BMS background data diagnosis method and system based on MRPCA and storage medium

Abstract

The invention relates to the technical field of battery management systems, and discloses a BMS background data diagnosis method and system based on MRPCA and a storage medium, and the method comprises the following steps: 1, obtaining the interval training data of a BMS system; 2, decomposing the interval training data into a midpoint matrix and a radius matrix; step 3, establishing an MRPCA model based on the uncertain covariance matrix, and calculating an SPE statistical magnitude control limit and a T2 statistical magnitude control limit; step 4, acquiring test data, and calculating SPE statistics and T2 statistics of the test data by using the MRPCA model; according to the method, the MRPCA model is introduced, information of interval data is fully utilized, the center trend of variables is considered, uncertain fluctuation of the variables is also considered, and therefore adaptability to a complex nonlinear structure is improved; compared with a traditional PCA method, the MRPCA can more accurately capture potential abnormal distribution in the data, and is more stable in performance especially in the environment of a battery management system which is large in dynamic change.

Description

Technical Field

[0001] This invention relates to the field of battery management system technology, specifically to a BMS background data diagnostic method, system, and storage medium based on MRPCA. Background Technology

[0002] Battery Management Systems (BMS) play a crucial role in electric vehicles and energy storage systems. With the increasing prevalence of new energy vehicles, massive amounts of BMS operational data are uploaded to the cloud in real time for health status analysis, anomaly detection, and lifespan assessment. Currently, anomaly detection for BMS backend data primarily employs the following methods: 1. Threshold-based rule method: This method determines whether a parameter exceeds a certain threshold by manually setting thresholds for various parameters (such as voltage, current, and temperature). While simple to implement and easy to deploy, this method is ineffective for handling multi-dimensional data anomalies and cannot cope with dynamic factors such as equipment aging or environmental changes.

[0003] 2. Statistical Methods and Principal Component Analysis (PCA): PCA is used to reduce the dimensionality of high-dimensional data, and anomaly detection is performed based on the principal component reconstruction error. This method is widely used in industry and is suitable for processing large-scale data, but its linear assumption limits its performance in complex nonlinear distributions.

[0004] 3. Kernel Principal Component Analysis (KPCA): This method introduces a kernel function to map the data to a high-dimensional feature space before performing PCA, thereby enhancing its ability to handle nonlinear structures. While this method performs well on some datasets, it is prone to misclassification when multiple outlier patterns or uneven class distributions exist due to a lack of geometric modeling capabilities for the data structure.

[0005] 4. Machine Learning and Deep Learning Methods: In recent years, some researchers have attempted to use methods such as autoencoders, support vector machines (SVM), random forests, and LSTM to model and predict BMS data. While these methods offer high performance, they place higher demands on large amounts of labeled data, computational resources, and model interpretability, and their deployment is also more challenging, making them unsuitable for most traditional vehicle and maintenance platforms.

[0006] Although the above methods have improved the anomaly detection capabilities of BMS data to some extent, they still have the following shortcomings: 1. Poor nonlinear adaptability: Traditional PCA and its improved methods are unable to fully exploit the potential abnormal distributions in complex nonlinear structures, especially exhibiting instability under different operating conditions.

[0007] 2. Insufficient intra-class structure modeling: Although methods such as KPCA introduce kernel mapping, they usually only focus on the global distribution and lack description of the geometric boundaries of normal data groups, making it impossible to accurately distinguish between "reasonable deviation" and "true anomalies".

[0008] 3. Sensitive to outliers: In the context of unstructured modeling, noisy data such as outliers and measurement errors are easily misjudged as anomalies, leading to an increased false alarm rate.

[0009] 4. Weak ability to identify multiple anomaly patterns: Existing methods usually classify all anomalies as "abnormal" and lack further classification of anomaly types, which is not conducive to refined fault tracing and operation and maintenance decisions.

[0010] 5. Insufficient model generalization ability: Deep models are highly dependent on training data, and their performance is prone to degrade when generalized to other vehicle models, battery types or operating conditions, resulting in high maintenance costs.

[0011] Therefore, this application proposes a BMS back-end data diagnostic method, system, and storage medium based on MRPCA. Summary of the Invention

[0012] The purpose of this invention is to provide a BMS backend data diagnostic method, system, and storage medium based on MRPCA to solve the problems mentioned in the background art.

[0013] Firstly, a BMS backend data diagnostic method based on MRPCA is provided, including the following steps: Step 1: Obtain the interval training data of the BMS system; Step 2: Decompose the interval training data into a midpoint matrix. and radius matrix And based on the midpoint matrix and radius matrix Construct an uncertain covariance matrix; Step 3: Establish an MRPCA model based on the aforementioned uncertain covariance matrix, and calculate the control limits of the SPE statistic. and T2 statistic control limits ; Step 4: Obtain test data and use the MRPCA model to calculate the SPE statistic and T2 statistic of the test data; Step 5: Based on the fact that the SPE statistic and T2 statistic exceed the corresponding control limits, determine whether a fault exists; Step 6: If a fault exists, calculate the proportion of each variable in the SPE and T2 contribution maps respectively, merge them to obtain the comprehensive contribution proportion, and determine the fault variable based on the comprehensive contribution proportion.

[0014] Furthermore, the interval training data forms the original sample matrix X, represented as: ; In the formula, n is the number of samples, i.e. how many time points of data were collected; m is the number of process variables, i.e. the number of BMS parameters monitored. Represents m distinct process variables; This indicates that the matrix is ​​an n-row, m-column real matrix; Each variable It can be represented as an interval: ; The midpoint value is: ; The radius value is: ; In the formula, This represents the interval measurement value of the i-th process variable at the k-th sampling time. This indicates the lower limit of the range of values; This indicates the upper limit of the range; This represents the midpoint of the interval measurement of the i-th variable at time k; This represents the radius of the interval measurement of the i-th variable at time k; Let be the midpoint matrix, consisting of all composition; Let be the radius matrix, consisting of all composition.

[0015] Furthermore, each variable in the interval training data is represented as follows: ; In the formula, This represents the interval measurement value of the i-th process variable at the k-th sampling time. This represents the midpoint of the interval measurement of the i-th variable at time k; Let represent the radius of the interval measurement value of the i-th variable at time k.

[0016] Furthermore, the formula for constructing the uncertain covariance matrix is ​​as follows: ; In the formula, R represents the uncertain covariance matrix, which is the basis of the MRPCA model; n is the number of samples. It is a standardization factor used to calculate the average value; Midpoint matrix transpose; It is the standard covariance matrix of the midpoint matrix, reflecting the central tendency correlation between variables; Radius matrix transpose; It is the covariance matrix of the radius matrix, reflecting the fluctuation correlation of the uncertainty of the variables; and The absolute value of the interaction term between the midpoint and the radius is used to quantify the cross-effect of uncertainty and avoid interference from negative values; and These are two standard variance-covariance matrices, which respectively represent the partial contributions to the total variance-covariance matrix given by the midpoint and radius.

[0017] Furthermore, the formula for calculating the overall contribution ratio is as follows: ; In the formula, The overall contribution of the i-th variable is the sum of the contribution percentages of SPE and T2, and is used for fault location. Let represent the percentage contribution of the i-th variable to the SPE statistic. This represents the percentage contribution of the i-th variable to the T2 statistic.

[0018] Furthermore, the interval training data includes battery external voltage, battery internal voltage, charging and discharging current, temperature, SOC, and internal resistance.

[0019] Furthermore, the battery's external voltage, internal voltage, charging / discharging current, temperature, state of charge (SOC), and internal resistance constitute six variables, and the contribution value of each variable is denoted as follows: , And calculate the proportion of each variable using the following formula; ; ; in, Let represent the percentage contribution of the i-th variable to the SPE statistic. This represents the percentage contribution of the i-th variable to the T2 statistic. This represents the raw contribution value of the i-th variable to the SPE statistic of the current fault sample; This represents the raw contribution value of the i-th variable to the T2 statistic of the current fault sample; and These represent the sum of the original SPE contribution values ​​of all 6 variable stars and the sum of the original T2 contribution values, i.e., the total contribution value; variable indices i=1 to 6, corresponding to the 6 key variables of the BMS system: external battery voltage, internal battery voltage, charging and discharging current, temperature, SOC, and internal resistance.

[0020] Secondly, a BMS backend data diagnostic system based on MRPCA is provided for performing the aforementioned MRPCA-based BMS backend data diagnostic method, including: The data acquisition module is used to acquire interval training data and test data of the BMS system; The model building module is used to decompose the interval training data into a midpoint matrix and a radius matrix, construct an uncertain covariance matrix, and build an MRPCA model. The control limit calculation module is used to calculate the control limits for the SPE statistic and the T2 statistic; The fault detection module is used to determine whether a fault exists based on the SPE and T2 statistics of the test data. The fault location module is used to calculate the overall contribution ratio of each variable to determine the fault variable when a fault exists.

[0021] Furthermore, the fault location module is specifically used for: Calculate the percentage of each variable in the SPE contribution plot. ; Calculate the proportion of each variable in the T2 contribution plot. ; pass Calculate the overall contribution ratio; The source of the fault is determined based on the variable with the highest overall contribution.

[0022] Thirdly, a computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a processor, implements the aforementioned MRPCA-based BMS background data diagnostic method.

[0023] Compared with the prior art, the beneficial effects of the present invention are: This invention, by introducing the MRPCA model, makes full use of the information in interval data, taking into account not only the central trend of variables but also their uncertain fluctuations, thereby improving its adaptability to complex nonlinear structures. Compared with the traditional PCA method, MRPCA can more accurately capture potential abnormal distributions in the data, especially in dynamic environments such as battery management systems, where it performs more stably.

[0024] Furthermore, by constructing an uncertain covariance matrix, this invention effectively quantifies the interaction between the midpoint and the radius, avoids negative value interference, and further improves the robustness of the model. In terms of fault detection and localization, this invention calculates the proportion of each variable in the SPE and T2 contribution maps and merges them to obtain the comprehensive contribution proportion, which can accurately determine the fault variables and provide strong support for refined fault tracing and operation and maintenance decisions. Attached Figure Description

[0025] Figure 1 This is a schematic diagram of the BMS backend data diagnostic processing flow of the present invention. Detailed Implementation

[0026] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Example 1:

[0027] Please see Figure 1 A BMS backend data diagnostic method based on MRPCA includes the following steps: Step 1: Obtain the interval training data of the BMS system; Step 2: Decompose the interval training data into a midpoint matrix. and radius matrix And based on the midpoint matrix and radius matrix Construct an uncertain covariance matrix; Step 3: Establish an MRPCA model based on the aforementioned uncertain covariance matrix, and calculate the control limits of the SPE statistic. and T2 statistic control limits ; Step 4: Obtain test data and use the MRPCA model to calculate the SPE statistic and T2 statistic of the test data; Step 5: Based on the fact that the SPE statistic and T2 statistic exceed the corresponding control limits, determine whether a fault exists; Step 6: If a fault exists, calculate the proportion of each variable in the SPE and T2 contribution maps respectively, merge them to obtain the comprehensive contribution proportion, and determine the fault variable based on the comprehensive contribution proportion.

[0028] Specifically, in step one, acquiring interval training data of the BMS system is the foundation of the entire diagnostic method. This data covers key parameters such as external battery voltage, internal battery voltage, charge / discharge current, temperature, SOC (state of charge), and internal resistance. Each parameter is represented in interval form, including midpoint and radius values, thus providing a more comprehensive reflection of the actual operating state of the battery system. In step two, decomposing the interval training data into a midpoint matrix and a radius matrix is ​​a crucial step in constructing the MRPCA model. The midpoint matrix reflects the central trend of each variable, while the radius matrix reflects the fluctuations in uncertainty of each variable. By constructing an uncertain covariance matrix, the interaction between the midpoint and radius is taken into account, effectively avoiding negative value interference and improving the robustness of the model. In step three, an MRPCA model is established based on the uncertain covariance matrix, and the SPE and T2 statistic control limits are calculated. The MRPCA model can more accurately capture potential abnormal distributions in the data, while the SPE and T2 statistic control limits are used for subsequent fault detection. By setting reasonable control limits, it is possible to effectively determine whether there are anomalies in the test data. In step four, test data is acquired, and the SPE and T2 statistics of the test data are calculated using the established MRPCA model. This step applies the model to actual data and evaluates the operating status of the battery system by calculating statistics. In step five, based on whether the SPE and T2 statistics exceed their corresponding control limits, it is determined whether a fault exists. If any statistic exceeds its control limit, it indicates that the battery system may have a fault, requiring further fault location. In step six, if a fault exists, the proportion of each variable in the SPE and T2 contribution maps is calculated separately and then merged to obtain the comprehensive contribution proportion. Through the comprehensive contribution proportion, the fault variables can be accurately identified, providing strong support for refined fault tracing and operation and maintenance decisions. This step is the core of the entire diagnostic method and can effectively improve the accuracy and efficiency of fault location.

[0029] Furthermore, the interval training data forms the original sample matrix X, represented as: ; In the formula, n is the number of samples, i.e. how many time points of data were collected; m is the number of process variables, i.e. the number of BMS parameters monitored. Represents m distinct process variables; This indicates that the matrix is ​​an n-row, m-column real matrix; Each variable It can be represented as an interval: ; The midpoint value is: ; The radius value is: ; In the formula, This represents the interval measurement value of the i-th process variable at the k-th sampling time. This indicates the lower limit of the range of values; This indicates the upper limit of the range; This represents the midpoint of the interval measurement of the i-th variable at time k; This represents the radius of the interval measurement of the i-th variable at time k; Let be the midpoint matrix, consisting of all composition; Let be the radius matrix, consisting of all composition.

[0030] Specifically, interval training data can be acquired in real time using sensors built into the BMS system. These sensors can accurately measure key parameters such as external battery voltage, internal battery voltage, charging and discharging current, temperature, SOC, and internal resistance, and output these parameters in interval form. The interval data includes midpoint and radius values, providing rich information for subsequent data processing. After acquiring the interval training data, it needs to be organized into the form of an original sample matrix X to facilitate subsequent data decomposition and model construction. Each row of the original sample matrix X represents data at a time point, and each column represents a process variable. This structure makes the data clearer and easier to process. When constructing the uncertain covariance matrix, the information from the midpoint and radius matrices needs to be fully utilized. The midpoint matrix reflects the central trend of each variable and is one of the main characteristics of the data; while the radius matrix reflects the fluctuations in the uncertainty of each variable and is an important supplementary information to the data. By constructing the uncertain covariance matrix, the interaction between the midpoint and the radius can be taken into account, thereby more accurately describing the distribution characteristics of the data.

[0031] When establishing an MRPCA model, calculations based on an uncertain covariance matrix are required. The MRPCA model is an improved principal component analysis method that can more accurately capture potential anomalies in the data, making it particularly suitable for dynamically changing environments like battery management systems. Calculating the control limits for the SPE and T2 statistics provides a valid basis for subsequent fault detection. After acquiring test data, the established MRPCA model is used to calculate the SPE and T2 statistics. This step is crucial for applying the model to real-world data, using the calculated statistics to assess the battery system's operating status. If the SPE or T2 statistics exceed their corresponding control limits, it indicates a potential fault in the battery system, requiring further fault localization. During the fault localization phase, the proportion of each variable in the SPE and T2 contribution maps needs to be calculated separately and then merged to obtain the overall contribution proportion.

[0032] Furthermore, each variable in the interval training data is represented as: ; In the formula, This represents the interval measurement value of the i-th process variable at the k-th sampling time. This represents the midpoint of the interval measurement of the i-th variable at time k; Let represent the radius of the interval measurement value of the i-th variable at time k.

[0033] Furthermore, the formula for constructing the uncertain covariance matrix is: ; In the formula, R represents the uncertain covariance matrix, which is the basis of the MRPCA model; n is the number of samples. It is a standardization factor used to calculate the average value; Midpoint matrix transpose; It is the standard covariance matrix of the midpoint matrix, reflecting the central tendency correlation between variables; Radius matrix transpose; It is the covariance matrix of the radius matrix, reflecting the fluctuation correlation of the uncertainty of the variables; and The absolute value of the interaction term between the midpoint and the radius is used to quantify the cross-effect of uncertainty and avoid interference from negative values; and These are two standard variance-covariance matrices, which respectively represent the partial contributions to the total variance-covariance matrix given by the midpoint and radius.

[0034] Specifically, the transpose of the midpoint and radius matrices is a crucial step in constructing the uncertain covariance matrix, revealing the central tendency correlation and uncertainty fluctuation correlation among variables, respectively. Introducing a standardization factor ensures numerical stability during the calculation process, enabling effective comparisons of variables with different dimensions within the same framework. The absolute value processing of the midpoint-radius interaction term not only quantifies the cross-influence between the two but also cleverly avoids interference from negative values, further enhancing the model's robustness. Ultimately, the construction of the uncertain covariance matrix provides a solid data foundation for the MRPCA model, enabling it to more accurately capture the underlying structure and outlier distributions in the data.

[0035] Furthermore, the formula for calculating the overall contribution ratio is as follows: ; In the formula, The overall contribution of the i-th variable is the sum of the contribution percentages of SPE and T2, and is used for fault location. Let represent the percentage contribution of the i-th variable to the SPE statistic. This represents the percentage contribution of the i-th variable to the T2 statistic.

[0036] Specifically, when calculating the overall contribution ratio, it is first necessary to determine the contribution ratio of each variable in the SPE statistic and T2 statistic. The SPE statistic mainly reflects the degree of deviation of the data in the residual space, while the T2 statistic reflects the outliers of the data in the principal component space. By calculating the contribution ratio of each variable in these two statistics, the degree of influence of the variables on the system anomalies can be comprehensively assessed.

[0037] Specifically, for the i-th variable, its contribution percentage in the SPE statistic can be determined by its percentage in the SPE contribution plot. This percentage reflects the relative contribution of the variable to the SPE statistic. Similarly, the contribution percentage of the variable in the T2 statistic can also be obtained by its percentage in the T2 contribution plot. These two percentages characterize the impact of the variables on system anomalies from different perspectives, providing a foundation for subsequent calculation of the overall contribution percentage.

[0038] After obtaining the contribution percentage of each variable in the SPE and T2 statistics, the overall contribution of the i-th variable can be obtained through a simple summation operation. This overall contribution takes into account the variable's outlier behavior in both the residual space and the principal component space, thus more accurately reflecting the variable's actual contribution to the system failure. During the failure localization phase, by comparing the overall contribution of each variable, the failure variable can be quickly identified, providing strong support for operational and maintenance decisions.

[0039] Furthermore, it is worth noting that in calculating the overall contribution ratio, it is necessary to ensure that the control limits of the SPE and T2 statistics used are reasonable and effective. These control limits are typically set based on historical or simulated data to distinguish between normal and abnormal data. By setting reasonable control limits, the accuracy and reliability of fault detection can be ensured, avoiding false alarms and missed alarms.

[0040] Furthermore, the interval training data includes battery external voltage, battery internal voltage, charging and discharging current, temperature, SOC, and internal resistance.

[0041] Furthermore, the battery's external voltage, internal voltage, charging / discharging current, temperature, state of charge (SOC), and internal resistance constitute six variables, and the contribution values ​​of each variable are denoted as follows: , And calculate the proportion of each variable using the following formula; ; ; in, Let represent the percentage contribution of the i-th variable to the SPE statistic. This represents the percentage contribution of the i-th variable to the T2 statistic. This represents the raw contribution value of the i-th variable to the SPE statistic of the current fault sample; This represents the raw contribution value of the i-th variable to the T2 statistic of the current fault sample; and These represent the sum of the original SPE contribution values ​​of all 6 variable stars and the sum of the original T2 contribution values, i.e., the total contribution value; variable indices i=1 to 6, corresponding to the 6 key variables of the BMS system: external battery voltage, internal battery voltage, charging and discharging current, temperature, SOC, and internal resistance.

[0042] Specifically, when calculating the proportion of each variable, it is first necessary to clarify the original contribution value of each variable to the SPE and T2 statistics. These original contribution values ​​are automatically calculated by the MRPCA model when processing the data, and they reflect the direct role of each variable in anomaly detection. For the SPE statistic, the original contribution value of each variable represents the contribution of the degree of deviation of that variable in the residual space to the overall SPE value; while for the T2 statistic, the original contribution value of each variable reflects the contribution of the abnormal performance of that variable in the principal component space to the overall T2 value.

[0043] After obtaining the raw contribution value for each variable, the next step is to calculate the total contribution value for all variables. This includes the sum of the raw contributions of all variables in the SPE statistic and the sum of the raw contributions of all variables in the T2 statistic. These two total contribution values ​​are used as denominators in subsequent calculations of the contribution percentage of each variable.

[0044] In specific calculations, for the i-th variable, its contribution percentage in the SPE statistic is obtained by dividing its original SPE contribution value by the sum of the original SPE contributions of all variables; similarly, its contribution percentage in the T2 statistic is calculated by dividing its original T2 contribution value by the sum of the original T2 contributions of all variables. These two percentages reflect the degree of influence of the variable on system anomalies in the residual space and principal component space, respectively.

[0045] Finally, by summing the contribution percentages of each variable in the SPE and T2 statistics, the overall contribution percentage of that variable is obtained. This comprehensive indicator takes into account the variable's outlier behavior in two different spaces, thus providing a more comprehensive assessment of the variable's actual contribution to system failures. During the fault localization phase, operations and maintenance personnel can quickly identify the faulty variable based on the overall contribution percentage of each variable, thereby taking targeted operational and maintenance measures to improve the efficiency and accuracy of fault handling. Example 2:

[0046] This embodiment provides a BMS backend data diagnostic system based on MRPCA, used to execute a BMS backend data diagnostic method based on MRPCA according to Embodiment 1 above, including: The data acquisition module is used to acquire interval training data and test data from the BMS system. It is responsible for collecting interval training data and test data from the BMS system in real time. These data cover key parameters such as battery external voltage, battery internal voltage, charging and discharging current, temperature, SOC and internal resistance. Each parameter is presented in interval form, including midpoint value and radius value, which provides a comprehensive and accurate information foundation for subsequent data processing and analysis.

[0047] The model building module decomposes the interval training data into a midpoint matrix and a radius matrix, constructs an uncertain covariance matrix, and establishes an MRPCA model. This module first decomposes the interval training data into a midpoint matrix and a radius matrix, which respectively reflect the central tendency and uncertainty fluctuations of each variable. Then, by constructing the uncertain covariance matrix, the interaction between the midpoint and radius is taken into account, thus more accurately describing the distribution characteristics of the data. Based on this, an MRPCA model is established, which can more effectively capture potential anomalies in the data, providing strong support for subsequent fault detection.

[0048] The control limit calculation module is used to calculate the control limits for the SPE statistic and the T2 statistic. Based on the established MRPCA model, it calculates the control limits for the SPE statistic and the T2 statistic. These two control limits serve as thresholds for subsequent fault detection, and their rationality and effectiveness directly affect the accuracy and reliability of fault detection. Therefore, this module must fully consider the characteristics of historical or simulated data during the calculation process to ensure that the control limits are set neither too lenient, leading to missed detections, nor too strict, leading to false alarms.

[0049] The fault detection module determines the presence of faults based on the SPE and T2 statistics of test data. It utilizes an established MRPCA model and calculated control limits to process and analyze the test data in real time. By calculating the SPE and T2 statistics of the test data and comparing them with the corresponding control limits, this module can quickly determine whether a fault exists in the battery system. Once a fault is detected, the system immediately triggers the subsequent fault localization process to ensure that the fault is addressed promptly.

[0050] The fault location module calculates the overall contribution ratio of each variable to identify the faulty variable when a fault exists. In the presence of a fault, this module calculates the proportion of each variable in the SPE and T2 contribution graphs separately and merges them to obtain the overall contribution ratio. This comprehensive indicator takes into account the abnormal behavior of variables in both the residual space and principal component space, thus more accurately reflecting the actual contribution of variables to system faults. By comparing the overall contribution of each variable, operations and maintenance personnel can quickly identify the faulty variable and take targeted operational and maintenance measures, improving the efficiency and accuracy of fault handling.

[0051] Furthermore, the fault location module is specifically used for: Calculate the percentage of each variable in the SPE contribution plot. ; Calculate the proportion of each variable in the T2 contribution plot. ; pass Calculate the overall contribution ratio; The source of the fault is determined based on the variable with the highest overall contribution.

[0052] Specifically, when performing its functions, the fault location module first calculates the proportion of each variable in the SPE contribution plot using the constructed MRPCA model on the acquired test data. This step quantifies the relative contribution of each variable to the SPE statistic by analyzing the deviation of the data in the residual space. Next, the module similarly calculates the proportion of each variable in the T2 contribution plot, which reflects the contribution of the variable's abnormal behavior in the principal component space to the T2 statistic. After obtaining these two proportions, the fault location module uses a specific formula to merge them, thus obtaining the comprehensive contribution proportion of each variable. This comprehensive indicator takes into account the abnormal behavior of the variable in two different spaces, thus more accurately reflecting the variable's actual contribution to the system fault. Finally, based on the comprehensive contribution proportion, the module quickly identifies the fault source, that is, finds the variable that contributes the most to the system fault, providing strong support for subsequent operation and maintenance decisions. Example 3:

[0053] This embodiment provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements a BMS background data diagnostic method based on MRPCA according to Embodiment 1.

[0054] Specifically, this computer-readable storage medium can be applied to various computing devices related to battery management systems, such as servers and dedicated data processing terminals. When the processor executes the computer program stored on this medium, it first initiates a data acquisition process, collecting interval training data and test data from the BMS system. This data includes key parameters such as external battery voltage, internal battery voltage, charging and discharging current, temperature, SOC, and internal resistance, presented in interval form to provide comprehensive information for subsequent analysis. Next, the program calls model building instructions to decompose the interval training data into a midpoint matrix and a radius matrix, thereby constructing an uncertain covariance matrix and establishing an MRPCA model to accurately describe the data distribution characteristics and capture potential anomalies. Subsequently, the program executes control limit calculation instructions, calculating the SPE statistic control limit and T2 statistic control limit based on the established MRPCA model, setting reasonable thresholds for fault detection. After acquiring the test data, the program uses the established model and control limits to calculate the SPE statistic and T2 statistic of the test data through fault detection instructions, compares them with the control limits, and determines whether a fault exists in the battery system. If a fault is detected, the program will initiate a fault location process, calculate the proportion of each variable in the SPE and T2 contribution graphs respectively, and merge them to obtain the comprehensive contribution proportion. Finally, the source of the fault is determined based on the variable with the highest comprehensive contribution proportion, providing strong support for operation and maintenance decisions.

[0055] In summary, this application provides a BMS backend data diagnostic method, system, and storage medium based on MRPCA. By introducing interval data and an uncertain covariance matrix, it significantly improves the accuracy and reliability of battery management system fault detection. This method not only effectively captures potential anomalies in the data but also accurately identifies fault variables through comprehensive contribution proportions, providing strong support for operation and maintenance decisions. In practical applications, this method can be widely used in electric vehicles, energy storage systems, and other fields, helping operation and maintenance personnel quickly locate and resolve battery system faults, improving system safety and stability. Simultaneously, the implementation of this method promotes the development of intelligent and refined operation and maintenance of battery management systems, laying a solid foundation for the sustainable development of the battery industry.

[0056] The various embodiments in this specification are described in a progressive manner, with each example focusing on the differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. The methods disclosed in the embodiments are described simply because they correspond to the methods disclosed in the embodiments; relevant parts can be found in the method section.

[0057] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A BMS backend data diagnostic method based on MRPCA, characterized in that, Includes the following steps: Step 1: Obtain the interval training data of the BMS system; Step 2: Decompose the interval training data into a midpoint matrix. and radius matrix And based on the midpoint matrix and radius matrix Construct an uncertain covariance matrix; Step 3: Establish an MRPCA model based on the aforementioned uncertain covariance matrix, and calculate the control limits of the SPE statistic. and T2 statistic control limits ; Step 4: Obtain test data and use the MRPCA model to calculate the SPE statistic and T2 statistic of the test data; Step 5: Based on the fact that the SPE statistic and T2 statistic exceed the corresponding control limits, determine whether a fault exists; Step 6: If a fault exists, calculate the proportion of each variable in the SPE and T2 contribution maps respectively, merge them to obtain the comprehensive contribution proportion, and determine the fault variable based on the comprehensive contribution proportion.

2. The BMS backend data diagnostic method based on MRPCA according to claim 1, characterized in that: The interval training data forms the original sample matrix X, represented as: ； In the formula, n is the number of samples, that is, how many time points of data were collected; m represents the number of process variables, i.e., the number of BMS parameters monitored. Represents m distinct process variables; This indicates that the matrix is ​​an n-row, m-column real matrix; Each variable It can be represented as an interval: ； The midpoint value is: ; The radius value is: ; In the formula, This represents the interval measurement value of the i-th process variable at the k-th sampling time. This indicates the lower limit of the range of values; This indicates the upper limit of the range; This represents the midpoint of the interval measurement of the i-th variable at time k; This represents the radius of the interval measurement of the i-th variable at time k; Let be the midpoint matrix, consisting of all composition; Let be the radius matrix, consisting of all composition.

3. The BMS backend data diagnostic method based on MRPCA according to claim 2, characterized in that: Each variable in the interval training data is represented as follows: ； In the formula, This represents the interval measurement value of the i-th process variable at the k-th sampling time. This represents the midpoint of the interval measurement of the i-th variable at time k; Let represent the radius of the interval measurement value of the i-th variable at time k.

4. The BMS backend data diagnostic method based on MRPCA according to claim 1, characterized in that: The formula for constructing the uncertain covariance matrix is ​​as follows: ； In the formula, R represents the uncertain covariance matrix, which is the basis of the MRPCA model; n is the number of samples. It is a standardization factor used to calculate the average value; Midpoint matrix transpose; It is the standard covariance matrix of the midpoint matrix, reflecting the central tendency correlation between variables; Radius matrix transpose; It is the covariance matrix of the radius matrix, reflecting the fluctuation correlation of the uncertainty of the variables; and The absolute value of the interaction term between the midpoint and the radius is used to quantify the cross-effect of uncertainty and avoid interference from negative values; and These are two standard variance-covariance matrices, which respectively represent the partial contributions to the total variance-covariance matrix given by the midpoint and radius.

5. The BMS backend data diagnostic method based on MRPCA according to claim 1, characterized in that: The formula for calculating the overall contribution ratio is as follows: ； In the formula, The overall contribution of the i-th variable is the sum of the contribution percentages of SPE and T2, and is used for fault location. Let represent the percentage contribution of the i-th variable to the SPE statistic. This represents the percentage contribution of the i-th variable to the T2 statistic.

6. The BMS backend data diagnostic method based on MRPCA according to claim 5, characterized in that: The interval training data includes battery external voltage, battery internal voltage, charging and discharging current, temperature, SOC, and internal resistance.

7. A BMS backend data diagnostic method based on MRPCA according to claim 6, characterized in that: The battery's external voltage, internal voltage, charging / discharging current, temperature, state of charge (SOC), and internal resistance constitute six variables, and the contribution value of each variable is denoted as follows: , And calculate the proportion of each variable using the following formula; ； ； in, Let represent the percentage contribution of the i-th variable to the SPE statistic. This represents the percentage contribution of the i-th variable to the T2 statistic. This represents the raw contribution value of the i-th variable to the SPE statistic of the current fault sample; This represents the raw contribution value of the i-th variable to the T2 statistic of the current fault sample; and These represent the sum of the original SPE contribution values ​​of all 6 variable stars and the sum of the original T2 contribution values, i.e., the total contribution value; variable indices i=1 to 6, corresponding to the 6 key variables of the BMS system: external battery voltage, internal battery voltage, charging and discharging current, temperature, SOC, and internal resistance.

8. A BMS backend data diagnostic system based on MRPCA, characterized in that, A BMS backend data diagnostic method based on MRPCA as described in any one of claims 1-7 includes: The data acquisition module is used to acquire interval training data and test data of the BMS system; The model building module is used to decompose the interval training data into a midpoint matrix and a radius matrix, construct an uncertain covariance matrix, and build an MRPCA model. The control limit calculation module is used to calculate the control limits for the SPE statistic and the T2 statistic. The fault detection module is used to determine whether a fault exists based on the SPE and T2 statistics of the test data. The fault location module is used to calculate the overall contribution ratio of each variable to determine the fault variable when a fault exists.

9. A BMS backend data diagnostic system based on MRPCA according to claim 8, characterized in that: The fault location module is specifically used for: Calculate the percentage of each variable in the SPE contribution plot. ; Calculate the proportion of each variable in the T2 contribution plot. ; pass Calculate the overall contribution ratio; The source of the fault is determined based on the variable with the highest overall contribution.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements a BMS background data diagnostic method based on MRPCA as described in any one of claims 1-7.

Images