Battery intelligent maintenance work order generation and management method

Abstract

The invention provides a battery intelligent maintenance work order generation and management method. The method comprises the following steps: collecting multi-dimensional data such as battery foundation, association, operation and historical maintenance to construct a data set, and carrying out standardization processing; inputting the data into a parallel execution multi-dimensional association judgment model, identifying a battery to be maintained, aiming at the battery to be maintained, determining a weight coefficient by constructing a judgment matrix in which each dimension is compared pairwise and executing consistency verification, calculating a priority score and dividing a processing level; and automatically generating a plan work order containing the object, the reason, the expert database matching content and the execution time limit based on the judgment result and the priority. In addition, the system iteratively optimizes the trigger threshold, the weight coefficient and the precursor feature library according to the maintenance feedback data. According to the method, the technical problems of potential fault judgment omission and low operation and maintenance resource scheduling efficiency caused by single data dimension, inaccurate judgment logic and lack of a hierarchical linkage mechanism in an existing battery maintenance mode can be solved.

Description

Technical Field

[0001] This invention relates to the field of energy storage equipment maintenance technology, specifically to a method for generating and managing intelligent battery maintenance work orders. Background Technology

[0002] With the rapid development of the new energy industry, power batteries and energy storage batteries are increasingly widely used in industrial and commercial energy storage systems, new energy commercial vehicles, and special equipment. To ensure the safe operation of battery systems and extend their service life, intelligent maintenance throughout their entire lifecycle has become a core requirement of the industry. However, existing battery maintenance and work order management technologies still have the following prominent problems in practical applications: First, traditional maintenance methods rely primarily on fixed time periods or single-parameter alarm thresholds. This approach fails to adequately incorporate the early evolution characteristics of battery failure and does not consider the recurrence frequency of similar faults. Due to the lack of a multi-dimensional parallel judgment mechanism, healthy batteries are often over-maintained while sub-healthy batteries are missed, making accurate prediction before failures occur difficult.

[0003] Furthermore, conventional maintenance processes lack standardized hierarchical and collaborative mechanisms. Existing systems typically only provide simple alarms upon detecting anomalies, lacking priority ranking logic based on multi-dimensional weights. This results in the inability to scientifically allocate maintenance resources when dealing with large-scale battery clusters. Simultaneously, the generated maintenance plans are vague, failing to accurately map fault types to repair actions, heavily relying on manual experience, and exhibiting poor execution timeliness.

[0004] In summary, how to accurately determine and prioritize battery maintenance needs through multi-dimensional correlated data is a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention proposes a method for generating and managing intelligent battery maintenance work orders, in order to solve the technical problems of single data dimensions and severe fragmentation in traditional maintenance methods.

[0006] The technical solution adopted in this invention is a method for generating and managing intelligent battery maintenance work orders, comprising: Collect multi-dimensional data of the battery and retrieve historical maintenance data to construct a multi-dimensional dataset; The multidimensional dataset is classified, encoded, and normalized to generate a standardized dataset. The standardized dataset is input into the association determination model for maintenance association determination. In response to any dimension of the standardized dataset satisfying the trigger threshold of the maintenance association determination, the maintenance requirement determination result of the battery is obtained, and the battery is added to the list of batteries requiring maintenance. The correlation determination model performs maintenance correlation determination in parallel for the following determination dimensions: component supplier correlation determination, production batch correlation determination, system solution correlation determination, fault precursor correlation determination, cycle life correlation determination, and recurring fault correlation determination. For batteries included in the maintenance requirement list, a weighted calculation is performed based on the maintenance requirement determination results to obtain a priority score, and the maintenance priority level is determined based on the priority score. Based on the maintenance requirement determination results and the priority level, a maintenance plan work order is generated.

[0007] Furthermore, the criteria for determining the association between the component suppliers are as follows: Statistics on the failure rate and repeat repair rate of components from all suppliers over the past three months; In response to the failure rate being greater than a set threshold and the repeat repair rate being greater than a set threshold, all batteries using components from the current supplier are included in the battery list requiring maintenance.

[0008] Furthermore, the determination criteria for the production batch association are as follows: Statistics were compiled on the failure rate and concentration rate of all batches of batteries within the past month. The formula for calculating the concentration ratio of the same type is: in, Indicates the concentration rate of the same type. B represents the number of faults of the same type, B represents batch information, and T represents the fault type. Indicates the number of faults in the batch; In response to a failure rate of less than 10% or a concentration rate of the same type of battery greater than 60%, all non-faulty batteries in the current batch are included in the battery list requiring maintenance.

[0009] Furthermore, the determination criteria for the system scheme association are as follows: Calculate the mean time between battery failures (MTBF) for all system configurations over the past two months: in, Indicates the total number of operating days. Indicates the number of days a single battery can operate. Indicates the total number of failures; In response to a system failure count of 5 or more or a battery average failure interval of less than 90 days, all batteries under the current system scheme will be included in the battery list requiring maintenance.

[0010] Furthermore, the determination criteria for the correlation between the fault precursors are as follows: The original feature set is constructed by extracting multidimensional historical feature vectors from historical fault samples. The K-means++ clustering algorithm was used to perform cluster analysis on the original feature set, extract the typical feature vectors corresponding to each type of fault, and construct a precursor feature library; Extract the multidimensional real-time feature vector from the real-time operating data of the target battery, and calculate its cosine similarity with the typical feature vectors described in the precursor feature library: in, Represents cosine similarity. This represents a multidimensional historical feature vector representing 72 hours of operation data for a single battery. This represents a typical feature vector in the precursor feature library; The multidimensional historical feature vector and typical feature vector are both composed of battery operating parameters and their corresponding mean, peak value, standard deviation and duration statistics. If the cosine similarity is greater than or equal to 80%, or if the real-time monitoring parameters of the target battery exceed a preset single parameter threshold, the target battery will be added to the list of batteries requiring maintenance.

[0011] Furthermore, the determination criteria for the cycle life correlation are as follows: Get the running days for all batteries; The lifespan percentage of all batteries is calculated based on the ratio of the current number of battery cycles to the designed number of battery cycles. In response to the target battery's operating days being greater than or equal to the designed maintenance cycle threshold, or its lifespan percentage being greater than or equal to the lifespan threshold, the target battery is added to the list of batteries requiring maintenance.

[0012] Furthermore, the determination criteria for the recurring fault association are as follows: Statistics were compiled on the number of times the same type of fault occurred and the performance recovery rate after repair for all batteries over the past 6 months. The formula for calculating the performance recovery rate after repair is as follows: in, Indicates the performance recovery rate after repair. This indicates the actual health status of the battery after repair. This indicates the actual battery health status before repair; If the number of recurrences of the same type of fault is greater than or equal to 2 or the performance recovery rate after repair is less than 85%, the target battery will be added to the list of batteries requiring maintenance.

[0013] Furthermore, for batteries included in the maintenance requirement list, a weighted calculation is performed based on the maintenance requirement determination results to obtain a priority score, and the maintenance priority level is determined based on the priority score, including: For each battery included in the maintenance list, in response to the trigger threshold of the maintenance association determination under a certain determination dimension, the state value of that dimension is mapped to a first preset score; otherwise, the state value of that dimension is mapped to a second preset score. For each battery included in the maintenance list, a judgment matrix is ​​constructed based on the scale values ​​generated by pairwise comparisons of the maintenance impact degree of all its judgment dimensions. Calculate the largest eigenvector of the judgment matrix and normalize it to obtain the initial weights of each judgment dimension; Calculate the consistency ratio of the judgment matrix; if the consistency ratio of the judgment matrix is ​​less than 0.1, then determine the initial weight as the weight coefficient. The overall score for each battery is calculated based on the state values ​​of each dimension and their corresponding weighting coefficients. The calculation formula is as follows: in, This represents the overall score. This represents the state value of the fault precursor dimension. This represents the status value at the component supplier level. This represents the status value at the production batch level. Represents the state values ​​of the system solution dimension. This represents the state value in the lifecycle dimension. This represents the state value of the fault precursor dimension. , , , , , These represent the weight coefficients corresponding to the state values ​​of each dimension; In response to the overall score being greater than or equal to the emergency priority threshold, the target battery corresponding to the current overall score is included in the emergency processing list of the first priority. In response to the overall score being less than the emergency priority threshold but greater than or equal to the normal priority threshold, the target battery corresponding to the current overall score is included in the normal processing list of the second priority. In response to the overall score being less than the conventional priority threshold, the target battery corresponding to the current overall score is included in the third priority plan processing list.

[0014] Furthermore, based on the maintenance requirement determination result and the priority level, a maintenance plan work order is generated, including: By associating the battery ID with its model, application scenario, and address information, and then concatenating the fields, maintenance object information is generated. Retrieve the dimension name that triggered the judgment and its corresponding key real-time data or statistical values ​​to generate maintenance reason information; Based on the determined fault type, the system retrieves the preset maintenance measures expert database, matches the corresponding maintenance actions, torque parameters, and tool and spare parts requirements to generate maintenance content information; Based on the priority level mapping preset response time, the execution time limit information of the batteries in the emergency processing list is set to a first preset time, the execution time limit information of the batteries in the regular processing list is set to a second preset time, and the execution time limit information of the batteries in the planned processing list is set to a third preset time.

[0015] Furthermore, it also includes iterative adjustments to the trigger threshold for the maintenance association determination, the weighting coefficient, and the precursor feature library: The accuracy rate, missed rate, and actual high-risk percentage of the emergency response list are calculated based on the collected maintenance feedback data. If the determination accuracy is less than the accuracy threshold, it is determined that the sensitivity is too high, and the trigger threshold of the dimension corresponding to the current determination accuracy is increased by a preset step size of 0.5%. In response to the missed detection rate being greater than the missed detection rate threshold, the trigger threshold of the dimension corresponding to the current missed detection rate is adjusted down by a preset step size of 0.5%. If the actual high-risk percentage of the emergency response list is less than the weight matching threshold, it is determined that the weight allocation is unreasonable. The scale value in the judgment matrix is ​​adjusted so that the weight coefficient of the corresponding dimension is corrected by ±5%. For new fault cases that do not match known fault precursors, extract their multidimensional feature vectors before the fault, perform incremental clustering using the K-means++ algorithm, and update the set of typical feature vectors in the precursor feature library.

[0016] As can be seen from the above technical solution, the beneficial technical effects of the present invention are as follows: 1. It changes the previous maintenance mode that relied on a single parameter threshold or fixed cycle. By integrating multi-dimensional correlation data of batteries, it can identify common risks caused by batch defects or systemic design problems. Combined with the fault precursor feature library extracted based on K-means++, it improves the false negative rate and identification accuracy of potential faulty batteries.

[0017] 2. By constructing pairwise comparison judgment matrices for the judgment dimensions and performing consistency ratio verification, the decision-making logic is transformed into a mathematical weight model, which can accurately distinguish between three different levels of urgency: urgent, routine, and planned maintenance needs. This allows maintenance resources to be prioritized for high-risk batteries, significantly optimizing the scheduling efficiency of manpower and spare parts while ensuring the safety of the battery system cluster and reducing the overall maintenance cost.

[0018] 3. The system can automatically calculate the accuracy of judgments and the rationality of weight allocation based on actual maintenance feedback data, and accordingly perform step-by-step iterative optimization of the trigger threshold, comparison matrix scale value, and feature library. This adaptive mechanism enables the system to continuously correct the model as the battery life evolves, effectively solving the problem that static judgment rules are difficult to adapt to the nonlinear degradation of the battery. Attached Figure Description

[0019] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0020] Figure 1 This is a flowchart of a method according to an embodiment of the present invention; Figure 2 This is a schematic diagram illustrating the feature set structure of an embodiment of the present invention. Detailed Implementation

[0021] The embodiments of the technical solution of the present invention will now be described in detail with reference to the accompanying drawings. These embodiments are merely illustrative of the technical solution of the present invention and are therefore intended to limit the scope of protection of the present invention.

[0022] It should be noted that, unless otherwise stated, the technical or scientific terms used in this application should have the ordinary meaning as understood by one of ordinary skill in the art to which this invention pertains.

[0023] Example 1 This embodiment discloses a method and system for generating and managing intelligent battery maintenance work orders. This method is compatible with lithium iron phosphate and ternary lithium-ion batteries with single-cell voltages of 2.5V-4.2V and capacities of 50Ah-500Ah, aiming to achieve intelligent maintenance processes through end-to-end data integration. The working principle of Embodiment 1 is explained in detail below: The method flowchart of this embodiment is as follows: Figure 1 As shown, it includes: Multi-dimensional data of the battery is collected through BMS, and historical maintenance data is retrieved from the database to construct a multi-dimensional dataset. Classify, encode, and normalize multidimensional datasets to generate standardized datasets; The standardized dataset is input into the association determination model for maintenance association determination. In response to any dimension of the standardized dataset meeting the trigger threshold for maintenance association determination, the maintenance requirement determination result of the battery is obtained, and the battery is added to the list of batteries requiring maintenance. For batteries included in the maintenance requirement list, a weighted calculation is performed based on the maintenance requirement assessment results to obtain a priority score, and the maintenance priority level is determined based on the priority score. Based on the maintenance needs assessment and priority level, a maintenance plan work order is generated.

[0024] In this embodiment, further, during the data acquisition phase, the system constructs a multi-dimensional data acquisition system, which includes basic identity data, associated attribute data, dynamic operation data, and historical maintenance data.

[0025] Basic identification data includes battery ID, model, installation date, design life, and fixed maintenance cycle; The associated attribute data covers core component suppliers, production batches, system solution numbers, and battery lists for the same batch / solution. Dynamic operating data includes acquiring cell voltage, charge / discharge current, and battery surface temperature at a frequency of once per second, and calculating voltage fluctuation rate, temperature change rate, and SOH at a frequency of once per hour, while recording the number of cycles with a discharge depth ≥80%; Historical maintenance data records the time of failure, type of failure, repair measures, and number of recurring failures. This multi-dimensional data is categorized and stored, with real-time data stored in InfluxDB and static and historical data stored in MySQL. A global association index is established using the battery ID to ensure that the data coverage is comprehensive across all dimensions.

[0026] In this embodiment, the collected multi-dimensional dataset further enters the preprocessing stage. The system encodes the categorical data, for example, supplier A=1, B=2…, batch 202305-LFP-01=101, 202306-LFP-01=102… Furthermore, for numerical data, a normalization formula is used. Normalization is performed to map the data to the [0,1] interval, eliminating the influence of units and generating a standardized dataset.

[0027] For the preprocessed dataset, the system uses a multidimensional association determination model for parallel analysis. The association determination model performs maintenance association determination in parallel for the following determination dimensions: component supplier association determination, production batch association determination, system solution association determination, fault precursor association determination, cycle life association determination, and recurring fault association determination.

[0028] In determining the association between parts suppliers and components, the failure rate and repeat repair rate of all suppliers' parts over the past three months are calculated using the following formula: in, Indicates the percentage of failures. This indicates the number of failures of component C from supplier S. This indicates the total number of components C installed (excluding new batteries installed less than 7 days ago to avoid break-in period malfunctions). This indicates the rate of repeated repairs. This indicates the number of times component C requires repeated maintenance.

[0029] If the failure rate is greater than 5% and the repeat repair rate is greater than the repeat repair rate threshold of 8%, all batteries using components from the current supplier will be included in the list of batteries requiring maintenance.

[0030] In this embodiment, the further determination criteria for production batch association are as follows: Statistics were compiled on the failure rate and concentration rate of all batches of batteries within the past month. The formula for calculating the failure rate is: in, Indicates the failure rate. Indicates the number of faults in the batch. This indicates the total number of batteries in the batch.

[0031] The formula for calculating the concentration ratio of the same type is: in, Indicates the concentration rate of the same type. B represents the number of faults of the same type, B represents batch information, and T represents the fault type. This indicates the number of faults in the batch.

[0032] In response to a failure rate of less than 10% or a concentration rate of the same type of battery greater than 60%, all non-faulty batteries in the current batch will be included in the list of batteries requiring maintenance.

[0033] In this embodiment, the further determination condition for system scheme association is: Calculate the mean time between battery failures (MTBF) for all system configurations over the past two months: in, Indicates the total number of operating days. Indicates the number of days a single battery can operate. Indicates the total number of failures; In response to a system failure count of 5 or more or a battery average failure interval of less than 90 days, all batteries under the current system scheme will be included in the battery list requiring maintenance.

[0034] In this embodiment, the fault precursor association determination is further performed using a feature database for deep identification. The system first extracts 128-dimensional feature vectors from 1500 historical fault samples to form an original feature set, which is structured as follows: Figure 2 As shown, each feature vector consists of the mean, peak value, standard deviation, and duration statistics corresponding to 32 parameters such as voltage, temperature, SOC, and current, and clearly distinguishes between static and dynamic features.

[0035] For the original feature set, the system uses an improved K-means++ clustering algorithm to perform analysis. The specific steps include: First, feature weights are adaptively allocated, and the priority of each feature dimension is determined through a combination of logic: a 1-9 scaling method plus fault correlation correction. The system follows the analytic hierarchy process (AHP) standard to establish initial scaling values ​​(1 represents equal importance, 3, 5, 7, and 9 represent increasing importance, 2, 4, 6, and 8 are intermediate values, and the reciprocal is used for inverse importance). Then, mutual information is used to quantify the correlation strength between each feature and the fault category. If the difference in mutual information between two features is greater than 0.3, the scaling value is increased by one level; if the difference is less than 0.1, it is decreased by one level. This constructs a judgment matrix, where elements... This indicates the degree of influence of the i-th dimension relative to the j-th dimension.

[0036] Furthermore, the system performs consistency checks and weight determination on the judgment matrix, and solves for the largest eigenvalue of the judgment matrix using the eigenvalue method. and corresponding feature vectors: in, Represents the judgment matrix. express , Represents the eigenvalues ​​of the judgment matrix; Furthermore, the consistency index is calculated: in, This indicates the order of the decision matrix, i.e., the number of decision dimensions; Finally, calculate the consistency ratio. The formula is in, This represents the average random consistency index, and its value is related to... Relevant values ​​can be directly referenced from existing standard values: =1 =0, =2 =0, =3 =0.52, =4 =0.89, =5 o'clock =1.12,…, =10 o'clock =1.49.

[0037] In response to If the deviation is ≥0.1, the consistency ratio does not meet the requirements. First, locate the element with the largest deviation in the judgment matrix, and then calculate the consistency deviation value corresponding to each element. Deviation value: The element with the largest deviation value is the key adjustment item; Prioritize adjusting the scale value corresponding to the element, with the adjustment range controlled within ±1 level, to avoid excessive deviation from the original comparison logic; If CR is still ≥0.1 after a single adjustment, then the scale values ​​of the 2-3 elements with the second largest deviation values ​​are adjusted simultaneously. Repeat the above steps until CR < 0.1. At this point, normalize the corresponding feature vectors and use them as the initial weights for each decision dimension. Then, make fine adjustments in increments of ±5% to finally determine the weight coefficients. For key features with mutual information values ​​greater than 0.6, the final weight is maintained at 1.2-1.5 times. For weakly correlated features with mutual information values ​​less than 0.2, the final weight is maintained at 0.5-0.8 times to reduce the interference of irrelevant features on the clustering results.

[0038] In the data processing and center calibration stages, the system uses min-max normalization for static features and Z-score standardization for dynamic features to eliminate dimensional differences. During cluster center initialization, the system presets a K value of 8 to 12 based on the fault type and compares the initial center with typical samples in the historical feature library using Euclidean distance. If the Euclidean distance between the initial center and the nearest typical sample is greater than 0.3, the center is automatically corrected to the feature vector of a known typical sample to avoid cluster center shift risks.

[0039] Finally, the system refines the model by fusing clustering and convergence criteria. The preprocessed static and dynamic features are fused into a unified matrix, input into the improved K-means++ algorithm, and subjected to 50 iterations of clustering using the aforementioned weight coefficients. Clustering stops when the iteration error reaches a threshold of 1e-4, ultimately outputting five typical feature vectors corresponding to each fault class.

[0040] During real-time determination, the system extracts the multi-dimensional historical feature vector A from the target battery's 72-hour data and compares it using the cosine similarity formula, which is as follows: in, Represents cosine similarity. This represents a multidimensional historical feature vector representing 72 hours of operation data for a single battery. This represents a typical feature vector in the precursor feature library; Both the multidimensional historical feature vector and the typical feature vector are composed of battery operating parameters and their corresponding mean, peak value, standard deviation and duration statistics. The target battery will be added to the list of batteries requiring maintenance if any of the following criteria are met: Cosine similarity greater than or equal to 80%; The voltage parameters include a difference in individual unit voltages greater than 0.15V; Battery state of health (SOH) is less than 80%; The surface temperature included in the temperature parameters is greater than 55°C.

[0041] In this embodiment, the further determination condition for cycle life correlation is: Get the running days for all batteries; The lifespan percentage of all batteries is calculated based on the ratio of the current number of battery cycles to the designed number of battery cycles. If the number of operating days of the target battery is greater than or equal to the design maintenance cycle threshold (180 days for industrial and commercial energy storage scenarios, and 90 days for commercial vehicle scenarios due to complex operating conditions), or if the lifespan percentage is greater than or equal to 80%, the target battery will be included in the list of batteries requiring maintenance.

[0042] In this embodiment, the further determination condition for the recurring fault association is: Statistics were compiled on the number of times the same type of fault occurred and the performance recovery rate after repair for all batteries over the past 6 months. The formula for calculating the performance recovery rate after repair is: in, Indicates the performance recovery rate after repair. This indicates the actual health status of the battery after repair. This indicates the actual battery health status before repair; If the number of recurrences of the same type of fault is greater than or equal to 2 or the performance recovery rate after repair is less than 85%, the target battery will be added to the list of batteries requiring maintenance.

[0043] For batteries included in the maintenance list, the system uses the Analytic Hierarchy Process (AHP) to prioritize them.

[0044] In response to the trigger threshold of the maintenance association judgment under a certain judgment dimension, the state value of that dimension is mapped to 100 points; otherwise, the state value of that dimension is mapped to 0 points. Furthermore, for the battery mentioned above, a judgment matrix is ​​constructed based on the scale values ​​generated by pairwise comparison of the impact of each dimension on maintenance. The maximum eigenvector is calculated using the eigenvector method and normalized to obtain the initial weights of each dimension.

[0045] The system calculates the consistency ratio in real time. If the consistency ratio of the judgment matrix is ​​less than 0.1, the initial weights are determined as the weight coefficients, and the judgment matrix is ​​deemed valid, thus determining the final weight coefficients. The comprehensive score for each battery is calculated using the comprehensive scoring formula: in, This represents the overall score. This represents the state value of the fault precursor dimension. This represents the status value at the component supplier level. This represents the status value at the production batch level. Represents the state values ​​of the system solution dimension. This represents the state value in the lifecycle dimension. This represents the state value of the fault precursor dimension. , , , , , These represent the weight coefficients corresponding to the state values ​​of each dimension; Based on the score P, the target battery corresponding to the current comprehensive score will be included in three priorities: the first priority emergency handling list (P≥80), the second priority routine handling list (50≤P<80), and the third priority planned handling list (P<50).

[0046] In this embodiment, the system further performs refined field concatenation and mapping when generating maintenance plan work orders. Maintenance object information is constructed by concatenating the model, application scenario, and installation address fields from the battery ID association database; maintenance reason information automatically traces back the dimension name that triggered the judgment and the corresponding real-time abnormal value; maintenance content information is retrieved from the expert database based on the identified fault type, matches specific operating procedures, and lists the required tools and spare parts models; execution time limit information is mapped to a preset response time based on priority level. Finally, the system pushes the work order through the maintenance APP and includes an overtime SMS reminder mechanism.

[0047] In this embodiment, a mechanism for comparing work order judgment results with actual maintenance feedback is further established by automatically checking feedback data on the 1st of each month; the accuracy of the judgment is calculated based on the collected maintenance feedback data, and the specific steps include: The accuracy rate, false negative rate, and actual high-risk percentage of the emergency response list are calculated based on the collected maintenance feedback data. If the accuracy is less than 85%, the model is considered too sensitive, and the trigger threshold for the corresponding dimension will be automatically increased by 0.5% increment. If the false negative rate is greater than 5%, a reduction of 0.5% will be implemented.

[0048] To ensure the rationality of priority allocation, the system calculates the actual high-risk percentage in emergency work orders. If this percentage is below 90%, the weighting is deemed unreasonable, and the system automatically fine-tunes the pairwise comparison scale values ​​in the judgment matrix to keep the weight fluctuation within ±5%. Furthermore, the system extracts feature vectors from new fault cases with unmatched features and performs incremental clustering using the K-means++ algorithm, thereby achieving dynamic updates to the precursor feature library and continuous optimization of its completeness.

[0049] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be covered within the scope of the claims and specification of the present invention.

Claims

1. A method for generating and managing intelligent battery maintenance work orders, characterized in that, include: Collect multi-dimensional data of the battery and retrieve historical maintenance data to construct a multi-dimensional dataset; The multidimensional dataset is classified, encoded, and normalized to generate a standardized dataset. The standardized dataset is input into the association determination model for maintenance association determination. In response to any dimension of the standardized dataset satisfying the trigger threshold of the maintenance association determination, the maintenance requirement determination result of the battery is obtained, and the battery is added to the list of batteries requiring maintenance. The correlation determination model performs maintenance correlation determination in parallel for the following determination dimensions: component supplier correlation determination, production batch correlation determination, system solution correlation determination, fault precursor correlation determination, cycle life correlation determination, and recurring fault correlation determination. For batteries included in the maintenance requirement list, a weighted calculation is performed based on the maintenance requirement determination results to obtain a priority score, and the maintenance priority level is determined based on the priority score. Based on the maintenance requirement determination results and the priority level, a maintenance plan work order is generated.

2. The method for generating and managing intelligent battery maintenance work orders according to claim 1, characterized in that, The criteria for determining the association between the component suppliers are as follows: Statistics on the failure rate and repeat repair rate of components from all suppliers over the past three months; In response to the failure rate being greater than a set threshold and the repeat repair rate being greater than a set threshold, all batteries using components from the current supplier are included in the battery list requiring maintenance.

3. The method for generating and managing intelligent battery maintenance work orders according to claim 1, characterized in that, The determination criteria for the production batch association are as follows: Statistics were compiled on the failure rate and concentration rate of all batches of batteries within the past month. The formula for calculating the concentration ratio of the same type is: in, Indicates the concentration rate of the same type. B represents the number of faults of the same type, B represents batch information, and T represents the fault type. Indicates the number of faults in the batch; In response to a failure rate of less than 10% or a concentration rate of the same type of battery greater than 60%, all non-faulty batteries in the current batch are included in the battery list requiring maintenance.

4. The method for generating and managing intelligent battery maintenance work orders according to claim 1, characterized in that, The determination criteria for the system scheme association are as follows: Calculate the mean time between battery failures (MTBF) for all system configurations over the past two months: in, Indicates the total number of operating days. Indicates the number of days a single battery can operate. Indicates the total number of failures; In response to a system failure count of 5 or more or a battery average failure interval of less than 90 days, all batteries under the current system scheme will be included in the battery list requiring maintenance.

5. The method for generating and managing intelligent battery maintenance work orders according to claim 1, characterized in that, The determination criteria for the correlation between the fault precursors are as follows: Extract feature vectors from historical fault samples containing both static and dynamic battery features to form the original feature set; Calculate the mutual information values ​​between each feature dimension and the fault category in the original feature set, and construct a judgment matrix by combining the relative importance between each dimension; A consistency check is performed on the judgment matrix, and the weights are corrected based on the mutual information value to obtain the weight coefficients of each feature dimension. After preprocessing and cluster center calibration of the original feature set, multimodal fusion clustering is performed using the improved K-means++ algorithm to extract typical feature vectors corresponding to each type of fault and construct a precursor feature library. Extract the multidimensional real-time feature vector from the real-time operating data of the target battery, and calculate its cosine similarity with the typical feature vectors described in the precursor feature library: in, Represents cosine similarity. This represents a multidimensional historical feature vector representing 72 hours of operation data for a single battery. This represents a typical feature vector in the precursor feature library; The multidimensional historical feature vector and typical feature vector are both composed of battery operating parameters and their corresponding mean, peak value, standard deviation and duration statistics. If the cosine similarity is greater than or equal to 80%, or if the real-time monitoring parameters of the target battery exceed a preset single parameter threshold, the target battery will be added to the list of batteries requiring maintenance.

6. The method for generating and managing intelligent battery maintenance work orders according to claim 1, characterized in that, The determination criteria for the cycle life correlation are as follows: Get the running days for all batteries; The lifespan percentage of all batteries is calculated based on the ratio of the current number of battery cycles to the designed number of battery cycles. In response to the target battery's operating days being greater than or equal to the designed maintenance cycle threshold, or its lifespan percentage being greater than or equal to the lifespan threshold, the target battery is added to the list of batteries requiring maintenance.

7. The method for generating and managing intelligent battery maintenance work orders according to claim 1, characterized in that, The determination criteria for the recurring fault association are as follows: Statistics were compiled on the number of times the same type of fault occurred and the performance recovery rate after repair for all batteries over the past 6 months. The formula for calculating the performance recovery rate after repair is as follows: in, Indicates the performance recovery rate after repair. This indicates the actual health status of the battery after repair. This indicates the actual battery health status before repair; If the number of recurrences of the same type of fault is greater than or equal to 2 or the performance recovery rate after repair is less than 85%, the target battery will be added to the list of batteries requiring maintenance.

8. The method for generating and managing intelligent battery maintenance work orders according to claim 1, characterized in that, For batteries included in the maintenance requirement list, a weighted calculation is performed based on the maintenance requirement determination results to obtain a priority score, and the maintenance priority level is determined based on the priority score, including: For each battery included in the maintenance list, in response to the trigger threshold of the maintenance association determination under a certain determination dimension, the state value of that dimension is mapped to a first preset score; otherwise, the state value of that dimension is mapped to a second preset score. For each battery included in the maintenance list, a judgment matrix is ​​constructed based on the scale values ​​generated by pairwise comparisons of the maintenance impact degree of all its judgment dimensions. Calculate the largest eigenvector of the judgment matrix and normalize it to obtain the initial weights of each judgment dimension; Calculate the consistency ratio of the judgment matrix; if the consistency ratio of the judgment matrix is ​​less than 0.1, then determine the initial weight as the weight coefficient. The overall score for each battery is calculated based on the state values ​​of each dimension and their corresponding weighting coefficients. The calculation formula is as follows: in, This represents the overall score. This represents the state value of the fault precursor dimension. This represents the status value at the component supplier level. This represents the status value at the production batch level. Represents the state values ​​of the system solution dimension. This represents the state value in the lifecycle dimension. This represents the state value of the fault precursor dimension. , , , , , These represent the weight coefficients corresponding to the state values ​​of each dimension; In response to the overall score being greater than or equal to the emergency priority threshold, the target battery corresponding to the current overall score is included in the emergency processing list of the first priority. In response to the overall score being less than the emergency priority threshold but greater than or equal to the normal priority threshold, the target battery corresponding to the current overall score is included in the normal processing list of the second priority. In response to the overall score being less than the conventional priority threshold, the target battery corresponding to the current overall score is included in the third priority plan processing list.

9. A method for generating and managing intelligent battery maintenance work orders according to claim 8, characterized in that, Based on the maintenance requirement assessment results and the priority level, a maintenance plan work order is generated, including: By associating the battery ID with its model, application scenario, and address information, and then concatenating the fields, maintenance object information is generated. Retrieve the dimension name that triggered the judgment and its corresponding key real-time data or statistical values ​​to generate maintenance reason information; Based on the determined fault type, the system retrieves the preset maintenance measures expert database, matches the corresponding maintenance actions, torque parameters, and tool and spare parts requirements to generate maintenance content information; Based on the priority level mapping preset response time, the execution time limit information of the batteries in the emergency processing list is set to a first preset time, the execution time limit information of the batteries in the regular processing list is set to a second preset time, and the execution time limit information of the batteries in the planned processing list is set to a third preset time.

10. A method for generating and managing intelligent battery maintenance work orders according to claim 5, characterized in that, It also includes iterative adjustments to the trigger threshold for the maintenance association determination, the weighting coefficient, and the precursor feature library: The accuracy rate, missed rate, and actual high-risk percentage of the emergency response list are calculated based on the collected maintenance feedback data. If the determination accuracy is less than the accuracy threshold, it is determined that the sensitivity is too high, and the trigger threshold of the dimension corresponding to the current determination accuracy is increased by a preset step size of 0.5%. In response to the missed detection rate being greater than the missed detection rate threshold, the trigger threshold of the dimension corresponding to the current missed detection rate is adjusted down by a preset step size of 0.5%. If the actual high-risk percentage of the emergency response list is less than the weight matching threshold, it is determined that the weight allocation is unreasonable. The scale value in the judgment matrix is ​​adjusted so that the weight coefficient of the corresponding dimension is corrected by ±5%. For new fault cases that do not match known fault precursors, extract their multidimensional feature vectors before the fault, perform incremental clustering using the K-means++ algorithm, and update the set of typical feature vectors in the precursor feature library.

Images