An end-cloud cooperative supported time delay drift energy storage master-slave BMS mismatch detection method

Abstract

The application relates to the technical field of electric power energy storage device control and discloses a time delay drift energy storage master-slave BMS mismatch detection method supported by terminal cloud cooperation, the method comprising the following steps: collecting operation state data of a slave board in a master-slave BMS and performing time sequence reconstruction processing, extracting an enhanced trajectory anchor point corresponding to a working condition change edge in a slave board feature time sequence reconstruction sequence; calculating a time delay drift caused by asynchronous awakening of the slave board based on the enhanced trajectory anchor point; performing time alignment compensation on the slave board feature time sequence reconstruction sequence to generate slave board consistency comparison data; and generating a mismatch type of the master-slave BMS according to a time delay drift evaluation result and the slave board consistency comparison data. Through the steps of slave board state collection of the master-slave BMS, time sequence reconstruction and enhanced trajectory anchor point extraction, time delay drift quantitative evaluation and consistency comparison are completed, the mismatch type can be automatically identified and early warning control can be realized, and the time sequence of the master-slave BMS is unified and data reliability is ensured.

Description

Technical Field

[0001] This invention relates to the field of power energy storage device control, and more particularly to an intelligent sensing system, specifically a time-delay drift master-slave BMS mismatch detection method supported by end-cloud collaboration. Background Technology

[0002] With the widespread application of large-scale energy storage systems in new energy grid connection, grid peak shaving, and distributed energy networks, battery management systems (BMS) are gradually evolving from a single-unit centralized architecture to a master-slave distributed architecture. In this architecture, the main board is responsible for global strategy and coordination control, while the slave boards are responsible for data acquisition and status monitoring of local battery cells.

[0003] In master-slave energy storage BMS, slave boards are typically deployed close to the battery modules. They are responsible for periodically sampling raw signals such as local cell voltage and temperature, and performing filtering, statistical feature extraction, anomaly detection, and first-level alarms on the sampled sequences based on a sliding window. The compressed feature quantities, status quantities, or event information are then uploaded to the master control board, which combines data from multiple slave boards to perform cross-module consistency analysis, mismatch identification, and system-level safety decisions. This hierarchical processing mechanism effectively reduces communication pressure, improves system real-time performance, and adapts to the management needs of massive sampling points in large-scale energy storage devices. However, as the operating scenarios of energy storage systems become increasingly complex, traditional mismatch detection methods that rely solely on the master control board to perform horizontal comparisons of identical features uploaded from each slave board are gradually revealing their inadequacy in adapting to distributed time base differences.

[0004] Specifically, the asynchronous wake-up of the slave board in the master-slave BMS causes misalignment of feature windows, misjudgment of cell mismatch, and frequent triggering of false alarms, resulting in multiple serious negative impacts. On the one hand, the proliferation of false alarms can easily lead to alarm fatigue among maintenance personnel, causing real battery anomalies to be masked and faults to be missed, making it impossible to deal with safety hazards such as cell overvoltage and overheating in a timely manner, and significantly increasing the risk of thermal runaway and safety accidents. On the other hand, distorted mismatch judgments can cause disorder in the equalization control logic and failure of battery cluster consistency management, accelerating cell aging and degradation, and reducing the lifespan and energy utilization rate of the energy storage system. At the same time, invalid alarms consume a large amount of communication and computing resources, reduce the real-time performance of edge-cloud collaboration, affect the stable execution of grid-connected functions such as peak shaving and frequency regulation of the energy storage system, and seriously restrict the safe and reliable operation of large-scale energy storage systems.

[0005] Existing research has proposed solutions based on multi-source data fusion and dynamic scheduling to address the optimization problem of energy storage battery management. For example, patent CN118944160A proposes a method for managing energy storage batteries in distributed energy networks. This method collects energy supply task characteristic information, battery characteristic information arrays, and environmental characteristic information arrays, and introduces a dynamic adjustment mechanism for collection accuracy parameters and transmission delay parameters to correct the battery characteristic information, thereby generating an optimal energy supply scheduling scheme. This method, by constructing a closed-loop optimization mechanism of "collection-correction-scheduling," solves to some extent the problems of inaccurate battery status collection and unbalanced scheduling under complex energy supply tasks, improving the accuracy and efficiency of energy storage battery management. However, this method does not pay enough attention to the timing drift problem between master and slave BMS caused by asynchronous wake-up, communication jitter, and multi-level processing delays.

[0006] To address this issue, this invention proposes a time-delay drift master-slave BMS mismatch detection method supported by edge-cloud collaboration. This method can identify characteristic pseudo-differences caused by time base offset after sleep wake-up, thereby improving the reliability of multi-slave board comparison from the source. This will significantly enhance the robustness of the master-slave BMS under low-power operation, multi-node asynchronous recovery, and dynamic operating conditions. Summary of the Invention

[0007] This invention proposes a time-drift mismatch detection method for master-slave BMS in energy storage with edge-cloud collaborative support. Step S1 collects slave board operating status data and performs time-series reconstruction to accurately extract enhanced trajectory anchor points at the edges of operating condition changes, thereby enhancing the ability to locate state transition features and solving the technical problems of chaotic timing of master-slave BMS status data, difficulty in capturing operating condition transition points, and ambiguous marking of key change trajectories. Step S2 maps slave board timing to a unified time axis of the main board and quantifies the time-drift evaluation results based on trajectory anchor points, establishing a unified time base evaluation standard and solving the technical problems of inconsistent time bases among multiple slave boards, inability to quantify asynchronous wake-up drift, and lack of basis for drift evaluation. Step S3 implements time alignment compensation based on the time-drift evaluation results, eliminating data misalignment caused by time base offset and solving the technical problems of large deviation in master-slave BMS data comparison benchmarks, distortion in consistency comparison, and high synchronization judgment errors. Step S4 identifies the mismatch type through time-drift and consistency data, enabling rapid response and closed-loop adjustment for anomalies and solving the technical problem of the inability to automatically identify master-slave BMS mismatch.

[0008] To achieve the above objectives, this invention provides a method for detecting latency drift in master-slave BMS for energy storage with edge-cloud collaborative support, comprising the following steps: S1: Collect the running status data of the slave board in the master-slave BMS, wherein the running status data includes sleep wake-up timing, sliding window parameters, sampling feature messages and alarm status. Perform timing reconstruction processing on the running status data to form a slave board feature timing reconstruction sequence, and extract the enhanced trajectory anchor points corresponding to the working condition change edges in the slave board feature timing reconstruction sequence. S2: Map the reconstructed timing sequence of the slave board features to the unified reference time axis of the motherboard, calculate the time delay drift caused by the asynchronous wake-up of the slave board based on the enhanced trajectory anchor point, and generate the time delay drift evaluation result; S3: Based on the time delay drift evaluation results, perform time alignment compensation on the slave board feature time sequence reconstruction sequence to generate slave board consistency comparison data; S4: Based on the latency drift evaluation results and the slave board consistency comparison data, generate the mismatch type of the master-slave BMS.

[0009] As a further improvement of the present invention: Further, in step S1, the operating status data undergoes time-series reconstruction processing to form a time-series reconstruction sequence of board features, including: S11: Based on the sliding window parameters, the sampling feature messages and alarm status in the running status data are represented by sliding windows to obtain window sampling feature messages and window alarm status of multiple sliding windows, and the start time of the sliding window is recorded. The start time, window sampling feature messages and window alarm status are used as the sliding features of the sliding window, wherein the window sampling feature messages include window voltage features and window temperature features. S12: Sort the sliding features of the sliding window according to the order of their appearance at the start time to obtain a sorted sliding feature sequence; S13: The sleep-wake sequence is concatenated with the sorted sliding feature sequence to form the reconstructed sequence of the board feature sequence, wherein the sleep-wake sequence includes the wake-up trigger time, the sampling recovery time, and the sleep entry time.

[0010] Furthermore, step S1, which extracts the enhanced trajectory anchor points corresponding to the working condition change edges in the time-series reconstruction sequence of plate features, also includes: S14: Extract the sorted sliding feature sequence from the time-series reconstruction sequence of the plate features, and calculate the edge strength of the sliding features in the sorted sliding feature sequence. The formula for calculating the edge strength of the sliding features is as follows: ; ; ; in, This represents the edge strength of the sliding feature associated with the nth sliding window. Indicates the sliding step size. This represents the window sampling feature message of the (n-1)th sliding window. This represents the window sampling feature message of the nth sliding window. This represents the window sampling feature message of the (n+1)th sliding window. Indicates the parameters of the calculation process. Indicates window sampling feature message Window voltage characteristics in Indicates window sampling feature message Window temperature characteristics, Indicates window sampling feature message Window voltage characteristics in Indicates window sampling feature message Window temperature characteristics, Indicates window sampling feature message Window voltage characteristics in Indicates window sampling feature message Window temperature characteristics, All represent intensity weighting coefficients. Both represent the weighting of voltage and temperature control; S15: Mark the sliding feature whose edge strength is higher than the preset strength threshold as a sliding enhancement feature, and extract the start time of the sliding window associated with the sliding enhancement feature as the enhancement trajectory anchor point.

[0011] Further, in step S2, mapping the time-series reconstruction sequence from board features to the unified reference time axis of the motherboard includes: Based on the motherboard's local absolute timestamp, a globally unified motherboard reference timeline is generated at fixed time intervals; Extract the time information from the reconstructed sequence of the slave board features, map the extracted time information to the nearest reference time point in the unified reference time axis of the motherboard according to the nearest time mapping method, and synchronously merge the sliding features associated with the time information to the mapped reference time point. The time information includes the start time of the sliding window and the sleep-wake time of the slave board.

[0012] Furthermore, step S2, which calculates the delay drift caused by asynchronous wake-up from the board based on the enhanced trajectory anchor point and generates a delay drift evaluation result, also includes: S21: Based on the reference time point obtained by mapping the enhanced trajectory anchor point on the motherboard unified reference time axis, calculate the global unified reference anchor point, and calculate the difference between the enhanced trajectory anchor point and the global unified reference anchor point as the time delay drift of the enhanced trajectory anchor point. Specifically, the calculation formula for the globally unified reference anchor point is as follows: ; in, This represents the m-th globally unified reference anchor point. Indicates the selection of a set the median of This represents the mapping result of the m-th enhanced trajectory anchor point in the b-th group of time-series reconstructions from board features onto the unified reference time axis of the motherboard. B represents the total number of time-series reconstructions from board features, and M represents the total number of enhanced trajectory anchor points in the time-series reconstructions from board features. The time delay drift of the m-th enhanced trajectory anchor point in the time-series reconstruction sequence of the plate features in group b is: ; S22: Calculate the starting offset of the sliding window in the time-series reconstruction sequence of the plate features based on the time delay drift of all enhanced trajectory anchor points in the time-series reconstruction sequence of the plate features; Specifically, the formula for calculating the starting offset of the nth sliding window in the time-series reconstruction sequence of the b-th plate feature is as follows: ; ; in, This represents the starting offset of the nth sliding window in the time-series reconstruction sequence of the b-th plate feature. This represents the fitting weight of the m-th enhanced trajectory anchor point to the n-th sliding window in the b-th group of time-series reconstruction sequences from plate features. This represents the edge strength of the sliding feature of the nth sliding window in the time-series reconstruction sequence of the b-th plate feature. The sliding window number represents the sequence number of the b-th group of time-series reconstruction sequences of the plate features corresponding to the m-th enhanced trajectory anchor point, where the sliding window number ranges from 1 to N; S23: Calculate the temporal drift change of the reconstructed sequence of plate features based on the initial offset of the sliding window in the reconstructed sequence of plate features; Specifically, the formula for calculating the temporal drift change of the reconstructed sequence from the plate features in group b is as follows: ; in, This represents the temporal drift change of the b-th group of reconstructed sequences from plate features. This represents the mean starting offset of the sequence reconstructed from the plate feature time series in group b; S24: Extract the temporal drift change of the reconstructed sequence of plate features and the starting offset of the sliding window in the reconstructed sequence of plate features, and encapsulate them as the time delay drift evaluation result of the reconstructed sequence of plate features.

[0013] Further, step S3 involves performing time alignment compensation on the time-series reconstruction sequence of the board features, including: Extract the time-series drift change from the time-delay drift evaluation result and the starting offset of the sliding window in the time-series reconstruction sequence of the board features. Correct the starting time of any sliding window in the time-series reconstruction sequence of the board features to obtain the time-aligned compensated time-series reconstruction sequence of the board features.

[0014] Furthermore, step S3, which reconstructs the slave board feature time series based on the time-aligned and compensated sequence, to generate slave board consistency comparison data, also includes: S31: Divide the motherboard's unified reference time axis into multiple standard windows of equal length and non-overlapping, and calculate the sliding feature mean of the time sequence reconstruction of the motherboard features after time alignment compensation and the time sequence reconstruction of the motherboard features before alignment compensation within the standard windows respectively. S32: Calculate the difference in sliding feature mean values ​​of different slave board feature time-series reconstructed sequences before and after time alignment compensation within the same standard window, and construct slave board consistency comparison data before and after time alignment compensation.

[0015] Furthermore, the mismatch type generated in step S4 for the master-slave BMS includes: Extract the time delay drift evaluation results of the time series reconstructed from the plate feature, and calculate the average starting offset of the time series reconstructed from the plate feature in the sliding window. Based on the average starting offset of the time series reconstructed from the plate feature in the sliding window and the time drift change, generate the time base bias intensity of the time series reconstructed from the plate feature, where the time base bias intensity includes high bias, medium bias and low bias. Based on the slave board consistency comparison data before and after time alignment compensation, the degree of improvement of the slave board consistency comparison data after time alignment is calculated, wherein the degree of improvement includes high improvement, medium improvement and low improvement; Based on the time base deviation intensity of the slave board characteristic time series reconstruction sequence and the degree of improvement in slave board consistency comparison data after time alignment, the mismatch type of the slave board associated with the slave board characteristic time series reconstruction sequence is generated.

[0016] Compared with existing technologies, this invention proposes a time-delay drift master-slave BMS mismatch detection method supported by edge-cloud collaboration. This technology has the following beneficial effects: First, this invention constructs edge strength by combining first-order and second-order differential methods with window alarm status, accurately locating the edges of changing operating conditions and forming enhanced trajectory anchor points that are anti-interference and highly recognizable. Compared to the traditional single-point threshold method, the enhanced trajectory anchor points can stably identify key timing positions such as charge / discharge switching, load abrupt changes, temperature jumps, and wake-up recovery, providing a stable and reliable benchmark for timing alignment. Furthermore, the preset strength threshold enables automatic anchor point selection, reducing computational complexity and improving engineering practicality. The enhanced trajectory anchor points can serve as a unified timing benchmark for the entire system, effectively offsetting timing offsets caused by asynchronous wake-up from slave boards, clock deviations, and cache reconstruction, achieving global positioning.

[0017] Meanwhile, this invention enhances trajectory anchor point matching and integrates a globally unified reference anchor point to accurately quantify timing deviations caused by asynchronous wake-up from slave boards, effectively solving the sliding window misalignment problem caused by time base asynchrony after master-slave BMS sleep wake-up. Specifically, this invention uses the median to calculate the globally unified reference anchor point, which can suppress abnormal anchor point interference and improve baseline stability. It also calculates the initial offset of the sliding window through weighted fitting, allowing high-edge strength anchor points to exert a stronger constraint effect and improving drift estimation accuracy. Linear fitting is used to obtain the timing drift change, which can intuitively reflect the drift development trend and effectively distinguish between transient and persistent deviations. The latency drift evaluation results can quantitatively describe the magnitude, development speed, and stability of timing deviations on each slave board, providing a precise basis for subsequent time alignment compensation. This fundamentally reduces false mismatch misjudgments caused by window misalignment, significantly improving the detection robustness and mismatch detection reliability of master-slave BMS in low-power operation and asynchronous wake-up scenarios. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating a delay drift detection method for master-slave BMS in energy storage with edge-cloud collaborative support, provided in one embodiment of the present invention.

[0019] Figure 2 A flowchart of master-slave BMS hibernation management provided in an embodiment of the present invention; Figure 3 This is an experimental comparison diagram provided for one embodiment of the present invention. Detailed Implementation

[0020] The realization of the objectives, functional characteristics, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0021] This invention provides a method for detecting latency drift in a master-slave BMS (Blockchain Management System) for energy storage, supported by edge-cloud collaboration. The execution entity of this method includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this invention: a server, a terminal, etc. In other words, the method can be executed by software or hardware installed on a terminal device or a server device, where the software may be a blockchain platform. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster.

[0022] Reference Figure 1 as well as Figure 2 Embodiment 1 of the present invention is as follows: A method for detecting latency drift in master-slave BMS for energy storage supported by edge-cloud collaboration, the method comprising: S1: Collect the running status data of the slave board in the master-slave BMS, including sleep wake-up timing, sliding window parameters, sampling feature messages and alarm status. Perform time-series reconstruction processing on the running status data to form a slave board feature time-series reconstruction sequence, and extract the enhanced trajectory anchor points corresponding to the working condition change edges in the slave board feature time-series reconstruction sequence.

[0023] Specifically, step S1 involves performing time-series reconstruction processing on the operating status data to form a time-series reconstruction sequence of board features, including: As an embodiment of the present invention, the BMS (Battery Management System) is the core control unit of the energy storage / power battery, responsible for collecting voltage, temperature, and current, estimating power, protecting the battery, and ensuring the safe, stable, and long-life operation of the battery. The master-slave BMS adopts a motherboard-slave board structure, including a motherboard and several slave boards. The motherboard is used to aggregate all slave board data and perform time delay drift calculation, slave board consistency comparison, and mismatch detection as in steps S2 to S4. The slave boards are deployed near the battery module and are used for local sampling, time series reconstruction, and enhanced trajectory anchor point extraction. The slave board feature time series reconstruction sequence and enhanced trajectory anchor points are uploaded to the motherboard. Furthermore, refer to, for example Figure 2 The flowchart of master-slave BMS hibernation management shows that when the energy storage / power battery is in standby, light load, or no charging / discharging action, the slave board enters low power mode, starts hibernation, and stops local sampling and communication uploading; when the slave board receives a command from the motherboard, the energy storage / power battery starts charging / discharging, or the timed wake-up condition is reached, the slave board exits hibernation and restarts local sampling. When the slave board starts local sampling, it synchronously collects the slave board's operating status data, which includes sleep wake-up timing, sliding window parameters, sampling feature messages, and alarm status. As another embodiment of the present invention, the collected operating status data is synchronously uploaded to the cloud server. The cloud server performs calculation processing as in steps S1 to S4 on the operating status data to generate mismatch types, thereby realizing the time-delay drift energy storage master-slave BMS mismatch detection method under end-cloud collaboration. The sleep / wake-up sequence includes the wake-up trigger time, the sampling recovery time, and the sleep entry time; the sliding window parameters include the window length and the sliding step size; the sampling feature message is the battery module sampling data and the sampling time corresponding to the battery module sampling data, including battery voltage and temperature, with the default time interval between adjacent sampling times set to 0.2 seconds; the alarm status includes the alarm status identifier and the alarm time. Specifically, the slave board completes the preliminary abnormality judgment of the battery locally based on the sampling feature message. If it is determined that the single cell voltage is too high / too low, or the temperature is too high / too low, an alarm status identifier is generated and the alarm time is recorded, wherein the alarm status identifier is represented in the form of 1; S11: Based on the sliding window parameters, the sampling feature messages and alarm status in the running status data are represented by sliding windows to obtain window sampling feature messages and window alarm status of multiple sliding windows, and the start time of the sliding window is recorded. The start time, window sampling feature messages and window alarm status are used as the sliding features of the sliding window, wherein the window sampling feature messages include window voltage features and window temperature features. Specifically, the window length in the sliding window parameters is 5 seconds by default, and the sliding step size is 1 second by default; As an embodiment of the present invention, the calculation formulas for the window sampling feature message and the window alarm status of the sliding window are as follows: ; ; ; ; in, This represents the window sampling feature message of the nth sliding window. Indicates window sampling feature message Window voltage characteristics in Indicates window sampling feature message Window temperature characteristics, Indicates the nth sliding window. The battery voltage in each sampled feature message, where L represents the number of sampled feature messages within the sliding window. Indicates the nth sliding window. Battery temperature in each sampled feature message, This represents the average battery voltage within the nth sliding window. This represents the average battery temperature within the nth sliding window. Indicates the selection of a set The maximum value in, Indicates the selection of a set The maximum value in, This represents the voltage deviation fluctuation weighting coefficient. This represents the weighting coefficient for temperature deviation fluctuations; the default setting is... , Both are 0.2. This indicates the alarm status of the nth sliding window. This represents the sum of alarm status indicators within the nth sliding window. Indicates the number of sliding windows; S12: Sort the sliding features of the sliding window according to the order of their appearance at the start time to obtain a sorted sliding feature sequence; S13: The sleep-wake sequence is concatenated with the sorted sliding feature sequence to form the reconstructed sequence of the board feature sequence, wherein the sleep-wake sequence includes the wake-up trigger time, the sampling recovery time, and the sleep entry time.

[0024] It should be noted that this invention achieves noise reduction, compression, and feature enhancement of battery module sampling data through sliding window processing, window sampling feature messages, and window alarm status calculation, thereby improving anomaly sensitivity while reducing communication volume. Specifically, this invention adopts an innovative feature structure of mean superposition and maximum deviation, which reflects the overall state of the battery module and sensitively captures voltage and temperature fluctuations and jumps, improving the distinction between timing misalignment and real faults. The sleep-wake-up timing sequence and sliding features are spliced ​​together to form a board feature timing reconstruction sequence, completely preserving the timing relationship of the entire process of wake-up, sampling recovery, and window reconstruction, providing a true and reliable timing basis for subsequent delay drift estimation.

[0025] Step S1, which extracts the enhanced trajectory anchor points corresponding to the working condition change edges in the time-series reconstruction sequence of plate features, also includes: S14: Extract the sorted sliding feature sequence from the time-series reconstruction sequence of the plate features, and calculate the edge strength of the sliding features in the sorted sliding feature sequence. The formula for calculating the edge strength of the sliding features is as follows: ; ; ; in, This represents the edge strength of the sliding feature associated with the nth sliding window. Indicates the sliding step size. This represents the window sampling feature message of the (n-1)th sliding window. This represents the window sampling feature message of the nth sliding window. This represents the window sampling feature message of the (n+1)th sliding window. Indicates the parameters of the calculation process. Indicates window sampling feature message Window voltage characteristics in Indicates window sampling feature message Window temperature characteristics, Indicates window sampling feature message Window voltage characteristics in Indicates window sampling feature message Window temperature characteristics, Indicates window sampling feature message Window voltage characteristics in Indicates window sampling feature message Window temperature characteristics, All represent intensity weighting coefficients, default settings. The values ​​are 0.4, 0.3, and 0.3 respectively. Both represent the voltage and temperature control weights, and are the default settings. The values ​​are 0.6 and 0.4 respectively. Furthermore, the operating condition change edge indicates the moment when the battery's operating state suddenly changes, that is, the moment when the voltage, temperature, and current show obvious jumps, inflection points, or sudden changes in slope. S15: Mark the sliding feature whose edge strength is higher than the preset strength threshold as a sliding enhancement feature, and extract the start time of the sliding window associated with the sliding enhancement feature as the enhancement trajectory anchor point.

[0026] Specifically, the preset intensity threshold is set to 0.5 by default.

[0027] S2: Map the reconstructed timing sequence of the slave board features to the unified reference time axis of the motherboard, calculate the delay drift caused by asynchronous wake-up of the slave board based on the enhanced trajectory anchor point, and generate the delay drift evaluation result.

[0028] Specifically, step S2, which maps the time-series reconstruction sequence from board features to the unified reference time axis of the motherboard, includes: Based on the motherboard's local absolute timestamp, a globally unified motherboard reference timeline is generated at fixed time intervals; Specifically, the motherboard unified reference time axis includes several consecutive reference time points, and the time interval between adjacent reference time points is 100 milliseconds by default. Extract the time information from the reconstructed sequence of the slave board features, map the extracted time information to the nearest reference time point in the unified reference time axis of the motherboard according to the nearest time mapping method, and synchronously merge the sliding features associated with the time information to the mapped reference time point. The time information includes the start time of the sliding window and the sleep-wake time of the slave board.

[0029] As an embodiment of the present invention, for the motherboard numbered BMS-MAIN-01 and the 16 slave boards numbered BMS-SLAVE-01 to BMS-SLAVE-16 participating in data acquisition, a globally unified motherboard unified reference time axis is generated at a fixed interval of 100 milliseconds, based on the local absolute timestamp of the motherboard BMS-MAIN-01, forming a continuous and ordered reference time point sequence. The time information in the characteristic time sequence reconstruction sequence of each slave board is extracted respectively. The time information includes the wake-up trigger time, sampling recovery time, sleep entry time in the slave board sleep wake-up time sequence, and the start time of each sliding window. By employing the nearest-neighbor time mapping method, the aforementioned time information is matched one by one to the nearest reference time point in the motherboard's unified reference time axis. The sliding features corresponding to the time information, including window sampling feature messages, window alarm status, sliding enhancement features, and enhancement trajectory anchor points, are synchronously merged into the mapped reference time point, completing the mapping of single-slave board timing data to the motherboard's unified time base. After mapping processing, the timing and feature data of all 16 slave boards are aligned to the same time base, providing a standard timing basis for subsequent delay drift calculation and cross-slave board consistency comparison.

[0030] The S2 step, which calculates the time delay drift caused by asynchronous wake-up from the board based on the enhanced trajectory anchor point and generates a time delay drift evaluation result, also includes: S21: Based on the reference time point obtained by mapping the enhanced trajectory anchor point on the motherboard unified reference time axis, calculate the global unified reference anchor point, and calculate the difference between the enhanced trajectory anchor point and the global unified reference anchor point as the time delay drift of the enhanced trajectory anchor point. Specifically, the calculation formula for the globally unified reference anchor point is as follows: ; in, This represents the m-th globally unified reference anchor point. Indicates the selection of a set the median of This represents the mapping result of the m-th enhanced trajectory anchor point in the b-th group of time-series reconstructions from board features onto the unified reference time axis of the motherboard. B represents the total number of time-series reconstructions from board features, and M represents the total number of enhanced trajectory anchor points in the time-series reconstructions from board features. The time delay drift of the m-th enhanced trajectory anchor point in the time-series reconstruction sequence of the plate features in group b is: ; S22: Calculate the starting offset of the sliding window in the time-series reconstruction sequence of the plate features based on the time delay drift of all enhanced trajectory anchor points in the time-series reconstruction sequence of the plate features; Specifically, the formula for calculating the starting offset of the nth sliding window in the time-series reconstruction sequence of the b-th plate feature is as follows: ; ; in, This represents the starting offset of the nth sliding window in the time-series reconstruction sequence of the b-th plate feature. This represents the fitting weight of the m-th enhanced trajectory anchor point to the n-th sliding window in the b-th group of time-series reconstruction sequences from plate features. This represents the edge strength of the sliding feature of the nth sliding window in the time-series reconstruction sequence of the b-th plate feature. The sliding window number represents the sequence number of the b-th group of time-series reconstruction sequences of the plate features corresponding to the m-th enhanced trajectory anchor point, where the sliding window number ranges from 1 to N; S23: Calculate the temporal drift change of the reconstructed sequence of plate features based on the initial offset of the sliding window in the reconstructed sequence of plate features; Specifically, the formula for calculating the temporal drift change of the reconstructed sequence from the plate features in group b is as follows: ; in, This represents the temporal drift change of the b-th group of reconstructed sequences from plate features. This represents the mean starting offset of the sequence reconstructed from the plate feature time series in group b; S24: Extract the temporal drift change of the reconstructed sequence of plate features and the starting offset of the sliding window in the reconstructed sequence of plate features, and encapsulate them as the time delay drift evaluation result of the reconstructed sequence of plate features.

[0031] It should be noted that this invention, by enhancing trajectory anchor point matching and fusing with a globally unified reference anchor point, accurately quantifies the timing deviation caused by asynchronous wake-up of slave boards, effectively solving the sliding window misalignment problem caused by time base asynchrony after master-slave BMS sleep wake-up. Specifically, this invention uses the median to calculate the globally unified reference anchor point, which can suppress abnormal anchor point interference and improve baseline stability. It also calculates the initial offset of the sliding window through weighted fitting, allowing high-edge strength anchor points to exert a stronger constraint effect and improving drift estimation accuracy. Linear fitting is used to obtain the timing drift change, which can intuitively reflect the drift development trend and effectively distinguish between transient and persistent deviations. The latency drift evaluation results can quantitatively describe the magnitude, development speed, and stability of timing deviations of each slave board, providing a precise basis for subsequent time alignment compensation. This fundamentally reduces false mismatch misjudgments caused by window misalignment, significantly improving the detection robustness and mismatch detection reliability of master-slave BMS in low-power operation and asynchronous wake-up scenarios.

[0032] Furthermore, the time delay drift represents the deviation between the enhanced trajectory anchor point in the plate and the globally unified reference anchor point characterizing the actual operating condition of the battery. The higher the time delay drift, the more severe the timing misalignment of the plate. The change in timing drift represents the rate and trend of change of the timing misalignment of the plate as the sliding window progresses.

[0033] S3: Based on the time delay drift evaluation results, perform time alignment compensation on the slave board feature time sequence reconstruction sequence to generate slave board consistency comparison data.

[0034] Specifically, step S3 involves performing time alignment compensation on the time-series reconstruction sequence of the board features, including: Extract the time-series drift change from the time-delay drift evaluation result and the starting offset of the sliding window in the time-series reconstruction sequence of the board features. Correct the starting time of any sliding window in the time-series reconstruction sequence of the board features to obtain the time-aligned compensated time-series reconstruction sequence of the board features.

[0035] Specifically, the formula for correcting the starting time of the nth sliding window in the time-series reconstruction sequence of the b-th plate feature is as follows: ; in, This represents the starting time of the nth sliding window in the time-series reconstruction sequence of the b-th plate feature. Indicates the start time The correction results This represents the time interval between the start time of the nth sliding window and the start time of the 1st sliding window.

[0036] It should be noted that this invention achieves precise time alignment of the sliding window's starting moment through static correction of the initial offset and dynamic correction of the timing drift change. Based on the correction formula, static deviation elimination and trend deviation correction are completed in one go, strictly pulling the misaligned window back to the motherboard's unified reference time axis, synchronizing the boundaries of each slave board. After compensation, the timing data eliminates the effects of time base drift, wake-up lag, and window reconstruction misalignment, ensuring that cross-slave board features are comparable within the same time period.

[0037] Step S3, which reconstructs the time-series sequence of the slave board features based on time alignment compensation, generates slave board consistency comparison data, and also includes: S31: Divide the motherboard's unified reference time axis into multiple standard windows of equal length and non-overlapping, and calculate the sliding feature mean of the time sequence reconstruction of the motherboard features after time alignment compensation and the time sequence reconstruction of the motherboard features before alignment compensation within the standard windows respectively. S32: Calculate the difference in sliding feature mean values ​​of different slave board feature time-series reconstructed sequences before and after time alignment compensation within the same standard window, and construct slave board consistency comparison data before and after time alignment compensation.

[0038] As an embodiment of the present invention, the set of sliding feature mean values ​​of different time-series reconstruction sequences of board features before and after time alignment compensation within the same standard window is as follows: ; in, This represents the set of sliding feature means of different time-series reconstructed sequences from different plate features within the same standard window before time alignment compensation. This represents the set of sliding feature means of different time-series reconstructed sequences from different plate features within the same standard window before time alignment compensation. This represents the mean sliding feature of the b-th group of time-series reconstructed sequences from plate features in the h-th standard window before time alignment compensation. B represents the number of standard windows, and B represents the number of sequences reconstructed from the board feature time series. This represents the mean sliding feature of the b-th group of time-series reconstructed sequences from board features in the h-th standard window before time alignment compensation; According to the set of sliding characteristic means as well as The consistency comparison data of the slave board before and after time alignment compensation was calculated: ; ; in, This represents the slave board consistency comparison data before time alignment compensation. This represents the consistency comparison data of the slave board after time alignment compensation. Represents the set of moving characteristic means The mean of the absolute values ​​of the differences between any two sliding feature means in the h-th standard window. Represents the set of moving characteristic means The mean of the absolute values ​​of the differences between any two sliding feature means in the h-th standard window is given. The lower the value of the consistency comparison data, the lower the difference between the sliding feature means of any two slave boards under the same standard window, and the higher the consistency of the slave boards. Optionally, the difference between any two sliding feature means can be calculated using the L2 norm.

[0039] It should be noted that this invention constructs slave board consistency comparison data that can be directly used to determine the nature of mismatch by dividing the standard window, calculating the mean of sliding features, and comparing the difference before and after compensation. Specifically, by comparing the difference before and after time alignment compensation, the degree to which the time-series pseudo-differences are eliminated can be intuitively reflected, and the source of mismatch can be initially distinguished. The lower the consistency comparison data, the better the real consistency between slave boards, and vice versa.

[0040] S4: Based on the latency drift evaluation results and the slave board consistency comparison data, generate the mismatch type of the master-slave BMS.

[0041] Specifically, the mismatch type generated in step S4 for the master-slave BMS includes: Extract the time delay drift evaluation results of the time series reconstructed from the plate feature, and calculate the average starting offset of the time series reconstructed from the plate feature in the sliding window. Based on the average starting offset of the time series reconstructed from the plate feature in the sliding window and the time drift change, generate the time base bias intensity of the time series reconstructed from the plate feature, where the time base bias intensity includes high bias, medium bias and low bias. Specifically, if the average initial offset is higher than the preset offset threshold and the change in time series drift is higher than the preset offset change threshold, the time base bias intensity is high; if the average initial offset is not higher than the preset offset threshold and the change in time series drift is higher than the preset offset change threshold, the time base bias intensity is medium; the time base bias intensity in other cases is set to low; the preset offset threshold is set to 500 milliseconds by default, and the preset offset change threshold is set to 20 milliseconds by default. Based on the slave board consistency comparison data before and after time alignment compensation, the degree of improvement of the slave board consistency comparison data after time alignment is calculated, wherein the degree of improvement includes high improvement, medium improvement and low improvement; Specifically, the formula for calculating the degree of improvement is as follows: If the degree of improvement is higher than the first improvement threshold (default setting is 0.4), the degree of improvement is high; if the degree of improvement is not higher than the first improvement threshold (default setting is 0.4) but higher than the second improvement threshold (default setting is 0.2), the degree of improvement is medium; if the degree of improvement is not higher than the second improvement threshold (default setting is 0.2), the degree of improvement is low. Based on the time base deviation intensity of the slave board characteristic time series reconstruction sequence and the degree of improvement in slave board consistency comparison data after time alignment, the mismatch type of the slave board associated with the slave board characteristic time series reconstruction sequence is generated.

[0042] Specifically, if the time base deviation intensity of the reconstructed sequence from the plate feature time series is medium / high deviation and the improvement degree of the plate consistency comparison data after time alignment is high improvement, then the mismatch type of the plate associated with the reconstructed sequence from the plate feature time series is time series pseudo mismatch, which is manifested as mismatch difference being completely / mainly caused by asynchronous wake-up of the plate, window misalignment, and time base drift, with no obvious abnormality in the battery itself, which is a false alarm; If the time base deviation intensity of the reconstructed sequence from the plate feature time series is low and the improvement degree of the plate consistency comparison data after time alignment is medium / high, then the mismatch type of the plate associated with the plate feature time series reconstructed sequence is true module mismatch, indicating that there are differences in the battery modules near the plate deployment, which manifests as problems such as aging, internal resistance differences, leakage, and undervoltage in the battery modules. If the time base deviation intensity of the sequence reconstructed from the board feature time series and the degree of improvement of the board consistency comparison data after time alignment are other combinations, the mismatch type is uniformly set to transient drift mismatch, no alarm is required, only delayed review is needed; Optionally, the present invention performs early warning control and time base synchronization according to the mismatch type. Specifically, for pseudo-time mismatch, the current alarm is directly suppressed, and the host computer and audible and visual alarms are not triggered. Only the time drift log is recorded. For transient drift mismatch, a delayed verification mechanism is implemented. The comparison is performed again after the drift converges. During this period, temporary alarms are blocked. For real module mismatch, the alarm suppression is immediately lifted, and real fault reporting, audible and visual reminders and background early warning are triggered.

[0043] It should be noted that this invention uses a dual-dimensional approach—time base deviation intensity and consistency improvement degree—to automatically classify timing pseudo-mismatches, genuine module mismatches, and transient drift mismatches. Specifically, it implements tiered early warning and control based on the mismatch type: pseudo-mismatches are directly suppressed with alarms, genuine mismatches are reported immediately, and transient drifts are reviewed with a delay. This reduces false alarms while ensuring that genuine faults are not missed. The threshold-based judgment logic is simple, efficient, and suitable for embedded real-time operation. This mechanism addresses the pain point of false alarms caused by sleep / wake-up issues at the source, improving the early warning accuracy and operational efficiency of master-slave BMS under low power consumption, large-scale, and dynamic operating conditions, and ensuring the safe and stable operation of energy storage systems.

[0044] Example 2: This invention collects real-world operating data from a 16-channel (i.e., 16 slave boards) energy storage master-slave BMS (covering typical scenarios such as asynchronous sleep / wake-up, charge / discharge switching, load surges, low-temperature startup, and light-load standby) to construct a test dataset containing timing drift, false mismatch, real cell mismatch, and mixed anomalies. The performance of the time-delay drift energy storage master-slave BMS mismatch detection method supported by edge-cloud collaboration described in this invention is compared with that of traditional BMS consistency comparison methods in four core indicators: false mismatch identification accuracy, real mismatch detection rate, false alarm rate, and timing alignment error. Experimental results show that this invention, by enhancing trajectory anchor point timing reconstruction, time-delay drift quantization compensation, and true / false mismatch classification mechanisms, refers to... Figure 3 The experimental comparison diagram shows that, compared with traditional methods, the accuracy of pseudo-mismatch identification is improved by 38.6%, the detection rate of true mismatch is improved by 15.3%, the false alarm rate of early warning is reduced by 57.2%, and the timing alignment error is reduced by 81.4%. It can effectively eliminate the timing drift interference caused by asynchronous wake-up, accurately distinguish between pseudo-mismatch and real fault, and has significant technical advantages and practical value in large-scale low-power energy storage BMS early warning scenarios.

[0045] Furthermore, traditional BMS consistency comparison methods only perform simple horizontal difference comparisons on the feature data uploaded by each slave board on the motherboard side, relying entirely on the message arrival time for timing alignment, without considering the time base differences and timing drift caused by slave board sleep wake-up, asynchronous startup, and window reconstruction.

[0046] It should be noted that the terms "comprising," "including," or any other variations thereof used herein are intended to cover non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, apparatus, article, or method. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, apparatus, article, or method that includes that element.

[0047] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0048] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A method for detecting time-delay drift in master-slave BMS for energy storage with edge-cloud collaborative support, characterized in that, The method includes: S1: Collect the running status data of the slave board in the master-slave BMS, wherein the running status data includes sleep wake-up timing, sliding window parameters, sampling feature messages and alarm status. Perform timing reconstruction processing on the running status data to form a slave board feature timing reconstruction sequence, and extract the enhanced trajectory anchor points corresponding to the working condition change edges in the slave board feature timing reconstruction sequence. S2: Map the reconstructed timing sequence of the slave board features to the unified reference time axis of the motherboard, calculate the time delay drift caused by the asynchronous wake-up of the slave board based on the enhanced trajectory anchor point, and generate the time delay drift evaluation result; S3: Based on the time delay drift evaluation results, perform time alignment compensation on the slave board feature time sequence reconstruction sequence to generate slave board consistency comparison data; S4: Based on the latency drift evaluation results and the slave board consistency comparison data, generate the mismatch type of the master-slave BMS.

2. The method for detecting latency drift in master-slave BMS for energy storage with edge-cloud collaborative support as described in claim 1, characterized in that, Step S1 involves performing time-series reconstruction processing on the operating status data to form a time-series reconstruction sequence based on board features, including: S11: Based on the sliding window parameters, the sampling feature messages and alarm status in the running status data are represented by sliding windows to obtain window sampling feature messages and window alarm status of multiple sliding windows, and the start time of the sliding window is recorded. The start time, window sampling feature messages and window alarm status are used as the sliding features of the sliding window, wherein the window sampling feature messages include window voltage features and window temperature features. S12: Sort the sliding features of the sliding window according to the order of their appearance at the start time to obtain a sorted sliding feature sequence; S13: The sleep-wake sequence is concatenated with the sorted sliding feature sequence to form the reconstructed sequence of the board feature sequence, wherein the sleep-wake sequence includes the wake-up trigger time, the sampling recovery time, and the sleep entry time.

3. The method for detecting latency drift in master-slave BMS for energy storage with edge-cloud collaborative support as described in claim 2, characterized in that, Step S1, which extracts the enhanced trajectory anchor points corresponding to the working condition change edges in the time-series reconstruction sequence of plate features, also includes: S14: Extract the sorted sliding feature sequence from the time-series reconstruction sequence of the plate features, and calculate the edge strength of the sliding features in the sorted sliding feature sequence. The formula for calculating the edge strength of the sliding features is as follows: ； ； ； in, This represents the edge strength of the sliding feature associated with the nth sliding window. Indicates the sliding step size. This represents the window sampling feature message of the (n-1)th sliding window. This represents the window sampling feature message of the nth sliding window. This represents the window sampling feature message of the (n+1)th sliding window. Indicates the parameters of the calculation process. Indicates window sampling feature message Window voltage characteristics in Indicates window sampling feature message Window temperature characteristics, Indicates window sampling feature message Window voltage characteristics in Indicates window sampling feature message Window temperature characteristics, Indicates window sampling feature message Window voltage characteristics in Indicates window sampling feature message Window temperature characteristics, Both represent intensity weighting coefficients. Both represent the weighting of voltage and temperature control; S15: Mark the sliding feature whose edge strength is higher than the preset strength threshold as a sliding enhancement feature, and extract the start time of the sliding window associated with the sliding enhancement feature as the enhancement trajectory anchor point.

4. The method for detecting latency drift in master-slave BMS for energy storage with edge-cloud collaborative support as described in claim 1, characterized in that, Step S2, which maps the time-series reconstruction sequence from board features to the unified reference time axis of the motherboard, includes: Based on the motherboard's local absolute timestamp, a globally unified motherboard reference timeline is generated at fixed time intervals; Extract the time information from the reconstructed sequence of the slave board features, map the extracted time information to the nearest reference time point in the unified reference time axis of the motherboard according to the nearest time mapping method, and synchronously merge the sliding features associated with the time information to the mapped reference time point. The time information includes the start time of the sliding window and the sleep-wake time of the slave board.

5. The method for detecting latency drift in master-slave BMS energy storage with edge-cloud collaborative support as described in claim 4, characterized in that, The S2 step, which calculates the time delay drift caused by asynchronous wake-up from the board based on the enhanced trajectory anchor point and generates a time delay drift evaluation result, also includes: S21: Based on the reference time point obtained by mapping the enhanced trajectory anchor point on the motherboard unified reference time axis, calculate the global unified reference anchor point, and calculate the difference between the enhanced trajectory anchor point and the global unified reference anchor point as the time delay drift of the enhanced trajectory anchor point. S22: Calculate the starting offset of the sliding window in the time-series reconstruction sequence of the plate features based on the time delay drift of all enhanced trajectory anchor points in the time-series reconstruction sequence of the plate features; S23: Calculate the temporal drift change of the reconstructed sequence of plate features based on the initial offset of the sliding window in the reconstructed sequence of plate features; S24: Extract the temporal drift change of the reconstructed sequence of plate features and the starting offset of the sliding window in the reconstructed sequence of plate features, and encapsulate them as the time delay drift evaluation result of the reconstructed sequence of plate features.

6. The method for detecting latency drift in master-slave BMS for energy storage with edge-cloud collaborative support as described in claim 1, characterized in that, Step S3 involves performing time alignment compensation on the time-series reconstruction sequence of the board features, including: Extract the time-series drift change from the time-delay drift evaluation result and the starting offset of the sliding window in the time-series reconstruction sequence of the board features. Correct the starting time of any sliding window in the time-series reconstruction sequence of the board features to obtain the time-aligned compensated time-series reconstruction sequence of the board features.

7. The method for detecting latency drift in master-slave BMS for energy storage with edge-cloud collaborative support as described in claim 6, characterized in that, Step S3, which reconstructs the time-series sequence of the slave board features based on time alignment compensation, generates slave board consistency comparison data, and also includes: S31: Divide the motherboard's unified reference time axis into multiple standard windows of equal length and non-overlapping, and calculate the sliding feature mean of the time sequence reconstruction of the motherboard features after time alignment compensation and the time sequence reconstruction of the motherboard features before alignment compensation within the standard windows respectively. S32: Calculate the difference in sliding feature mean values ​​of different slave board feature time-series reconstructed sequences before and after time alignment compensation within the same standard window, and construct slave board consistency comparison data before and after time alignment compensation.

8. The method for detecting latency drift in master-slave BMS for energy storage with edge-cloud collaborative support as described in claim 1, characterized in that, The mismatch type generated in step S4 for the master-slave BMS includes: Extract the time delay drift evaluation results of the time series reconstructed from the plate feature, and calculate the average starting offset of the time series reconstructed from the plate feature in the sliding window. Based on the average starting offset of the time series reconstructed from the plate feature in the sliding window and the time drift change, generate the time base bias intensity of the time series reconstructed from the plate feature, where the time base bias intensity includes high bias, medium bias and low bias. Based on the slave board consistency comparison data before and after time alignment compensation, the degree of improvement of the slave board consistency comparison data after time alignment is calculated, wherein the degree of improvement includes high improvement, medium improvement and low improvement; Based on the time base deviation intensity of the slave board characteristic time series reconstruction sequence and the degree of improvement in slave board consistency comparison data after time alignment, the mismatch type of the slave board associated with the slave board characteristic time series reconstruction sequence is generated.

Images