State monitoring and fault diagnosis method of battery management system

Abstract

The application discloses a state monitoring and fault diagnosis method of a battery management system, and relates to the technical field of battery management.The steps are as follows: S1, collecting battery basic parameters in real time based on a sensor, and integrating the parameters into a data set after preprocessing; S2, respectively calculating the state of charge value, the health state value and the function state value of the battery, performing weighted summation on the obtained states, comprehensively obtaining the battery state value, and feeding back the battery state condition based on the battery state value; and S3, obtaining the average value of the state of the historical battery based on big data, comparing the obtained battery state value with the average value, and identifying abnormal values by using the central tendency and the dispersion degree of data.The application accurately locates the fault root of an abnormal battery by voltage, capacity and internal resistance testing, adopts differentiated maintenance measures, avoids blind replacement, reduces maintenance cost, prolongs the overall service life of the battery, and facilitates the management and maintenance of the battery system.

Description

Technical Field

[0001] This invention relates to the field of battery management technology, specifically to a method for state monitoring and fault diagnosis of a battery management system. Background Technology

[0002] In an era where batteries are widely used in various devices and systems, from electric vehicles to energy storage power stations, the performance, safety, and reliability of batteries are of paramount importance. As a result, battery management systems (BMS) have emerged to become the core of ensuring the efficient and stable operation of batteries.

[0003] Current battery management systems face numerous challenges in state monitoring and fault diagnosis. On one hand, traditional methods for collecting basic battery parameters rely on limited information such as voltage, current, and temperature monitored by a small number of sensors. This is insufficient to comprehensively reflect the complex internal state changes of the battery, and the data accuracy is easily affected by environmental interference. This results in poor accuracy of the calculated state of charge (SOC), state of health (SOH), and state of function (SOF), failing to provide a reliable basis for battery performance evaluation. For example, in high-speed driving or extreme temperature environments, inaccurate SOC estimations may lead drivers to misjudge the remaining range, affecting travel plans and driving safety. On the other hand, existing fault diagnosis methods often rely on single-dimensional data or simple threshold judgments. When a battery malfunctions, alarms are often triggered solely based on voltage or current exceeding preset ranges, easily leading to numerous false alarms and missed alarms.

[0004] Chinese patent CN118611199A discloses a fault diagnosis system and method for energy storage battery equalization management. The disclosed technology is as follows: providing a fault diagnosis system and method for energy storage battery equalization management, wherein the fault diagnosis system detects the working status of the equalization unit through a detection feedback unit and uploads it to the AFE unit (analog front-end unit), and then the AFE unit transmits it to the MCU unit, and the MCU unit determines whether the equalization unit is faulty based on the working status of the equalization unit detected by the detection feedback unit; however, it does not involve the monitoring of the charge status, health status and functional status, resulting in poor accuracy. Summary of the Invention

[0005] To address the aforementioned technical problems, this technical solution provides a method for state monitoring and fault diagnosis of a battery management system, thus resolving the aforementioned issues.

[0006] To achieve the above objectives, the technical solution adopted by this invention is: a method for state monitoring and fault diagnosis of a battery management system, comprising the following steps:

[0007] S1. Based on real-time acquisition of basic battery parameters by sensors, the data is preprocessed and integrated into a dataset;

[0008] S2. Calculate the battery's state of charge, health, and functional states respectively, sum the weighted values, and obtain the overall battery state value. Feedback on the battery state status is based on the battery state value.

[0009] S3. Based on big data, obtain the historical battery status average value, compare the obtained battery status value with the average value, use the central tendency and dispersion of the data to identify outliers, judge the degree of deviation between the battery status value and the average value, and filter out batteries that show abnormal battery status values.

[0010] S4. Conduct performance tests on abnormal batteries, including voltage, capacity and internal resistance tests, to find the cause of battery failure and perform maintenance on the battery.

[0011] Preferably, the basic battery parameters in step S1 include voltage, current, and capacity; preprocessing includes data cleaning, data standardization, and data alignment.

[0012] Data cleaning includes noise reduction, missing value handling, and outlier correction.

[0013] Data standardization is performed through Z-score standardization, and data alignment is performed by aligning the sampled data from different sensors in the time dimension.

[0014] Preferably, in step S2, the battery's state of charge is calculated based on the ampere-hour integral method. The set time interval is determined based on the data sampling frequency, the change in charge within the time interval is calculated, the current at each moment is multiplied by the time interval and then summed up, and the current value is obtained by subtracting the ratio of the change in charge within the time interval to the actual capacity from the initial state of charge.

[0015] Preferably, in step S2, the battery health status value is calculated by determining the initial rated capacity of the battery, continuously collecting current, voltage and temperature data during the charging and discharging process through sensors, and recording the battery's usage time and cycle number information.

[0016] Discharge the battery completely to the cutoff voltage, then charge it to full capacity using a standard constant current, and record the total amount of charge received as the current actual capacity.

[0017] The health status value is obtained by dividing the current actual capacity by the initial rated capacity.

[0018] Preferably, in step S2, the functional state value calculation is based on the preprocessed dataset, extracting the current state of charge, health status, real-time temperature, voltage, current, and recent charge / discharge history to construct a power limit model. Combined with the battery equivalent circuit model, the upper and lower voltage limits are determined according to the state of charge, the internal resistance parameters are corrected according to the health status, and the power threshold is adjusted through the temperature coefficient. Considering the continuous discharge time, the voltage drop at different durations is calculated based on the current current and internal resistance to ensure that the voltage does not exceed the safe range, and the maximum output power for the corresponding duration is obtained. The power limit is compared with the battery's rated power, and the ratio is the functional state value.

[0019] Preferably, the step in step S2 to obtain the battery state value is as follows:

[0020] The weight values ​​of each state are determined, and the weight coefficients of the state of charge, state of health and functional state values ​​are set by using the analytic hierarchy process (AHP) in combination with the battery application scenario and safety requirements.

[0021] The state of charge, health, and functional states are uniformly mapped to the interval (0, 1);

[0022] The battery state value is obtained by combining the weighted summation formula.

[0023] Preferably, step S3 specifically includes:

[0024] Collect historical state values ​​of batteries in the same group, clean them, calculate the average value as the normal benchmark, and calculate the standard deviation to reflect the fluctuation range.

[0025] By setting a threshold based on the data distribution characteristics, the real-time status value of the battery to be tested is obtained, and the absolute deviation between the value and the historical average is calculated. The status is judged by comparing the deviation value with the threshold. A deviation ≤ 2σ is normal, 2σ < deviation ≤ 3σ is generally abnormal, and a deviation > 3σ is severely abnormal.

[0026] The selected abnormal batteries are subjected to secondary verification, the calculation logic is reviewed, and the true abnormality is confirmed by combining multiple consecutive abnormalities, and a list of abnormal batteries is generated.

[0027] Preferably, the standard deviation is calculated to reflect the fluctuation range. Specifically, the screening criteria for batteries in the same group are defined, and the target group is determined based on the model, usage scenario and usage duration. The historical comprehensive state value of the batteries in this group per unit time is extracted from the big data platform to form the raw dataset. After preprocessing, the historical state average is calculated, and the effective data is arithmetically averaged. The result is used as the normal state benchmark for the batteries in this group. The standard deviation is calculated, and based on the obtained mean, the dispersion of the state value within the group is quantified to obtain an index reflecting the fluctuation range of the battery state value under normal conditions.

[0028] Preferably, the threshold is set based on the standard deviation of historical data, and three levels of judgment thresholds are set, including normal threshold, general abnormal threshold and severe abnormal threshold. The judgment standard for normal threshold is the historical mean ± 2 times the standard deviation; the judgment standard for general abnormal threshold is between the historical mean ± 2 times the standard deviation and ± 3 times the standard deviation; and the judgment standard for severe abnormal threshold is outside the historical mean ± 3 times the standard deviation.

[0029] Preferably, in step S4, the performance testing and maintenance of abnormal batteries should involve locating faults through voltage, capacity, and internal resistance tests, and then taking targeted measures; for batteries with low charge and poor consistency, slow charging and equalization maintenance measures should be taken; for batteries with severe internal resistance surges and unrecoverable short circuits, safe scrapping and cell replacement should be carried out.

[0030] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0031] This invention achieves a comprehensive quantitative assessment of the current performance of a battery by collecting real-time parameters and comprehensively calculating its charge, health, and functional status. This avoids misjudgments based on a single indicator, provides a reliable basis for usage decisions, and scientifically screens batteries with abnormal conditions by combining the average and fluctuation characteristics of historical big data. This allows for the early detection of potential faults and reduces the risk of equipment downtime or safety accidents caused by battery failure. By accurately locating the root cause of abnormal battery faults through voltage, capacity, and internal resistance testing, differentiated maintenance measures can be taken to avoid blind replacement, reduce maintenance costs, extend the overall battery life, and facilitate the management and maintenance of the battery system. Attached Figure Description

[0032] Figure 1 This is a flowchart of the steps of the present invention. Detailed Implementation

[0033] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0034] Reference Figure 1 As shown, the battery management system's state monitoring and fault diagnosis method comprises the following steps:

[0035] S1. Based on real-time acquisition of basic battery parameters by sensors, the data is preprocessed and integrated into a dataset;

[0036] S2. Calculate the battery's state of charge, health, and functional states respectively, sum the weighted values, and obtain the overall battery state value. Feedback on the battery state status is based on the battery state value.

[0037] S3. Based on big data, obtain the historical battery status average value, compare the obtained battery status value with the average value, use the central tendency and dispersion of the data to identify outliers, judge the degree of deviation between the battery status value and the average value, and filter out batteries that show abnormal battery status values.

[0038] S4. Conduct performance tests on abnormal batteries, including voltage, capacity and internal resistance tests, to find the cause of battery failure and perform maintenance on the battery.

[0039] At the data level, real-time sensor acquisition and preprocessing ensure the accuracy of basic data, providing high-quality input for subsequent analysis and avoiding judgment bias caused by data errors. In terms of state assessment, it integrates and weights three key state values—charge, health, and function—to overcome the limitations of single indicators and comprehensively reflect the current performance of the battery. This allows users to clearly understand the battery's real-time availability, aging level, and power output capability. In the anomaly identification stage, it uses historical averages and discrete characteristics (such as standard deviation) of big data for comparison, enabling the scientific screening of abnormal batteries rather than relying on subjective experience. This allows for early detection of potential faults and reduces the risk of sudden equipment shutdowns or safety accidents. At the maintenance level, it conducts specialized tests on voltage, capacity, and internal resistance for abnormal batteries, which can accurately locate the root cause of the fault (such as low charge, sudden increase in internal resistance, and capacity decay). Then, targeted measures (such as recharging, equalization maintenance, and individual cell replacement) can be taken, avoiding blindly replacing the entire battery pack, significantly reducing maintenance costs, and extending the overall lifespan of the battery.

[0040] In step S1, the basic battery parameters include voltage, current, and capacity; preprocessing includes data cleaning, data standardization, and data alignment.

[0041] Data cleaning includes noise reduction, missing value handling, and outlier correction.

[0042] Data standardization is performed through Z-score standardization, and data alignment is performed by aligning the sampled data from different sensors in the time dimension.

[0043] From the perspective of parameter acquisition, this application focuses on three core indicators: voltage, current, and capacity. These indicators comprehensively reflect the battery's energy state, charge / discharge dynamics, and storage capacity, providing crucial raw data for subsequent calculations of state of charge and health status. This ensures the relevance and accuracy of the state assessment. The data cleaning process, through noise reduction (filtering sensor fluctuations, environmental interference, and other noise), eliminates invalid interference signals, ensuring data authenticity. Missing value handling (such as interpolation completion) avoids the impact of data fragmentation on analysis, ensuring dataset integrity. Outlier correction (removing or correcting extreme values ​​outside the reasonable range) reduces the misleading influence of erroneous data on subsequent calculations, improving data quality from the source.

[0044] In step S2, the battery's state of charge (SOC) is calculated using the ampere-hour integral method. By determining the set time interval based on the data sampling frequency, the change in charge within the time interval is calculated. The current at each moment is multiplied by the time interval and then summed up. The initial SOC is then subtracted from the ratio of the change in charge over the time interval to the actual capacity to obtain the current SOC value.

[0045] Step S2 of this application uses the ampere-hour integration method to calculate the battery's state of charge (SOC). Its principle is based on the direct correlation between "current-time-charge." By matching the time interval of the sampling frequency, the product of current and time is accumulated to quantify the change in charge. The current value is then calculated by combining the initial SOC and the actual capacity. The logic is intuitive and easy to understand, facilitating engineering implementation. Relying on real-time current data, it dynamically updates according to the sampling frequency, quickly responding to state fluctuations during charging and discharging, meeting the real-time requirements for SOC. Furthermore, it is not limited by the battery chemical system, is applicable to various battery types, requires only basic parameters for calculation, has low hardware computing power requirements, and is easy to deploy in various battery management systems.

[0046] The steps for calculating the battery health status value in step S2 are as follows:

[0047] The initial rated capacity of the battery is clearly defined as the benchmark. This value is the standard capacity of the battery when it leaves the factory (e.g., the rated capacity of a certain lithium battery is 100Ah). It is the original reference for measuring the health status. Based on this, key parameters during the charging and discharging process are collected in real time by sensors: current (reflecting the charging and discharging intensity), voltage (reflecting the internal electrochemical state of the battery), and temperature (affecting the rate of capacity decay). At the same time, the cumulative usage time of the battery (e.g., total operating hours) and the number of cycles (one complete charge and discharge cycle) are recorded. These data provide environmental and historical background information for subsequent capacity assessment.

[0048] Discharge the battery completely to the cutoff voltage (e.g., 3.0V for lithium-ion batteries) using a standard discharge current (e.g., 0.5C) to ensure that the remaining charge inside the battery is depleted; then charge it to full capacity using the same standard current (0.5C) at a constant current (reaching the rated voltage and the current dropping below the threshold). Record the total charge received, which is the current actual capacity. If a battery is charged to 85Ah, its actual capacity is 85Ah.

[0049] The health status value is calculated using the formula: "Current actual capacity ÷ Initial rated capacity × 100%".

[0050] Full charge-discharge testing directly obtains the actual capacity, ensuring reliable and accurate results and avoiding errors from indirect estimation. Combining parameters such as current, voltage, and temperature eliminates interference from abnormal charge-discharge environments (such as temporary capacity reduction caused by low temperatures). Recording usage time and cycle count helps analyze the causes of aging (such as excessive cycle count or long-term high-temperature storage), providing a basis for dynamic correction of subsequent health status. In this way, the SOH value can not only intuitively reflect the degree of battery aging but also provide a quantitative indicator for battery maintenance and replacement decisions, making it a core basis for evaluating long-term battery performance.

[0051] In step S2, the functional state value is calculated by extracting the current state of charge, health status, real-time temperature, voltage, current, and recent charge / discharge history from the preprocessed dataset to construct a power limit model. Combined with the battery equivalent circuit model, the upper and lower voltage limits are determined based on the state of charge, the internal resistance parameters are corrected based on the health status, and the power threshold is adjusted through the temperature coefficient. Considering the continuous discharge time, the voltage drop at different durations is calculated based on the current current and internal resistance to ensure that the voltage does not exceed the safe range. The maximum output power for the corresponding duration is obtained, and the power limit is compared with the battery's rated power; the ratio is the functional state value.

[0052] From the perspective of parameter fusion, this application integrates multi-dimensional data such as state of charge (SOC), state of health (SOH), temperature, voltage, current, and charge / discharge history, breaking through the limitations of single parameters and comprehensively covering key factors affecting battery power output (such as low SOC limiting power and low temperature reducing allowable current), providing a complete basis for power limit assessment. Combined with the equivalent circuit model, the voltage safety boundary is determined by SOC (avoiding overcharging and over-discharging), the internal resistance is corrected based on SOH (increased internal resistance in aging batteries leads to decreased power), and the threshold is dynamically adjusted with the help of the temperature coefficient (such as actively reducing the power limit in low-temperature environments), ensuring that the model output closely matches the actual physical characteristics of the battery.

[0053] The steps in step S2 to obtain the battery state value are as follows:

[0054] The determination of weighting coefficients needs to be closely integrated with the battery's application scenarios and safety requirements, and quantitative decision-making should be achieved through the Analytic Hierarchy Process (AHP).

[0055] The "Comprehensive Battery State Assessment" is used as the target layer, the State of Charge (SOC), State of Health (SOH), and State of Function (SOF) are used as the criteria layer, and the core requirements of different application scenarios (such as the power safety of electric vehicles and the capacity stability of energy storage systems) are used as the solution layer.

[0056] Pairwise comparison and judgment matrix construction: Battery experts were invited to compare the importance of criteria-level indicators in pairs (e.g., "In electric vehicle scenarios, SOF (power output capability) is more important than SOC"). Subjective judgments were transformed into quantitative values ​​(1-9 scale method) to form a judgment matrix.

[0057] The weights of each indicator are calculated using matrix eigenvalues, and a consistency check (CR < 0.1) is performed simultaneously to ensure logical consistency. Finally, weight coefficients that meet the requirements of the scenario are output.

[0058] Prioritizing power safety, the weights are set as follows: SOC weight 0.3, SOH weight 0.2, and SOF weight 0.5.

[0059] Focusing on capacity and lifetime, the weights are set as follows: SOC weight 0.4, SOH weight 0.4, and SOF weight 0.2.

[0060] Step S3 specifically includes:

[0061] Collect historical state values ​​of batteries in the same group, clean them, calculate the average value as the normal benchmark, and calculate the standard deviation to reflect the fluctuation range.

[0062] By setting a threshold based on the data distribution characteristics, the real-time status value of the battery to be tested is obtained, and the absolute deviation between the value and the historical average is calculated. The status is judged by comparing the deviation value with the threshold. A deviation ≤ 2σ is normal, 2σ < deviation ≤ 3σ is generally abnormal, and a deviation > 3σ is severely abnormal.

[0063] The selected abnormal batteries are subjected to secondary verification, the calculation logic is reviewed, and the true abnormality is confirmed by combining multiple consecutive abnormalities, and a list of abnormal batteries is generated.

[0064] This application calculates the mean and standard deviation based on historical data of the same group, providing an objective benchmark for anomaly identification. By cleaning historical data to eliminate interference, the mean can truly reflect the normal state of the group, while the standard deviation quantifies the reasonable range of fluctuations. This avoids the bias of relying on subjective experience to set standards, making the definition of "normal" and "abnormal" more scientific.

[0065] Secondly, by setting three threshold levels (≤2σ for normal, 2σ-3σ for general abnormality, and >3σ for severe abnormality) based on the data distribution characteristics, a refined classification of abnormal states is achieved. This classification method not only covers the vast majority of normal data (e.g., 2σ covers 95% of reasonable fluctuations), but also distinguishes risk levels based on the degree of deviation, providing a clear basis for subsequent differentiated treatment measures (e.g., key monitoring or emergency maintenance).

[0066] The standard deviation is calculated to reflect the fluctuation range. Specifically, by clarifying the screening criteria for batteries in the same group, the target group is determined based on model, usage scenario, and usage duration. The historical comprehensive state values ​​of batteries in this group per unit time are extracted from the big data platform to form the raw dataset. After preprocessing, the historical state average is calculated. The effective data are then arithmetically averaged, and the result is used as the normal state benchmark for the batteries in this group. The standard deviation is calculated, and based on the obtained mean, the dispersion of state values ​​within the group is quantified to obtain an index reflecting the fluctuation range of battery state values ​​under normal conditions.

[0067] This application ensures the comparability of historical data by clearly defining the screening criteria for batteries in the same group (consistent model, usage scenario, and usage duration). Batteries in the same group operate under the same conditions, and the fluctuation of their state values ​​can better reflect the performance differences of the batteries themselves, rather than interference caused by differences in external environment or equipment. This allows the standard deviation calculated subsequently to truly represent the normal fluctuation characteristics of this type of battery, laying a foundation of common source data for anomaly judgment.

[0068] Secondly, the preprocessed historical data is averaged to obtain a baseline for normal conditions. Then, the dispersion is quantified using the standard deviation, transforming the "fluctuation range" from a vague concept into a quantifiable indicator. The size of the standard deviation directly reflects the consistency of battery state within the group—a small standard deviation indicates stable group performance and a narrow fluctuation range; a large standard deviation indicates significant individual differences and a wider reasonable fluctuation range. This quantification avoids the subjectivity of judging "normal fluctuations" based on experience, providing clear data support for setting abnormal thresholds (such as 2σ and 3σ).

[0069] Thresholds are set based on the standard deviation of historical data, and three levels of judgment thresholds are set, including normal threshold, general abnormal threshold and severe abnormal threshold. The judgment standard for normal threshold is the historical mean ± 2 standard deviations; the judgment standard for general abnormal threshold is between the historical mean ± 2 standard deviations and ± 3 standard deviations; and the judgment standard for severe abnormal threshold is outside the historical mean ± 3 standard deviations.

[0070] In step S4, the performance testing and maintenance of abnormal batteries requires locating faults through voltage, capacity, and internal resistance tests, and then taking targeted measures; for batteries with low charge and poor consistency, slow charging and equalization maintenance measures are taken; for batteries with severe internal resistance surges and unrecoverable short circuits, safe scrapping and cell replacement are carried out.

[0071] This application focuses on core performance testing indicators such as voltage, capacity, and internal resistance, which can accurately locate the type of fault and avoid blind repairs. Voltage testing can quickly identify problems such as low charge and short circuit; capacity testing reflects the degree of battery aging; and internal resistance testing reveals the internal electrochemical state, providing a clear direction for maintenance.

[0072] Maintenance measures are categorized according to fault type. For recoverable faults such as low charge and poor consistency, slow charging and equalization maintenance are used to restore battery performance at low cost, extend service life, and avoid resource waste. For unrecoverable faults such as severe internal resistance increase and short circuit, timely and safe scrapping or individual cell replacement can eliminate safety hazards and prevent the fault from spreading and affecting the overall system.

[0073] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for state monitoring and fault diagnosis of a battery management system, characterized in that, The steps are as follows: S1. Based on real-time acquisition of basic battery parameters by sensors, the data is preprocessed and integrated into a dataset; S2. Calculate the battery's state of charge, health, and functional states respectively, sum the weighted values, and obtain the overall battery state value. Feedback on the battery state status is based on the battery state value. In step S2, the functional state value is calculated by extracting the current state of charge, health status, real-time temperature, voltage, current, and recent charge / discharge history from the preprocessed dataset to construct a power limit model. Combined with the battery equivalent circuit model, the upper and lower voltage limits are determined based on the state of charge, the internal resistance parameters are corrected based on the health status, and the power threshold is adjusted through the temperature coefficient. Considering the continuous discharge time, the voltage drop at different durations is calculated based on the current current and internal resistance to ensure that the voltage does not exceed the safe range. The maximum output power for the corresponding duration is obtained, and the power limit is compared with the battery's rated power; the ratio is the functional state value. The steps in step S2 to obtain the battery state value are as follows: The weight values ​​of each state are determined, and the weight coefficients of the state of charge, state of health and functional state values ​​are set by using the analytic hierarchy process (AHP) in combination with the battery application scenario and safety requirements. The state of charge, health, and functional states are uniformly mapped to the interval (0, 1); The battery state value is obtained by combining the weighted summation formula; The comprehensive battery state assessment is used as the target layer, the state of charge, state of health and functional state are used as the criteria layer, and the core requirements of different application scenarios are used as the solution layer. Pairwise comparison and judgment matrix construction: Experts in the battery field were invited to compare the importance of the criteria layer indicators pairwise, transforming subjective judgments into quantitative values ​​to form a judgment matrix. The weights of each indicator are calculated using matrix eigenvalues, and consistency checks are performed simultaneously to ensure logical consistency. Finally, weight coefficients that meet the requirements of the scenario are output. Prioritizing power safety, the weights are set as follows: SOC weight 0.3, SOH weight 0.2, and SOF weight 0.

5. Focusing on capacity and lifetime, the weights are set as follows: SOC weight 0.4, SOH weight 0.4, and SOF weight 0.

2. S3. Based on big data, obtain the historical battery status average value, compare the obtained battery status value with the average value, use the central tendency and dispersion of the data to identify outliers, judge the degree of deviation between the battery status value and the average value, and filter out batteries that show abnormal battery status values. Step S3 specifically includes: Collect historical state values ​​of batteries in the same group, clean them, calculate the average value as the normal benchmark, and calculate the standard deviation to reflect the fluctuation range. By setting a threshold based on the data distribution characteristics, the real-time status value of the battery to be tested is obtained, and the absolute deviation between the value and the historical average is calculated. The status is judged by comparing the deviation value with the threshold. A deviation ≤ 2σ is normal, 2σ < deviation ≤ 3σ is generally abnormal, and a deviation > 3σ is severely abnormal. The selected abnormal batteries are subjected to secondary verification, the calculation logic is reviewed, and the true abnormality is confirmed by combining multiple consecutive abnormalities to generate a list of abnormal batteries. The standard deviation is calculated to reflect the fluctuation range. Specifically, by clarifying the screening criteria for batteries in the same group, the target group is determined based on model, usage scenario, and usage duration. The historical comprehensive state values ​​of batteries in this group per unit time are extracted from the big data platform to form a raw dataset. After preprocessing, the historical state average is calculated. The effective data are then arithmetically averaged, and the result is used as the normal state benchmark for the batteries in this group. The standard deviation is calculated, and based on the obtained mean, the dispersion of state values ​​within the group is quantified to obtain an index reflecting the fluctuation range of battery state values ​​under normal conditions. S4. Conduct performance tests on abnormal batteries, including voltage, capacity and internal resistance tests, to find the cause of battery failure and perform maintenance on the battery.

2. The battery management system state monitoring and fault diagnosis method according to claim 1, characterized in that: In step S1, the basic battery parameters include voltage, current, and capacity; preprocessing includes data cleaning, data standardization, and data alignment. Data cleaning includes noise reduction, missing value handling, and outlier correction. Data standardization is performed through Z-score standardization, and data alignment is performed by aligning the sampled data from different sensors in the time dimension.

3. The battery management system state monitoring and fault diagnosis method according to claim 1, characterized in that: In step S2, the battery's state of charge (SOC) is calculated using the ampere-hour integral method. By determining the set time interval based on the data sampling frequency, the change in charge within the time interval is calculated. The current at each moment is multiplied by the time interval and then summed up. The initial SOC is then subtracted from the ratio of the change in charge over the time interval to the actual capacity to obtain the current SOC value.

4. The battery management system state monitoring and fault diagnosis method according to claim 1, characterized in that: In step S2, the battery health status value is calculated by determining the initial rated capacity of the battery, continuously collecting current, voltage and temperature data during the charging and discharging process through sensors, and recording the battery's usage time and cycle number information. Discharge the battery completely to the cutoff voltage, then charge it to full capacity using a standard constant current, and record the total amount of charge received as the current actual capacity. The health status value is obtained by dividing the current actual capacity by the initial rated capacity.

5. The battery management system state monitoring and fault diagnosis method according to claim 1, characterized in that: In step S4, the performance testing and maintenance of abnormal batteries requires locating faults through voltage, capacity, and internal resistance tests, and then taking targeted measures; for batteries with low charge and poor consistency, slow charging and equalization maintenance measures are taken; for batteries with severe internal resistance surges and unrecoverable short circuits, safe scrapping and cell replacement are carried out.

Images