio (IUU) of the predicted and estimated ellipsoids. When measurement data is attacked or affected by sensor malfunctions, the estimated ellipsoid deviates from the predicted ellipsoid, causing a decrease in the IUU, thus timely detecting data anomalies. Upon detecting anomalies, the uncontaminated predicted ellipsoid is used instead of the estimated ellipsoid for confidence interval calculation, avoiding the impact of abnormal data on state estimation and solving the problem of the lack of an effective data anomaly detection mechanism in existing technologies.

[0009] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention are realized and obtained in accordance with the structures particularly pointed out in the description, claims and drawings.

[0010] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0011] Figure 1 This is a schematic diagram of the first embodiment of the battery remote monitoring method in this invention; Figure 2 This is a schematic diagram of a second embodiment of the battery remote monitoring method in this invention; Figure 3 This is a schematic diagram of one embodiment of the battery remote monitoring device in this invention; Figure 4 This is a schematic diagram of one embodiment of the battery remote monitoring device in this invention. Detailed Implementation

[0012] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions 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, 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.

[0013] The terms "comprising" and "having," and any variations thereof, used in the embodiments of this invention are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units is not limited to the steps or units listed, but may optionally include other steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or devices.

[0014] To facilitate understanding of this embodiment, a battery remote monitoring method disclosed in this invention will first be described in detail. For example... Figure 1 As shown, this method includes the following steps: 101. Dimensionally reduce the pre-defined state space model of the target battery to obtain the dimensionality-reduced state equation and measurement equation; In this embodiment, the step of dimensionality reduction processing of the preset state space model of the target battery to obtain the dimensionality-reduced state equation and measurement equation includes: obtaining the original state variables in the preset state space model of the target battery, the original state variables including charge state variables and multiple polarization voltage variables; merging the multiple polarization voltage variables by weighted summation to obtain a comprehensive polarization voltage variable; constructing a dimensionality-reduced projection matrix based on the charge state variables and the comprehensive polarization voltage variable; and transforming the state transition matrix, input matrix, and observation matrix of the state space model through the dimensionality-reduced projection matrix to obtain the dimensionality-reduced state equation and measurement equation.

[0015] Specifically, when performing dimensionality reduction on the pre-defined state-space model of the target battery, it is first necessary to understand the meaning of the state-space model. A state-space model is a mathematical model used to describe dynamic processes, capable of characterizing the battery's operating state at different times through a set of state variables. In this embodiment, the model is based on a second-order equivalent circuit and is used to describe the battery's electrochemical characteristics and dynamic response behavior.

[0016] After obtaining the state-space model, it is necessary to extract the primitive state variables. Primitive state variables refer to the basic physical quantities in the model used to fully describe the battery state, including the state of charge variable, the first polarization voltage variable, and the second polarization voltage variable. The state of charge variable represents the proportion of the battery's current remaining charge to its total capacity; the first and second polarization voltage variables correspond to two electrochemical polarization processes at two different time scales within the battery, with the first polarization voltage variable reflecting rapid electrochemical reactions and the second polarization voltage variable reflecting slow concentration polarization. Since these three variables work together to accurately describe the complete state of the battery, they are called primitive state variables.

[0017] After identifying the initial state variables, multiple polarization voltage variables need to be merged. Since both the first and second polarization voltage variables physically reflect the polarization phenomenon inside the battery and are strongly correlated, they can be merged into a single comprehensive polarization voltage variable through weighted summation. Specifically, correlation analysis is used to assess the degree of correlation between the numerical changes of the two polarization voltage variables, and their respective weighting coefficients are determined based on the strength of the correlation. The weighting coefficients reflect the proportion of each polarization voltage variable's contribution to the comprehensive polarization voltage variable. By multiplying each polarization voltage variable by its corresponding weighting coefficient and then summing them, the comprehensive polarization voltage variable can be obtained. This merging process reduces the number of state variables from three to two while preserving the main electrochemical characteristics.

[0018] After obtaining the integrated polarization voltage variable, a dimension-reduced projection matrix is ​​constructed based on the retained state-of-charge (POC) variable and the newly generated integrated polarization voltage variable. The dimension-reduced projection matrix is ​​a linear transformation tool used to establish a mapping relationship between the original state space and the dimension-reduced state space. The construction of this matrix follows these principles: fully preserving the information of the POC variable, as the POC is the most critical parameter in battery management; and linearly combining the two polarization voltage variables according to weighted coefficients to form a single integrated polarization voltage variable. The dimension-reduced projection matrix constructed in this way can reduce computational complexity while preserving as much of the physical meaning and dynamic characteristics of the original model as possible.

[0019] After constructing the dimension-reduced projection matrix, this matrix is ​​used to transform the key matrices in the state-space model. The state transition matrix describes the evolution of the battery state over time, the input matrix describes how external inputs affect the battery state, and the observation matrix describes how measurable output quantities are derived from state variables. By performing a linear transformation on these matrices using the dimension-reduced projection matrix, the dynamic equations originally defined in the three-dimensional state space can be converted into equivalent equations defined in the two-dimensional state space, thus obtaining the dimension-reduced state equations and measurement equations. These dimension-reduced equations significantly reduce the computational cost of subsequent ellipsoidal ensemble estimation while maintaining the accuracy of the battery's dynamic characteristics description, enabling the method to run in real time on computationally limited battery management units.

[0020] 102. Calculate the center and shape matrix of the predicted ellipsoid based on the reduced state equation, and calculate the center and shape matrix of the estimated ellipsoid based on the reduced measurement equation and measurement data, to obtain the predicted ellipsoid and the estimated ellipsoid respectively; In this embodiment, ellipsoidal ensemble estimation is a state estimation method based on set theory. Unlike traditional point estimation methods, this method can provide a confidence region that includes the true state. The ellipsoid, as a geometric shape, is fully described by its center and shape matrix. The center represents the optimal estimate of the state, while the shape matrix describes the distribution characteristics of the state uncertainty. The size and orientation of the ellipsoid are determined by the shape matrix.

[0021] Specifically, in estimating the ellipsoid set, the first step is to calculate the predicted ellipsoid based on the reduced-dimensional state equations. The predicted ellipsoid is derived from the battery physical model and the state information from the previous time step, using state transition relationships to deduce the possible regions where the battery state may exist at the current time step. This process does not depend on the measurement data at the current time step; therefore, the predicted ellipsoid reflects the state deduction results purely based on the physical model. When calculating the center of the predicted ellipsoid, a linear transformation is performed using the state transition matrix and the estimated ellipsoid center from the previous time step. When calculating the shape matrix of the predicted ellipsoid, the propagation of uncertainties during the state transition process and the influence of process noise need to be considered.

[0022] Subsequently, an estimated ellipsoid is calculated based on the dimension-reduced measurement equations and the actual measurement data. The measurement data includes directly measurable physical quantities such as battery terminal voltage and operating current. The estimated ellipsoid is obtained by fusing the predicted ellipsoid and measurement information, reflecting the integration of physical model predictions and actual measurement data. When calculating the center of the estimated ellipsoid, a Kalman filter-like update mechanism is used to correct for deviations between the prediction results and measurement data. When calculating the shape matrix of the estimated ellipsoid, the effects of prediction uncertainty and measurement noise need to be comprehensively considered. Through the above processing, the predicted ellipsoid reflecting the model prediction and the estimated ellipsoid fused with measurement information are obtained, respectively.

[0023] 103. Calculate the intersection-union ratio (IUR) of the predicted ellipsoid and the estimated ellipsoid, and compare the IUR with a threshold to obtain the anomaly detection result; In this embodiment, calculating the intersection-union ratio (IUR) of the predicted ellipsoid and the estimated ellipsoid, and comparing the IUR with a threshold to obtain an anomaly detection result includes: generating a first set of sampling points using a random sampling method based on the center and shape matrix of the predicted ellipsoid, and generating a second set of sampling points using a random sampling method based on the center and shape matrix of the estimated ellipsoid; calculating the Mahalanobis distance of each sampling point in the first set of sampling points based on the center and shape matrix of the estimated ellipsoid, counting the number of sampling points whose Mahalanobis distance satisfies the ellipsoid constraints, and obtaining the number of intersection samples; calculating the IUR of the predicted ellipsoid and the estimated ellipsoid based on the total number of the first set of sampling points, the total number of the second set of sampling points, and the number of intersection samples; comparing the IUR with a threshold, determining that an anomaly has been detected when the IUR is less than the threshold, and determining that no anomaly has been detected when the IUR is greater than or equal to the threshold, thus obtaining an anomaly detection result.

[0024] Specifically, when calculating the intersection-union ratio (IUR) between the predicted and estimated ellipsoids, it's essential to first understand its meaning. The IUR is the ratio of the volume of the intersection to the volume of the union of the two ellipsoids, used to measure the degree of overlap. When the predicted and estimated ellipsoids highly overlap, the IUR is close to 1, indicating good agreement between the prediction based on the physical model and the estimation based on the fused measurement data. When the overlap is low, the IUR is close to 0, indicating a significant deviation between the prediction and estimation results, potentially due to abnormal interference with the measurement data.

[0025] In this embodiment, since directly calculating the exact intersection-union ratio of two ellipsoids in high-dimensional space involves complex computational geometry problems, the computational load is extremely large, and it is difficult to obtain an analytical solution, the Monte Carlo sampling method is used for approximate calculation. The Monte Carlo method is a numerical calculation method based on random sampling, which uses the statistical properties of a large number of random samples to approximate the solution of complex mathematical problems.

[0026] Specifically, a first set of sampling points is generated based on the center and shape matrix of the predicted ellipsoid. This process is achieved through random sampling using a multivariate normal distribution, where the distribution characteristics of the sampling points are jointly determined by the center position and shape matrix of the predicted ellipsoid. The center position determines the central distribution area of ​​the sampling points, while the shape matrix determines the degree of dispersion of the sampling points in different directions. By generating several random sampling points, these discrete point sets can be used to approximate the spatial distribution of the predicted ellipsoid. Similarly, a second set of sampling points is generated based on the center and shape matrix of the estimated ellipsoid, which is used to characterize the spatial distribution of the estimated ellipsoid. In one embodiment, the number of sampling points in the first and second sets can be set to the same value, for example, one hundred sampling points each, to ensure a balance between computational accuracy and computational efficiency.

[0027] After generating the first and second sampling point sets, it is necessary to calculate the intersection of the two ellipsoids. Specifically, for each sampling point in the first sampling point set, it is determined whether the sampling point falls within the estimated ellipsoid. The determination method uses the Mahalanobis distance criterion. Mahalanobis distance is a distance metric that considers the data distribution characteristics and can effectively measure the normalized distance of a sampling point relative to the center of the ellipsoid. For any sampling point in the first sampling point set, the Mahalanobis distance from the sampling point to the center of the estimated ellipsoid is calculated based on the center and shape matrix of the estimated ellipsoid. When the Mahalanobis distance is less than or equal to 1, it indicates that the sampling point satisfies the ellipsoid constraint, i.e., the sampling point falls within the estimated ellipsoid; when the Mahalanobis distance is greater than 1, it indicates that the sampling point falls outside the estimated ellipsoid. By traversing all sampling points in the first sampling point set and counting the number of sampling points that satisfy the ellipsoid constraint, the number of intersection samples can be obtained. The number of intersection samples reflects the number of sampling points contained in the overlapping part of the predicted ellipsoid and the estimated ellipsoid.

[0028] After obtaining the number of intersection samples, the intersection-union ratio (IUR) between the predicted ellipsoid and the estimated ellipsoid is calculated based on the total number of samples in the first sampling point set, the total number of samples in the second sampling point set, and the number of intersection samples. Specifically, the intersection is approximated by the number of intersection samples, the union is approximated by the sum of the total number of samples in the first sampling point set and the total number of samples in the second sampling point set, and the IUR is equal to the intersection divided by the union. This calculation process utilizes the statistical characteristics of the sampling points to approximate the volume relationship of the ellipsoid, avoiding complex analytical calculations.

[0029] After calculating the intersection-union ratio (IUR), this value is compared with a preset threshold to determine if any anomalies exist. The threshold is set based on the typical overlap between the predicted and estimated ellipsoids under normal operating conditions. When the IUR is less than the threshold, it indicates that the overlap between the two ellipsoids is too low, an anomaly is detected, and an anomaly detection result is output. When the IUR is greater than or equal to the threshold, it indicates that the overlap between the two ellipsoids is normal, no anomaly is detected, and a normal detection result is output. This anomaly detection mechanism can promptly detect the impact of network attacks or sensor malfunctions on measurement data.

[0030] 104. When the anomaly detection result does not indicate that an anomaly has been detected, the battery state is calculated in intervals based on the estimated ellipsoid. When the anomaly detection result indicates that an anomaly has been detected, the predicted ellipsoid is used to calculate in intervals to obtain the confidence interval of the battery state.

[0031] In this embodiment, the step of calculating the battery state interval based on the estimated ellipsoid when the anomaly detection result does not indicate that an anomaly has been detected, and calculating the interval using the predicted ellipsoid when the anomaly detection result indicates that an anomaly has been detected, to obtain the confidence interval of the battery state includes: determining whether an anomaly has been detected based on the anomaly detection result; selecting the estimated ellipsoid as the target ellipsoid when no anomaly has been detected, and selecting the predicted ellipsoid as the target ellipsoid when an anomaly has been detected; extracting the center vector and shape matrix of the target ellipsoid, and calculating the variance value of the battery state variable direction based on the diagonal elements of the shape matrix in the direction of the battery state variable; calculating the standard deviation of the battery state variable based on the statistical distribution critical value corresponding to the preset confidence level and the variance value; and calculating the upper and lower bound values ​​of the battery state based on the battery state estimate in the center vector and the standard deviation to obtain the confidence interval of the battery state.

[0032] Specifically, when calculating the confidence interval for battery state based on anomaly detection results, it is first necessary to understand the meaning of the confidence interval. A confidence interval is a numerical range used to represent the region where the battery's true state may exist. Unlike traditional methods that only provide a single estimate, the confidence interval provides upper and lower bounds for the state estimate, allowing users to understand the reliability and uncertainty range of the estimation results.

[0033] In this embodiment, anomaly detection results determine whether an anomaly has been detected, and a suitable ellipsoid is selected for confidence interval calculation accordingly. When the anomaly detection result does not indicate an anomaly, the measurement data is considered reliable. In this case, the estimated ellipsoid, which incorporates measurement information, is selected as the target ellipsoid because it combines physical model predictions and actual measurement data, providing a more accurate state estimate. When the anomaly detection result indicates an anomaly, the measurement data may be unreliable due to network attacks, sensor malfunctions, or communication interference. In this case, the predicted ellipsoid, based solely on the physical model, is selected as the target ellipsoid. The predicted ellipsoid does not depend on the current measurement data and is therefore not contaminated by anomalous measurement data, providing a more reliable state estimate. This adaptive selection mechanism ensures that reliable battery state information is obtained under all circumstances.

[0034] After determining the target ellipsoid, it is necessary to extract the key parameters required for the confidence interval from it. Specifically, this involves extracting the center vector and shape matrix of the target ellipsoid. The center vector contains the optimal estimates of each state variable, such as the estimates of the state of charge and the combined polarization voltage. The shape matrix is ​​a symmetric positive definite matrix that describes the distribution characteristics of state uncertainty in different directions. The diagonal elements of the shape matrix reflect the magnitude of the uncertainty of each state variable in its own direction; the larger the value of the diagonal element, the greater the uncertainty in that direction, and the wider the extension of the ellipsoid in that direction.

[0035] After extracting the center vector and shape matrix, the variance is calculated based on the diagonal elements of the shape matrix in the direction of the battery state variable. Variance is a statistic describing the degree of data dispersion, and in this embodiment, it is used to quantify the uncertainty of the battery state estimation. Specifically, for the state of charge variable, the diagonal elements of the first row and first column of the shape matrix are extracted as the variance value in the direction of the state of charge. This variance value directly reflects the degree of uncertainty in the state of charge estimation; the larger the variance value, the more uncertain the state of charge estimation.

[0036] After obtaining the variance, the standard deviation needs to be calculated based on the statistical distribution critical value corresponding to the preset confidence level. The confidence level represents the probability that the confidence interval contains the true state; commonly used confidence levels include 95% or 99%. The statistical distribution critical value is obtained from the chi-square distribution table based on the confidence level and the dimension of the state space. The chi-square distribution is an important distribution in probability theory, often used to describe the distribution characteristics of the sum of squares of a multidimensional normally distributed variable. In this embodiment, since the state space after dimensionality reduction is two-dimensional, the corresponding critical value is obtained by looking up the chi-square distribution table based on the two-dimensional degrees of freedom and the preset confidence level. Multiplying the variance by this critical value and then taking the square root yields the standard deviation of the battery state variable. The standard deviation is the square root of the variance; compared to the variance, it has the same dimensions as the original data, making it easier to understand and apply.

[0037] After calculating the standard deviation, the upper and lower bounds of the confidence interval are calculated based on the battery state estimate and the standard deviation in the center vector. Specifically, the value corresponding to the position of the state of charge variable is extracted from the center vector; this value is the optimal estimate of the state of charge. The lower bound is obtained by subtracting the standard deviation from this estimate, and the upper bound is obtained by adding the standard deviation to this estimate. To ensure that the confidence interval conforms to physical reality, physical constraints need to be checked on the upper and lower bounds. For the state of charge, its physical range is 0% to 100%. When the calculated lower bound is less than 0%, it is corrected to 0%; when the calculated upper bound is greater than 100%, it is corrected to 100%. The closed interval formed by the upper and lower bounds is the confidence interval for the battery state. This confidence interval can provide a reliable reference for battery management decisions, such as prompting users to charge when the lower bound is low, and prompting battery health checks when the confidence interval is wide.

[0038] Furthermore, the step of calculating the upper and lower bounds of the battery state based on the battery state estimate in the center vector and the standard deviation to obtain the confidence interval of the battery state includes: extracting the values ​​at the corresponding positions of the battery state variables from the center vector to obtain the battery state estimate; subtracting the standard deviation from the battery state estimate to obtain the lower bound; adding the standard deviation to the battery state estimate to obtain the upper bound; and forming a closed interval based on the lower bound and the upper bound to obtain the confidence interval of the battery state.

[0039] Specifically, when calculating the confidence interval of the battery state, the battery state estimate first needs to be extracted from the center vector. The center vector is a column vector, where each element corresponds to an estimate of a state variable. In this embodiment, the dimensionality-reduced center vector contains two elements: the first element corresponds to the estimate of the state of charge variable, and the second element corresponds to the estimate of the comprehensive polarization voltage variable. Since the state of charge is the most critical parameter in battery management, the first element of the center vector is extracted as the battery state estimate. For example, if the center vector is [65, 0.15], then the first element 65 is extracted as the state of charge estimate, indicating that the battery state of charge at the current moment is 65%. This estimate represents the most likely value of the battery state of charge at the current moment and is the optimal estimate obtained based on the physical model and measurement data.

[0040] After extracting the battery state estimate, the confidence interval boundaries need to be calculated based on the estimate and the standard deviation. The standard deviation reflects the degree of uncertainty in the battery state estimate; a larger standard deviation indicates lower confidence in the estimate and a wider confidence interval. Specifically, subtracting the standard deviation from the battery state estimate yields the initial lower bound, and adding the standard deviation to the battery state estimate yields the initial upper bound. For example, if the battery state estimate is 65% and the standard deviation is 3%, then the initial lower bound is 62% and the initial upper bound is 68%. This calculation process is based on the statistical properties of the normal distribution. Under the assumption of a normal distribution, the probability that the true value falls within the range of the estimate plus or minus one standard deviation is approximately 68%, and the probability that it falls within the range of plus or minus two standard deviations is approximately 95%. By combining the standard deviation with the statistical distribution critical value corresponding to the pre-set confidence level, the confidence interval boundaries that meet the specified confidence level requirements can be obtained.

[0041] It's important to note that the physical constraints of the battery state variables must be considered when calculating the initial lower and upper bounds. For the state of charge (SBC), the physical meaning is the percentage of the battery's current remaining charge relative to its total capacity; therefore, its value must range from 0% to 100%. When the calculated initial lower bound is less than 0%, it indicates that the lower bound derived from the statistical distribution exceeds the physically feasible range. In this case, the lower bound needs to be corrected to 0% to ensure that the lower bound of the confidence interval conforms to physical reality. For example, if the calculated initial lower bound is -5%, it should be corrected to 0%. Similarly, when the calculated initial upper bound is greater than 100%, it needs to be corrected to 100%. For example, if the battery state estimate is 98% and the standard deviation is 4%, the initial upper bound is 102%, which needs to be corrected to 100%. This physical constraint correction mechanism avoids generating unreasonable confidence intervals and improves the practicality of the estimation results.

[0042] After obtaining the lower and upper bounds of the constraints, a closed interval is formed using these two boundary values, which is the confidence interval for the battery state. A closed interval is an interval containing the boundary points, usually represented mathematically by square brackets. For example, [62%, 68%] indicates a confidence interval for the state of charge (SOC) of 62% to 68%. This confidence interval not only provides the optimal estimate of the battery SOC (65%) but also gives the range of possible states, allowing users to understand the reliability of the estimation results. In practical applications, the width of the confidence interval is an important indicator. The width of the confidence interval equals the upper bound minus the lower bound; for example, the width of the confidence interval mentioned above is 6%. A smaller width indicates a more accurate estimate, while a larger width indicates higher uncertainty. When the width of the confidence interval exceeds a preset threshold, such as 10%, it may indicate a decrease in the reliability of the battery state estimation, requiring further inspection or calibration.

[0043] Furthermore, confidence intervals can support a variety of practical applications. In remaining range prediction, a conservative estimate of the remaining range can be calculated based on a lower bound of 62%, ensuring that the vehicle can reach its destination even under the worst-case scenario. In charging decisions, the width of the confidence interval can be used to determine whether charging is necessary; if the lower bound is below the safety threshold by 20% or the confidence interval is wide, charging is recommended to ensure electrical safety. In battery health assessment, if the confidence interval width for the state of charge continuously increases from 3% to over 15%, it may indicate battery performance degradation or decreased sensor accuracy, requiring professional testing. This intuitive representation of confidence intervals transforms complex ellipsoidal ensemble estimation results into a form that is easy for users to understand and apply, significantly improving the intelligence level of battery management and the user experience.

[0044] In this embodiment, the state-space model of the target battery is reduced in dimensionality to obtain the reduced state equation and measurement equation. The center and shape matrices of the predicted ellipsoid are calculated based on the state equation, and the center and shape matrices of the estimated ellipsoid are calculated based on the measurement equation and measurement data, resulting in the predicted ellipsoid and the estimated ellipsoid, respectively. The intersection-union ratio (IUR) of the predicted and estimated ellipsoids is calculated and compared with a threshold to obtain the anomaly detection result. When the anomaly detection result is obtained, the battery state is calculated based on either the estimated or predicted ellipsoid to obtain the confidence interval for the battery state. This invention provides the confidence interval for the battery state through an ellipsoid set estimation method and performs data anomaly detection based on the IUR of the predicted and estimated ellipsoids, improving the reliability and security of remote battery monitoring.

[0045] Please see Figure 2 Another embodiment of the battery remote monitoring method in this application includes: 201. Dimensionally reduce the pre-defined state space model of the target battery to obtain the dimensionality-reduced state equation and measurement equation; In this embodiment, step 201 is similar to step 101 in the first embodiment, and will not be described again here.

[0046] 202. Calculate the predicted ellipsoid center at the current moment based on the estimated ellipsoid center at the previous moment and the state transition matrix in the reduced state equation. In this embodiment, the calculation of the predicted ellipsoid center is based on the state transition relation. The state transition matrix describes the evolution of the battery state over time and reflects the dynamic characteristics of the internal electrochemical processes of the battery. Specifically, the estimated ellipsoid center at the previous time step contains the optimal estimate of the battery state at that time step, for example, a state of charge of 70% and a combined polarization voltage of 0.12 volts at the previous time step. By performing a matrix multiplication operation between the estimated ellipsoid center at the previous time step and the state transition matrix, the state transition result based on the physical model can be obtained. The elements of the state transition matrix are determined by the equivalent circuit parameters of the battery, such as the time constant and internal resistance of the RC circuit. During the matrix multiplication operation, the row vector of the state transition matrix is ​​multiplied by the column vector of the estimated ellipsoid center to obtain the transitioned state components. In addition, the influence of external inputs, such as the battery's operating current, needs to be considered. The reduced-dimensional input matrix is ​​multiplied by the current operating current to obtain the state change caused by the input. The state transition result is added to the state change caused by the input to obtain the predicted ellipsoid center at the current time step. For example, if the state of charge was 70% at the previous moment, after state transition and current input, the predicted state of charge at the current moment may become 68%. The center of this predicted ellipsoid reflects the state evolution result derived solely from the physical model and does not include measurement information at the current moment.

[0047] 203. Calculate the predicted ellipsoid shape matrix based on the estimated ellipsoid shape matrix of the previous time step, the state transition matrix, the process noise covariance matrix, and the preset dilation factor. In this embodiment, calculating the center and initial shape matrix of the estimated ellipsoid based on the Kalman gain matrix, the measurement data, and the predicted ellipsoid includes: calculating the predicted measurement value based on the observation matrix in the dimensionality-reduced measurement equation and the center of the predicted ellipsoid; calculating the difference between the measurement data and the predicted measurement value to obtain a measurement residual vector, and performing matrix multiplication on the Kalman gain matrix and the measurement residual vector to obtain a state correction; adding the center of the predicted ellipsoid to the state correction to obtain the center of the estimated ellipsoid, and constructing a difference matrix by subtracting the product of the Kalman gain matrix and the observation matrix from the identity matrix; and performing matrix multiplication on the difference matrix and the shape matrix of the predicted ellipsoid to obtain the initial shape matrix of the estimated ellipsoid.

[0048] Specifically, calculating the predicted ellipsoid shape matrix requires considering not only the propagation of uncertainty during state transitions but also the impact of process noise and conservative estimation strategies. Specifically, the estimated ellipsoid shape matrix from the previous time step describes the uncertainty distribution of the previous state estimate. During state transitions, this uncertainty propagates and changes as the state evolves. The propagation process of uncertainty is calculated using matrix similarity transformation techniques. This involves multiplying the state transition matrix with the estimated ellipsoid shape matrix from the previous time step, and then multiplying it again with the transpose of the state transition matrix to obtain the shape matrix change caused by the state transition. Matrix similarity transformation is an important concept in linear algebra, used to describe the influence of linear transformations on matrix characteristics; in this embodiment, it is used to characterize the impact of state transitions on the uncertainty distribution.

[0049] In addition to the uncertainties caused by state transitions, the impact of process noise must also be considered. Process noise refers to random disturbances present during the battery's state evolution, which may originate from the randomness of internal electrochemical reactions, temperature fluctuations, modeling errors, and other factors. The process noise covariance matrix describes the statistical characteristics of process noise, with its diagonal elements reflecting the intensity of process noise in each state variable direction. Adding the shape matrix caused by state transitions to the process noise covariance matrix yields the initial shape matrix considering the influence of process noise. This initial shape matrix comprehensively reflects the impact of both state transitions and process noise on prediction uncertainty.

[0050] To ensure that the predicted ellipsoid reliably contains the true state, this embodiment employs a conservative estimation strategy to expand the initial shape matrix. Specifically, each element of the initial shape matrix is ​​multiplied by a preset dilation factor to obtain the final predicted ellipsoid shape matrix. The dilation factor is a value greater than 1, such as 1.2 or 1.5, which moderately expands the range of the predicted ellipsoid, increasing the probability that the ellipsoid contains the true state. The selection of the dilation factor needs to comprehensively consider the estimation accuracy and inclusion requirements. A larger dilation factor results in a larger range of the predicted ellipsoid and a higher probability of containing the true state, but the estimation accuracy will decrease accordingly; a smaller dilation factor results in higher estimation accuracy, but may not guarantee that the ellipsoid always contains the true state. In one embodiment, the dilation factor can be dynamically adjusted according to the uncertainty of the process noise. A larger dilation factor is used when the process noise is high, and a smaller dilation factor is used when the process noise is low. Through this conservative expansion strategy, the obtained predicted ellipsoid shape matrix can provide the most accurate uncertainty description possible while ensuring inclusion.

[0051] It's important to note that the shape matrix, as a symmetric positive definite matrix, determines the geometric characteristics of the ellipsoid through its mathematical properties. The eigenvalues ​​of the shape matrix determine the radii of the ellipsoid along each principal axis, while the eigenvectors determine the orientation of these axes. After dilation, the eigenvalues ​​of the shape matrix increase proportionally to the dilation factor, meaning the ellipsoid's radii in each direction expand accordingly. For example, if the original shape matrix has eigenvalues ​​of 2 and 5 and a dilation factor of 1.2, the dilated eigenvalues ​​become 2.4 and 6, corresponding to ellipsoid radii increasing from approximately 1.41 and 2.24 to approximately 1.55 and 2.45. This geometric expansion ensures that the predicted ellipsoid more reliably covers regions where the actual state might exist, providing a solid foundation for subsequent anomaly detection and confidence interval calculation.

[0052] 204. Calculate the Kalman gain matrix based on the dimension-reduced measurement equation, the predicted ellipsoid, and the measurement noise covariance matrix; In this embodiment, the Kalman gain matrix is ​​a key parameter for fusing measurement information, balancing the weights of physical model predictions and actual measurement data. The calculation of the Kalman gain matrix is ​​based on the minimum variance estimation criterion, determining the optimal fusion weights by optimizing the variance of the state estimates. Specifically, firstly, the observation matrix is ​​extracted based on the dimensionality-reduced measurement equation. This observation matrix describes the linear relationship between state variables and measurement outputs. Then, the shape matrix of the predicted ellipsoid is multiplied by the transpose of the observation matrix to obtain an intermediate calculation result. This intermediate result reflects the projection of prediction uncertainty onto the measurement space. Next, the observation matrix is ​​multiplied by the intermediate result, and the measurement noise covariance matrix is ​​added to obtain the innovation covariance matrix. The innovation covariance matrix comprehensively reflects the overall impact of prediction uncertainty and measurement uncertainty on the measurement space. The measurement noise covariance matrix describes the characteristics of random errors present in the measurement process, with its diagonal elements reflecting the noise intensity of each measurement quantity. Finally, the intermediate result obtained in the first step is multiplied by the inverse of the innovation covariance matrix to obtain the Kalman gain matrix. Inverse matrix operations are important in linear algebra, and matrix inversion can be used to solve systems of linear equations. The size of the elements of the Kalman gain matrix reflects the reliability of the measurement information. When the prediction uncertainty is large or the measurement noise is small, the Kalman gain is large, indicating that it relies more on the measurement information; conversely, it relies more on the model prediction.

[0053] 205. Based on the Kalman gain matrix, the measurement data, and the predicted ellipsoid, calculate the center and initial shape matrix of the estimated ellipsoid; In this embodiment, the predicted measurement value is calculated using the observation matrix and the predicted ellipsoid center. The difference between the actual measurement data and the predicted measurement value is used to obtain the measurement residual vector. This residual vector is multiplied by the Kalman gain matrix to obtain the state correction. The predicted ellipsoid center is added to the state correction to obtain the estimated ellipsoid center. When updating the shape matrix, a difference matrix is ​​constructed by subtracting the product of the Kalman gain matrix and the observation matrix from the identity matrix. This difference matrix is ​​multiplied by the predicted ellipsoid shape matrix to obtain the estimated initial shape matrix of the ellipsoid. It should be noted that this calculation process follows the update mechanism of standard Kalman filtering, but within the ellipsoid ensemble estimation framework, not only is the estimated state center value updated, but the shape matrix describing the uncertainty distribution is also updated simultaneously. The shape matrix update process reflects the constraint effect of measurement information on uncertainty; the more reliable the measurement information, the more significant the shape matrix shrinkage. Furthermore, in actual calculations, it is necessary to ensure the symmetry and positive definiteness of the shape matrix. Symmetry ensures the geometric symmetry of the ellipsoid, and positive definiteness ensures the actual existence of the ellipsoid.

[0054] 206. Multiply the initial shape matrix of the estimated ellipsoid by a shrinkage factor to obtain the shape matrix of the estimated ellipsoid; In this embodiment, shrinkage processing is a crucial step in ellipsoid estimation optimization, aiming to further improve estimation accuracy while ensuring the ellipsoid contains the true state. Specifically, each element of the initial shape matrix of the estimated ellipsoid is multiplied by a shrinkage factor to obtain the final estimated ellipsoid shape matrix. The shrinkage factor is a positive number less than 1, typically ranging from 0.8 to 0.95. Determining the shrinkage factor requires comprehensive consideration of the statistical characteristics of measurement noise and the accuracy requirements of the practical application. In one embodiment, the shrinkage factor can be dynamically determined based on the trace of the measurement noise covariance matrix. The trace is the sum of the diagonal elements of the matrix, reflecting the magnitude of overall uncertainty. When the trace of the measurement noise covariance matrix is ​​small, it indicates that the measurement is relatively reliable, and a smaller shrinkage factor, such as 0.85, can be used for further shrinkage; when the trace is large, a larger shrinkage factor, such as 0.95, should be used to avoid over-shrinkage. Through shrinkage processing, the range of the estimated ellipsoid is moderately reduced, which means that the confidence interval width of the state estimation is reduced, providing a more accurate state estimation result. It is important to note that the shrinkage factor cannot be set too small, otherwise the ellipsoid may fail to contain the true state, violating the basic requirements of ensemble estimation methods. In practical applications, appropriate shrinkage factor values ​​can be calibrated through offline simulation experiments or historical data statistical analysis to ensure reliable estimation performance under various operating conditions.

[0055] 207. Calculate the intersection-union ratio (IUR) of the predicted ellipsoid and the estimated ellipsoid, and compare the IUR with a threshold to obtain the anomaly detection result; 208. When the anomaly detection result does not indicate that an anomaly has been detected, the battery state is calculated in intervals based on the estimated ellipsoid. When the anomaly detection result indicates that an anomaly has been detected, the predicted ellipsoid is used to calculate the intervals to obtain the confidence interval of the battery state.

[0056] In this embodiment, steps 207-208 are similar to steps 103-104 in the first embodiment, and will not be described again here.

[0057] In this embodiment, the state-space model of the target battery is reduced in dimensionality to obtain the reduced state equation and measurement equation. The center and shape matrices of the predicted ellipsoid are calculated based on the state equation, and the center and shape matrices of the estimated ellipsoid are calculated based on the measurement equation and measurement data, resulting in the predicted ellipsoid and the estimated ellipsoid, respectively. The intersection-union ratio (IUR) of the predicted and estimated ellipsoids is calculated and compared with a threshold to obtain the anomaly detection result. When the anomaly detection result is obtained, the battery state is calculated based on either the estimated or predicted ellipsoid to obtain the confidence interval for the battery state. This invention provides the confidence interval for the battery state through an ellipsoid set estimation method and performs data anomaly detection based on the IUR of the predicted and estimated ellipsoids, improving the reliability and security of remote battery monitoring.

[0058] The battery remote monitoring method in the embodiments of the present invention has been described above. The battery remote monitoring device in the embodiments of the present invention will be described below. Please refer to [link to relevant documentation] for details on this battery remote monitoring device. Figure 3 One embodiment of the battery remote monitoring device in this invention includes: The dimension reduction processing module 301 is used to perform dimension reduction processing on the preset state space model of the target battery to obtain the dimension-reduced state equation and measurement equation. Ellipsoid estimation module 302 is used to calculate the center and shape matrix of the predicted ellipsoid based on the reduced state equation, and to calculate the center and shape matrix of the estimated ellipsoid based on the reduced measurement equation and measurement data, so as to obtain the predicted ellipsoid and the estimated ellipsoid respectively. Anomaly detection module 303 is used to calculate the intersection-union ratio (IUR) between the predicted ellipsoid and the estimated ellipsoid, and compare the IUR with a threshold to obtain anomaly detection results; The interval calculation module 304 is used to perform interval calculation on the battery state based on the estimated ellipsoid when the anomaly detection result does not indicate that an anomaly has been detected, and to perform interval calculation using the predicted ellipsoid when the anomaly detection result indicates that an anomaly has been detected, so as to obtain the confidence interval of the battery state.

[0059] In this embodiment of the invention, the battery remote monitoring device operates the aforementioned battery remote monitoring method. The device performs dimensionality reduction processing on a preset state-space model of the target battery to obtain a dimensionality-reduced state equation and measurement equation. Based on the state equation, it calculates the center and shape matrix of a predicted ellipsoid, and based on the measurement equation and measurement data, it calculates the center and shape matrix of an estimated ellipsoid, obtaining a predicted ellipsoid and an estimated ellipsoid, respectively. It calculates the intersection-union ratio (IUR) of the predicted and estimated ellipsoids and compares it with a threshold to obtain an anomaly detection result. When the anomaly detection result is obtained, it performs interval calculations on the battery state based on either the estimated or predicted ellipsoid to obtain a confidence interval for the battery state. This invention provides a confidence interval for the battery state through an ellipsoid set estimation method and performs data anomaly detection based on the IUR of the predicted and estimated ellipsoids, improving the reliability and security of battery remote monitoring.

[0060] above Figure 3 The battery remote monitoring device in this embodiment of the invention will be described in detail from the perspective of unitized functional entities. The battery remote monitoring device in this embodiment of the invention will be described in detail from the perspective of hardware processing.

[0061] Figure 4 This is a schematic diagram of the structure of a battery remote monitoring device 300 provided in an embodiment of the present invention. The battery remote monitoring device 300 can vary significantly due to different configurations or performance characteristics. It may include one or more central processing units (CPUs) 410 (e.g., one or more processors) and a memory 420, and one or more storage media 430 (e.g., one or more mass storage devices) for storing application programs 333 or data 432. The memory 420 and storage media 430 can be temporary or persistent storage. The program stored in the storage media 430 may include one or more units (not shown in the diagram), each unit may include a series of instruction operations on the battery remote monitoring device 400. Furthermore, the processor 410 may be configured to communicate with the storage media 430 and execute a series of instruction operations in the storage media 430 on the battery remote monitoring device 400 to implement the steps of the aforementioned battery remote monitoring method.

[0062] The battery remote monitoring device 400 may also include one or more power supplies 440, one or more wired or wireless network interfaces 450, one or more input / output interfaces 460, and / or one or more operating systems 431, such as Windows Server, Mac OS X, Unix, Linux, FreeBSD, etc. Those skilled in the art will understand that... Figure 4The battery remote monitoring device structure shown does not constitute a limitation on the battery remote monitoring device provided by the present invention. It may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0063] The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium, wherein the computer-readable storage medium stores instructions that, when executed on a computer, cause the computer to perform the steps of the battery remote monitoring method.

[0064] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the system, device, or unit described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0065] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0066] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail 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 of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for remote monitoring of a battery, characterized in that, The battery remote monitoring method includes: The pre-defined state-space model of the target battery is reduced in dimension to obtain the reduced state equation and measurement equation. The center and shape matrices of the predicted ellipsoid are calculated based on the reduced state equation, and the center and shape matrices of the estimated ellipsoid are calculated based on the reduced measurement equation and measurement data, thus obtaining the predicted ellipsoid and the estimated ellipsoid, respectively. The intersection-union ratio (IUU) of the predicted ellipsoid and the estimated ellipsoid is calculated, and the IUU is compared with a threshold to obtain the anomaly detection result. When the anomaly detection result does not indicate that an anomaly has been detected, the battery state is calculated in intervals based on the estimated ellipsoid. When the anomaly detection result indicates that an anomaly has been detected, the predicted ellipsoid is used to calculate in intervals to obtain the confidence interval of the battery state.

2. The battery remote monitoring method according to claim 1, characterized in that, The step of performing dimensionality reduction processing on the preset state-space model of the target battery to obtain the dimensionality-reduced state equation and measurement equation includes: Obtain the original state variables in the preset state space model of the target battery, the original state variables including the state of charge variables and multiple polarization voltage variables; The multiple polarization voltage variables are combined by weighted summation to obtain a comprehensive polarization voltage variable; A dimension-reduced projection matrix is ​​constructed based on the charged state variables and the integrated polarization voltage variables; The state transition matrix, input matrix, and observation matrix of the state space model are transformed by the dimensionality-reduced projection matrix to obtain the dimensionality-reduced state equation and measurement equation.

3. The battery remote monitoring method according to claim 1, characterized in that, The step of calculating the center and shape matrix of the predicted ellipsoid based on the reduced state equation, and calculating the center and shape matrix of the estimated ellipsoid based on the reduced measurement equation and measurement data, to obtain the predicted ellipsoid and the estimated ellipsoid respectively, includes: The predicted ellipsoid center at the current moment is calculated based on the estimated ellipsoid center at the previous moment and the state transition matrix in the reduced state equation. The predicted ellipsoid shape matrix is ​​calculated based on the estimated ellipsoid shape matrix from the previous time step, the state transition matrix, the process noise covariance matrix, and the preset dilation factor. Calculate the Kalman gain matrix based on the dimension-reduced measurement equation, the predicted ellipsoid, and the measurement noise covariance matrix. Based on the Kalman gain matrix, the measurement data, and the predicted ellipsoid, calculate the center and initial shape matrix of the estimated ellipsoid; The initial shape matrix of the estimated ellipsoid is multiplied by a shrinkage factor to obtain the shape matrix of the estimated ellipsoid.

4. The battery remote monitoring method according to claim 3, characterized in that, The step of calculating the center and initial shape matrix of the estimated ellipsoid based on the Kalman gain matrix, the measurement data, and the predicted ellipsoid includes: The predicted measurement value is calculated based on the observation matrix in the reduced-dimensional measurement equation and the center of the predicted ellipsoid; The difference between the measured data and the predicted measured value is calculated to obtain the measurement residual vector, and the Kalman gain matrix and the measurement residual vector are multiplied to obtain the state correction amount; The center of the predicted ellipsoid is added to the state correction to obtain the center of the estimated ellipsoid, and the difference matrix is ​​constructed by subtracting the product of the Kalman gain matrix and the observation matrix from the identity matrix. The difference matrix is ​​multiplied by the shape matrix of the predicted ellipsoid to obtain the initial shape matrix of the estimated ellipsoid.

5. The battery remote monitoring method according to claim 1, characterized in that, The calculation of the intersection-union ratio (IUR) between the predicted ellipsoid and the estimated ellipsoid, and the comparison of the IUR with a threshold to obtain the anomaly detection result, includes: Based on the center and shape matrix of the predicted ellipsoid, a first set of sampling points is generated using a random sampling method, and based on the center and shape matrix of the estimated ellipsoid, a second set of sampling points is generated using a random sampling method. For each sampling point in the first sampling point set, the Mahalanobis distance of the sampling point is calculated based on the center and shape matrix of the estimated ellipsoid. The number of sampling points whose Mahalanobis distance satisfies the ellipsoid constraint is counted to obtain the number of intersection samples. The intersection-union ratio of the predicted ellipsoid and the estimated ellipsoid is calculated based on the total number of the first sampling point set, the total number of the second sampling point set, and the number of intersection samples. The crossover-union ratio is compared with a threshold. When the crossover-union ratio is less than the threshold, an anomaly is detected. When the crossover-union ratio is greater than or equal to the threshold, no anomaly is detected, and an anomaly detection result is obtained.

6. The battery remote monitoring method according to claim 1, characterized in that, When the anomaly detection result does not indicate that an anomaly has been detected, the battery state is calculated in intervals based on the estimated ellipsoid; when the anomaly detection result indicates that an anomaly has been detected, the predicted ellipsoid is used to calculate the intervals, resulting in confidence intervals for the battery state including: Based on the anomaly detection results, it is determined whether an anomaly has been detected. If no anomaly is detected, the estimated ellipsoid is selected as the target ellipsoid. If an anomaly is detected, the predicted ellipsoid is selected as the target ellipsoid. Extract the center vector and shape matrix of the target ellipsoid, and calculate the variance of the battery state variable direction based on the diagonal elements of the shape matrix in the direction of the battery state variable. Calculate the standard deviation of the battery state variables based on the statistical distribution critical value corresponding to the preset reliability and the variance value; Based on the battery state estimate in the center vector and the standard deviation, the upper and lower bounds of the battery state are calculated to obtain the confidence interval of the battery state.

7. The battery remote monitoring method according to claim 6, characterized in that, The step of calculating the upper and lower bounds of the battery state based on the battery state estimate in the center vector and the standard deviation, to obtain the confidence interval of the battery state, includes: The battery state estimate is obtained by extracting the values ​​of the corresponding positions of the battery state variables from the central vector. Subtracting the standard deviation from the battery state estimate yields the lower bound, and adding the standard deviation to the battery state estimate yields the upper bound. The confidence interval for the battery state is obtained by constructing a closed interval based on the lower bound value and the upper bound value.

8. A battery remote monitoring device, characterized in that, The battery remote monitoring device includes: The dimension reduction module is used to reduce the dimension of the preset state space model of the target battery to obtain the reduced state equation and measurement equation. The ellipsoid estimation module is used to calculate the center and shape matrix of the predicted ellipsoid based on the reduced state equation, and to calculate the center and shape matrix of the estimated ellipsoid based on the reduced measurement equation and measurement data, so as to obtain the predicted ellipsoid and the estimated ellipsoid respectively. An anomaly detection module is used to calculate the intersection-union ratio (IUR) between the predicted ellipsoid and the estimated ellipsoid, and compare the IUR with a threshold to obtain an anomaly detection result. The interval calculation module is used to perform interval calculation on the battery state based on the estimated ellipsoid when the anomaly detection result does not indicate that an anomaly has been detected, and to perform interval calculation using the predicted ellipsoid when the anomaly detection result indicates that an anomaly has been detected, so as to obtain the confidence interval of the battery state.

9. A battery remote monitoring device, characterized in that, The battery remote monitoring device includes: a memory and at least one processor, wherein the memory stores instructions; The at least one processor invokes the instructions in the memory to cause the battery remote monitoring device to perform the steps of the battery remote monitoring method as described in any one of claims 1-7.

10. A computer-readable storage medium storing instructions thereon, characterized in that, When the instruction is executed by the processor, it implements the steps of the battery remote monitoring method as described in any one of claims 1-7.

Images