Storage battery degradation evaluation method based on threshold self-consistency of normal distribution

Abstract

The invention belongs to the technical field of electrical variable measurement, and particularly relates to a normal distribution-based threshold self-consistent storage battery degradation evaluation method, which comprises the following steps of: acquiring and cleaning single voltage data of a lead-acid storage battery; estimating a mean value and a standard deviation of voltage distribution on line by adopting a recursive moment estimation algorithm; dynamically generating an adaptive threshold in combination with the confidence coefficient, the historical attenuation correction and the sensitivity factor; a degradation index is calculated by integrating the voltage over-limit proportion, the distribution center drift and the dispersion expansion degree; constructing a five-dimensional tensor from a provincial level to a monomer level, and positioning the first K hidden danger paths through tensor decomposition; and fusing the histogram, the normal fitting curve and the hidden danger mark to carry out visual display, and triggering grading alarm. According to the technical scheme, accurate quantification, efficient positioning and intelligent alarm of the deterioration state of the storage battery are achieved, and the evaluation automation level and the operation and maintenance efficiency are remarkably improved.

Description

Technical Field

[0001] This invention belongs to the field of electrical variable measurement technology, specifically relating to a threshold self-consistent method for assessing battery degradation based on normal distribution. Background Technology

[0002] With the continuous expansion of communication infrastructure, batteries, as the core backup power source for data center power systems, directly impact network reliability and maintenance costs. Current mainstream battery degradation monitoring methods generally rely on preset static voltage thresholds for anomaly detection. These methods use fixed upper and lower limits throughout the battery's lifespan, ignoring the objective law that voltage distribution parameters continuously drift with the state of health (SOH) during battery aging. This leads to a significant disconnect between the assessment results and the actual degree of degradation. Especially in large-scale deployment scenarios, millions of individual batteries exhibit significant individual differences and time-varying characteristics, placing higher demands on the monitoring system's adaptability, quantification accuracy, and positioning efficiency.

[0003] Among these, voltage data modeling based on normal distribution characteristics has become a key path to achieving refined assessment. Under steady-state float charging conditions, the voltage of individual lead-acid batteries typically exhibits a statistical characteristic of approximately a normal distribution, while the degradation process often manifests as a dual effect of distribution center shift and increased dispersion. However, existing technologies lack effective means to quantify the voltage distribution pattern, and it is difficult to distinguish between normal fluctuations and early degradation signals by simply judging exceeding limits. At the same time, traditional solutions often use battery packs as the smallest monitoring unit, which cannot achieve individual-level hazard identification, and they heavily rely on manually maintained rule bases for alarm logic configuration. Faced with large-scale networks with diverse equipment types and complex geographical environments, rule updates are lagging and maintenance costs are high.

[0004] In existing technologies, fixed threshold mechanisms cannot dynamically respond to the drift of distributed parameters caused by battery aging, resulting in a large number of false alarms or missed alarms; voltage data is only used for point-to-point comparisons, without establishing interpretable correlation indicators between distribution patterns and the degree of degradation; hazard location is limited to a coarse-grained level, making it difficult to support efficient operation and maintenance decisions; the rule base requires regular manual calibration, making it difficult to adapt to the system evolution during long-term operation. These shortcomings are particularly prominent in provincial and even national power monitoring systems, urgently requiring a battery degradation assessment method that integrates online statistical modeling, dynamic threshold self-consistency, multi-dimensional correlation analysis, and visual interaction to achieve high-precision, low-false-alarm, fine-grained, and maintenance-free intelligent diagnostic capabilities. Summary of the Invention

[0005] The purpose of this invention is to provide a battery degradation assessment method based on a threshold self-consistency of normal distribution, which can effectively solve the problems in the background art.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A threshold-consistent method for assessing battery degradation based on normal distribution includes the following specific steps: S1: Collect and preprocess battery voltage data, filter preset target data through the southbound acquisition cluster, and remove outliers based on rated voltage level; S2: Estimate the voltage distribution parameters of the preset target data online, and use the recursive moment estimation algorithm to calculate the mean and standard deviation in real time for the individual voltage sequence of a specific voltage level, and use the current mean and current standard deviation as distribution coupling parameters; S3: Dynamically generate adaptive thresholds, and calculate the upper limit and lower limit of the dynamic threshold based on the distribution coupling parameters, the standard normal quantiles corresponding to the confidence coefficients, the historical attenuation correction amount and the sensitivity factor. S4: Calculate the battery degradation index. Based on the distribution coupling parameters, dynamic threshold, and new battery calibration parameters, calculate the quantitative degradation index value by considering the proportion of voltage exceeding the dynamic threshold, the degree of drift of the distribution center relative to the calibration center, and the degree of expansion of the distribution dispersion relative to the calibration dispersion. S5: Perform multidimensional correlation analysis and hazard location, construct a multi-level multidimensional tensor data structure, mark the individual with the deterioration index value exceeding the preset threshold as an abnormal state, use tensor decomposition algorithm to extract the main components, and realize the location and drilling analysis of the first K hazard paths. S6: Generates visualization results and alarm information, integrates and displays the voltage histogram, fitted normal distribution curve and dynamic threshold upper and lower limits, highlights batteries that exceed the standard, interactively jumps to the hidden danger details view, and writes the exceeding record to the hidden danger details table to trigger the corresponding level of alarm.

[0007] Furthermore, the data cleaning in S1 also includes performing sliding window mid-range filtering on the original voltage data, with a window size of 5 data points, and using linear interpolation to fill in consecutive missing data points.

[0008] Furthermore, the recursive moment estimation algorithm in S2 adopts a sliding window mechanism. When the window is full and new data points arrive, the influence of the oldest data points on the mean and auxiliary variables is first removed by reverse recursion, and then the distribution coupling parameters are updated by adding new data points by forward recursion.

[0009] Furthermore, in S3, the confidence coefficient is 0.95, corresponding to a standard normal quantile of 1.96, the smoothing coefficient of the historical attenuation correction is 0.2, and the sensitivity factor is configured differently according to the rated voltage: 0.15 for 2-volt batteries, 0.40 for 6-volt batteries, and 0.60 for 12-volt batteries.

[0010] Furthermore, in S4, the new battery calibration parameters are set for a 2-volt battery with a mean of 2.25 volts and a standard deviation of 0.02 volts. The degradation index is calculated using the following formula:

[0011] μ0,ω0 are the calibration values ​​of the new battery; N out The number of voltage samples exceeding the dynamic threshold; N: Total number of samples in the query window; N out / N: Exceedance rate, reflecting the proportion of abnormal voltage readings; X c Voltage distribution center; By integrating information from three dimensions—frequency of voltage anomalies, mean drift, and fluctuation changes—a unified degradation index is generated, providing a quantitative basis for battery health status assessment and early warning. A DI value greater than 0.15 triggers a level 2 alarm, and a DI value greater than 0.30 triggers a level 1 alarm.

[0012] Furthermore, the size of each dimension of the five-dimensional tensor in S5 is dynamically determined according to the actual monitoring network scale. The tensor cell stores the degradation index value or the binary anomaly flag. An anomaly flag of 1 indicates that the degradation index exceeds 0.15, and 0 indicates that it is normal.

[0013] Furthermore, the tensor decomposition algorithm in S5 adopts normal decomposition, decomposing the five-dimensional tensor into the sum of multiple rank tensors, and extracting the top 20 components with the largest weights to identify the 20 most significant abnormal propagation paths from the provincial level to the individual level.

[0014] Furthermore, the hidden danger details table in S6 includes fields such as province name, city name, site identifier, computer room identifier, battery rated capacity, deterioration index value, dynamic threshold version number, and data timestamp, which together uniquely identify a hidden danger record.

[0015] Furthermore, in S6, the front-end interface displays the voltage statistical histogram, the normal distribution fitting curve drawn based on the estimated parameters, and the upper and lower limits of the dynamic threshold in different colors and line types. Exceeding batteries are highlighted on the histogram. When the user clicks on the highlighted area, a statistical count of abnormal cells will pop up. Further clicking will jump to a detailed cell information list page and supports exporting to a standard format table file.

[0016] Furthermore, the entire evaluation method is implemented based on a distributed streaming data processing framework. Data acquisition and distribution are completed through message queues. The distributed coupled analysis engine and threshold self-consistent optimizer are deployed in the form of microservices, and the battery voltage data is processed in real time at fixed time intervals or sliding time windows.

[0017] Compared with the prior art, the present invention has the following beneficial effects: By introducing a normal distribution coupling model for online parameter estimation and dynamic threshold generation of battery voltage distribution, the shortcomings of fixed threshold methods in adapting to the drift of distribution parameters caused by battery aging are effectively overcome, significantly reducing false alarm and false negative rates. By designing a degradation index that comprehensively considers exceedance rate, center drift, and dispersion, a refined quantitative assessment and ranking of battery health status is achieved. By constructing a multidimensional tensor and applying a tensor decomposition algorithm, key hidden danger paths can be quickly located from a massive number of battery cells, greatly improving the efficiency of operation and maintenance location. By integrating visualization and interactive design at the front end, the efficiency of human-machine collaborative analysis is improved. Through a threshold self-consistent optimization mechanism and online update capability, the workload of manual rule maintenance is greatly reduced, realizing the automation, intelligence, and efficiency of battery degradation assessment. Attached Figure Description

[0018] Figure 1 The flowchart illustrates a threshold self-consistent battery degradation assessment method based on normal distribution, as claimed in an embodiment of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0020] The terms "first," "second," and "third" in this application are for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first," "second," or "third" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified. All directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of this application are only used to explain the relative positional relationships and movements between components in a specific orientation (as shown in the figures). If the specific orientation changes, the directional indications also change accordingly. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or devices.

[0021] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a mutually exclusive, independent, or alternative embodiment. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0022] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0023] Currently, with the continuous expansion of communication infrastructure, batteries, as the core backup power source for data center power systems, directly impact network reliability and maintenance costs. Existing mainstream battery degradation monitoring methods generally rely on preset static voltage thresholds for anomaly detection, neglecting the objective law that voltage distribution parameters continuously drift with health status during battery aging, leading to a significant disconnect between assessment results and actual degradation levels. Especially in large-scale deployment scenarios, millions of individual batteries exhibit significant individual differences and time-varying characteristics, placing higher demands on the adaptive capabilities, quantification accuracy, and location efficiency of monitoring systems. To address these technical problems, this invention proposes a threshold-self-consistent battery degradation assessment method based on normal distribution. This method estimates voltage distribution parameters online, dynamically generates adaptive thresholds, calculates a comprehensive degradation index, performs multi-dimensional correlation analysis and hazard location, and generates visualized results and alarm information. This achieves high-precision, low-false-alarm, fine-grained, and maintenance-free intelligent diagnostic capabilities, and is applied to the threshold-self-consistent battery degradation assessment method based on normal distribution.

[0024] Reference Appendix Figure 1 The overall technical solution of the present invention includes: S1: Collect and preprocess battery voltage data, filter preset target data through the southbound acquisition cluster, and remove outliers based on rated voltage level; S2: Estimate the voltage distribution parameters of the preset target data online, and use the recursive moment estimation algorithm to calculate the mean and standard deviation in real time for the individual voltage sequence of a specific voltage level, and use the current mean and current standard deviation as distribution coupling parameters; S3: Dynamically generate adaptive thresholds, and calculate the upper limit and lower limit of the dynamic threshold based on the distribution coupling parameters, the standard normal quantiles corresponding to the confidence coefficients, the historical attenuation correction amount and the sensitivity factor. S4: Calculate the battery degradation index. Based on the distribution coupling parameters, dynamic threshold, and new battery calibration parameters, calculate the quantitative degradation index value by considering the proportion of voltage exceeding the dynamic threshold, the degree of drift of the distribution center relative to the calibration center, and the degree of expansion of the distribution dispersion relative to the calibration dispersion. S5: Perform multidimensional correlation analysis and hazard location, construct a multi-level multidimensional tensor data structure, mark the individual with the deterioration index value exceeding the preset threshold as an abnormal state, use tensor decomposition algorithm to extract the main components, and realize the location and drilling analysis of the first K hazard paths. S6: Generates visualization results and alarm information, integrates and displays the voltage histogram, fitted normal distribution curve and dynamic threshold upper and lower limits, highlights batteries that exceed the standard, interactively jumps to the hidden danger details view, and writes the exceeding record to the hidden danger details table to trigger the corresponding level of alarm.

[0025] In the aforementioned battery degradation assessment method based on a normal distribution and threshold self-consistency, step S1 involves collecting and preprocessing battery voltage data. Specifically, performance telemetry data of lead-acid batteries is collected in real time from power environment monitoring systems in various provinces across the country via a southward acquisition cluster. The data filtering criteria are strictly limited to: equipment type code 07, equipment type is lead-acid battery, signal standard name is single-cell voltage, signal code is 007303, and signal type is telemetry. This filtering mechanism ensures a high degree of consistency and accuracy of the data source for subsequent processing, eliminating interference from other types of equipment or signals.

[0026] During the data cleaning phase, the effective voltage range is first defined based on the battery's rated voltage level. For a single cell with a rated voltage of 2 volts, the effective voltage range is defined as 1 volt to 3 volts; for a 6-volt single cell, the effective range is 4 volts to 9 volts; and for a 12-volt single cell, the effective range is 10 volts to 18 volts. Any voltage data point outside the corresponding effective range is identified as a hard outlier and immediately removed from the data sequence to prevent extreme noise from having a destructive impact on subsequent statistical modeling.

[0027] In addition, to improve data quality, the cleaning process also includes sliding window filtering. A sliding window with a length of 5 data points is used, and a medium-range filtering algorithm is applied to smooth the original voltage sequence. This operation can effectively remove impulse noise caused by external electromagnetic interference or momentary sensor failures, while preserving the true trend of the voltage signal to the greatest extent. For continuous data gaps caused by communication interruptions, the system uses a linear interpolation method to fill in the gaps. Specifically, if N consecutive data points are missing (N=2), the nearest valid data points before and after the missing segment are linearly fitted to generate intermediate missing values, thereby ensuring the continuity and integrity of the voltage data sequence in the time dimension and providing a reliable basis for subsequent online parameter estimation.

[0028] In the aforementioned threshold-consistent battery degradation assessment method based on normal distribution, step S2 involves online estimation of voltage distribution parameters. Specifically, for each preprocessed individual cell voltage data sequence at a specific voltage level, the system employs a recursive moment estimation algorithm to calculate the mean μ and standard deviation σ of its voltage distribution in real time. The core advantage of this algorithm lies in its low computational complexity and small memory footprint, making it highly suitable for deployment on resource-constrained edge computing nodes or large-scale stream processing environments.

[0029] Using recursive moment estimation: ; u k : Estimate the sample mean based on the first k data points; u k-1 Estimate the sample mean based on the first k-1 data points; v k The voltage measurement value at time k; k: The number of data points currently processed (counting from 1); Recurrence formula for auxiliary variables: ; S k : Auxiliary variables used to calculate variance (corrected sum of squares); S k-1 Auxiliary variables from the previous time step; During initialization, the initial state is manually set as k=0, μ0=0, s0=0, and S1=0 is defined by the Welford algorithm. k=2, let μ1=v1, s1=0; Formula for calculating standard deviation: ; ; : Sample standard deviation based on the first k data points; k-1: Bessel correction factor (degrees of freedom); The recursive range is k=2,...,n, and the recursive calculation starts from the second data point; The final distributed coupling parameters are: ; X c Voltage distribution center parameter (mean); Voltage distribution width parameter (standard deviation); In battery monitoring scenarios, voltage data is continuously generated streaming data. Traditional methods for batch calculating the mean (μ) and standard deviation (σ) require storing all historical data, which can lead to significant storage and computational burdens when the data volume is large. The Welford algorithm is an online, recursive algorithm used to calculate the sample mean and sample variance / standard deviation.

[0030] No need to store historical data: Each calculation only needs to retain the results of the previous calculation (mean M, auxiliary variable S) and the current data count n.

[0031] High numerical stability: Directly using the batch calculation formula Σ(x-μ)^2 can easily lose precision due to rounding errors when calculating large amounts of data. The Welford algorithm effectively alleviates this problem through a recursive approach.

[0032] High computational efficiency: When a new data point arrives, only a few addition, subtraction, multiplication, and division operations are needed to update the estimated value, making it very suitable for real-time monitoring systems.

[0033] During algorithm initialization, k=0, μ0=0, s0=0. When the first voltage data point v1 arrives, k=1, μ1=v1, and s1=0. Starting from the second data point (k=2), the calculation is performed strictly according to the above recursive formula.

[0034] Window Mechanism Explanation: The above recursion applies to all historical data. In practical applications, to better reflect the latest state of battery voltage, a sliding window mechanism can be used. For example, only data from the most recent N = 720 hours (30 days) can be used. The implementation can be as follows: Maintain a queue of length N. Each time new data is added, if the queue is full, first remove the oldest data point, and subtract its impact on the mean and S value using a similar recursive method, before adding the new data point for a positive update.

[0035] In the aforementioned battery degradation assessment method based on a normal distribution and a self-consistent threshold, step S3 dynamically generates an adaptive threshold. Specifically, based on the current mean μ and current standard deviation σ of the distribution coupling parameters obtained in S2, a confidence coefficient α is defined with a default value of 0.95, and the upper limit is... and lower limit The calculations are as follows: ; in: (Standard normal quantiles); The historical decay correction is given by the exponentially weighted moving average (EWMA): ,

[0036] Smoothing coefficient, with a value of 0.2, controls the decay rate of historical information. The mean change reflects the drift trend of the voltage center. Current monitoring status adjustment amount : The amount of monitoring status adjustment in the previous moment; Based on real-time estimated voltage distribution parameters and historical degradation trends, an adaptive threshold is dynamically generated, which can automatically adjust as the battery ages, improving the accuracy of anomaly detection.

[0037] β is the sensitivity factor, calibrated via grid search: β=0.15 for 2V battery, β=0.40 for 6V battery, and β=0.60 for 12V battery.

[0038] By introducing the historical decay correction Δ_decay and the sensitivity factor η, the dynamic threshold of this invention is not only based on the current statistical distribution, but also incorporates historical evolution trends and prior knowledge of the device, achieving true "threshold self-consistency". That is, the threshold can be automatically and reasonably adjusted as the battery ages, avoiding a large number of missed reports caused by distribution drift in the later stages of the battery life cycle by the traditional fixed threshold method.

[0039] In the aforementioned threshold-consistent battery degradation assessment method based on normal distribution, step S4 involves calculating the battery degradation index (DI). Specifically, this step aims to transform complex distribution patterns into a single, quantifiable numerical indicator for easy ranking, comparison, and alerting. The degradation index DI is calculated by comprehensively considering three core dimensions: the proportion of voltage exceeding the dynamic threshold, the degree of drift of the distribution center relative to the calibration center, and the extent of expansion of the distribution dispersion relative to the calibration dispersion.

[0040] The new battery calibration parameters are set for a 2-volt battery with a mean of 2.25 volts and a standard deviation of 0.02 volts. The degradation index is calculated using the following formula:

[0041] μ0,ω0 are the calibration values ​​of the new battery; N out The number of voltage samples exceeding the dynamic threshold; N: Total number of samples in the query window; N out / N: Exceedance rate, reflecting the proportion of abnormal voltage readings; X c Voltage distribution center; By integrating information from three dimensions—frequency of voltage anomalies, mean drift, and fluctuation changes—a unified degradation index is generated, providing a quantitative basis for battery health status assessment and early warning. A DI value greater than 0.15 triggers a level 2 alarm, and a DI value greater than 0.30 triggers a level 1 alarm.

[0042] This embodiment organically integrates information from three dimensions, enabling the DI value to comprehensively and objectively reflect the overall degradation level of the battery. The system sets a tiered alarm strategy based on the DI value: when DI exceeds 0.15, a level 2 alarm is triggered, alerting maintenance personnel; when DI exceeds 0.30, a level 1 alarm is triggered, indicating a serious risk of battery degradation requiring immediate attention.

[0043] In the aforementioned battery degradation assessment method based on a normal distribution and threshold self-consistency, step S5 involves performing multidimensional correlation analysis and hazard localization. Specifically, a five-dimensional tensor is established, encompassing the individual unit, data center, site, city, and province. The dimensions correspond to provincial, municipal, site, data center, and individual unit levels, respectively; batteries with DI>0.15 are set to 1, and the rest are set to 0; the maximum component is extracted by CP decomposition to achieve rapid drilling of Top-K hidden danger paths K=20.

[0044] Where: N prov Provincial level, N city Dimensions of specific cities under the provincial level, N site Dimensions of a specific site, N room Dimensions of a specific data center, N batt The dimension of a single battery cell. Tensor values, each cell stores the DI value of that cell or a binary flag: 1 for abnormal and 0 for normal.

[0045] The size of each dimension of the tensor is dynamically determined based on the actual scale of the monitoring network: D1 is the number of provinces, for example, 31; D2 is the set of cities under the jurisdiction of each province; D3 is the set of communication sites contained in each city; D4 is the set of equipment rooms contained in each site; and D5 is the set of battery cells contained in each equipment room. Each cell T[i,j,k,l,m] in the tensor T stores the degradation index DI value of the m-th cell located in the i-th province, j-th city, k-th site, and l-th equipment room. Alternatively, for efficient anomaly pattern mining, the system can also binarize this tensor to generate a new binary tensor B, where B[i,j,k,l,m] = 1 indicates an anomaly if DI > 0.15, and otherwise 0 indicates normal.

[0046] In the aforementioned battery degradation assessment method based on a normal distribution and threshold self-consistency, step S6 generates visual results and alarm information. Specifically, this invention designs a front-end interface that integrates display and interactive analysis, aiming to improve human-machine collaboration efficiency.

[0047] At the visualization level, the system overlays three core elements in the same view: First, a statistical histogram of the battery cell voltage, which visually displays the actual distribution of the voltage data; second, a normal distribution fitting curve plotted based on the mean μ and standard deviation σ estimated in S2, used to verify whether the data conforms to the normality assumption and to show the theoretical distribution; and finally, the dynamic threshold upper and lower limits UCL and LCL determined in S3, usually marked with prominent dashed or solid lines. For batteries exceeding the standard identified in S4, i.e., cells with DI values ​​exceeding the limit, the system will highlight them with a bright color such as red in the corresponding voltage range of the histogram.

[0048] In terms of user interaction, when a user clicks on a highlighted area, a floating window pops up displaying the total number of abnormal units within that voltage range. If the user clicks on this count, the system will redirect to a detailed list of potential hazards. This list displays specific information about all relevant abnormal units and supports exporting the list data to standard table formats such as CSV or Excel for offline analysis or work order dispatch.

[0049] Regarding alarm information generation, the system writes each identified record of exceeding the degradation index into a structured hazard details table. This table includes key fields such as: province name, city name, site identifier, data center identifier, battery rated capacity, calculated degradation index value, dynamic threshold version number used, and data timestamp. These fields collectively uniquely identify a battery degradation hazard record and serve as the basis for subsequent alarm notifications, work order generation, and historical tracing. Alarm information triggers different levels of notifications based on the DI value, ensuring that maintenance resources are allocated accurately and efficiently.

[0050] Building upon Example 1, and considering the significant impact of different geographical environments such as high temperature, high humidity, and high altitude on the aging rate of batteries, this invention can be further extended into an environmentally adaptive degradation assessment method. Specifically, in the online parameter estimation process of S2, in addition to calculating the global mean μ and standard deviation σ, the system also performs environmental correction on the distributed parameters based on the latitude and longitude of the individual battery location and the real-time environmental data (temperature, humidity, and air pressure) provided by the local weather station.

[0051] The calibration model can be a pre-trained regression model, such as a gradient boosting tree (GBDT) or a small neural network, whose input is an environmental feature vector and output is a correction value for the offset of the mean and standard deviation. The distribution parameters after environmental calibration better reflect the "normal" state of the battery under specific environmental conditions, making the dynamic threshold generated in S3 and the degradation index calculated in S4 more regionally adaptable, further reducing the false alarm rate caused by environmental differences. This embodiment is particularly suitable for large-scale communication networks spanning multiple regions and climate zones, enabling truly refined and intelligent operation and maintenance.

[0052] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

[0053] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces, or indirect coupling or communication connection between apparatuses or units, and may be electrical, mechanical, or other forms.

[0054] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated units described above can be implemented in hardware or as software functional units. The above are merely embodiments of this application and do not limit the patent scope of this application. Any equivalent structural or procedural transformations made based on the description and drawings of this application, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

[0055] The specific embodiments of the invention have been described in detail above, but they are only examples, and this application is not limited to the specific embodiments described above. For those skilled in the art, any equivalent modifications or substitutions to the invention are also within the scope of this application. Therefore, all equivalent changes, modifications, and improvements made without departing from the spirit and principles of this application should be covered within the scope of this application.

Claims

1. A self-consistent threshold assessment method for battery degradation based on normal distribution, characterized in that, The specific steps include the following: S1: Collect and preprocess battery voltage data, filter preset target data through the southbound acquisition cluster, and remove outliers based on rated voltage level; S2: Estimate the voltage distribution parameters of the preset target data online, and use the recursive moment estimation algorithm to calculate the mean and standard deviation in real time for the individual voltage sequence of a specific voltage level, and use the current mean and current standard deviation as distribution coupling parameters; S3: Dynamically generate adaptive thresholds, and calculate the upper limit and lower limit of the dynamic threshold based on the distribution coupling parameters, the standard normal quantiles corresponding to the confidence coefficients, the historical attenuation correction amount and the sensitivity factor. S4: Calculate the battery degradation index. Based on the distribution coupling parameters, dynamic threshold, and new battery calibration parameters, calculate the quantitative degradation index value by considering the proportion of voltage exceeding the dynamic threshold, the degree of drift of the distribution center relative to the calibration center, and the degree of expansion of the distribution dispersion relative to the calibration dispersion. S5: Perform multidimensional correlation analysis and hazard location, construct a multi-level multidimensional tensor data structure, mark the individual with the deterioration index value exceeding the preset threshold as an abnormal state, use tensor decomposition algorithm to extract the main components, and realize the location and drilling analysis of the first K hazard paths. S6: Generates visualization results and alarm information, integrates and displays the voltage histogram, fitted normal distribution curve and dynamic threshold upper and lower limits, highlights batteries that exceed the standard, interactively jumps to the hidden danger details view, and writes the exceeding record to the hidden danger details table to trigger the corresponding level of alarm.

2. The battery degradation assessment method based on a normal distribution and threshold self-consistency as described in claim 1, characterized in that... The data cleaning in S1 also includes performing sliding window mid-range filtering on the original voltage data, with a window size of 5 data points, and using linear interpolation to fill in consecutive missing data points.

3. The battery degradation assessment method based on a normal distribution and threshold self-consistency as described in claim 1, characterized in that, The recursive moment estimation algorithm in S2 adopts a sliding window mechanism. When the window is full and new data points arrive, the influence of the oldest data points on the mean and auxiliary variables is first removed by reverse recursion, and then the distribution coupling parameters are updated by adding new data points by forward recursion.

4. The battery degradation assessment method based on a threshold self-consistency of normal distribution according to claim 1, characterized in that, The confidence coefficient in S3 is 0.95, corresponding to a standard normal quantile of 1.

96. The smoothing coefficient of the historical attenuation correction is 0.

2. The sensitivity factor is configured differently according to the rated voltage: 0.15 for 2-volt batteries, 0.40 for 6-volt batteries, and 0.60 for 12-volt batteries.

5. The battery degradation assessment method based on a normal distribution threshold self-consistency according to claim 1, characterized in that, The new battery calibration parameters in S4 are set for a 2-volt battery with a mean of 2.25 volts and a standard deviation of 0.02 volts. The degradation index is calculated using the following formula: μ0,ω0 are the calibration values ​​of the new battery; N out The number of voltage samples exceeding the dynamic threshold; N: Total number of samples in the query window; N out / N: Exceedance rate, reflecting the proportion of abnormal voltage readings; X c Voltage distribution center; By integrating information from three dimensions—frequency of voltage anomalies, mean drift, and fluctuation changes—a unified degradation index is generated, providing a quantitative basis for battery health status assessment and early warning. A DI value greater than 0.15 triggers a level 2 alarm, and a DI value greater than 0.30 triggers a level 1 alarm.

6. The battery degradation assessment method based on a threshold self-consistency of normal distribution according to claim 1, characterized in that, The size of each dimension of the five-dimensional tensor in S5 is dynamically determined according to the actual monitoring network scale. The tensor cell stores the degradation index value or the binary anomaly flag. An anomaly flag of 1 indicates that the degradation index exceeds 0.15, and 0 indicates that it is normal.

7. The battery degradation assessment method based on a normal distribution and threshold self-consistency as described in claim 6, characterized in that, The tensor decomposition algorithm in S5 uses normal decomposition to decompose the five-dimensional tensor into the sum of multiple rank tensors, and extracts the top 20 components with the largest weights to identify the 20 most significant abnormal propagation paths from the provincial level to the individual level.

8. The battery degradation assessment method based on a threshold self-consistency of normal distribution according to claim 1, characterized in that, The hidden danger details table in S6 includes the following fields: province name, city name, site identifier, computer room identifier, battery rated capacity, deterioration index value, dynamic threshold version number, and data timestamp, which together uniquely identify a hidden danger record.

9. The battery degradation assessment method based on a threshold self-consistency of normal distribution according to claim 1, characterized in that, In the S6 interface, the voltage statistical histogram, the normal distribution fitting curve drawn based on the estimated parameters, and the upper and lower limits of the dynamic threshold are displayed in superimposed with different colors and line types. Exceeding batteries are highlighted on the histogram. When the user clicks on the highlighted area, the abnormal cell count will pop up. Further clicking will jump to the detailed cell information list page and can be exported as a standard format table file.

10. The battery degradation assessment method based on a normal distribution and threshold self-consistency as described in claim 1, characterized in that, The entire evaluation method is implemented based on a distributed streaming data processing framework. Data acquisition and distribution are completed through message queues. The distributed coupled analysis engine and threshold self-consistent optimizer are deployed in the form of microservices, and the battery voltage data is processed in real time at fixed time intervals or sliding time windows.

Images