Energy storage battery fault monitoring management method and system based on big data

Abstract

The invention discloses an energy storage battery fault monitoring management method and system based on big data, and the method comprises the steps: calculating the change rate of a battery monomer between each window and an adjacent window, and recognizing an energy output inconsistent time period; identifying adjacent monomer power compensation behaviors in the energy output inconsistent time period; the load allocation rate, the deviation ratio and the heat consumption change amplitude of the compensated single battery in the compensation process are obtained; predicting the damage probability of the single battery in a plurality of periods in the future, and generating a fault probability curve; constructing a masked fault identification model, and identifying the fault state of the single battery; evaluating the fault level of the single battery, and constructing a fault level index table; and evaluating the validity of the fault early warning and maintenance strategy, and updating the fault early warning strategy and the health maintenance strategy. Through the technical method provided by the invention, accurate monitoring and intelligent management of the energy storage battery pack can be realized, the accuracy and timeliness of battery fault early warning are remarkably improved, and efficient and safe operation of an energy storage system is ensured.

Description

Technical Field

[0001] The present invention relates to the field of new generation information technology, and in particular to a method and system for monitoring and managing energy storage battery faults based on big data. Background Art

[0002] With the rapid development of renewable energy and the gradual adoption of energy storage technologies, energy storage batteries, as a crucial component of power grid systems, play a vital role. Energy storage battery packs are used to balance power supply and demand, optimize load scheduling, and provide support during peak power demand. However, over long-term operation, the performance of energy storage battery packs gradually degrades. Especially in large-scale applications, cell damage and aging are becoming increasingly prominent. These issues directly impact the stability, efficiency, and overall lifespan of energy storage systems. Therefore, cell health monitoring and fault detection have become crucial tasks in battery management systems. Currently, battery management systems widely monitor battery parameters such as voltage, current, and temperature. However, in practical applications, the limitations of traditional monitoring methods have become increasingly apparent. For one thing, cell power variations and string load fluctuations are often difficult to detect promptly through simple parameter monitoring. When a cell fails or ages, the battery's power output often exhibits inconsistent variations. This can be particularly true when cell power fluctuations are masked by the compensating behavior of neighboring cells, potentially overlooking the fault signal. Since compensation often partially balances the overall load of the battery pack, some potential faults may not be identified promptly through conventional monitoring. In addition, because the power compensation behavior between different battery cells is often dynamic and affects each other, it is difficult to determine the true situation of the fault by monitoring only a single battery cell. Battery cells with frequent and large compensation behaviors may be mistakenly judged as healthy during long-term operation, resulting in the failure to be discovered in time. In addition, existing monitoring systems usually lack quantitative assessment of the risk of battery cell damage. Although battery cell damage may be exposed through signals such as power changes and temperature anomalies, these signals often lack sufficient accuracy and predictability and cannot provide valuable information about the degree of battery cell damage and the probability of future failure. Battery management systems usually find it difficult to provide accurate damage probability predictions, and it is even more difficult to take effective early warning and maintenance measures in advance before a battery cell fails. Therefore, how to accurately identify potential faults, conduct accurate health assessments, and timely adjust early warning mechanisms during the operation of energy storage battery packs has become a technical problem that needs to be solved urgently. Summary of the Invention

[0003] The present invention addresses the above-mentioned problems in the prior art and provides a method and system for monitoring and managing energy storage battery faults based on big data.

[0004] The first embodiment of the present invention provides a method for monitoring and managing energy storage battery faults based on big data, which mainly includes:

[0005] The battery management system obtains the operating parameters of each battery cell in the energy storage battery string, and uses a sliding window method to calculate the battery cell change rate between each window and the adjacent windows to identify periods of inconsistent energy output;

[0006] Based on the power changes, string load fluctuations, and temperature distribution of the battery cells under normal operating conditions, the power fluctuations of the battery cells during periods of inconsistent energy output are predicted, and whether there is power compensation behavior by adjacent cells during periods of inconsistent energy output.

[0007] Based on the power distribution data of the total load of the energy storage battery string during the period when the adjacent battery power compensation behavior occurs, combined with the voltage and current change data obtained for each battery cell, the load sharing rate, deviation ratio, and heat consumption change range of the compensated battery cell during the compensation process are determined;

[0008] Based on the characteristic offset values, historical abnormal trigger records, and working cycle numbers of battery cells, a damage probability model is constructed using the Bayesian estimation method to predict the damage probability of battery cells in the next several cycles and generate a failure probability curve for the battery cells in the next several cycles.

[0009] Through battery monitoring data, we can obtain damage probability curves, compensation behavior frequency and compensation amplitude data, mark masked fault cases, build masked fault identification models, identify the fault status of battery cells, and formulate fault warning and health maintenance strategies for battery cells.

[0010] Based on the compensation mode, failure probability trend, and location relationship of battery cells identified as masked faults, the fault level of the battery cells is assessed. A fault level index table is constructed using search tags containing location numbers, compensation path numbers, and failure risk weights. Battery maintenance strategies and early warning mechanisms are then developed.

[0011] Based on the real-time power changes, fault detection results and compensation behavior data of battery cells, the effectiveness of fault warning and maintenance strategies is evaluated, and the fault warning strategy and health maintenance strategy are updated.

[0012] Furthermore, the method of obtaining the operating parameters of each battery cell in the energy storage battery string through the battery management system, calculating the battery cell change rate between each window and the adjacent windows using a sliding window method, and identifying the energy output inconsistent period includes:

[0013] The battery management system obtains the operating parameters and timestamp information of each battery cell in the energy storage battery string, and stores the operating parameter data of the battery cell in a time series in the battery operation monitoring database. The operating parameters include instantaneous voltage, current, power output and temperature; the load distribution information of each battery cell is obtained by statistically analyzing the voltage, current, power output and temperature data; based on the preset window length and the number of sampling points, a sliding window method is used to compare the voltage, current, power output and temperature changes of each cell point by point, calculate the battery cell change rate between each window and the adjacent window, and calculate the deviation amplitude of its maximum, minimum and average change rate; the historical steady-state operation data of the energy storage battery is obtained through the battery operation monitoring database, and the Euclidean distance algorithm is used to calculate the difference between the battery cell change rate and the change rate in the historical steady-state operation data of the energy storage battery; if there is a period where the difference value is greater than the preset difference value threshold, the period is judged to be an energy output inconsistent period.

[0014] Furthermore, the method of predicting battery cell power fluctuations during a period of inconsistent energy output based on power changes, string load fluctuations, and temperature distribution of the battery cells under normal operating conditions, and identifying whether adjacent cell power compensation behavior occurs during the period of inconsistent energy output, includes:

[0015] The total load fluctuation information of the string is obtained through the battery management system, and combined with the ambient temperature sensor data, the power change of each battery cell, the total load fluctuation data of the string and the system temperature distribution data are obtained; the power change, string load fluctuation and temperature distribution of the battery cell under normal operating conditions are obtained through the battery operation monitoring database, and the long short-term memory network is used for model training to predict the power fluctuation of the battery cell in several future time periods; based on the power change of the battery cell during the energy output inconsistency period, the total load fluctuation data of the string and the system temperature distribution data, the trained long short-term memory network model is used to predict the power fluctuation of the battery cell during the energy output inconsistency period; the mean square error is used as the evaluation standard to calculate the deviation between the predicted battery cell power value and the actual observed value during the energy output inconsistency period; if the deviation value exceeds the set deviation value threshold, the period is marked as an abnormal period; the power data of the battery cell during the abnormal period is obtained, and the compensation link coupling coefficient formula is used Calculate the compensation link coupling coefficient between the battery cell and the adjacent battery cell, where CLC ij is the compensation link coupling coefficient between battery cell i and battery cell j, P i (t) is the power value of battery cell i at time t, is the average power value of battery cell i, P j (t) is the power value of battery cell j at time t, is the average power value of battery cell j, T is the length of the time series, indicating the number of data sampling points; if the absolute value of the compensation link coupling coefficient between a battery cell and its adjacent battery cells is greater than the preset coefficient threshold, it is determined that adjacent battery cell power compensation behavior exists between the battery cell and its adjacent battery cells; by identifying the compensation link, the power compensation relationship between different battery cells is determined, and the impact range is determined based on the occurrence frequency and duration of the adjacent battery cell power compensation behavior.

[0016] Furthermore, the power distribution data of the total load of the energy storage battery string during the period when the adjacent battery power compensation behavior occurs is combined with the voltage and current change data obtained for each battery cell to determine the load sharing rate, deviation ratio, and heat consumption change range of the compensated battery cell during the compensation process, including:

[0017] The power distribution data of the total load of the energy storage battery string is obtained according to the battery management system. The voltage and current change data of each battery cell during the period when the power compensation behavior of the adjacent cells occurs are combined to calculate the load sharing rate of each battery cell. The load sharing rate is the ratio of the power change of each battery cell to the total power change of the string. The deviation ratio is calculated by comparing the power output of the battery cell under normal operating conditions and during the compensation period. The deviation ratio is the relative change in power output. The load sharing rate of the compensated battery cell during the compensation process is determined by calculating the maximum, minimum, average and change rate of the battery cell power change, and its deviation ratio is calculated to obtain the degree of deviation in each time period. In combination with the temperature sensor data, the heat consumption change in the period is obtained, and the heat change amplitude of each cell is recorded.

[0018] Furthermore, the damage probability model is constructed using the Bayesian estimation method based on the characteristic offset value, historical abnormal trigger record and working cycle number of the battery cell to predict the damage probability of the battery cell in the next several cycles, and generate the failure probability curve of the battery cell in the next several cycles, including:

[0019] A battery management system is used to obtain the offset values ​​of various performance indicators of battery cells during operation, including the amplitude of voltage, current and heat changes. Historical abnormality trigger records are obtained in combination with system logs, recording the time, type and corresponding working conditions of each abnormality triggering, the number of working cycles of the battery cells, and the number of each charge and discharge cycle. A damage probability model is constructed based on the characteristic offset values, historical abnormality triggering records and working cycle numbers of the battery cells. The degree of deviation is used as the prior distribution, and the prior distribution and the posterior distribution are updated through the Bayesian theorem to obtain the damage probability distribution of the battery cells under different working conditions. The Monte Carlo sampling method is used for multiple sampling to randomly extract characteristic offset values, working cycle numbers and historical abnormality triggering records to obtain the damage probability of the battery cells in the future cycles. The damage probability is updated after each sampling through the damage probability model to obtain the failure probability of the battery cells being damaged in the future cycles, and a failure probability curve of the battery cells in the future cycles is generated.

[0020] Furthermore, the battery monitoring data is used to obtain damage probability curves, compensation behavior frequency and compensation amplitude data, mark masked fault cases, build a masked fault identification model, identify the fault status of battery cells, and formulate fault warning and health maintenance strategies for battery cells, including:

[0021] Through battery monitoring data, the compensation behavior frequency data is obtained, the number and duration of compensation behavior of each cell are recorded, and the compensation amplitude information is obtained. The compensation amplitude is the amplitude of the power change of the battery cell during the compensation behavior, and is stored in the battery operation monitoring database; the battery operation monitoring database is used to obtain historical damage probability curves, compensation behavior frequency and compensation amplitude data, and the corresponding masked fault cases in the data are marked. The decision tree algorithm is used for model training to construct a masked fault identification model to identify the fault status of the battery cell, which includes masked faults and no faults; based on the fault status identification results, a fault warning and health maintenance strategy for the battery cell is formulated, including arranging detailed inspections or early warning system monitoring for battery cells that are determined to have masked faults, and regular health assessments and performance inspections for battery cells without faults.

[0022] Furthermore, the method evaluates the fault level of the battery cell based on the compensation mode, fault probability trend, and location relationship of the battery cell determined to have a masked fault, constructs a fault level index table based on the search tags including the location number, compensation path number, and failure risk weight, and formulates a battery maintenance strategy and early warning mechanism, including:

[0023] Through the battery management system, the compensation mode, fault probability trend, and position relationship of the battery cells that are judged to be masked faults are obtained, and a retrieval tag is assigned to each single cell. Each retrieval tag contains a position number, a compensation path number, and a failure risk weight. The compensation mode includes whether the cell has compensation behavior and the compensation amplitude. The fault probability trend includes the damage probability distribution of the cell under different working conditions. The position relationship includes the physical position of the cell in the battery pack. The position number indicates the specific position of the cell in the battery pack. The compensation path number indicates the path of the compensation behavior. The failure risk weight indicates the failure probability of the battery cell in the future operation. Patterns, failure probability trends, and location relationship data are used to train the model using a decision tree algorithm to build a battery cell fault level assessment model to evaluate the fault level of the battery cell, which includes severe, moderate, and minor. Based on the fault level of the battery cell and combined with retrieval tags, a number is generated for each cell, and the location, compensation path, and failure risk weight of the battery cell with masked faults are determined to build a fault level index table. According to the fault level, location, and compensation path of the battery cell, a battery maintenance strategy and early warning mechanism are formulated, battery cells with severe fault levels are given priority, and corresponding inspection and maintenance cycles are formulated for each battery cell.

[0024] Furthermore, the effectiveness of the fault warning and maintenance strategies is evaluated based on the real-time power changes of the battery cells, the fault detection results, and the compensation behavior data, and the fault warning and health maintenance strategies are updated, including:

[0025] Real-time feedback data is obtained through the battery management system, including power changes of battery cells, fault detection results, compensation behavior data, and fault level information; by comparing the operating data of battery cells after fault processing with the expected status, the effectiveness of the current fault warning and health maintenance strategy is evaluated. If the effectiveness is lower than the preset effect requirement, the warning threshold and monitoring frequency are adjusted, and the fault warning and health maintenance strategy of the battery cell is updated until the effectiveness is higher than the preset effect requirement; based on the updated fault warning and health maintenance strategy, a new battery cell health assessment and maintenance plan is implemented, and feedback data is obtained for continuous optimization.

[0026] The second embodiment of the present invention provides a big data-based energy storage battery fault monitoring and management system, which mainly includes:

[0027] The cell energy output analysis module is used to obtain the operating parameters of each cell in the energy storage battery string through the battery management system, and uses the sliding window method to calculate the cell change rate between each window and the adjacent windows to identify the period of inconsistent energy output;

[0028] The adjacent cell power compensation behavior identification model is used to predict the battery cell power fluctuation during the energy output inconsistency period based on the power changes, string load fluctuations and temperature distribution of the battery cell under normal operating conditions, and to identify whether adjacent cell power compensation behavior exists during the energy output inconsistency period;

[0029] The battery cell power analysis module is used to determine the load sharing rate, deviation ratio, and heat consumption change range of the compensated battery cell during the compensation process based on the power distribution data of the total load of the energy storage battery string during the period when the adjacent battery cell power compensation behavior occurs, combined with the voltage and current change data obtained for each battery cell;

[0030] The failure probability prediction module is used to build a damage probability model based on the characteristic offset value of the battery cell, historical abnormal trigger records and the number of working cycles using the Bayesian estimation method to predict the damage probability of the battery cell in the next few cycles and generate the failure probability curve of the battery cell in the next few cycles;

[0031] The masked fault identification module is used to obtain damage probability curves, compensation behavior frequency and compensation amplitude data through battery monitoring data, mark masked fault cases, build a masked fault identification model, identify the fault status of battery cells, and formulate fault warning and health maintenance strategies for battery cells;

[0032] The fault level assessment module is used to evaluate the fault level of battery cells based on the compensation mode, fault probability trend, and location relationship of battery cells that are determined to have concealed faults. It also constructs a fault level index table based on search tags including location number, compensation path number, and failure risk weight, and formulates battery maintenance strategies and early warning mechanisms.

[0033] The health maintenance strategy optimization module is used to evaluate the effectiveness of fault warning and health maintenance strategies based on the real-time power changes of battery cells, fault detection results, and compensation behavior data, and to update the fault warning and health maintenance strategies.

[0034] The technical solution provided by the embodiment of the present invention may have the following beneficial effects:

[0035] This invention provides a method and system for monitoring and managing energy storage battery faults based on big data. This method uses a battery management system to obtain real-time operating parameters of battery cells and employs a sliding window method to calculate the rate of change of battery cells, accurately identifying periods of inconsistent energy output. By combining the power changes of battery cells, string load fluctuations, and temperature distribution, this method can effectively predict and identify power fluctuations and compensation behaviors during periods of inconsistent energy output, further accurately determining the occurrence and impact of compensation behaviors. This method utilizes battery cell voltage and current change data and power compensation behaviors to quantify load sharing ratios, deviation ratios, and heat loss changes, providing a detailed battery health assessment. By constructing a damage probability model and combining it with historical abnormality records and operating cycle counts, this method can predict battery cell damage risks in advance and generate a failure probability curve, providing a basis for maintenance decision-making. Based on compensation behavior frequency, damage probability data, and compensation amplitude, this method can identify masked faults, promptly detect potential problems, and issue early warnings. This method optimizes battery maintenance strategies and early warning mechanisms based on the fault level, location, and failure risk weight of each battery cell, ensuring that high-risk batteries receive priority attention and improving the stability and safety of energy storage battery packs. The energy storage battery fault monitoring and management method and system based on big data provided by the present invention can achieve precise monitoring and intelligent management of energy storage battery packs, significantly improve the accuracy and timeliness of battery fault warnings, and ensure the efficient and safe operation of the energy storage system. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 This is a flow chart of a method for monitoring and managing energy storage battery faults based on big data according to the present invention;

[0037] Figure 2 Schematic diagram of a method for monitoring and managing energy storage battery faults based on big data according to the present invention;

[0038] Figure 3 This is a schematic diagram of a big data-based energy storage battery fault monitoring and management system of the present invention. DETAILED DESCRIPTION

[0039] In order to make the objectives, technical solutions and advantages of the present invention more clear, the present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0040] like Figure 1-2 In this embodiment, a method and system for monitoring and managing energy storage battery faults based on big data may specifically include:

[0041] Step S101: obtaining the operating parameters of each battery cell in the energy storage battery string through the battery management system, calculating the battery cell change rate between each window and the adjacent windows using a sliding window method, and identifying the energy output inconsistency period.

[0042] The battery management system obtains the operating parameters and timestamp information of each battery cell in the energy storage battery string, and stores the operating parameter data of the battery cell in a time series in the battery operation monitoring database. The operating parameters include instantaneous voltage, current, power output and temperature. By statistically analyzing the voltage, current, power output and temperature data, the load distribution information of each battery cell is obtained. According to the preset window length and the number of sampling points, the sliding window method is used to compare the voltage, current, power output and temperature changes of each cell point by point, calculate the battery cell change rate between each window and the adjacent window, and calculate the deviation amplitude of the maximum, minimum and average change rate. The historical steady-state operation data of the energy storage battery is obtained through the battery operation monitoring database, and the Euclidean distance algorithm is used to calculate the difference between the battery cell change rate and the change rate in the historical steady-state operation data of the energy storage battery. If there is a period where the difference value is greater than the preset difference value threshold, the period is judged to be a period of inconsistent energy output.

[0043] For example, during the use of a battery pack, the operating parameters of the battery cells, including voltage, current, power output, and temperature, are regularly collected and statistically analyzed to obtain load distribution information. For example, at a certain moment, the voltage of battery cell A is 3.7V, the current is 5A, the power output is 18.5W, and the temperature is 25°C. These data are stored in a time series in the battery operation monitoring database and can then be processed. Using a sliding window method, an appropriate window length, such as 30 minutes, and the number of sampling points are set, such as collecting data every 5 minutes, and calculating the instantaneous rate of change of the battery cell voltage and current. If, within the first time window, the voltage of battery cell A changes from 3.7V to 3.8V, the current changes from 5A to 5.2A, the power output increases from 18.5W to 19W, and the temperature rises from 25°C to 26°C, the data is analyzed. Based on this data, the battery cell change rate within this time window is calculated: the voltage change rate is (3.8-3.7) / 3.7≈0.027, the current change rate is (5.2-5) / 5=0.04, the power output change rate is (19-18.5) / 18.5≈0.027, and the temperature change rate is (26-25) / 25=0.04. The sliding window continues to move forward, calculating the battery cell voltage, current, power output, and temperature change rates in adjacent time windows, and calculating the maximum, minimum, and average change rates for each window. If the change rate of a particular period deviates significantly from the preceding and following windows, for example, if the battery cell current change rate suddenly rises to 0.15 during a period, exceeding the 0.04 deviation from the average change rate, then this period is considered to be a period of battery cell anomaly. To determine whether a period of inconsistent energy output occurs, these calculated rates of change are compared with the energy storage battery's historical steady-state operating data. If, in the historical steady-state data, the current rate of change of a battery cell is typically between 0.02 and 0.05, and the temperature rate of change is between 0.01 and 0.03, then if during a certain period, the current rate of change of battery cell A is 0.12, exceeding the range of the historical steady-state data, and the rate of change difference during that period is greater than a preset threshold of 0.1, then that period will be marked as inconsistent energy output.

[0044] Step S102 , predicting the battery cell power fluctuation during the energy output inconsistency period based on the power variation, string load fluctuation, and temperature distribution of the battery cell under normal operating conditions, and identifying whether there is adjacent cell power compensation behavior during the energy output inconsistency period.

[0045] The total load fluctuation information of the energy storage battery string is obtained through the battery management system, and combined with the ambient temperature sensor data, the power change of each battery cell, the total load fluctuation data of the string, and the system temperature distribution data are obtained. Through the battery operation monitoring database, the power change, string load fluctuation and temperature distribution of the battery cell under normal operation are obtained, and the long short-term memory network is used for model training to predict the power fluctuation of the battery cell in several future time periods. According to the power change of the battery cell during the period of inconsistent energy output, the total load fluctuation data of the string, and the system temperature distribution data, the trained long short-term memory network model is used to predict the power fluctuation of the battery cell during the period of inconsistent energy output. The mean square error is used as the evaluation criterion to calculate the deviation between the predicted battery cell power value and the actual observed value during the period of inconsistent energy output. If the deviation value exceeds the set deviation value threshold, the period is marked as an abnormal period. Obtain the power data of the battery cell during the abnormal period, and use the compensation link coupling coefficient formula Calculate the compensation link coupling coefficient between the battery cell and the adjacent battery cell, where CLC ij is the compensation link coupling coefficient between battery cell i and battery cell j, P i (t) is the power value of battery cell i at time t, is the average power value of battery cell i, P j (t) is the power value of battery cell j at time t, is the average power value of battery cell j, and T is the length of the time series, representing the number of data sampling points. If the absolute value of the compensation link coupling coefficient between a battery cell and its neighboring battery cells exceeds the preset coefficient threshold, it is determined that neighboring cell power compensation occurs between the battery cell and its neighboring battery cells. By identifying the compensation link, the power compensation relationship between different battery cells is determined, and the impact range is determined based on the frequency and duration of neighboring cell power compensation.

[0046] For example, consider an energy storage battery pack containing 10 cells. The power data for each cell is recorded at one-minute intervals. The power of cell A experiences significant fluctuations within a certain period. For example, within a 5-minute window, at 1 minute, cell A's power is 50W, and the battery temperature is 25°C. At 2 minutes, cell A's power is 55W, and the battery temperature is 26°C. At 3 minutes, cell A's power is 53W, and the battery temperature is 27°C. At 4 minutes, cell A's power is 60W, and the battery temperature is 28°C. At 5 minutes, cell A's power is 58W, and the battery temperature is 29°C. The string load fluctuates from 200W to 220W. Based on the power variation data of the cells under normal operating conditions, the string load fluctuation, and the temperature distribution, a long short-term memory network is used to train a model to predict the power fluctuations of the cells over several future time periods. If the power fluctuation of battery cell A in a certain period deviates significantly from the value predicted by the model, this period will be marked as an energy output inconsistent period. For example, if the long short-term memory network model predicts that the power of battery cell A should be around 55W, and the actual monitored power is 60W, the deviation is 5W, and the set deviation threshold is 4W, then this period will be marked as an abnormal period. By compensating the link coupling coefficient formula Calculate the compensation link coupling coefficient between battery cell A and adjacent battery cell B, where CLC ij is the compensation link coupling coefficient between battery cell i and battery cell j, P i (t) is the power value of battery cell i at time t, is the average power value of battery cell i, P j (t) is the power value of battery cell j at time t, is the average power value of battery cell j, T is the length of the time series, and represents the number of sampling points of the data. If the power of battery cell A at a certain moment is 60W and the average power is 57W, the power of battery cell B at the same moment is 58W and the average power is 55W, and the number of sampling points T is 5, then the CLC is calculated by the compensation link coupling coefficient formula AB =0.85, if the preset coefficient threshold is 0.8, due to CLC AB If the value is >0.8, it is determined that adjacent cell power compensation behavior exists between battery cells A and B. By analyzing multiple compensation links, the power compensation relationship between different battery cells can be determined, and the impact range can be assessed based on the frequency and duration of adjacent cell power compensation behavior. For example, if the compensation link behavior between battery cells A and B occurs frequently and each time lasts for a long time, this may indicate that battery cells A and B are dependent on each other during the power compensation process, which may affect the stability and health of the entire battery pack.

[0047] Step S103 , based on the power distribution data of the total load of the energy storage battery string during the period when the adjacent battery cell power compensation behavior occurs, combined with the voltage and current change data obtained for each battery cell, the load sharing rate, deviation ratio, and heat consumption change range of the compensated battery cell during the compensation process are determined.

[0048] The power distribution data of the total load of the energy storage battery string is obtained according to the battery management system. The voltage and current change data of each battery cell during the period when the power compensation behavior of the adjacent cells occurs are combined to calculate the load sharing rate of each battery cell. The load sharing rate is the ratio of the power change of each single cell to the total power change of the string. By comparing the power output of the battery cell under normal operating conditions and during the compensation period, the deviation ratio is calculated. The deviation ratio is the relative change in power output. By calculating the maximum, minimum, average and change rate of the battery cell power change, the load sharing rate of the compensated battery cell during the compensation process is determined, and its deviation ratio is calculated to obtain the degree of deviation in each time period. Combined with the temperature sensor data, the heat consumption change within the period is obtained, and the heat change amplitude of each cell is recorded.

[0049] For example, during a certain period of time, the total power change of the energy storage battery group is 200W. If the total load power of the string increases from 1000W to 1200W during this period, a change of 200W occurs. At the same time, the power change of battery cell C during this period is 20W, the power change of battery cell D is 30W, the power change of battery cell E is 50W, and the power change of battery cell F is 100W. Based on these data, the load sharing rate of each battery cell can be calculated. The load sharing rate is the ratio between the power change of each battery cell and the total power change of the string. For battery cell C, the load sharing rate can be calculated as C's load sharing rate = battery cell C's power change / total string power change = 20W / 200W = 0.1, which means that battery cell C is responsible for 10% of the total power change of the string. Using the same method, the load sharing rates of battery cells D, E, and F are calculated to be 0.15, 0.25, and 0.5, respectively. The deviation ratio is the relative change in power output, indicating the change in power output of each battery cell during the compensation period relative to normal operation. For example, if battery cell C's power during normal operation is 50W, and during the compensation period, its power reaches 70W, then the deviation ratio of battery cell C = (compensation period power - normal operation power) / normal operation power = 20W / 50W = 0.4, indicating that battery cell C's power output has increased by 40% relative to normal operation. If the power of battery cells D, E, and F changes by 5W, 10W, and 15W, respectively, during the compensation period, while their normal operation powers are 60W, 80W, and 100W, respectively, then the deviation ratios for battery cells D, E, and F are 0.0833, 0.125, and 0.15, respectively. These deviation ratios can be used to analyze whether the power changes of each battery cell meet expectations and, therefore, determine whether abnormal load sharing has occurred. If the power of battery cell C varies from 50W to 70W, its maximum power change is 70W, its minimum power change is 50W, and its average power change is 60W. The rate of change for C = (maximum power - minimum power) / minimum power = 20W / 50W = 0.4. This data can be used to further assess the load sharing of the battery cells during the compensation process. Large power changes or excessively high deviation ratios may indicate overcompensation, leading to uneven load sharing and potentially requiring further adjustment. Combined with temperature sensor data, the heat dissipation changes during this period are captured and the magnitude of the heat dissipation change for each battery cell is recorded. For example, if the temperature of battery cell C rises from 25°C to 30°C during the compensation period, the heat dissipation change for battery cell C is 5°C. If the temperature changes for battery cell D are 3°C, E is 4°C, and F is 6°C, these heat dissipation changes can be used to assess the thermal load of the battery cells and determine their proper operating status.

[0050] Step S104 , based on the characteristic offset value of the battery cell, historical abnormal trigger records and the number of working cycles, a damage probability model is constructed using the Bayesian estimation method to predict the damage probability of the battery cell in the next several cycles and generate a failure probability curve of the battery cell in the next several cycles.

[0051] The battery management system uses a battery management system to obtain the deviation values ​​of various battery cell performance indicators during operation, including the amplitude of voltage, current, and heat fluctuations. Historical anomaly trigger records are collected through system logs, recording the time, type, and corresponding operating conditions of each anomaly trigger. The number of operating cycles of the battery cell is also obtained, recording the number of charge and discharge cycles. Based on the characteristic deviation values, historical anomaly trigger records, and operating cycle count of the battery cell, a damage probability model is constructed using the Bayesian estimation method. Using the degree of deviation as the prior distribution, the prior and posterior distributions are updated using the Bayesian theorem to obtain the damage probability distribution of the battery cell under different operating conditions. Monte Carlo sampling is used to perform multiple samplings, randomly extracting characteristic deviation values, operating cycle counts, and historical anomaly trigger records to determine the damage probability of the battery cell over several future cycles. The damage probability model updates the damage probability after each sampling step, determining the failure probability of the battery cell in the future cycle and generating a failure probability curve for the battery cell over several future cycles.

[0052] For example, the operation of a single energy storage battery cell is being monitored and various performance data, including voltage, current, and thermal fluctuations, is being collected. The battery management system (BMS) reports that over the past week, the voltage of the battery cell fluctuated from 3.6V to 3.9V, the current from 5A to 7A, and the battery temperature increased from 25°C to 30°C. This data is used to determine the performance excursion of the battery cell during this period. System log data is then used to determine whether the battery cell has experienced any abnormal triggering events during past usage. For example, historical system logs indicate that the battery cell experienced three abnormal events over the past six months. The first occurred when the battery temperature exceeded the set threshold of 40°C, the second occurred when the battery current suddenly surged to 8A, and the third occurred when the battery cell voltage dropped below 3.4V. The specific trigger time and event type for each abnormal event are recorded. The number of charge and discharge cycles of the battery cell is also recorded, including 250 cycles in the past year. Each charge and discharge cycle contributes to battery aging and increases the risk of damage. The system obtained the number of battery cell operating cycles and found that the battery had reached 50% of its expected lifespan, indicating that the battery's damage probability gradually increases with increasing usage. A damage probability model was constructed using the Bayesian estimation method, taking the battery cell's performance offset, the number of operating cycles, and historical anomaly triggering records as inputs. Using the Bayesian theorem, a prior distribution was constructed to represent the battery's damage probability under normal conditions. The performance offset was used as the prior distribution, and the posterior distribution was updated based on the battery's historical anomaly triggering records and the number of operating cycles. Ultimately, the damage probability distribution of the battery cell under different operating conditions was obtained. Multiple sampling was performed using the Monte Carlo method, randomly extracting data such as the battery cell's characteristic offset, the number of operating cycles, and historical anomaly triggering records to generate multiple samples. The damage probability model was then updated after each sampling period, resulting in a failure probability curve for the battery cell over the next several cycles. Simulations revealed that the damage probability of the battery cell over the next three cycles was 0.15, 0.18, and 0.20, respectively, indicating that the damage probability of the battery cell gradually increases over time. This analysis method based on Bayesian estimation and Monte Carlo sampling can accurately predict the failure probability of battery cells and generate a failure probability curve for battery cells in future cycles. This curve can identify potential problems with battery cells in advance and provide data support for subsequent maintenance and replacement.

[0053] Step S105: Obtain damage probability curve, compensation behavior frequency and compensation amplitude data through battery monitoring data, mark masked fault cases, build a masked fault identification model, identify the fault status of battery cells, and formulate fault warning and health maintenance strategies for battery cells.

[0054] Through battery monitoring data, compensation behavior frequency data is obtained, the number and duration of compensation behaviors for each battery cell are recorded, and compensation amplitude information is obtained. The compensation amplitude is the amplitude of the power change of the battery cell during the compensation behavior, and is stored in the battery operation monitoring database. Using the battery operation monitoring database, historical damage probability curves, compensation behavior frequency, and compensation amplitude data are obtained, and corresponding masked fault cases in the data are annotated. A decision tree algorithm is used for model training to construct a masked fault identification model to identify the fault status of battery cells, which includes masked faults and no faults. Based on the fault status identification results, fault warning and health maintenance strategies for battery cells are formulated, including arranging detailed inspections or early warning system monitoring for battery cells determined to have masked faults, and regular health assessments and performance inspections for battery cells without faults.

[0055] For example, a battery storage system is monitoring an energy storage battery pack. The battery management system collects operational data about the battery cells, including information such as power changes, compensation frequency, and compensation amplitude for each cell. Through real-time monitoring, we recorded that battery cell A had undergone multiple compensation events over the past three months. During this period, battery cell A underwent 15 compensation events, 10 of which lasted longer than 5 minutes, while the remaining 5 were shorter, approximately 2 minutes. The amplitude of each compensation event, i.e., the magnitude of the battery cell power change, varied under different circumstances, with the maximum compensation amplitude being 25W and the minimum being 5W. This data is stored in real time in the battery operation monitoring database. From the database, we obtained a historical damage probability curve for battery cell A, which reflects the health status of battery cell A over time. We found that over the past two months, the damage probability of battery cell A increased from 10% to 18%, indicating that the damage risk of this cell increases with operating time. Furthermore, the increase in compensation frequency and amplitude also indirectly indicates that the cell may have some degree of damage. A decision tree algorithm is trained to construct a masked fault identification model, combining compensation behavior frequency, compensation amplitude, and historical damage probability data. This model can identify battery cells with frequent and large compensation behaviors that may harbor faults that go undetected by conventional monitoring methods. For example, in the case of battery cell A, the decision tree model identifies a high frequency of compensation behaviors and a large power fluctuation amplitude during each compensation behavior. Combined with the upward trend in its damage probability curve, this predicts that this battery cell is at risk of a masked fault. Based on this identification result, a detailed fault warning and health maintenance strategy is developed. For battery cell A identified as having a masked fault, the system automatically triggers an alert and schedules a detailed inspection of the battery. This inspection includes examining the battery's internal chemical reactions, temperature fluctuations, and power output to ensure timely diagnosis of the battery's health. Simultaneously, the battery management system strengthens real-time monitoring of this battery cell to ensure early detection of potential faults during subsequent operation. For battery cells that are judged to be fault-free, such as battery cell B, regular health assessments and performance checks are recommended. For example, battery cell B has not experienced abnormal compensation behavior in the past three months, its damage probability curve remains stable, and the compensation amplitude is small. Therefore, it is recommended to conduct a routine inspection of the battery every two months to ensure its normal performance and to detect potential problems in a timely manner.

[0056] Step S106, based on the compensation mode, fault probability trend, and position relationship of the battery cell determined to be a masked fault, the fault level of the battery cell is evaluated, and a fault level index table is constructed in combination with the search tags including the position number, compensation path number, and failure risk weight, and a battery maintenance strategy and early warning mechanism are formulated.

[0057] The battery management system (BMS) obtains the compensation mode, failure probability trend, and location relationship of battery cells identified as masked faults. Each cell is assigned a search tag, each containing a location number, compensation path number, and failure risk weight. The compensation mode includes whether the cell exhibits compensation behavior and the compensation magnitude. The failure probability trend includes the cell's damage probability distribution under different operating conditions. The location relationship includes the cell's physical location within the battery pack. The location number indicates the specific location of the cell within the pack, the compensation path number indicates the path of compensation behavior, and the failure risk weight indicates the probability of failure of the cell in future operation. Based on the compensation mode, failure probability trend, and location relationship data, a decision tree algorithm is used for model training to construct a battery cell failure level assessment model. This model assesses the failure level of each cell, which is categorized as severe, moderate, or minor. Based on the cell failure level and combined with the search tag, a number is generated for each cell. The location, compensation path, and failure risk weight of the cell with masked faults are determined, and a failure level index table is constructed. Based on the fault level, location and compensation path of the battery cells, a battery maintenance strategy and early warning mechanism are formulated, battery cells with serious fault levels are given priority, and corresponding inspection and maintenance cycles are formulated for each battery cell.

[0058] For example, a large energy storage battery pack is being monitored, which contains multiple battery cells. The battery management system collects the operating data of these battery cells through real-time monitoring, among which battery cell H is determined to have a concealed fault. According to the records of the battery management system, this battery cell has undergone multiple compensation behaviors in the past three months, each compensation amplitude is about 20W, and the duration ranges from 5 minutes to 10 minutes. The failure probability trend of battery cell H shows that its damage probability under different working conditions gradually increases over time. For example, when the load increases, the damage probability of battery cell H increases from 10% to 30% within a week. These changes indicate that battery cell H may have potential failure risks, but because the compensation behavior masks these problems, the fault does not appear directly. To further manage these cells, each cell is assigned a search tag. The search tag contains multiple pieces of information about the cell, such as its location number, compensation path number, and failure risk weight. For example, the search tag for cell H includes the location number, indicating its specific location in the battery pack. For example, it is located in the 5th column, 3rd row of the battery pack. Its search tag also includes the compensation path number, indicating that its compensation path number is path-03, indicating that it is in an area with frequent compensation. Its search tag also includes a failure risk weight, indicating that its failure risk weight is 0.35, indicating a relatively high probability of failure in future operation. Combining compensation patterns, failure probability trends, and location relationship data, a decision tree algorithm is used to train a model to identify the failure level of each cell. The result is that the failure level of cell H is assessed as severe because its compensation behavior is frequent and large, with a significant upward trend in failure probability and a high failure risk weight. A number is generated for each cell, and its location, compensation path, and failure risk weight are determined. Finally, a failure level index table is generated. The index record of battery cell H includes the number 00123, the position number 5-3, the compensation path number path-03, the fault level is severe, and the failure risk weight is 0.35. According to the index table, maintenance strategies and early warning mechanisms can be formulated according to the fault level, location and compensation path, including for battery cell H, because its fault level is severe, the system will give priority to its inspection and conduct a detailed battery health assessment and maintenance within the next month. In addition, the system will also set a shorter maintenance cycle for battery cell H, such as checking the battery performance and fault warning data every two weeks. For other battery cells, such as battery cells I and J, their fault levels may be assessed as medium or mild, which means that their failure risks are relatively low. For example, the compensation amplitude of battery cell I is small and the failure probability is low. Its fault level is medium and the failure risk weight is 0.1.In this case, battery cell 1 may be scheduled for regular inspections, such as a health assessment every three months.

[0059] Step S107 , evaluating the effectiveness of the fault warning and maintenance strategies based on the real-time power changes of the battery cells, the fault detection results, and the compensation behavior data, and updating the fault warning strategies and health maintenance strategies.

[0060] The battery management system obtains real-time feedback data, including battery cell power changes, fault detection results, compensation behavior data, and fault level information. By comparing the operating data of the battery cells after fault resolution with the expected status, the effectiveness of the current fault warning and health maintenance strategies is evaluated. If the effectiveness is lower than the preset effectiveness requirements, the warning threshold and monitoring frequency are adjusted, and the battery cell fault warning and health maintenance strategies are updated until the effectiveness exceeds the preset effectiveness requirements. Based on the updated fault warning and health maintenance strategies, new battery cell health assessments and maintenance plans are implemented, and feedback data is collected for continuous optimization.

[0061] For example, the battery management system collects real-time data on battery cell power variations, fault detection results, compensation behavior data, and fault level information. Using this data, the system effectively monitors the battery's health and adjusts fault warning and health maintenance strategies based on actual operating conditions. For example, if battery cell A has recently undergone three compensation events, each with a compensation amplitude of approximately 15W and durations of 6, 8, and 10 minutes, the fault detection results for that battery cell indicate an increase in its damage probability from 10% to 20%. Based on this data, battery cell A's fault level is determined to be moderate, and a two-week health assessment and maintenance plan is scheduled. This plan includes checking parameters such as battery cell temperature, voltage, and power output to ensure proper operation. After these two weeks of maintenance, operational data for battery cell A was collected again. The new data indicated that battery cell A experienced significant power variations and frequent compensation events, and the damage probability remained around 20%. Based on this feedback, the battery management system determined that the current fault warning and health maintenance strategies were failing to effectively reduce the damage risk of battery cell A. Therefore, when evaluating the effectiveness of the current strategy, it was found to fall short of the pre-determined performance requirements and to have failed to significantly improve the health of the battery cell. Based on this, the fault warning threshold and monitoring frequency needed to be adjusted. First, the fault warning threshold was lowered so that a warning would be triggered when the power fluctuation of a battery cell exceeded 12W. The monitoring frequency was increased to every two days to more promptly detect potential fault signs. The new warning strategy means that battery cell A will be monitored more frequently during large power fluctuations, allowing for timely maintenance or repairs. After implementing the updated warning and maintenance strategy, the system began to reacquire feedback data. Over the next three days, the power fluctuation of battery cell A decreased, and the frequency and magnitude of compensation actions also decreased significantly. Simultaneously, the damage probability of battery cell A gradually decreased, returning to around 10%. After this period of feedback, the system evaluated the effectiveness of the updated fault warning and health maintenance strategy and found that it had met the preset performance requirements. To ensure the continuous optimization of the overall health of the battery pack, the battery management system continued to implement the new health assessment and maintenance plan. Monitoring data from each battery cell was continuously fed back and adjusted to further optimize the fault warning and health maintenance strategy. This continuous cycle ensures that the battery cells remain healthy throughout their use and enables the early detection and repair of potential faults, thereby improving the efficiency and safety of the battery pack.

[0062] like Figure 3 In this embodiment, a big data-based energy storage battery fault monitoring and management system may specifically include:

[0063] The single cell energy output analysis module is used to obtain the operating parameters of each battery cell in the energy storage battery string through the battery management system, and uses the sliding window method to calculate the battery cell change rate between each window and the adjacent windows to identify the period of inconsistent energy output.

[0064] The adjacent cell power compensation behavior identification module is used to predict the battery cell power fluctuations during the energy output inconsistency period based on the power changes, string load fluctuations and temperature distribution of the battery cell under normal operating conditions, and to identify whether there is adjacent cell power compensation behavior during the energy output inconsistency period.

[0065] The battery cell power analysis module is used to determine the load sharing rate, deviation ratio, and heat consumption change range of the compensated battery cell during the compensation process based on the power distribution data of the total load of the energy storage battery string during the period when the adjacent battery cell power compensation behavior occurs, combined with the voltage and current change data obtained for each battery cell.

[0066] The failure probability prediction module is used to build a damage probability model based on the characteristic offset value of the battery cell, historical abnormal trigger records and the number of working cycles using the Bayesian estimation method to predict the damage probability of the battery cell in the next few cycles and generate the failure probability curve of the battery cell in the next few cycles.

[0067] The masked fault identification module is used to obtain damage probability curves, compensation behavior frequency and compensation amplitude data through battery monitoring data, mark masked fault cases, build a masked fault identification model, identify the fault status of battery cells, and formulate fault warning and health maintenance strategies for battery cells.

[0068] The fault level assessment module is used to evaluate the fault level of battery cells based on the compensation mode, fault probability trend, and position relationship of battery cells that are determined to have concealed faults. It also constructs a fault level index table based on the search tags containing the position number, compensation path number, and failure risk weight, and formulates battery maintenance strategies and early warning mechanisms.

[0069] The health maintenance strategy optimization module is used to evaluate the effectiveness of fault warning and health maintenance strategies based on the real-time power changes of battery cells, fault detection results, and compensation behavior data, and to update the fault warning and health maintenance strategies.

[0070] The above description is merely a preferred embodiment of the present application and an illustration of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in this application is not limited to the technical solutions formed by the specific combination of the above-mentioned technical features, but also encompasses other technical solutions formed by any combination of the above-mentioned technical features or their equivalents without departing from the concept of this application. For example, a technical solution formed by replacing the above-mentioned features with (but not limited to) technical features with similar functions disclosed in this application.

Claims

1. A method for monitoring and managing energy storage battery faults based on big data, characterized in that: The method comprises: The battery management system obtains the operating parameters of each battery cell in the energy storage battery string, and uses a sliding window method to calculate the battery cell change rate between each window and the adjacent windows to identify periods of inconsistent energy output; Based on the power changes, string load fluctuations, and temperature distribution of the battery cells under normal operating conditions, the power fluctuations of the battery cells during periods of inconsistent energy output are predicted, and whether there is power compensation behavior by adjacent cells during periods of inconsistent energy output. Based on the power distribution data of the total load of the energy storage battery string during the period when the adjacent battery power compensation behavior occurs, combined with the voltage and current change data obtained for each battery cell, the load sharing rate, deviation ratio, and heat consumption change range of the compensated battery cell during the compensation process are determined; Based on the characteristic offset values, historical abnormal trigger records, and working cycle numbers of battery cells, a damage probability model is constructed using the Bayesian estimation method to predict the damage probability of battery cells in the next several cycles and generate a failure probability curve for the battery cells in the next several cycles. Through battery monitoring data, we can obtain damage probability curves, compensation behavior frequency and compensation amplitude data, mark masked fault cases, build masked fault identification models, identify the fault status of battery cells, and formulate fault warning and health maintenance strategies for battery cells. Based on the compensation mode, failure probability trend, and location relationship of battery cells identified as masked faults, the fault level of the battery cells is assessed. A fault level index table is constructed using search tags containing location numbers, compensation path numbers, and failure risk weights. Battery maintenance strategies and early warning mechanisms are then developed. Based on the real-time power changes, fault detection results and compensation behavior data of battery cells, the effectiveness of fault warning and maintenance strategies is evaluated, and the fault warning strategy and health maintenance strategy are updated.

2. The method according to claim 1, wherein The method of obtaining the operating parameters of each battery cell in the energy storage battery string through the battery management system, calculating the battery cell change rate between each window and the adjacent windows using a sliding window method, and identifying the period of inconsistent energy output includes: The battery management system obtains the operating parameters and timestamp information of each battery cell in the energy storage battery string, and stores the operating parameter data of the battery cell in a time series in the battery operation monitoring database. The operating parameters include instantaneous voltage, current, power output and temperature; the load distribution information of each battery cell is obtained by statistically analyzing the voltage, current, power output and temperature data; based on the preset window length and the number of sampling points, a sliding window method is used to compare the voltage, current, power output and temperature changes of each cell point by point, calculate the battery cell change rate between each window and the adjacent window, and calculate the deviation amplitude of its maximum, minimum and average change rate; the historical steady-state operation data of the energy storage battery is obtained through the battery operation monitoring database, and the Euclidean distance algorithm is used to calculate the difference between the battery cell change rate and the change rate in the historical steady-state operation data of the energy storage battery; if there is a period where the difference value is greater than the preset difference value threshold, the period is judged to be an energy output inconsistent period.

3. The method according to claim 1, wherein The method predicts the power fluctuation of battery cells during the period of inconsistent energy output based on the power variation, string load fluctuation, and temperature distribution of the battery cells under normal operating conditions, and identifies whether there is power compensation behavior of adjacent cells during the period of inconsistent energy output, including: The battery management system obtains total string load fluctuation information and, combined with ambient temperature sensor data, obtains the power change of each battery cell, the total string load fluctuation data, and the system temperature distribution data. The battery operation monitoring database obtains the power change, string load fluctuation, and temperature distribution of the battery cells under normal operating conditions. A long-short-term memory network is used for model training to predict the power fluctuation of the battery cells in several future time periods. Based on the power change of the battery cells, the total string load fluctuation data, and the system temperature distribution data during the period of inconsistent energy output, the trained long-short-term memory network model is used to predict the power fluctuation of the battery cells during the period of inconsistent energy output. The mean square error is used as the evaluation criterion to calculate the deviation between the predicted battery cell power value during the period of inconsistent energy output and the actual observed value. If the deviation value exceeds the set deviation value threshold, the period is marked as an abnormal period; Obtain the power data of the battery cell during the abnormal period and use the compensation link coupling coefficient formula Calculate the compensation link coupling coefficient between the battery cell and the adjacent battery cell, where CLC ij is the compensation link coupling coefficient between battery cell i and battery cell j, P i (t) is the power value of battery cell i at time t, is the average power value of battery cell i, P j (t) is the power value of battery cell j at time t, is the average power value of battery cell j, T is the length of the time series, indicating the number of data sampling points; if the absolute value of the compensation link coupling coefficient between a battery cell and its adjacent battery cells is greater than the preset coefficient threshold, it is determined that adjacent battery cell power compensation behavior exists between the battery cell and its adjacent battery cells; by identifying the compensation link, the power compensation relationship between different battery cells is determined, and the impact range is determined based on the occurrence frequency and duration of the adjacent battery cell power compensation behavior.

4. The method according to claim 1, wherein The power distribution data of the total load of the energy storage battery string during the period when the adjacent battery power compensation behavior occurs is combined with the voltage and current change data obtained for each battery cell to determine the load sharing rate, deviation ratio, and heat consumption change range of the compensated battery cell during the compensation process, including: The power distribution data of the total load of the energy storage battery string is obtained according to the battery management system. The voltage and current change data of each battery cell during the period when the power compensation behavior of the adjacent cells occurs are combined to calculate the load sharing rate of each battery cell. The load sharing rate is the ratio of the power change of each battery cell to the total power change of the string. The deviation ratio is calculated by comparing the power output of the battery cell under normal operating conditions and during the compensation period. The deviation ratio is the relative change in power output. The load sharing rate of the compensated battery cell during the compensation process is determined by calculating the maximum, minimum, average and change rate of the battery cell power change, and its deviation ratio is calculated to obtain the degree of deviation in each time period. In combination with the temperature sensor data, the heat consumption change in the period is obtained, and the heat change amplitude of each cell is recorded.

5. The method according to claim 1, wherein The damage probability model is constructed using the Bayesian estimation method based on the characteristic offset value, historical abnormal trigger record and working cycle number of the battery cell to predict the damage probability of the battery cell in the next several cycles and generate the failure probability curve of the battery cell in the next several cycles, including: A battery management system is used to obtain the offset values ​​of various performance indicators of battery cells during operation, including the amplitude of voltage, current and heat changes. Historical abnormality trigger records are obtained in combination with system logs, recording the time, type and corresponding working conditions of each abnormality triggering, the number of working cycles of the battery cells, and the number of each charge and discharge cycle. A damage probability model is constructed based on the characteristic offset values, historical abnormality triggering records and working cycle numbers of the battery cells. The degree of deviation is used as the prior distribution, and the prior distribution and the posterior distribution are updated through the Bayesian theorem to obtain the damage probability distribution of the battery cells under different working conditions. The Monte Carlo sampling method is used for multiple sampling to randomly extract characteristic offset values, working cycle numbers and historical abnormality triggering records to obtain the damage probability of the battery cells in the future cycles. The damage probability is updated after each sampling through the damage probability model to obtain the failure probability of the battery cells being damaged in the future cycles, and a failure probability curve of the battery cells in the future cycles is generated.

6. The method according to claim 1, wherein The battery monitoring data is used to obtain damage probability curves, compensation behavior frequency and compensation amplitude data, mark masked fault cases, build a masked fault identification model, identify the fault status of battery cells, and formulate fault warning and health maintenance strategies for battery cells, including: Obtain compensation behavior frequency data through battery monitoring data, record the number and duration of compensation behavior for each cell, obtain compensation amplitude information, which is the amplitude of the battery cell power change during the compensation behavior, and store it in the battery operation monitoring database; Using the battery operation monitoring database, we obtain historical damage probability curves, compensation behavior frequency, and compensation amplitude data, and annotate the corresponding masked fault cases in the data. We use a decision tree algorithm for model training and build a masked fault recognition model to identify the fault status of battery cells, including masked faults and no faults. Based on the fault status identification results, formulate fault warning and health maintenance strategies for battery cells, including arranging detailed inspections or early warning system monitoring for battery cells that are determined to have concealed faults, and conducting regular health assessments and performance checks for battery cells without faults.

7. The method according to claim 1, wherein The method evaluates the fault level of the battery cell based on the compensation mode, fault probability trend, and location relationship of the battery cell determined to have a masked fault, constructs a fault level index table based on the search tags including the location number, compensation path number, and failure risk weight, and formulates a battery maintenance strategy and early warning mechanism, including: Through the battery management system, the compensation mode, fault probability trend, and position relationship of the battery cells that are determined to have masked faults are obtained, and a retrieval tag is assigned to each battery cell. Each retrieval tag contains a position number, a compensation path number, and a failure risk weight. The compensation mode includes whether the cell has compensation behavior and the compensation amplitude. The fault probability trend includes the damage probability distribution of the cell under different working conditions. The position relationship includes the physical position of the cell in the battery pack. The position number indicates the specific position of this type of cell in the battery pack. The compensation path number indicates the path of the compensation behavior. The failure risk weight indicates the failure probability of the battery cell in future operation. Based on the compensation mode, fault probability trend, and position relationship data, a decision tree algorithm is used for model training to construct a battery cell fault level assessment model to assess the fault level of the battery cell. The fault level includes severe, moderate, and minor. Based on the fault level of the battery cell and combined with the retrieval tag, a number is generated for each cell, and the position, compensation path, and failure risk weight of the battery cell with masked faults are determined to construct a fault level index table. Based on the fault level, location and compensation path of the battery cells, a battery maintenance strategy and early warning mechanism are formulated, battery cells with serious fault levels are given priority, and corresponding inspection and maintenance cycles are formulated for each battery cell.

8. The method according to claim 1, wherein The effectiveness of the fault warning and maintenance strategies is evaluated based on the real-time power changes, fault detection results, and compensation behavior data of the battery cells, and the fault warning and health maintenance strategies are updated, including: Real-time feedback data is obtained through the battery management system, including power changes of battery cells, fault detection results, compensation behavior data, and fault level information; by comparing the operating data of battery cells after fault processing with the expected status, the effectiveness of the current fault warning and health maintenance strategy is evaluated. If the effectiveness is lower than the preset effect requirement, the warning threshold and monitoring frequency are adjusted, and the fault warning and health maintenance strategy of the battery cell is updated until the effectiveness is higher than the preset effect requirement; based on the updated fault warning and health maintenance strategy, a new battery cell health assessment and maintenance plan is implemented, and feedback data is obtained for continuous optimization.

9. A big data-based energy storage battery fault monitoring and management system, which is implemented based on the big data-based energy storage battery fault monitoring and management method according to any one of claims 1 to 8, characterized in that: The system includes the following modules: The cell energy output analysis module is used to obtain the operating parameters of each cell in the energy storage battery string through the battery management system, and uses the sliding window method to calculate the cell change rate between each window and the adjacent windows to identify the period of inconsistent energy output; The adjacent cell power compensation behavior identification model is used to predict the battery cell power fluctuation during the energy output inconsistency period based on the power changes, string load fluctuations and temperature distribution of the battery cell under normal operating conditions, and to identify whether adjacent cell power compensation behavior exists during the energy output inconsistency period; The battery cell power analysis module is used to determine the load sharing rate, deviation ratio, and heat consumption change range of the compensated battery cell during the compensation process based on the power distribution data of the total load of the energy storage battery string during the period when the adjacent battery cell power compensation behavior occurs, combined with the voltage and current change data obtained for each battery cell; The failure probability prediction module is used to build a damage probability model based on the characteristic offset value of the battery cell, historical abnormal trigger records and the number of working cycles using the Bayesian estimation method to predict the damage probability of the battery cell in the next few cycles and generate the failure probability curve of the battery cell in the next few cycles; The masked fault identification module is used to obtain damage probability curves, compensation behavior frequency and compensation amplitude data through battery monitoring data, mark masked fault cases, build a masked fault identification model, identify the fault status of battery cells, and formulate fault warning and health maintenance strategies for battery cells; The fault level assessment module is used to evaluate the fault level of battery cells based on the compensation mode, fault probability trend, and location relationship of battery cells that are determined to have concealed faults. It also constructs a fault level index table based on search tags including location number, compensation path number, and failure risk weight, and formulates battery maintenance strategies and early warning mechanisms. The health maintenance strategy optimization module is used to evaluate the effectiveness of fault warning and health maintenance strategies based on the real-time power changes of battery cells, fault detection results, and compensation behavior data, and to update the fault warning and health maintenance strategies.

Images