A battery monitoring and early warning method and device based on parameter space statistical learning

Abstract

This application discloses a battery monitoring and early warning method and device based on parameter space statistical learning, relating to the field of electrochemical energy storage anomaly monitoring. The method includes: constructing a multi-dimensional operating condition feature vector and monitoring indicators for the battery, and discretizing the multi-dimensional operating condition space into uniquely identified subspace units. Based on historical normal operation data, the sample mean and standard deviation of the monitoring indicators in each subspace are statistically analyzed to establish its historical statistical distribution model. During real-time monitoring, the corresponding subspace is located according to the current operating condition. If the data volume is insufficient, the parent space is expanded level by level for retrieval. When the data is sufficient, the deviation of the current monitoring value is calculated to determine the anomaly point, and an early warning is issued using the anomaly frequency within a sliding window, while the statistical model is adaptively updated. This application reduces the error rate of battery cell anomaly identification and early warning, improves the adaptive capability and interpretability of the statistical distribution model, and can autonomously optimize and update the judgment criteria, thereby improving the system's self-learning capability.

Description

Technical Field

[0001] This application relates to the field of anomaly monitoring in electrochemical energy storage, and in particular to a battery monitoring and early warning method and device based on parameter space statistical learning. Background Technology

[0002] With the acceleration of the global energy transition, electrochemical energy storage systems have become a key infrastructure in areas such as smart grids, renewable energy consumption, and electric vehicles. As the core component of energy storage systems, the safety, reliability, and lifespan of batteries directly determine the performance and economy of the entire system. However, batteries are affected by complex external conditions (such as charge / discharge rate, ambient temperature, operating voltage, and state of charge (SOC)) and internal aging mechanisms during operation, exhibiting highly nonlinear evolution characteristics and making them highly susceptible to performance degradation and even serious failures such as thermal runaway. Currently, the industry commonly uses fixed threshold methods for early warning of battery anomalies. For example, setting an upper voltage limit of 3.65V, a lower limit of 2.8V, an upper temperature limit of 45℃, and an upper current limit of 1.5C. When the monitored data exceeds these preset thresholds, the system triggers an alarm. While this method is simple and direct, its inherent defects are becoming increasingly apparent, as detailed below.

[0003] 1) Poor adaptability to operating conditions: The performance boundaries of a battery are not fixed. For example, when charging at high rates at low temperatures, the battery is more prone to lithium plating, which changes the voltage plateau. If the voltage threshold at room temperature is still used for judgment, it may lead to false alarms or missed alarms. Similarly, at different SOCs, the battery's internal resistance and heat generation rate are also different, and fixed temperature or current thresholds cannot accurately reflect the true safety boundaries.

[0004] 2) Ignoring individual differences: Even batteries of the same model can exhibit varying performance parameters due to differences in production batches, raw materials, and initial capacity distribution. A uniform, fixed threshold cannot accommodate all individuals, potentially leading to overly stringent requirements for some batteries while being too lenient for others.

[0005] 3) High false alarm rate: Under complex operating conditions, the voltage, current, temperature and other signals of the battery may fluctuate normally. The fixed threshold method cannot distinguish between such normal fluctuations and real anomalies, which can easily cause false alarms and reduce the reliability of the monitoring system.

[0006] 4) Inability to learn and evolve: Battery performance degrades with age. Fixed thresholds cannot adaptively adjust to the battery's state of health (SOH), potentially failing to provide effective protection later in the battery's lifespan.

[0007] To overcome the limitations of the fixed threshold method, academia and industry have begun to explore advanced approaches, such as model-based methods (e.g., electrochemical models, equivalent circuit models) and data-driven methods (e.g., machine learning, deep learning). However, these methods often suffer from computational complexity, require precise parameters, rely on large amounts of fault data for training, and exhibit strong "black box" characteristics, making them difficult to deploy on a large scale in engineering practice.

[0008] Therefore, there is an urgent need for an adaptive early warning method for battery anomalies that can adapt to complex and variable working conditions and individual differences, while also being easy to implement in engineering, highly interpretable, and requiring no complex models. Summary of the Invention

[0009] The purpose of this application is to provide a battery monitoring and early warning method and device based on parameter space statistical learning, which can reduce the error rate of battery cell anomaly identification and early warning, improve the adaptability and interpretability of the statistical distribution model, and enable the system to autonomously optimize and update the judgment criteria, thereby improving the system's self-learning ability.

[0010] To achieve the above objectives, this application provides the following solution.

[0011] Firstly, this application provides a battery monitoring and early warning method based on parameter space statistical learning. The method includes: constructing a multidimensional feature vector and monitoring and early warning indicators for a target battery cell; the multidimensional feature vector includes multiple operating condition parameters; discretizing each operating condition parameter in the multidimensional feature vector into multiple levels based on the range and accuracy requirements of physical parameters; treating each level of each operating condition parameter as a subspace unit, and constructing a multidimensional operating condition parameter space based on all subspace units; wherein each subspace unit corresponds to a unique identifier; acquiring historical normal operation data of the target battery cell; the normal operation data is data that has not triggered monitoring and early warning; the historical normal operation data includes multidimensional feature vectors and monitoring data for multiple historical data points; and dividing the historical normal operation data using subspace units. The system collects historical data points falling into each subspace unit; calculates the sample mean and standard deviation of the monitoring data for each subspace unit to obtain the historical distribution of the monitoring and early warning indicators for each subspace unit, and constructs a statistical distribution model; it acquires the multidimensional feature vector and monitoring data of the target battery unit at the current moment in real time, and indexes the current subspace unit according to the identifier; it determines whether the sample of the current subspace unit is greater than the threshold. If not, it expands the search range to the parent space level by level and returns "determine whether the sample of the current subspace unit is greater than the threshold"; if so, it determines the anomaly by calculating the deviation between the monitoring data at the current moment and the historical distribution, and performs monitoring and early warning based on the frequency of anomalies within the sliding window; it stores the normal operation data at the current moment into the historical database and adaptively updates the statistical distribution model.

[0012] Secondly, this application provides a computer device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the battery monitoring and early warning method based on parameter space statistical learning described above.

[0013] Thirdly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the battery monitoring and early warning method based on parameter space statistical learning described above.

[0014] Fourthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the battery monitoring and early warning method based on parameter space statistical learning described above.

[0015] Based on the specific embodiments provided in this application, the following technical effects are disclosed.

[0016] This application constructs a multi-dimensional operating condition feature vector and monitoring indicators for batteries, and discretizes the multi-dimensional operating condition space into uniquely identified subspace units, achieving refined modeling of complex operating environments. Based on historical normal operation data, the sample mean and standard deviation of monitoring indicators within each subspace are statistically analyzed to establish its historical statistical distribution model. This process does not require complex electrochemical models or deep learning algorithms, reducing computational load and making it easy to implement. It can be embedded into edge devices of the battery management system for real-time operation. In the real-time monitoring phase, this application locates the corresponding subspace based on the current operating condition. If the data volume is insufficient, an extended retrieval mechanism is initiated, progressively expanding to the parent space for judgment, effectively solving the problem of judgment in sparse data regions and ensuring the completeness of the method. When data is sufficient, the standardized deviation of the current monitoring value relative to the historical distribution is calculated, and an early warning is issued based on the anomaly frequency within the sliding window. This method, through multi-dimensional operating condition matching and statistical distribution comparison, makes the early warning standard more scientific, significantly outperforming the fixed threshold method, achieving high accuracy and low false alarm rate. It also improves interpretability. Furthermore, this application automatically stores real-time normal data into a historical database and regularly updates the statistical distribution model, enabling the model to automatically adapt to different individual batteries, aging states, and external operating conditions, possessing strong adaptive capabilities and lifelong learning and evolutionary characteristics. Moreover, the parameters of this application, such as dimension selection, subspace granularity, deviation coefficient, and window length, can be flexibly configured according to specific applications, exhibiting high adaptability and scalability, providing a dynamic, accurate, and self-evolving intelligent early warning solution for battery status monitoring. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating a battery monitoring and early warning method based on parameter space statistical learning, provided as an embodiment of this application.

[0019] Figure 2 This is a three-dimensional schematic diagram of the multi-dimensional working condition parameter space provided in the embodiments of this application.

[0020] Figure 3 This is a schematic diagram of the voltage distribution in a subspace of operating parameters provided in an embodiment of this application.

[0021] Figure 4 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0022] 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 some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0023] This application aims to completely resolve the deficiencies in the existing technology, specifically addressing the following core pain points:

[0024] 1) Solve the problem that fixed thresholds cannot adapt to changing working conditions: By constructing a high-dimensional working condition parameter space, the anomaly judgment is changed from "fixed value" to "comparison with historical performance under the same working conditions", making the early warning standard dynamic and contextualized.

[0025] 2) Solving the problem of the inapplicability of a uniform threshold due to individual differences in batteries: This application essentially learns the historical behavior patterns of a specific battery or a group of similar batteries under specific working conditions, which naturally includes the statistical characteristics of individuals and groups, without the need to artificially set a uniform threshold.

[0026] 3) Reduce false alarm rate and improve early warning reliability: By introducing a mechanism that triggers an early warning only when "multiple consecutive sampling points are abnormal", single-point false anomalies caused by noise or transient interference can be effectively filtered out.

[0027] 4) Solving the problem of identifying sparse data regions: By designing a subspace expansion retrieval mechanism, even when there is little historical data for the current working condition, the statistical distribution of similar working conditions can be found for reference and judgment by expanding the retrieval scope, thus ensuring the completeness of the method.

[0028] 5) Achieve continuous knowledge accumulation and self-updating: As the system runs, it can automatically accumulate more data, continuously update and enrich the statistical distribution model of each working condition subspace, and achieve self-evolution of performance.

[0029] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0030] Example 1, as Figure 1 As shown in the figure, this embodiment provides a battery monitoring and early warning method based on parameter space statistical learning, the method including the following steps.

[0031] S1. Construct a multidimensional feature vector and monitoring and early warning indicators for the target battery cell; the multidimensional feature vector includes multiple operating condition parameters.

[0032] Furthermore, the operating parameters include at least one of the following: charge / discharge rate, ambient temperature, battery surface temperature, state of charge, open-circuit voltage, and number of cycles.

[0033] Furthermore, the monitoring and early warning indicators include at least one of the following: single-point voltage, voltage change rate, temperature, temperature change rate, current, and internal resistance.

[0034] In practical applications, step S1 is mainly used to define the M-dimensional parameter vector P (multidimensional feature vector) that characterizes the battery's operating conditions: .

[0035] in, These can be the standardized charge / discharge rates ( ), ambient temperature ( ), battery surface temperature ( ), state of charge (SOC), open circuit voltage (OCV), number of cycles, etc., each dimension can be selected and increased or decreased according to actual needs.

[0036] S2. Based on the range of physical parameters To meet accuracy requirements, each working condition parameter in the multidimensional feature vector... Discretized into multiple levels ( (Levels).

[0037] S3. For example Figures 2-3As shown, each level of each working condition parameter is treated as a subspace unit, and a multi-dimensional working condition parameter space is constructed based on all subspace units; where each subspace unit corresponds to a unique identifier.

[0038] In practical applications, step S3 divides the entire M-dimensional working space into an M-dimensional grid, which has a total of Subspace unit.

[0039] Each unit It can be uniquely identified by an M-dimensional index.

[0040] .

[0041] , .

[0042] S4. Obtain historical normal operation data of the target battery cell; the normal operation data is data that has not triggered monitoring and early warning; the historical normal operation data includes multi-dimensional feature vectors of multiple historical data points and monitoring data.

[0043] S5. Divide the historical normal operation data into subspace units and count the historical data points that fall into each subspace unit.

[0044] In practical applications, step S5 will iterate through all historical normal operation data. For each historical data point (current time) t, its operating condition is as follows: The monitoring value is ;according to Calculate its subspace unit For each subspace unit Collect monitoring values ​​of all historical data points falling into this subspace cell. .

[0045] S6. Calculate the sample mean and sample standard deviation of the monitoring data for each subspace unit to obtain the historical distribution of the monitoring and early warning indicators for each subspace unit, and construct a statistical distribution model.

[0046] In practical applications, the formula for the sample mean is: .

[0047] The formula for the sample standard deviation is: .

[0048] in, This represents the number of valid samples in the subspace.

[0049] S7. Real-time acquisition of the multidimensional feature vector and monitoring data of the target battery cell at the current moment, and indexing the current subspace cell according to the identifier.

[0050] In practical applications, step S7 will monitor time t in real time and obtain multidimensional feature vectors. and monitoring data Simultaneously locate the subspace: calculate using the identifier index. Corresponding subspace unit .

[0051] S8. Determine whether the sample size of the current subspace unit is greater than the threshold. If not, expand the search range to the parent space level by level and return "Determine whether the sample size of the current subspace unit is greater than the threshold". If yes, determine the anomaly by calculating the deviation between the current monitoring data and the historical distribution, and perform monitoring and early warning based on the frequency of anomalies within the sliding window.

[0052] Furthermore, the search scope is expanded level by level to the parent space, specifically including the following steps.

[0053] 1) Take the current subspace unit as the original unit, and set the expansion level starting from the original unit.

[0054] 2) Expand the indexes of each dimension of the original unit to both sides according to the expansion level to construct the parent space.

[0055] 3) Calculate the total number of samples in the parent space, determine whether the total number of samples is greater than the sample threshold. If not, increase the value of the preset unit by 1, update the preset unit, and return "Expand the original unit indexes of each dimension to both sides by the preset units to build the parent space". If the total number of samples is still less than the sample threshold when the maximum expansion level is reached, mark it as unknown working condition. If so, use the merged statistics to judge the anomaly.

[0056] In practical applications, the process of progressively expanding the search scope to the parent space (subspace expansion search mechanism) is as follows.

[0057] When the current subspace Data volume If insufficient, initiate an extended search.

[0058] Define extension levels , Represents the original unit.

[0059] For extended levels Expand the index range of each dimension of the original unit to both sides. Each unit forms a larger hypercube region (parent space).

[0060] Calculate the total number of samples in all subspace cells within the parent space. The combined mean of all monitored values and pooled standard deviation , and .

[0061] like Then use and Perform anomaly detection.

[0062] If the requirements are still not met, increase the expansion level. This continues until a parent space that meets the data volume requirements is found or the maximum expansion level is reached. If it reaches If no data is available, it can be marked as "Unknown operating condition" and a conservative fixed threshold method can be used, or "Cannot be determined" can be output.

[0063] Furthermore, the sliding window comprises multiple consecutive sampling points.

[0064] Furthermore, by calculating the deviation between the current monitoring data and the historical distribution, anomalies are identified, and monitoring and early warning are carried out based on the frequency of anomalies within the sliding window. Specifically, the following steps are included.

[0065] Calculate the deviation between the current monitoring data and the historical distribution.

[0066] When the deviation exceeds the deviation threshold, the sampling point is considered an outlier.

[0067] A sliding window is used to continuously identify outliers. When the number of outliers within the sliding window exceeds the outlier count threshold, a monitoring and early warning system is activated.

[0068] Furthermore, the formula for calculating the deviation is as follows.

[0069] .

[0070] in, for Deviation in time; for Real-time monitoring data; For subspace units The sample mean; For subspace units The sample standard deviation.

[0071] In practical applications, the process of checking data sufficiency is as follows: If the number of valid samples in the subspace... Greater than the preset threshold If the search is successful, proceed to the next step; otherwise, jump to expand the search scope level by level to the parent space.

[0072] Calculate deviation: Calculate the current value Relative to subspace Historical distribution Standardized deviation.

[0073] Preliminary assessment: If (in If the preset coefficient is used (usually 2.5, 3, or 3.5), then that moment is marked as an outlier.

[0074] Continuous assessment and early warning: Set a sliding window of length W (e.g., W=5). If the number of outliers exceeds a threshold in the most recent W consecutive sampling points... ( If the condition is met, a deterministic early warning message will be generated and reported.

[0075] S9. Store the current normal operation data in the historical database and adaptively update the statistical distribution model.

[0076] In practical applications, step S9 requires periodic or trigger-based updates to the statistical model to ensure it can track slow changes in battery performance. Normal data points that do not trigger warnings are added to the historical database. The statistical distribution model is updated periodically (e.g., daily, weekly), and all subspaces are recalculated. , , .

[0077] This application uses a multi-dimensional feature vector composed of charge / discharge rate, ambient temperature, and SOC as an example to illustrate the process of the entire battery monitoring and early warning method, as follows.

[0078] Implementation environment: Battery management system of a large-scale energy storage power station.

[0079] Monitoring objective: To provide early warning of "abnormal voltage" in a specific battery within a battery cluster.

[0080] Specific parameters: Operating condition parameter vector P: [charge / discharge rate ( ), ambient temperature ( SOC]->M=3-dimensional. Target index X: Battery voltage (V).

[0081] The spatial discretization process is as follows.

[0082] 1) :[0,2]C, discrete into 20 equal parts->ΔC=0.1C.

[0083] 2) :[-10,50]°C, discrete into 30 equal parts -> ΔT=2°C.

[0084] 3) SOC: [0,100]%, discrete into 20 equal parts -> ΔSOC=5%.

[0085] The total number of subspaces is 20 * 30 * 20 = 12,000.

[0086] Parameters to be determined: =50, K=3.0, W=5, =3.

[0087] Update strategy: Recalculate the statistics of all subspaces offline every 24 hours.

[0088] The battery monitoring and early warning process is as follows.

[0089] 1) Initialization: The system loads historical data of the power plant’s normal operation over the past year and constructs a statistical model containing 12,000 subspaces according to step three.

[0090] 2) Real-time monitoring: At 14:30 on a certain day, the system collected data from a battery: =0.8C, =28°C, SOC=65%, Voltage=3.45V.

[0091] 3) Spatial location: Calculate the subspace index to which the point belongs, as follows.

[0092] =floor(0.8 / 0.1)=8.

[0093] =floor((28-(-10)) / 2)=19.

[0094] =floor(65 / 5)=13.

[0095] Locate cell S(8,19,13).

[0096] Statistical query: The query reveals the information for this unit. =1200> (50), =3.52V, =0.03V.

[0097] Anomaly detection: Calculation =|3.45-3.52| / 0.03=2.33<3.0. Single-point judgment is normal.

[0098] Subsequent developments: From 14:30 to 14:34 (5 consecutive sampling points, 1 point per minute), the battery voltage was 3.45V, 3.42V, 3.41V, 3.40V, and 3.40V. The calculated... The values ​​are 2.33, 3.33, 3.67, 4.00, and 4.00, respectively.

[0099] Warning triggered: Within the sliding window (W=5), there are 4 points (the last 4) >3.0, exceeding =3. The system triggered a voltage anomaly warning at 14:34.

[0100] Action taken: After receiving the warning, the maintenance personnel inspected the battery unit and found that the connection strip was loose. They promptly addressed the issue, preventing a potential thermal runaway accident.

[0101] The technical advantages of this application are as follows.

[0102] 1) High accuracy and low false alarm rate: Through multi-dimensional working condition matching and statistical distribution comparison, the early warning standard is more scientific and significantly better than the fixed threshold method, greatly reducing false alarms and missed alarms.

[0103] 2) Strong adaptability: The method automatically adapts to different individual batteries, different aging states, and different external working conditions, making it highly versatile.

[0104] 3) Good interpretability: The cause of the warning can be traced back to "the current value X deviates from the historical normal level by K standard deviations under operating condition P", which is easy to understand and diagnose.

[0105] 4) Ease of engineering implementation: No complex electrochemical or deep learning models are required, the computational load is relatively small, and it is suitable for real-time operation in edge computing devices embedded in battery management systems.

[0106] 5) Self-learning and evolution: The system can continuously collect data, optimize itself, and update its judgment criteria, thus possessing lifelong learning capabilities.

[0107] 6) Flexible and configurable: Parameters such as dimension, subspace granularity, deviation coefficient K, and window length W can all be adjusted according to specific application scenarios, providing high flexibility.

[0108] This application can be widely used for health status monitoring, safety early warning and life management of various electrochemical energy storage systems such as lithium-ion batteries, sodium-ion batteries and flow batteries in scenarios such as new energy vehicles, energy storage power stations and backup power supplies.

[0109] Example 2: This application also provides a computer device, which may be a server or a terminal, and its internal structure diagram may be as follows. Figure 4As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores and processes data. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network connection. When the computer program is executed by the processor, it implements the methods described above.

[0110] Those skilled in the art will understand that Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0111] Example 3: This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the methods described above.

[0112] Example 4: This application also provides a computer program product, including a computer program that, when executed by a processor, implements the methods described above.

[0113] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0114] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0115] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A battery monitoring and early warning method based on parametric space statistical learning, characterized in that, The method comprises: constructing a multi-dimensional feature vector of a target battery unit and a monitoring early warning index; the multi-dimensional feature vector comprises a plurality of working condition parameters; discretizing each working condition parameter in the multi-dimensional feature vector into a plurality of levels based on a physical parameter range and an accuracy requirement; taking each level of each working condition parameter as a subspace unit, and constructing a multi-dimensional working condition parameter space based on all subspace units; each subspace unit corresponds to a unique identifier; obtaining historical normal operation data of the target battery unit; the normal operation data is data that has not triggered a monitoring early warning; the historical normal operation data comprises a multi-dimensional feature vector and monitoring data of a plurality of historical data points; dividing the historical normal operation data using the subspace units, and counting the historical data points falling into each subspace unit; calculating the sample mean and sample standard deviation of the monitoring data of each subspace unit, obtaining the historical distribution of the monitoring early warning index of each subspace unit, and constructing a statistical distribution model; obtaining the multi-dimensional feature vector and monitoring data of the target battery unit at a current time in real time, and indexing the current subspace unit according to the identifier; determining whether the sample of the current subspace unit is greater than a threshold value, if not, expanding the search range to the parent space level by level, and returning to "determining whether the sample of the current subspace unit is greater than a threshold value", if yes, determining an abnormal point by calculating the deviation degree of the monitoring data at the current time from the historical distribution, and performing monitoring early warning based on the frequency of the abnormal point in the sliding window; storing the normal operation data at the current time to a historical database, and adaptively updating the statistical distribution model. 2.The battery monitoring and early warning method based on parametric space statistical learning according to claim 1, wherein, The working condition parameters at least include any one of the following: charge-discharge rate, ambient temperature, battery surface temperature, state of charge, open circuit voltage, and cycle number. 3.The battery monitoring and warning method based on parameter space statistical learning according to claim 1, wherein, The monitoring early warning index at least includes any one of the following: single-point voltage, voltage change rate, temperature, temperature change rate, current, and internal resistance. 4.The battery monitoring and warning method based on parameter space statistical learning according to claim 1, wherein, The search range is expanded to the parent space level by level, specifically comprising: taking the current subspace unit as an original unit, setting an expansion level from the original unit; expanding each dimension index of the original unit to both sides according to the expansion level to construct a parent space; calculating the total sample number in the parent space, determining whether the total sample number is greater than a sample threshold value, if not, increasing the value of a preset unit by 1, updating the preset unit, and returning to "expanding each dimension index of the original unit to both sides by a preset unit to construct a parent space", if the total sample number is still less than the sample threshold value when the maximum expansion level is reached, marking the working condition as unknown, if yes, performing abnormality determination using a merged statistical quantity. 5.The battery monitoring and warning method based on parameter space statistical learning according to claim 1, wherein, The sliding window comprises a plurality of continuous sampling points. 6.The battery monitoring and warning method based on parameter space statistical learning according to claim 5, wherein, The abnormal point is determined by calculating the deviation degree of the monitoring data at the current time from the historical distribution, and monitoring early warning is performed based on the frequency of the abnormal point in the sliding window, specifically comprising: calculating the deviation degree of the monitoring data at the current time from the historical distribution; when the deviation degree is greater than a deviation degree threshold value, taking the sampling point as an abnormal point; continuously determining the abnormal point using the sliding window, and performing monitoring early warning when the number of abnormal points in the sliding window is greater than an abnormal point number threshold value. 7.The battery monitoring and warning method based on parametric space statistical learning according to claim 1, wherein, The calculation formula of the deviation degree is as follows: ； wherein, is a degree of deviation of the time instant; is monitoring data of the time instant; is a sample mean of the subspace unit ; is a sample standard deviation of the subspace unit .

8. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the battery monitoring and early warning method based on parameter space statistical learning according to any one of claims 1-7.

9. A computer readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the battery monitoring and early warning method based on parameter space statistical learning according to any one of claims 1-7.

10. A computer program product comprising a computer program, characterized in that, The computer program is executed by the processor to implement the battery monitoring and early warning method based on parameter space statistical learning according to any one of claims 1-7.

Images