Operation fault judgment method and system applied to energy storage equipment

Abstract

According to the operation fault judgment method and system applied to the energy storage equipment, the accuracy and efficiency of operation fault diagnosis of the energy storage equipment are remarkably improved by constructing the prior fault feature database and effectively matching the iteration fault sample event corresponding to the fault tag. According to the method, key fault features can be automatically extracted from a large amount of historical data, and the specific fault type of the current energy storage equipment in operation can be quickly positioned and identified based on the features. By extracting the candidate fault sample event highly matched with the to-be-diagnosed fault data and combining with the advanced fault diagnosis model, the accurate diagnosis data can be generated, and a timely and accurate fault judgment result is provided for operation and maintenance personnel. Therefore, the energy loss caused by fault shutdown is reduced, the manual diagnosis cost and time are reduced, and the reliability and maintenance efficiency of the energy storage equipment are improved.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to an operation fault identification method and system applied to energy storage equipment. Background Art

[0002] In the context of the widespread application of renewable energy, energy storage equipment is a key link in energy storage and release, and its stability and reliability are crucial to maintaining the balance of the power system. However, with the popularization of energy storage equipment and the continuous advancement of technology, the challenges encountered in operation and maintenance are also increasing.

[0003] At present, the operation fault identification of energy storage equipment mainly relies on traditional maintenance mode and operator's experience judgment. This method has several significant limitations: first, manual inspection is not only time-consuming and laborious, but also difficult to cover all possible fault conditions, especially in large-scale energy storage systems, the efficiency of fault detection is low; second, relying on the operator's experience judgment is often subjective, and the judgment criteria of different operators may be different, resulting in inconsistent fault identification results; third, facing new energy storage equipment or complex fault conditions, traditional methods are often difficult to respond quickly and lack fast and effective identification methods.

[0004] In addition, with the rapid development of energy storage technology, the types of equipment and fault types are increasing, which puts higher requirements on the diversity and adaptability of fault identification methods. When faced with diverse energy storage equipment and complex and changeable fault conditions, the traditional methods are limited in their identification ability and efficiency, and it is difficult to meet the high efficiency and accuracy requirements in actual operation.

[0005] Therefore, developing a method that can automatically, quickly and accurately identify operating failures of energy storage equipment is of great significance for improving the overall operating efficiency and stability of the energy storage system. Summary of the invention

[0006] In order to at least overcome the above-mentioned deficiencies in the prior art, an object of the embodiments of the present application is to provide a method and system for determining operating faults of an energy storage device.

[0007] According to one aspect of the present application, a method for determining an operation fault of an energy storage device is provided, the method comprising:

[0008] Acquire the energy storage device operation fault data to be diagnosed and the fault label corresponding to the energy storage device operation fault data;

[0009] Determine a fault feature database matching the fault label from a priori fault feature database; the fault feature database includes event encoding vector data corresponding to iterative fault sample events matching the fault label;

[0010] Based on the event coding vector data in the fault feature database, extracting X iterative fault sample events that match the event coding vector data of the energy storage device operation fault data from the iterative fault sample events corresponding to the event coding vector data in the fault feature database; X is a positive integer;

[0011] Based on the fault state chain vectors associated with each of the X iterative fault sample events, extracting, from the X iterative fault sample events, an iterative fault sample event that matches the fault state chain vector corresponding to the energy storage device operation fault data as a candidate fault sample event;

[0012] Based on the candidate fault sample events and the energy storage device operation fault data, model input data corresponding to a fault diagnosis model is generated, and using the fault diagnosis model, diagnostic data of the energy storage device operation fault data is generated based on the model input data, and a judgment result of the energy storage device operation fault data is determined based on the diagnostic data.

[0013] According to one aspect of the embodiments of the present application, a computer program product or a computer program is provided, the computer program product or the computer program includes computer instructions, and the computer instructions are stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the methods provided in various optional implementations of the above three aspects.

[0014] In the technical solutions provided in some embodiments of the present application, the accuracy and efficiency of energy storage equipment operation fault diagnosis are significantly improved by constructing a priori fault feature database and effectively matching iterative fault sample events corresponding to fault labels. The method can automatically extract key fault features from a large amount of historical data, and quickly locate and identify specific fault types in the current energy storage device operation based on these features. By extracting candidate fault sample events that are highly matched with the fault data to be diagnosed, combined with advanced fault diagnosis models, the present invention can generate accurate diagnostic data and provide timely and accurate fault identification results for operation and maintenance personnel. This not only helps to reduce energy losses caused by fault downtime, but also reduces manual diagnosis costs and time, and improves the reliability and maintenance efficiency of energy storage equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required to be used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be extracted in combination with these drawings without creative work.

[0016] Figure 1 A schematic diagram of a flow chart of a method for determining an operating fault of an energy storage device provided in an embodiment of the present application;

[0017] Figure 2 A schematic block diagram of the architecture of an operation fault identification system for energy storage equipment provided in an embodiment of the present application. DETAILED DESCRIPTION

[0018] The following description is intended to enable one of ordinary skill in the art to implement and combine the present application, and the description is provided in the context of a specific application scenario and its requirements. It will be apparent to one of ordinary skill in the art that various changes may be made to the disclosed embodiments, and that the general principles defined in the present application may be applied to other embodiments and application scenarios without departing from the principles and scope of the present application. Therefore, the present application is not limited to the described embodiments, but should be given the broadest scope consistent with the claims.

[0019] Figure 1 This is a flow chart of an operation fault determination method applied to an energy storage device provided by an embodiment of the present application. The operation fault determination method applied to the energy storage device is introduced in detail below.

[0020] Step S110, obtaining the energy storage device operation fault data to be diagnosed and the fault label corresponding to the energy storage device operation fault data.

[0021] In this embodiment, the server is responsible for diagnosing the faults of an energy storage system of a large data center. The energy storage system is composed of multiple battery packs, power conversion modules, battery management systems and other components.

[0022] At a certain moment, the server receives a notification from the energy storage device monitoring system, indicating that the energy storage device may have a fault. The server begins to obtain the operating fault data of the energy storage device to be diagnosed. For example, the server obtains data from the sensor network of the energy storage device, which is distributed in key parts such as battery packs and power conversion modules.

[0023] For battery packs, sensors may collect data such as battery voltage, current, and temperature. For example, the voltage of multiple battery cells in a battery pack fluctuates abnormally, the voltage of some batteries drops much faster than normal, and the overall current of the battery pack is unstable. Fault data obtained from the power conversion module may include information such as reduced conversion efficiency and mismatched input and output power. For example, the input power of the power conversion module is normal, but the output power is significantly lower than expected, and abnormal heat is generated during the conversion process.

[0024] These abnormal data about the operation of the energy storage device collected from various components constitute the energy storage device operation fault data to be diagnosed.

[0025] At the same time, the server also obtains the fault labels corresponding to these fault data. The fault label is a preliminary classification of the fault type. For example, according to the preliminary analysis of the fault phenomenon, if the battery pack voltage is abnormal and the power conversion module efficiency is reduced, it may be marked with the fault label of "low energy storage efficiency fault". This label is determined based on previous fault experience and some predefined rules. For example, in the server's fault classification rule base, if the average voltage of the battery pack is lower than 80% of the normal range, and the conversion efficiency of the power conversion module is lower than 90% (normal is above 95%), it will be marked as "low energy storage efficiency fault". In this way, the server completes step S110 and obtains the energy storage equipment operation fault data to be diagnosed and the corresponding fault label.

[0026] Step S120: Determine a fault feature database that matches the fault label from the priori fault feature database. The fault feature database includes event encoding vector data corresponding to the iterative fault sample event that matches the fault label.

[0027] In this embodiment, after obtaining the fault label "low energy storage efficiency fault", the server starts to search in the prior fault feature database stored in it. The prior fault feature database contains multiple sub-databases for different fault types, and each sub-database is a fault feature database matching a fault label.

[0028] This priori fault feature database is built through long-term data accumulation and analysis. For example, in the past operation process, the data center has experienced various energy storage equipment failures. Whenever a failure occurs, the technicians will conduct a detailed analysis of the failure, including information such as the status of each component when the failure occurs and the process of the failure development. This information is organized into iterative fault sample events and encoded to form event encoding vector data stored in the corresponding fault feature database.

[0029] For the label "low energy storage efficiency fault", the server found the corresponding fault feature database in the prior fault feature database. This database stores detailed information on previous "low energy storage efficiency faults". For example, one of the iterative fault sample events may be as follows: During a high temperature period in summer, due to a failure in the air conditioning system, the heat dissipation of the battery pack was poor, the battery temperature increased, and the internal resistance of the battery increased, and the voltage dropped. At the same time, the high temperature environment also affected the performance of the power conversion module, reducing the conversion efficiency. This event was recorded in detail and converted into event encoding vector data. The event encoding vector data may contain quantized encodings of parameters such as battery temperature, battery internal resistance, power conversion module temperature, input and output power, etc. Through this search process, the server determines the fault feature database that matches the current fault label, providing a data basis for subsequent fault diagnosis.

[0030] Step S130, based on the event coding vector data in the fault feature database, extract X iterative fault sample events that match the event coding vector data of the energy storage device operation fault data from the iterative fault sample events corresponding to the event coding vector data in the fault feature database, where X is a positive integer.

[0031] Assume that there are 100 iterative fault sample events related to "low energy storage efficiency fault" and their corresponding event encoding vector data in the fault feature database determined in step S120. The server now needs to extract X iterative fault sample events that match the event encoding vector data of the current energy storage device operation fault data from these 100 events, where X=5.

[0032] First, the server needs to determine the feature matching degree between the event coding vector data of the energy storage device operation fault data and the event coding vector data in the fault feature database. For example, for the current energy storage device operation fault data, its event coding vector data contains parameter codes such as the average voltage of the battery pack is 3.2V (normal is 3.6-3.8V), the conversion efficiency of the power conversion module is 85% (normal is more than 95%), and the average temperature of the battery pack is 40℃ (normal is 25-35℃).

[0033] The server will compare these parameters with each event encoding vector data in the fault feature database. For one of the iterative fault sample events, the battery pack voltage in its event encoding vector data is 3.1V, the power conversion module conversion efficiency is 83%, and the battery pack temperature is 42°C. The server will calculate the difference between these parameters and the current fault data parameters. For example, for the battery pack voltage, the calculated relative difference is (3.2-3.1) / 3.6≈0.028 (assuming the lower limit of the normal range of 3.6V is used as the benchmark); for the power conversion module conversion efficiency, the difference is (85-83) / 95≈0.021; for the battery pack temperature, the difference is (40-42) / 25=-0.08 (here the lower limit of the normal range of 25°C is used as the benchmark). Then, according to the pre-set algorithm, these differences are combined to obtain a feature matching degree. Assuming that the weight of each parameter is set to the same, the feature matching degree of this iterative fault sample event and the current fault data is obtained by a certain weighted average algorithm, which is 0.8 (the full score is 1.0, indicating a high match).

[0034] The server calculates all 100 event encoding vector data in the fault feature database in this way, and then extracts 5 event encoding vector data with a feature matching degree greater than the set matching degree (assuming the set matching degree is 0.7) from the fault feature database. For example, in addition to the event mentioned above, another 4 event encoding vector data are found, and the differences between their battery pack voltage, power conversion module conversion efficiency, battery pack temperature and the current fault data are within an acceptable range, and the feature matching degrees are 0.82, 0.75, 0.78 and 0.85 respectively.

[0035] Finally, based on the index relationship between the event encoding vector data and the iterative fault sample event, the iterative fault sample events that have index relationships with the five event encoding vector data are determined. Just like in a database, each event encoding vector data has a unique index pointing to a detailed description of the corresponding iterative fault sample event. Through this index relationship, the server accurately finds the five iterative fault sample events that match the event encoding vector data of the current energy storage device operation fault data. These iterative fault sample events will serve as the basis for further analysis.

[0036] Step S140: based on the fault state chain vectors associated with each of the X iterative fault sample events, extracting, from the X iterative fault sample events, an iterative fault sample event that matches the fault state chain vector corresponding to the energy storage device operation fault data as a candidate fault sample event.

[0037] In the previous step, the server has extracted 5 (X=5) iterative fault sample events. Now, the server needs to extract a fault state chain vector for each iterative fault sample event.

[0038] For the first iterative fault sample event, the server creates an independent storage space for it. This independent storage space is like a special folder, which is used to orderly store information such as logic vectors, fault penetration nodes, and trigger status points. At the same time, this iterative fault sample event is marked, for example, as "Event 1".

[0039] Then, the server starts to collect all operating status information related to "Event 1". For example, in this iterative fault sample event, a battery cell in the battery pack has a sharp rise in temperature due to an internal short circuit (this is the location information where the fault first occurred, which belongs to the fault propagation and penetration related information). As the heat is conducted, the surrounding battery cells are also affected, and the overall temperature of the battery pack rises (this is the path information of the fault propagating from one component to another). This increases the internal resistance of the battery pack, which in turn affects the operating mode of the entire energy storage device, from the normal high-efficiency energy storage mode to the low-efficiency mode (this is the range information of the impact of the fault on the overall operation of the device at different operating stages). In addition, the overall operating mode information of the energy storage device includes that it was in a high-load operating state at the time, which also affected the development of the fault. In terms of the operating process, due to the increase in the temperature of the battery pack, the battery management system triggered the overheating protection mechanism, which is a state transition information.

[0040] Based on this operating status information, the server constructs a logic vector. For example, the logic vector may contain logic elements such as battery cell short circuit (1 indicates occurrence, 0 indicates non-occurrence), battery pack temperature rise (1 indicates non-rise), and operating mode conversion (1 indicates occurrence, 0 indicates non-occurrence). Then, information related to fault propagation and penetration is filtered out from the operating status information to preliminarily determine the reference range of the fault penetration node. In this example, the reference range of the fault penetration node is preliminarily determined to be the battery pack component.

[0041] Next, within the reference range of the initially determined fault penetration node, the server further analyzes the performance characteristics and impact mechanism of "Event 1" within this range. It is found that the physical location of the fault penetration node of the battery cell short circuit is the battery cell in the 3rd row and 5th column of the battery pack, and its position in the operation logic is the key part of the power supply in the battery pack. The manifestation of the fault at this fault penetration node is that the internal short circuit generates high heat, and the fault has the potential direction of propagating to the surrounding battery cells, such as spreading heat and fault influence to adjacent battery cells. If the fault state chain vector of this iterative fault sample event contains a trigger state point, assuming that in this example, the trigger state point is the electrolyte leakage inside the battery cell causing the short circuit, the server searches for the target state point related to the trigger factor that caused the fault from the operation state information. Finally, the constructed logic vector, the relevant information of the determined and sorted fault penetration node, and the information of the target state point are combined in the set order to construct the fault state chain vector of "Event 1".

[0042] In the same way, the server also constructs fault state chain vectors for the other four iterative fault sample events.

[0043] Next, based on these fault state chain vectors, the server extracts the iterative fault sample events that match the fault state chain vectors corresponding to the energy storage device operation fault data from the five iterative fault sample events as candidate fault sample events.

[0044] First, based on the logic vector of the energy storage device operation fault data and the logic vectors associated with each of the five iterative fault sample events, the logical association between the energy storage device operation fault data and the five iterative fault sample events is determined. For example, for "Event 1", the logic vector of the energy storage device operation fault data and the logic vector of "Event 1" are parsed. The logic vector of the energy storage device operation fault data contains logic elements such as low battery pack voltage (1 for low, 0 for normal), low power conversion module efficiency (1 for low, 0 for normal). The logic vector of "Event 1" contains logic elements such as high battery pack temperature (1 for high, 0 for normal), and short circuit of battery cells (1 for short circuit, 0 for normal).

[0045] The logical vector of "Event 1" is matched with the logical vector of the energy storage device operation fault data one by one. For the logical element of low battery pack voltage, there is no directly corresponding logical element in "Event 1", but the high battery pack temperature may indirectly cause the low battery pack voltage, so it is judged to have similar logical meanings, but different logical states, and marked as unmatched logical elements. For the logical element of low power conversion module efficiency, there is no direct association in "Event 1" and it is marked as an unmatched logical element. For the logical element of battery cell short circuit, there is no directly corresponding logical element in the energy storage device operation fault data, and it is marked as an unmatched logical element. The number of successfully matched logical elements in the logical vector of "Event 1" is counted as 0, and the total number of logical elements is 4 (the four logical elements of high battery pack temperature, battery cell short circuit, low battery pack voltage, and low power conversion module efficiency). The first proportion value is 0 when the number of successfully matched logical elements is divided by the total number of logical elements.

[0046] Different weights are then assigned to different logic elements. Assume that the weight of low battery pack voltage is 0.3, the weight of low power conversion module efficiency is 0.3, the weight of high battery pack temperature is 0.2, and the weight of battery cell short circuit is 0.2. The first ratio value is adjusted according to the matching status of each logic element and the assigned weight. Since there is no successful matching logic element, the adjusted second ratio value is still 0.

[0047] Then, the hidden logic dependency relationship between the energy storage device operation fault data and the logic vector of "Event 1" is analyzed. For example, although the short circuit of the battery cell is not directly reflected in the fault data, it may be a potential cause of low battery pack voltage and low power conversion module efficiency. The second ratio value is increased or decreased according to this hidden logic dependency relationship. Assume that after analysis, the second ratio value is increased to 0.2, and the final target logic correlation is 0.2.

[0048] Using the same method, the server calculates the logical correlation between the other four iterative fault sample events and the energy storage device operation fault data.

[0049] Then, based on the fault permeation nodes of the energy storage device operation fault data and the fault permeation nodes associated with the five iterative fault sample events, the fault permeation node correlation degrees between the energy storage device operation fault data and the five iterative fault sample events are determined.

[0050] For "Event 1", the permeation time domain correlation is determined based on the fault permeation node of the energy storage device operation fault data and the fault permeation node corresponding to "Event 1". Assuming that the fault permeation node in the energy storage device operation fault data is the state of the battery pack at the current moment, the fault permeation node in "Event 1" is the time period from the short circuit of the battery cell to the increase in the temperature of the battery pack. By analyzing the overlap of the two time periods, the order of precedence and other factors, the permeation time domain correlation is determined to be 0.6.

[0051] The first operation area correlation is determined based on the logical operation area in the fault penetration node of the energy storage device operation fault data and the logical operation area in the fault penetration node corresponding to "Event 1". If the logical operation area in the energy storage device operation fault data is the energy supply link of the entire battery pack, and the logical operation area in "Event 1" is also the energy supply link of the battery pack, then the first operation area correlation is 1.0.

[0052] Based on the non-logical operation area in the fault penetration node of the energy storage device operation fault data (such as the physical layout area of ​​the battery pack) and the non-logical operation area in the fault penetration node corresponding to "Event 1", the second operation area correlation is determined. Assuming that in these two areas, the positions of the battery cells affected by the fault in the battery pack partially overlap, the second operation area correlation is 0.8.

[0053] The penetration time domain correlation, the first operation area correlation and the second operation area correlation are fused and calculated, for example, weighted averaged according to weights of 0.3, 0.4 and 0.3, to generate a fault penetration node correlation between the energy storage device operation fault data and "Event 1" of (0.6×0.3+1.0×0.4+0.8×0.3)=0.82.

[0054] Using the same method, the server calculates the fault penetration node correlation between the other four iterative fault sample events and the energy storage equipment operation fault data.

[0055] Finally, based on the logical correlation and fault penetration node correlation associated with each of the five iterative fault sample events, the iterative fault sample events that match the fault state chain vector corresponding to the energy storage device operation fault data are extracted from the five iterative fault sample events as candidate fault sample events. For example, if the comprehensive weights of the logical correlation and the fault penetration node correlation are set to 0.4 and 0.6 respectively, for "Event 1", the calculated comprehensive correlation is 0.2×0.4+0.82×0.6=0.572. The comprehensive correlations of the other four iterative fault sample events are calculated in this way. Assuming that after calculation, the comprehensive correlations of "Event 1" and "Event 3" are greater than the set threshold (assuming 0.5), then "Event 1" and "Event 3" are extracted as candidate fault sample events.

[0056] Step S150, based on the candidate fault sample events and the energy storage device operation fault data, generate model input data corresponding to a fault diagnosis model, use the fault diagnosis model to generate diagnostic data of the energy storage device operation fault data based on the model input data, and determine a judgment result of the energy storage device operation fault data based on the diagnostic data.

[0057] In the previous steps, the server identified "Event 1" and "Event 3" as candidate fault sample events. Now, the server needs to generate model input data corresponding to the fault diagnosis model based on these two candidate fault sample events and the energy storage device operation fault data.

[0058] For the candidate fault sample event "Event 1", the server organizes the logic vector, fault penetration node, and trigger state point information in its related fault state chain vector. For example, the information such as high battery pack temperature and battery cell short circuit in the logic vector, the physical location of the fault penetration node, the location in the operation logic, and the electrolyte leakage at the trigger state point. For the energy storage device operation fault data, the voltage, current, and temperature data of the battery pack, the input and output power of the power conversion module and other data are organized. Then combine this information in the format required by the fault diagnosis model to generate model input data. For example, the battery pack temperature data in "Event 1" and the current battery pack temperature data in the energy storage device operation fault data are arranged in correspondence, and the battery cell short circuit information is associated with the short circuit hidden danger information that may exist in the battery pack.

[0059] Similarly, similar operations are performed on the candidate fault sample event "Event 3", which is combined with the energy storage device operation fault data to generate another set of model input data.

[0060] The server uses pre-trained fault diagnosis models to generate diagnostic data of energy storage equipment operation fault data based on these model input data. This fault diagnosis model is trained based on a large amount of historical fault data and corresponding diagnostic results. For example, the model may be a neural network model that has learned the mapping relationship between different fault characteristics and diagnostic results.

[0061] When the model input data generated based on "Event 1" and the energy storage equipment operation fault data are input, the fault diagnosis model may output a series of diagnostic data on the possibility of faults, fault development trends, etc. For example, the diagnostic data output by the model indicates that there is an 80% probability of a short circuit fault in the battery pack, and the increase in battery pack temperature due to the short circuit may further reduce the life of the battery pack, and the energy storage efficiency of the battery pack is expected to drop by another 10% in the next 24 hours.

[0062] For the model input data generated based on "Event 3" and the energy storage equipment operation fault data, the fault diagnosis model will also output the corresponding diagnostic data. Assuming that "Event 3" corresponds to the aging of some components of the power conversion module, the model outputs that the probability of the power conversion module's efficiency being reduced due to component aging is 70%, and as time goes by, if no maintenance is carried out, the efficiency may be reduced by another 15% in the next week.

[0063] Then, the server determines the results of the energy storage device operation fault data based on these diagnostic data. Since the fault probability corresponding to "Event 1" (battery pack short circuit related) is 80% higher than the 70% corresponding to "Event 3" (power conversion module component aging related), the server determines that the main cause of the current operation failure of the energy storage device is the short circuit fault in the battery pack. At the same time, based on the model's prediction of the fault development trend, the server can formulate corresponding response strategies. For example, in the case of a battery pack short circuit failure, the server can notify maintenance personnel to give priority to a detailed inspection of the battery pack, find the specific location of the short circuit and repair it to prevent further damage to the battery pack and a continuous decrease in energy storage efficiency. This completes the entire fault diagnosis process. By acquiring, analyzing and diagnosing the energy storage equipment operation fault data, the main cause of the fault is finally determined and guidance is provided for subsequent maintenance work.

[0064] Therefore, by gradually analyzing the operating fault data of the energy storage equipment, matching information is found from the prior fault feature database. After multiple rounds of data matching and analysis, the fault diagnosis model is finally used to determine the fault judgment result, providing an effective method for fault diagnosis and maintenance of energy storage equipment.

[0065] Based on the above steps, the embodiment of the present application significantly improves the accuracy and efficiency of energy storage equipment operation fault diagnosis by constructing a priori fault feature database and effectively matching iterative fault sample events corresponding to fault labels. The method can automatically extract key fault features from a large amount of historical data, and quickly locate and identify specific fault types in the current energy storage device operation based on these features. By extracting candidate fault sample events that are highly matched to the fault data to be diagnosed, combined with advanced fault diagnosis models, the present invention can generate accurate diagnostic data and provide timely and accurate fault identification results for operation and maintenance personnel. This not only helps to reduce energy losses caused by fault downtime, but also reduces manual diagnosis costs and time, and improves the reliability and maintenance efficiency of energy storage equipment.

[0066] In a possible implementation, step S130 includes:

[0067] Step S131, determining a feature matching degree between event coding vector data of the energy storage device operation fault data and event coding vector data in the fault feature database.

[0068] Step S132: extracting X event encoding vector data with a feature matching degree greater than a set matching degree from the fault feature database.

[0069] Step S133: determining, based on the index relationship between the event encoding vector data and the iterative fault example events, the iterative fault example events having index relationships with the X event encoding vector data respectively.

[0070] Step S134: outputting the iterative fault sample events having index relationships with the X event encoding vector data respectively as X iterative fault sample events matching the event encoding vector data of the energy storage device operation fault data.

[0071] In this embodiment, in the energy storage system, operating fault data is obtained from the battery pack, and its event coding vector data includes parameter codes such as the average voltage of the battery pack is 3.0V (normal is 3.6-3.8V), the conversion efficiency of the power conversion module is 80% (normal is above 95%), and the average temperature of the battery pack is 45°C (normal is 25-35°C).

[0072] In the fault feature database, there are many iterative fault sample events related to "low energy storage efficiency fault" and their corresponding event coding vector data. The server starts to calculate the feature matching degree between the event coding vector data of the current energy storage device operation fault data and the data in the fault feature database.

[0073] For an iterative fault sample event in the fault feature database, the battery pack voltage in its event encoding vector data is 2.9V, the power conversion module conversion efficiency is 78%, and the battery pack temperature is 48°C. The server calculates the difference between these parameters and the current fault data parameters. For the battery pack voltage, the relative difference is calculated as (3.0-2.9) / 3.6≈0.028 (based on the lower limit of the normal range of 3.6V); for the power conversion module conversion efficiency, the difference is (80-78) / 95≈0.021; for the battery pack temperature, the difference is (45-48) / 25=-0.12 (based on the lower limit of the normal range of 25°C). Then, according to the pre-set algorithm (assuming a simple weighted average with the same weight for each parameter), these differences are combined to obtain a feature matching degree of 0.8 (full score 1.0, indicating a high degree of match) between this iterative fault sample event and the current fault data. The server calculates each event encoding vector data in the fault feature database in this way.

[0074] Assuming that the matching degree is set to 0.7, the server starts the extraction operation after completing the feature matching degree calculation for the event coding vector data in the fault feature database. For example, after calculation, it is found that the feature matching degree of 10 event coding vector data is greater than 0.7. If X=5 is set, the server selects the first 5 event coding vector data from these 10 event coding vector data according to the feature matching degree from high to low (or other predetermined rules).

[0075] In the fault signature database, each event encoding vector data has a unique index pointing to the detailed description of the corresponding iterative fault sample event, just like in a library, the book number (event encoding vector data) corresponds to a specific book (iterative fault sample event).

[0076] After selecting five event encoding vector data, the server uses this index relationship to search for the corresponding iterative fault sample event. For example, for the first selected event encoding vector data, its index points to an iterative fault sample event with a detailed description of "During a rainstorm, due to water leakage in the room where the energy storage equipment is located, the battery pack was damp, the internal resistance of the battery increased, and the voltage dropped. At the same time, some components of the power conversion module were damaged due to moisture, and the conversion efficiency was reduced." Through such an index search, the server accurately determines the iterative fault sample events corresponding to these five event encoding vector data.

[0077] The server will find these five iterative fault sample events through the index relationship as iterative fault sample event outputs that match the event encoding vector data of the current energy storage device operation fault data. These iterative fault sample events will serve as an important basis for subsequent fault diagnosis and analysis, for example, in further analysis of fault state chain vector matching and other operations.

[0078] For example, in a possible implementation, the step of extracting the fault state chain vectors associated with each of the X iterative fault sample events includes:

[0079] Step A110, create an independent storage space for each iterative fault sample event, the independent storage space is used to store the fault state chain vector related information to be extracted, the independent storage space is a data structure for orderly storing logic vectors, fault penetration nodes and trigger state points, and at the same time, each iterative fault sample event is marked.

[0080] In this embodiment, X iterative fault sample events (assuming X=3) matching the event encoding vector data of the energy storage device operation fault data are determined in the previous step. Now, the fault state chain vector is constructed for these three iterative fault sample events.

[0081] For the first iterative fault sample event, the server opens up an independent storage space in its storage system. This storage space is like a "folder" specially customized for this event. It is a data structure that can orderly store information such as the logic vector, fault penetration node, and trigger state point to be constructed later. At the same time, the server marks this iterative fault sample event as "Event A".

[0082] In the same way, create an independent storage space for the second iteration fault sample event and mark it as "Event B"; create an independent storage space for the third iteration fault sample event and mark it as "Event C".

[0083] Step A120, for each iterative fault sample event, collect all operating status information related to the iterative fault sample event, the operating status information including status information of each component of the energy storage device before and after the fault occurs, the overall operating mode information of the energy storage device, and the state transition information of the energy storage device in the operation process.

[0084] For example, for an iterative fault sample event marked as "Event A", the server starts to collect relevant operating status information. This event involves the battery pack and power conversion module of the energy storage device.

[0085] Before the failure occurred, the battery pack was in a normal charging state, the voltage of each battery cell was stable between 3.6-3.8V, the overall current of the battery pack was normal, the input and output power of the power conversion module matched, the conversion efficiency was above 95%, and the energy storage device as a whole was in a high-efficiency energy storage mode. In terms of operating procedures, the battery management system monitored the battery pack status normally.

[0086] When the fault occurs, a battery cell in the battery pack (assuming it is the battery cell in the 3rd column of the 2nd row) has an internal short circuit fault. This is the location information where the fault first occurred. Because this battery cell is short-circuited, its temperature rises sharply, the surrounding battery cells are affected, and the overall temperature of the battery pack begins to rise. This is the path information of the fault propagating from one component (a single battery cell) to another component (battery pack). As the temperature of the battery pack rises, the internal resistance of the battery increases, and the overall voltage of the battery pack drops, resulting in the input power of the power conversion module being normal, but the output power is significantly reduced, the conversion efficiency drops to about 80%, and the energy storage device as a whole is converted from a high-efficiency energy storage mode to a low-efficiency mode. This is the range information of the impact of the fault on the overall operation of the equipment at different operating stages. At the same time, the battery management system detects that the temperature of the battery pack is too high and triggers the overheating protection mechanism. This is the state transition information of the energy storage device in the operation process.

[0087] For another example, for "Event B", before the failure occurred, the energy storage device was in a high-load operation state. The voltage of the battery cells in the battery pack fluctuated within the normal range, but the overall temperature was high, close to the upper limit of the normal range (35°C). Due to long-term high-load operation, the power conversion module had a certain degree of aging signs, but the conversion efficiency was still above 90%.

[0088] When the fault occurred, due to the failure of the air conditioning and refrigeration system in the computer room, the ambient temperature rose and the battery pack had difficulty in dissipating heat. The temperature of the battery pack rose rapidly, which was the first location where the fault occurred (from the perspective of the impact of ambient temperature on the heat dissipation of the battery pack). The increase in battery pack temperature caused the internal resistance of the battery to increase, the voltage of multiple battery cells to drop, and the overall voltage of the battery pack to decrease. The input power of the power conversion module was normal, but the output power decreased due to the drop in battery pack voltage, and the conversion efficiency dropped to about 85%. The overall energy storage efficiency of the energy storage device was low, and it changed from a high-load normal operating mode to a low-efficiency operating mode. The battery management system adjusts the charging strategy to deal with the situation where the battery pack temperature is too high. This is the state transition information in the operation process.

[0089] For another example, in "Event C", before the fault occurred, some of the battery cells in the battery pack had poor contact. Although it did not affect the overall voltage, there was a slight increase in resistance. The power conversion module operated normally, with a conversion efficiency of more than 95%, and the overall operation mode of the energy storage device was normal.

[0090] When the fault occurred, due to a slight vibration (possibly caused by the start-up of nearby equipment), the line in the battery pack that was originally in poor contact was completely disconnected. This is the location information of the first occurrence of the fault. The voltage of the small module where this battery cell is located dropped to 0, which in turn affected the overall voltage balance of the battery pack. The overall voltage of the battery pack dropped. Although the input power of the power conversion module was normal, the output power decreased, and the conversion efficiency dropped to about 82%. The energy storage device as a whole entered a low-efficiency operation mode. The battery management system detected the abnormal voltage of the battery pack, issued an alarm and adjusted the operating parameters. This is the state transition information in the operation process.

[0091] Step A130, for each iterative fault sample event, construct a logic vector according to the operating status information of the iterative fault sample event, and filter out information related to fault propagation and penetration from the operating status information, wherein the information related to fault propagation and penetration includes location information where the fault first occurs, path information of the fault propagating from one component to another, and scope information of the impact of the fault on the overall operation of the device at different operating stages.

[0092] In this embodiment, a logic vector is constructed based on the operating status information of "Event A". For example, the logic vector includes logic elements such as battery cell short circuit (1 indicates occurrence, 0 indicates non-occurrence), battery pack temperature rise (1 indicates occurrence, 0 indicates non-occurrence), power conversion module efficiency reduction (1 indicates occurrence, 0 indicates non-occurrence), and operating mode conversion (1 indicates occurrence, 0 indicates non-occurrence). Therefore, for "Event A", its logic vector may be [1,1,1,1].

[0093] Then, the operating status information is filtered out to find out the location where the fault first occurred, which is the short circuit of the battery cell in the 2nd row and 3rd column of the battery pack; the path information of the fault propagating from this battery cell to the surrounding battery cells, thereby affecting the entire battery pack; and the range information of the fault affecting the battery pack voltage, thereby affecting the output power of the power conversion module, and ultimately leading to the conversion of the overall operating mode of the energy storage device.

[0094] For “Event B”, the logic vector may contain elements such as battery pack temperature increase (1), power conversion module efficiency decrease (1), and operation mode change (1). (Since the fault does not start from a battery cell short circuit, the battery cell short circuit element is 0.) The logic vector may be [0, 1, 1, 1].

[0095] The fault first occurred in the computer room air conditioning refrigeration system, which affected the heat dissipation of the battery pack. The fault affected the battery pack temperature from the ambient temperature, and then affected the battery pack voltage and the output power of the power conversion module, ultimately affecting the path and range information of the overall operating mode of the energy storage device.

[0096] In “Event C”, the logic vector includes elements such as line disconnection (1), battery pack voltage drop (1), power conversion module efficiency reduction (1), and operation mode conversion (1). The logic vector may be [1,1,1,1].

[0097] The fault first occurs when the connection lines of some battery cells in the battery pack are disconnected; the fault affects the voltage of the small module where it is located, which in turn affects the overall voltage of the battery pack, and then affects the output power of the power conversion module, ultimately leading to the path and range information of the overall operation mode conversion of the energy storage device.

[0098] Step A140, based on the information related to the fault propagation and penetration, preliminarily determine a reference range of the fault penetration node, where the reference range represents one or more components, an operating area, or an operation process segment.

[0099] In this embodiment, based on the information related to fault propagation and penetration in "Event A", the reference range of the fault penetration node is initially determined to be the battery pack component. This is because the fault starts from a single battery cell in the battery pack and then propagates inside the battery pack, affecting the voltage and other parameters of the battery pack, and then affecting the power conversion module.

[0100] For "Event B", the reference range of the fault penetration node is initially determined to be the ambient temperature control system of the battery pack and the energy storage device (because the increase in ambient temperature affects the heat dissipation of the battery pack, leading to a series of problems). This range includes the battery pack component and the operating area of ​​the external environmental control system that affects the heat dissipation of the battery pack.

[0101] In "Event C", the reference range of the fault penetration node is initially determined to be the battery pack, because the fault starts from the disconnection of the line inside the battery pack, propagates inside the battery pack, affects the overall voltage of the battery pack, and then affects other components.

[0102] Step A150, for each iterative fault sample event, within the reference range of the initially determined fault penetration node, analyze the performance characteristics and impact mechanism of the iterative fault sample event within the reference range to further define the fault penetration node, and then organize the relevant information of the further defined fault penetration node, the relevant information including the physical location of the fault penetration node, the location in the operating logic, the manifestation of the fault at the fault penetration node, and the potential direction of the fault propagation from the fault penetration node to other parts.

[0103] In this embodiment, within the reference range of the fault penetration node of the battery pack that has been preliminarily determined, the performance characteristics and impact mechanism of "Event A" are analyzed. It is found that the physical location of the fault penetration node (the battery cell in the 3rd column of the 2nd row) is clear, and its position in the operating logic is a key part of the power supply of the battery pack (because a short circuit of a battery cell will affect the voltage balance of the entire battery pack). The manifestation of the fault at this fault penetration node is high heat generated by an internal short circuit, and the fault has the potential direction of propagation to the surrounding battery cells, such as spreading heat and fault effects to adjacent battery cells. The relevant information of these fault penetration nodes is sorted out, including the physical location (the 3rd column of the 2nd row), the position in the operating logic (the key part of the power supply), the manifestation (internal short circuit generates high heat), and the potential direction of propagation (spreading to adjacent battery cells).

[0104] Within the reference range of the fault penetration node of "Event B" (battery pack and ambient temperature control system), for the battery pack part, the performance characteristic of the fault penetration node is that the battery pack has difficulty in dissipating heat, resulting in a rise in temperature. Its position in the operating logic is the core part of the energy storage of the energy storage device. It is manifested in that the temperature rise causes the internal resistance of the battery to increase. The impact is that it propagates to each battery cell inside the battery pack, causing the voltage to drop. For the ambient temperature control system, the physical location of the fault penetration node is the air conditioning refrigeration system. Its position in the operating logic is to control the operating environment temperature of the energy storage device. It is manifested in the failure of the refrigeration function. The impact is that it causes difficulty in dissipating heat for the battery pack. Organize these relevant information, such as the relevant information of the battery pack and the relevant information of the ambient temperature control system, separately.

[0105] Within the reference range of the fault penetration node of the battery pack of "Event C", the physical location of the fault penetration node (the part of the battery pack where the line is disconnected) is a specific battery cell connection line, and its location in the operating logic is the key link of power transmission within the battery pack. The manifestation is that the line is disconnected, causing the voltage to drop to 0. The potential direction of fault propagation is to affect the voltage balance of the surrounding battery cells, thereby affecting the voltage of the entire battery pack. Organize these relevant information, including physical location, location in the operating logic, manifestation, and potential direction of propagation.

[0106] Step A160, if the fault state chain vector of the iterative fault sample event contains a trigger state point, search the operating state information of the iterative fault sample event for a target state point related to the trigger factor that causes the fault to occur, wherein the trigger factor is an operation, an external event, or the internal state value of the energy storage device reaches a set critical value.

[0107] In this embodiment, it is assumed that the fault state chain vector of "Event A" contains a trigger state point, and the target state point related to the trigger factor that causes the fault is searched from the operating state information. In this event, the trigger state point is the short circuit caused by the leakage of electrolyte inside the battery cell. The server searches for this target state point by analyzing some weak abnormal data before the battery cell failure in the operating state information (such as the chemical composition monitoring data inside the battery cell).

[0108] For "Event B", assume that its fault state chain vector contains a trigger state point, and the trigger factor is that the compressor in the air conditioning refrigeration system of the computer room suddenly stops working. The server searches the operation log of the air conditioning system and searches for this target state point, that is, the compressor stops working, which causes the ambient temperature to rise, thereby affecting the heat dissipation of the battery pack.

[0109] In "Event C", it is assumed that the fault state chain vector contains a trigger state point, and the trigger factor is that the vibration generated by the startup of the nearby device exceeds the vibration threshold that the battery pack connection line can withstand. The server searches for this target state point by querying the startup records of the nearby devices and the vibration sensor data.

[0110] Step A170, combining the previously constructed logic vector, the determined and sorted related information of the fault penetration node, and the information of the target state point in a set order to construct the fault state chain vector.

[0111] Therefore, the previously constructed logic vector [1,1,1,1] of "Event A", the determined and sorted related information of the fault penetration node (the battery cell in the 3rd column of the 2nd row, the key part of the power supply, the internal short circuit generates high heat, which spreads to the adjacent battery cells) and the target state point (the electrolyte leakage inside the battery cell causes a short circuit) are combined in a set order (for example, the logic vector first, then the related information of the fault penetration node, and finally the target state point) to construct the fault state chain vector of "Event A".

[0112] For "Event B", the logic vector [0,1,1,1], the fault penetration node related information (battery pack: heat dissipation difficulties lead to temperature rise, the core part of energy storage, the temperature rise leads to increased internal resistance, which propagates to the internal battery cells; ambient temperature control system: air conditioning refrigeration system, controls the operating environment temperature, refrigeration function failure, resulting in difficulty in heat dissipation of the battery pack) and the target state point (air conditioning compressor stops working) are combined in the set order to construct the fault state chain vector of "Event B".

[0113] In "Event C", the logic vector [1,1,1,1], information related to the fault penetration node (a specific line in the battery pack, a key link for power transmission, where the line is disconnected and the voltage becomes 0, affecting the voltage balance of the surrounding battery cells), and the target state point (vibration generated by the startup of a nearby device exceeds the threshold) are combined in a set order to construct the fault state chain vector of "Event C".

[0114] In a possible implementation, the fault state chain vector corresponding to the iterative fault sample event includes a logic vector and a fault penetration node of the iterative fault sample event, and the fault state chain vector corresponding to the energy storage device operation fault data includes a logic vector and a fault penetration node of the energy storage device operation fault data.

[0115] Step S140 includes:

[0116] Step S141 : determining the logical association degree between the energy storage device operation fault data and the X iterative fault sample events respectively based on the logical vector of the energy storage device operation fault data and the logical vectors associated with each of the X iterative fault sample events.

[0117] Step S142, based on the fault permeation nodes of the energy storage device operation fault data and the fault permeation nodes associated with the X iterative fault sample events, determining the fault permeation node association degrees between the energy storage device operation fault data and the X iterative fault sample events.

[0118] Step S143: based on the logical associations and fault penetration node associations associated with the X iterative fault sample events, extracting, from the X iterative fault sample events, an iterative fault sample event that matches the fault state chain vector corresponding to the energy storage device operation fault data as a candidate fault sample event.

[0119] In a possible implementation, step S141 includes:

[0120] Step S1411, parsing the logic vector of the energy storage device operation fault data and the logic vector of each iterative fault sample event to determine the respective logic elements contained therein, wherein the logic elements represent different operation states, component states or logic information related to the operation logic of the energy storage device.

[0121] Step S1412, for each iterative fault sample event, the logic vector is matched with the logic vector of the energy storage device operation fault data one by one in terms of logical elements, and for each logical element, it is determined whether the logical element has the same or similar logical meaning, and if it has the same or similar logical meaning, it is further determined whether its logical state is the same, and if the logical state is the same, it is marked as a successfully matched logical element, and if the logical state is different, it is marked as an unmatched logical element, thereby counting the number of successfully matched logical elements and the total number of logical elements in the logic vector of each iterative fault sample event.

[0122] Step S1413: For each iterative fault sample event, a first ratio value is obtained by dividing the number of successfully matched logical elements by the total number of logical elements. The first ratio value reflects the degree of logical element matching between the iterative fault sample event and the energy storage device operation fault data at the logical vector level.

[0123] Step S1414, assigning different weights to different logical elements in the logical vector of the iterative fault sample event and the logical vector of the energy storage device operation fault data, and adjusting the first proportion value of each iterative fault sample event according to the matching condition of each logical element and the assigned weight, to obtain a second proportion value after weight adjustment.

[0124] Step S1415, analyzing the hidden logical dependency relationship existing in the logic vector of the energy storage device operation fault data and the iterative fault sample event, and increasing or decreasing the second proportion value after weight adjustment according to the hidden logical dependency relationship to obtain a final target logical correlation.

[0125] In a possible implementation, step S142 includes:

[0126] Step S1421, based on the fault penetration node of the energy storage device operation fault data and the fault penetration node corresponding to the ath iteration fault sample event, determine the penetration time domain correlation between the ath iteration fault sample event and the energy storage device operation fault data. a is a positive integer not greater than X.

[0127] Step S1422: determining a first operating area correlation between the a-th iteration fault sample event and the energy storage device operation fault data based on the logical operating area in the fault permeation node of the energy storage device operation fault data and the logical operating area in the fault permeation node corresponding to the a-th iteration fault sample event.

[0128] Step S1423, determining a second operating area correlation between the a-th iteration fault sample event and the energy storage device operation fault data based on the non-logical operating area in the fault penetration node of the energy storage device operation fault data and the non-logical operating area in the fault penetration node corresponding to the a-th iteration fault sample event.

[0129] Step S1424, performing a fusion calculation on the penetration time domain correlation degree corresponding to the a-th iterative fault sample event, the first operating area correlation degree, and the second operating area correlation degree, to generate the fault penetration node correlation degrees between the energy storage device operation fault data and the X iterative fault sample events.

[0130] In this embodiment, the server has constructed the logic vector of the energy storage device operation fault data and the logic vector of X iterative fault sample events (assuming X=3, which are event A, event B, and event C respectively).

[0131] For the energy storage equipment operation fault data, its logic vector includes logic elements such as low battery pack voltage (1 means low, 0 means normal), low power conversion module efficiency (1 means low, 0 means normal), and high battery pack temperature (1 means high, 0 means normal).

[0132] The logic vector for event A includes logic elements such as high battery pack temperature (1), battery cell short circuit (1), and operation mode conversion (1). The logic vector for event B includes logic elements such as high battery pack temperature (1), signs of aging of the power conversion module (1), and operation mode conversion (1). The logic vector for event C includes logic elements such as line disconnection (1), battery pack voltage drop (1), and operation mode conversion (1).

[0133] Next, the logical vector of event A is matched with the logical vector of the energy storage device operation fault data one by one. For the logical element of low battery pack voltage, there is no directly corresponding logical element in event A, but the high battery pack temperature may indirectly cause the low battery pack voltage, so it is judged to have similar logical meanings, but different logical states, and marked as unmatched logical elements. For the logical element of low power conversion module efficiency, there is no direct association in event A, and it is marked as an unmatched logical element. For the logical element of high battery pack temperature, the two logical states are the same, and it is marked as a successfully matched logical element. For the logical element of battery cell short circuit, there is no directly corresponding logical element in the energy storage device operation fault data, and it is marked as an unmatched logical element. Therefore, the number of successfully matched logical elements in the logical vector of event A is 1, and the total number of logical elements is 4 (the four logical elements of low battery pack voltage, low power conversion module efficiency, high battery pack temperature, and battery cell short circuit).

[0134] The same operation is performed on event B. For the logic element of low battery pack voltage, there is no direct correspondence in event B, but the high battery pack temperature may have an indirect impact, so it is marked as a mismatched element with similar logical meaning but different logical state. For the logic element of low power conversion module efficiency, there is a certain correlation with the aging signs of the power conversion module in event B, and it is judged as a mismatched element with similar logical meaning but different logical state. For the logic element of high battery pack temperature, the two have the same logical state, so it is marked as a successfully matched logic element. The number of successfully matched logic elements in the logic vector of event B is 1, and the total number of logic elements is 4.

[0135] For event C, the battery pack voltage drop has similar logical meaning and the same logical state as the battery pack voltage drop in the energy storage device operation fault data, and is marked as a successfully matched logical element. For the logic element of low power conversion module efficiency, there is no direct association in event C, and it is marked as an unmatched logical element. For the logic element of high battery pack temperature, there is no direct correspondence in event C, and it is marked as an unmatched logical element. The number of successfully matched logical elements in the logic vector of event C is 1, and the total number of logical elements is 4.

[0136] Next, the number of successfully matched logic elements is divided by the total number of logic elements. For event A, the first ratio value is 1 / 4=0.25. The first ratio value of event B is also 1 / 4=0.25. The first ratio value of event C is also 1 / 4=0.25.

[0137] Continuing to assume that the weight of low battery pack voltage is 0.3, the weight of low power conversion module efficiency is 0.3, the weight of high battery pack temperature is 0.2, and the weight of battery cell short circuit (for event A) or power conversion module aging signs (for event B) or line disconnection (for event C) is 0.2.

[0138] For event A, since the high battery pack temperature is the only successfully matched logical element with a weight of 0.2, the first ratio value is adjusted according to the matching situation. The sum of the weights of the unmatched logical elements is 0.8, so the adjusted second ratio value is 0.2×(1 / 0.2)×(1 / 4)=0.25.

[0139] In event B, high battery pack temperature is the only successfully matched logic element with a weight of 0.2, and the sum of weights of unsuccessfully matched logic elements is 0.8. The adjusted second ratio value is 0.2×(1 / 0.2)×(1 / 4)=0.25.

[0140] For event C, the battery pack voltage drop is a successfully matched logic element with a weight of 0.3, and the total weight of unsuccessfully matched logic elements is 0.7. The adjusted second ratio value is 0.3×(1 / 0.3)×(1 / 4)=0.25.

[0141] Then, the hidden logic dependency relationship in the logic vector of the energy storage device operation fault data and event A is analyzed. Although the short circuit of the battery cell is not directly reflected in the fault data, it may be a potential cause of low battery pack voltage and low power conversion module efficiency. The second ratio value is increased or decreased according to this hidden logic dependency relationship. Assume that after analysis, the second ratio value is increased to 0.3, and the final target logic correlation is 0.3.

[0142] In event B, although the aging signs of the power conversion module do not directly match the low efficiency of the power conversion module, aging may be an early manifestation of reduced efficiency. Based on this hidden logical dependency, the second ratio value is increased to 0.3, and the target logical correlation is obtained as 0.3.

[0143] For event C, line disconnection is the direct cause of the battery pack voltage drop and is closely logically associated with the low battery pack voltage. Based on this hidden logical dependency, the second ratio value is increased to 0.4, resulting in a target logical correlation of 0.4.

[0144] On this basis, it is assumed that the fault penetration node in the energy storage equipment operation fault data is the state of the battery pack at the current moment, and the fault penetration node in event A is the time period from the short circuit of the battery cell to the increase in the temperature of the battery pack. The server analyzes the overlapping parts of the two time periods, the order of precedence, and other factors. Since the increase in battery pack temperature caused by the short circuit of the battery cell is a continuous process and is directly related to the current state of the battery pack, the penetration time domain correlation is 0.8.

[0145] For event B, the fault penetration node in the energy storage equipment operation fault data is the current state of the battery pack, and the fault penetration node in event B is the time period from the failure of the air conditioning and refrigeration system in the computer room to the increase in the battery pack temperature. Analysis shows that these two time periods overlap partially, because the failure of the air conditioning and refrigeration system affects the battery pack temperature and then affects the current battery pack state. The penetration time domain correlation is 0.6.

[0146] In event C, the fault penetration node in the energy storage device operation fault data is the current state of the battery pack, and the fault penetration node in event C is the time period from line disconnection to battery pack voltage drop. The battery pack voltage drop caused by line disconnection is the direct cause of the current battery pack state, and the penetration time domain correlation is 0.9.

[0147] If the logical operation area in the energy storage device operation fault data is the energy supply link of the entire battery pack, and the logical operation area in event A is also the energy supply link of the battery pack, then the first operation area correlation is 1.0. For event B, if the logical operation area in the energy storage device operation fault data is the energy supply link of the entire battery pack, and the logical operation area in event B is also the energy supply link of the battery pack, the first operation area correlation is 1.0. In event C, assuming that the logical operation area in the energy storage device operation fault data is the energy supply link of the entire battery pack, and the logical operation area in event C is also the energy supply link of the battery pack, the first operation area correlation is 1.0.

[0148] Assuming that in event A, the non-logical operating area (e.g., the physical layout area of ​​the battery pack) in the fault penetration node in the energy storage device operation fault data and the non-logical operating area in the fault penetration node corresponding to event A, the position of the battery cells affected by the fault in the battery pack partially overlaps, and the correlation degree of the second operating area is 0.8. For event B, the non-logical operating area in the fault penetration node in the energy storage device operation fault data and the non-logical operating area in the fault penetration node corresponding to event B, assuming that the battery pack area affected by the fault partially overlaps, the correlation degree of the second operating area is 0.7. In event C, the non-logical operating area in the fault penetration node in the energy storage device operation fault data and the non-logical operating area in the fault penetration node corresponding to event C, assuming that the battery pack line position affected by the fault partially overlaps, the correlation degree of the second operating area is 0.8.

[0149] Next, the penetration time domain correlation, the first operation area correlation, and the second operation area correlation corresponding to event A are fused and calculated. For example, weighted average is performed according to the weights of 0.3, 0.4, and 0.3, and the fault penetration node correlation between the energy storage device operation fault data and event A is generated as (0.8×0.3+1.0×0.4+0.8×0.3)=0.88. For event B, weighted average is performed according to the same weight, and the fault penetration node correlation is (0.6×0.3+1.0×0.4+0.7×0.3)=0.79. In event C, the fault penetration node correlation is calculated to be (0.9×0.3+1.0×0.4+0.8×0.3)=0.89.

[0150] Further assume that the comprehensive weights of logical correlation and fault penetration node correlation are 0.4 and 0.6 respectively. For event A, the calculated comprehensive correlation is 0.3×0.4+0.88×0.6=0.668. For event B, the comprehensive correlation is 0.3×0.4+0.79×0.6=0.614. For event C, the comprehensive correlation is 0.4×0.4+0.89×0.6=0.71. Assume that the threshold of the comprehensive correlation is set to 0.6. Since the comprehensive correlation of event A and event C is greater than 0.6, event A and event C are extracted as candidate fault sample events. These candidate fault sample events will be used for subsequent operations such as input of the fault diagnosis model to further determine the judgment results of the energy storage equipment operation fault data.

[0151] In a possible implementation manner, the fault state chain vector corresponding to the iterative fault sample event further includes a trigger state point of the iterative fault sample event.

[0152] Step S143 includes:

[0153] Step S1431 : determining the correlation of the state points associated with the X iterative fault sample events based on the state point coincidence between the trigger state points associated with the X iterative fault sample events and the current state point.

[0154] Step S1432, perform fusion calculation on the state point correlation, logic correlation, and fault penetration node correlation corresponding to the a-th iteration fault sample event to generate a first event correlation between the a-th iteration fault sample event and the energy storage device operation fault data. a is a positive integer not greater than X.

[0155] Step S1433: outputting the iterative fault sample events whose first event correlation degree is greater than the first set correlation degree among the X iterative fault sample events as candidate fault sample events that match the fault state chain vector corresponding to the energy storage device operation fault data.

[0156] In a possible implementation, step S143 further includes:

[0157] The logic correlation and fault penetration node correlation corresponding to the a-th iteration fault sample event are fused and calculated to generate a second event correlation between the a-th iteration fault sample event and the energy storage device operation fault data. a is a positive integer not greater than X.

[0158] The iterative fault sample events whose second event correlation degree is greater than the second set correlation degree among the X iterative fault sample events are output as candidate fault sample events that match the fault state chain vector corresponding to the energy storage device operation fault data.

[0159] In this embodiment, three iterative fault sample events (X=3, respectively event A, event B, event C) have been determined before, and their fault state chain vectors have been constructed, which include trigger state points.

[0160] The triggering state point of event A is the short circuit caused by the leakage of electrolyte inside the battery cell. For the energy storage equipment operation failure data, the current state point analysis shows that the low voltage of the battery pack may be caused by problems inside the battery cell. The server compares these two state points and finds that there is a certain degree of overlap. Since the short circuit caused by electrolyte leakage is related to the internal problems of the battery cell and may be the cause of the low voltage of the battery pack, the state point correlation of event A is determined to be 0.8.

[0161] The triggering state point of event B is that the compressor in the air conditioning refrigeration system of the computer room suddenly stops working. The state points in the current energy storage device operation fault data show that the battery pack temperature is high and the power conversion module efficiency is low. The server analysis found that the compressor stopped working, causing the ambient temperature to change, which in turn affected the battery pack temperature, which was partially related to the current state point. After comprehensive judgment, it was determined that the state point correlation of event B was 0.6.

[0162] The trigger state point of event C is that the vibration generated by the startup of nearby equipment exceeds the vibration threshold that the battery pack connection line can withstand. The state points in the current energy storage equipment operation fault data include battery pack voltage drop and other situations. Server analysis shows that the vibration causes line connection problems and then causes the battery pack voltage to drop. The two are closely related, and the state point correlation of event C is determined to be 0.9.

[0163] On this basis, the previously calculated logical correlation of event A is 0.3, the fault penetration node correlation is 0.88, and the state point correlation is 0.8. Assume that the fusion calculation weights of the state point correlation, logical correlation, and fault penetration node correlation are 0.3, 0.3, and 0.4 respectively. The first event correlation of event A is calculated as: 0.8×0.3+0.3×0.3+0.88×0.4=0.714.

[0164] The logical correlation of event B is 0.3, the fault penetration node correlation is 0.79, and the state point correlation is 0.6. Calculated with the same weight, the first event correlation of event B is: 0.6×0.3+0.3×0.3+0.79×0.4=0.638.

[0165] The logical correlation of event C is 0.4, the fault penetration node correlation is 0.89, and the state point correlation is 0.9. The first event correlation of event C is calculated as: 0.9×0.3+0.4×0.3+0.89×0.4=0.826.

[0166] Therefore, it is assumed that the first set correlation degree is 0.7.

[0167] 1. The first event correlation of event A is 0.714, which is greater than 0.7.

[0168] 2. The first event correlation of event C is 0.826, which is greater than 0.7.

[0169] 3. The first event correlation of event B is 0.638, which is less than 0.7.

[0170] Therefore, event A and event C are output as candidate fault sample events that match the fault state chain vector corresponding to the energy storage device operation fault data.

[0171] Among them, the logical correlation of event A is 0.3, and the fault penetration node correlation is 0.88. Assume that the fusion calculation weights of the logical correlation and the fault penetration node correlation are 0.4 and 0.6 respectively. The second event correlation of event A is calculated as: 0.3×0.4+0.88×0.6=0.668.

[0172] The logical correlation of event B is 0.3, and the fault penetration node correlation is 0.79. According to the same weight calculation, the second event correlation of event B is: 0.3×0.4+0.79×0.6=0.614.

[0173] The logical correlation of event C is 0.4, and the fault penetration node correlation is 0.89. The second event correlation of event C is calculated as: 0.4×0.4+0.89×0.6=0.71.

[0174] Further, it is assumed that the second set correlation degree is 0.6.

[0175] 1. The second event correlation of event A is 0.668, which is greater than 0.6.

[0176] 2. The second event correlation of event C is 0.71, which is greater than 0.6.

[0177] 3. The second event correlation of event B is 0.614, which is greater than 0.6.

[0178] Therefore, events A, B, and C are all output as candidate fault sample events that match the fault state chain vector corresponding to the energy storage device operation fault data (here, this setting is followed. If different correlation thresholds are set in practice, the output results may be different). These candidate fault sample events will be used in subsequent operations such as generating model input data corresponding to the fault diagnosis model based on candidate fault sample events and energy storage device operation fault data.

[0179] In a possible implementation, the method further includes:

[0180] Step B110, obtaining the energy storage device operation update data to be analyzed and the characteristic attribute label corresponding to the energy storage device operation update data. The characteristic attribute label corresponds to a fault label.

[0181] In this embodiment, during daily operation, the server will periodically obtain energy storage device operation update data. For example, the energy storage device operation update data obtained this time covers the operation information of multiple components such as battery packs, power conversion modules, and battery management systems over a period of time (such as the past 24 hours).

[0182] At the same time, each data has a corresponding characteristic attribute label. Assume that this characteristic attribute label is "energy storage equipment operation data during the high temperature period in summer", and this label corresponds to a fault label "high temperature related fault". This label is determined based on the period of data collection (summer high temperature period) and the type of fault that may be related to this period (faults that may be caused by high temperature).

[0183] Step B120: Based on the characteristic attribute tag, the energy storage device operation update data is decomposed to generate Y iterative fault sample events, and event encoding vector data associated with each of the Y iterative fault sample events is generated. Y is a positive integer.

[0184] Step B130, loading the event encoding vector data associated with each of the Y iterative fault sample events into the fault feature database corresponding to the feature attribute label.

[0185] In this embodiment, assuming that five iterative fault sample events and their associated event encoding vector data are generated according to the operation period label "energy storage device operation data during summer high temperature period", the server will find the fault feature database corresponding to "high temperature related faults". This database is specifically used to store historical data and analysis results related to faults that may be caused by high temperature.

[0186] The server loads the event encoding vector data associated with each of the five iterative fault sample events into the fault feature database. For example, for the event encoding vector data [0.2, 0.15, 0.1] of the first iterative fault sample event (regarding the state of a specific battery cell in the battery pack during a high temperature period), the server stores it in the corresponding position in the database so that in the subsequent fault diagnosis process, if similar high temperature related fault conditions are encountered, these new data can be used for more accurate analysis and diagnosis.

[0187] Similarly, for the three iterative fault sample events and their event encoding vector data generated according to the device module label "battery module data", the server will also load them into the corresponding fault feature database (assuming that there is a fault feature database related to battery module faults). This continuous updating of the fault feature database can improve the accuracy and efficiency of fault diagnosis, because the database contains more fault samples and related data under different circumstances.

[0188] In a possible implementation, step B120 includes:

[0189] Step B121: when the characteristic attribute tag is an operating period tag, extracting operating period status data associated with each of the Z components from the energy storage device operation update data.

[0190] Step B122, decomposing the runtime status data associated with each of the Z components to generate runtime status sub-data associated with each of the Z components.

[0191] Step B123, based on the trigger state points associated with the runtime state sub-data of the Z components, integrate b adjacent runtime state sub-data of the runtime state sub-data associated with the Z components to generate Y integrated runtime state data. b is a positive integer.

[0192] Step B124, extracting features from the Y integrated runtime status data, and generating event coding vector data associated with each of the Y integrated runtime status data.

[0193] Step B125, outputting the Y integrated runtime status data as Y iterative fault sample events, and determining the event coding vector data associated with each of the Y integrated runtime status data based on the event coding vector data associated with each of the Y integrated runtime status data.

[0194] In this embodiment, when the characteristic attribute label is an operating period label, for example, assuming that the characteristic attribute label is "energy storage device operating data during high temperature period in summer" (operating period label), Z=3 (here it represents the three components of battery pack, power conversion module, and battery management system).

[0195] For battery packs, the server extracts the operating status data of the battery pack during the high temperature period in summer from the energy storage equipment operation update data. This includes the voltage, current, and temperature data of each battery cell, the overall charge and discharge status of the battery pack, the change in internal resistance, and other data. For example, the voltage of each battery cell in the battery pack fluctuates during the high temperature period, and the voltage of some battery cells drops significantly between 2 and 4 pm. The average temperature fluctuates between 38 and 42°C, and the whole is in a discharge state. The internal resistance increases relative to normal temperature.

[0196] For the power conversion module, the extracted operating time status data include input and output power, conversion efficiency, module temperature, etc. During the high temperature period in summer, the input power of the power conversion module is normal, but the output power decreases to a certain extent during the high temperature period, the conversion efficiency drops from the normal 95% to about 90%, and the module temperature rises to 50-55°C (normal is 40-45°C).

[0197] For the battery management system, the operating period status data includes the monitoring parameter settings, alarm thresholds, control strategy adjustments, etc. of the battery pack. During the high temperature period in summer, the battery management system adjusts the high temperature alarm threshold of the battery pack from 40°C to 42°C, and adjusts the charging current limit strategy to cope with high temperature conditions.

[0198] The operating period status data of the battery pack is further decomposed into sub-data of each battery cell. For example, the voltage data of the battery pack is decomposed into the voltage change curve of each battery cell. It is found that the voltage of the battery cell in the 3rd row and the 5th column has the largest voltage drop, reaching 0.5V (from 3.6V to 3.1V), and its temperature rise rate is faster than other cells, reaching 45°C. This is the operating period status sub-data of the battery cell.

[0199] For the power conversion module, its operating period status data is decomposed into input power sub-data, output power sub-data, conversion efficiency sub-data, etc. The input power sub-data shows a small fluctuation around 3 pm, the output power sub-data shows a continuous downward trend during the high temperature period, and the conversion efficiency sub-data shows a gradual downward trend as the temperature rises.

[0200] For the battery management system, its operating period status data is decomposed into alarm threshold sub-data and control strategy sub-data. The alarm threshold sub-data indicates the adjustment of the alarm threshold at different time points during the high temperature period, and the control strategy sub-data shows the specific adjustment values ​​of the charging current limit strategy in different temperature ranges.

[0201] Assume b = 2. For the battery pack, take the trigger state point of the voltage drop of the battery cell in the 3rd row and 5th column as an example, and integrate the operating time state sub-data of the two adjacent time periods (such as 3-4 pm and 4-5 pm). The integrated operating time state data includes information such as the voltage, temperature, and internal resistance changes of the battery cells in these two time periods, as well as the voltage difference with the adjacent battery cells. For example, the integrated voltage data shows that the voltage continues to drop from 3.3V to 3.1V, the temperature rises from 42℃ to 45℃, and the internal resistance increases from 0.1Ω to 0.12Ω in these two time periods.

[0202] For the power conversion module, taking the trigger state point of conversion efficiency decline as an example, the sub-data of input power, output power, conversion efficiency, module temperature, etc. in two adjacent time periods (such as 2-3 pm and 3-4 pm) are integrated. The integrated operation time state data shows that the input power is stable, but the output power continues to decline, the conversion efficiency drops from 92% to 90%, and the module temperature rises from 50°C to 52°C.

[0203] For the battery management system, taking the alarm threshold adjustment as an example, the alarm threshold sub-data and control strategy sub-data of two adjacent time periods (such as 10-11 am and 11-12 pm) are integrated. The integrated operation time period status data shows that the alarm threshold is adjusted from 40°C to 41°C, and the charging current limit strategy is reduced by 10% from the initial value.

[0204] After such integration operation, Y (assuming Y=5) pieces of integration running time period status data are generated.

[0205] Therefore, for the first integrated operating period status data (taking the battery pack as an example), the extracted features include the battery cell voltage drop amplitude (coded as 0.2, expressed as the decrease ratio relative to the normal voltage), the temperature rise amplitude (coded as 0.15, expressed as the increase ratio relative to the normal temperature range), the internal resistance increase ratio (coded as 0.1, expressed as the increase ratio relative to the normal internal resistance), etc., forming the event coding vector data [0.2, 0.15, 0.1].

[0206] For the integrated operation period status data of the power conversion module, features are extracted such as the conversion efficiency decrease (coded as 0.05, expressed as the decrease ratio relative to the normal conversion efficiency), the output power decrease ratio (coded as 0.1, expressed as the decrease ratio relative to the normal output power), and the module temperature increase (coded as 0.12, expressed as the increase ratio relative to the normal module temperature range). The event coding vector data is [0.05, 0.1, 0.12].

[0207] In the same way, corresponding event encoding vector data is generated for each integrated runtime status data.

[0208] Finally, the five integrated operation time state data are used as five iterative fault sample events. For example, the first iterative fault sample event is the integrated operation time state data about the voltage drop, temperature rise, and internal resistance increase of a specific battery cell in the battery pack during a specific period of time. In addition, based on the previously generated event coding vector data, the event coding vector data associated with each of the five iterative fault sample events is determined, and these event coding vector data will serve as an important basis for subsequent analysis and storage.

[0209] In a possible implementation, step B120 may further include:

[0210] Step B126: when the characteristic attribute tag is a device module tag, the energy storage device operation update data is decomposed into unit operation status data associated with each of Y device units.

[0211] Step B127: perform equivalent logic update on the initial unit state representation information of the cth device unit to generate updated state representation information of the cth device unit. c is a positive integer not greater than Y.

[0212] Step B128, generating event coding vector data of the initial unit state representation information of the c-th device unit and event coding vector data of the updated state representation information of the c-th device unit.

[0213] Step B129, output the event coding vector data of the initial unit state characterization information and the event coding vector data of the updated state characterization information of the c-th device unit as the event coding vector data corresponding to the c-th device unit, output the unit operation state data associated with each of the Y device units as Y iterative fault sample events, and determine the event coding vector data associated with each of the Y iterative fault sample events based on the event coding vector data corresponding to the Y device units.

[0214] In this embodiment, when the characteristic attribute label is a device module label, it is assumed that the characteristic attribute label is "battery pack module data" (device module label), Y=3 (here it is assumed that the battery pack is divided into 3 device units, for example, according to the physical partition or logical function partition of the battery pack).

[0215] For the first device unit, the server decomposes its unit operation status data from the energy storage device operation update data. This includes data such as the voltage distribution, current flow, and temperature distribution of the battery cells in the unit. For example, the average voltage of the battery cells in the device unit is 3.5V, the current is stable at 5A during charging, and the temperature is between 36-38℃. The temperature of some battery cells near the edge is slightly higher.

[0216] For the second equipment unit, the unit operation status data includes the health status assessment of the battery cells (e.g., through parameters such as internal resistance), charge and discharge balance, etc. It was found that a few battery cells in the unit had large internal resistance and a certain deviation in charge and discharge balance, requiring more charging time to reach the same power level as other battery cells.

[0217] For the third device unit, the unit operation status data includes connection status monitoring between battery cells, overall power storage status, etc. For example, the connection status is good, but the overall power storage is slightly lower than that of the other two device units.

[0218] For the first device unit (c=1), the initial unit state characterization information may be a simple parameter description, such as average voltage, current, temperature range, etc. When performing equivalent logic updates, the server converts these parameters into logic states based on the operating logic and fault analysis requirements of the battery pack. For example, the average voltage of 3.5V is compared with the normal voltage range (3.6-3.8V), and updated to a logic state of low voltage (indicated by 1); the current is stable at 5A and belongs to the normal range, and is updated to a logic state of normal current (indicated by 0); the temperature is between 36-38℃ and close to the normal upper limit, and is updated to a logic state of high temperature (indicated by 1), and the updated state characterization information is [1,0,1].

[0219] For the second device unit (c=2), the initial unit state characterization information contains a few battery cells with large internal resistance and charge-discharge balance deviation. After the equivalent logic is updated, the large internal resistance is updated to a logic state of poor battery health (indicated by 1), and the charge-discharge balance deviation is updated to a logic state of abnormal balance (indicated by 1), and the updated state characterization information is obtained as [1,1].

[0220] For the third device unit (c=3), the power storage capacity in the initial unit state characterization information is low, and after equivalent logic update, it is updated to a logic state of insufficient power (represented by 1), and the good connection state is updated to a logic state of normal connection (represented by 0), and the updated state characterization information is obtained as [1,0].

[0221] For the first device unit, the event coding vector data of the initial unit state characterization information can be based on the encoding of the original parameters. For example, the average voltage of 3.5V is encoded as 0.9 (expressed as a ratio relative to the lower limit of the normal voltage range), the current of 5A is encoded as 0 (indicating normal), and the temperature of 36-38℃ is encoded as 0.8 (expressed as a ratio relative to the upper limit of the normal temperature range), and the event coding vector data of the initial unit state characterization information is obtained as [0.9, 0, 0.8]. The event coding vector data of the updated state characterization information is encoded according to the logical state, the voltage is too low and encoded as 1, the current is normal and encoded as 0, and the temperature is too high and encoded as 1, and the event coding vector data of the updated state characterization information is obtained as [1, 0, 1].

[0222] In the same manner, event encoding vector data of initial unit state representation information and updated state representation information are generated for the second and third equipment units, respectively.

[0223] For the first device unit, the event coding vector data [0.9, 0, 0.8] of the initial unit state representation information and the event coding vector data [1, 0, 1] of the updated state representation information are combined and output as the event coding vector data corresponding to the first device unit, for example, [0.9, 0, 0.8, 1, 0, 1].

[0224] Similarly, corresponding event encoding vector data are generated for the second and third equipment units.

[0225] Therefore, the unit operation status data of each of the three equipment units are respectively used as three iterative fault sample events. For example, the first iterative fault sample event is the operation status data of the voltage, current, temperature, etc. of the first equipment unit. In addition, based on the event coding vector data corresponding to the equipment unit generated previously, the event coding vector data associated with each of the three iterative fault sample events are determined. These event coding vector data will help to analyze the faults of the equipment module and update the fault feature database.

[0226] Figure 2 The operation fault identification system 100 applied to the energy storage device shown includes: a processor 1001 and a memory 1003. Among them, the processor 1001 and the memory 1003 are connected, such as through a bus 1002. Optionally, the operation fault identification system 100 applied to the energy storage device may also include a transceiver 1004, and the transceiver 1004 can be used for data interaction between the server and other servers, such as data transmission and / or data reception. It should be noted that the transceiver 1004 is not limited to one in actual scheduling, and the structure of the operation fault identification system 100 applied to the energy storage device does not constitute a limitation on the embodiments of the present application.

[0227] Processor 1001 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array) or other programmable logic devices, transistor logic devices, hardware components or any combination thereof. It may implement or execute various exemplary logic blocks, modules and circuits described in conjunction with the disclosure of this application. Processor 1001 may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0228] The bus 1002 may include a path to transmit information between the above components. The bus 1002 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus. The bus 1002 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 2 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0229] The memory 1003 can be a ROM (Read Only Memory) or other types of static storage devices that can store static information and instructions, a RAM (Random Access Memory) or other types of dynamic storage devices that can store information and instructions, or an EEPROM (Electrically Erasable Programmable Read Only Memory), a CD-ROM (Compact Disc Read Only Memory) or other optical disk storage, optical disk storage (including compressed optical disk, laser disk, optical disk, digital versatile disk, Blu-ray disk, etc.), magnetic disk storage medium, other magnetic storage devices, or any other medium that can be used to carry or store program code and can be read by a computer, without limitation herein.

[0230] The memory 1003 is used to store program codes for executing the embodiments of the present application, and the execution is controlled by the processor 1001. The processor 1001 is used to execute the program codes stored in the memory 1003 to implement the steps shown in the above method embodiments.

[0231] In addition, an embodiment of the present application further provides a readable storage medium, in which computer executable instructions are pre-installed. When a processor executes the computer executable instructions, the operation fault identification method applied to the energy storage device as above is implemented.

[0232] Similarly, it should be noted that in order to simplify the description disclosed in this application, thereby helping to understand one or more embodiments of the invention, in the foregoing description of the embodiments of the present application, multiple features are sometimes combined into one embodiment, drawings, or descriptions thereof. Similarly, it should be noted that in order to simplify the description disclosed in this application, thereby helping to understand one or more embodiments of the invention, in the foregoing description of the embodiments of the present application, multiple features are sometimes combined into one embodiment, drawings, or descriptions thereof.

Claims

1. A method for determining an operating fault of an energy storage device, characterized in that: The method comprises: Acquire the energy storage device operation fault data to be diagnosed and the fault label corresponding to the energy storage device operation fault data; Determine a fault feature database matching the fault label from a priori fault feature database; the fault feature database includes event encoding vector data corresponding to iterative fault sample events matching the fault label; Based on the event coding vector data in the fault feature database, extracting X iterative fault sample events that match the event coding vector data of the energy storage device operation fault data from the iterative fault sample events corresponding to the event coding vector data in the fault feature database; X is a positive integer; Based on the fault state chain vectors associated with each of the X iterative fault sample events, extracting, from the X iterative fault sample events, an iterative fault sample event that matches the fault state chain vector corresponding to the energy storage device operation fault data as a candidate fault sample event; Based on the candidate fault sample events and the energy storage device operation fault data, model input data corresponding to a fault diagnosis model is generated, and using the fault diagnosis model, diagnostic data of the energy storage device operation fault data is generated based on the model input data, and a judgment result of the energy storage device operation fault data is determined based on the diagnostic data.

2. The method for determining an operating fault of an energy storage device according to claim 1, characterized in that: The extracting, based on the event coding vector data in the fault feature database, X iterative fault sample events that match the event coding vector data of the energy storage device operation fault data from iterative fault sample events corresponding to the event coding vector data in the fault feature database, comprises: Determine a feature matching degree between the event coding vector data of the energy storage device operation fault data and the event coding vector data in the fault feature database; Extracting X event encoding vector data whose feature matching degree is greater than a set matching degree from the fault feature database; Based on the index relationship between the event encoding vector data and the iterative fault sample event, determining the iterative fault sample events having index relationships with the X event encoding vector data respectively; The iterative fault sample events respectively having index relationships with the X event encoding vector data are output as X iterative fault sample events matching the event encoding vector data of the energy storage device operation fault data.

3. The method for determining an operating fault of an energy storage device according to claim 1, characterized in that: The fault state chain vector corresponding to the iterative fault sample event includes a logic vector and a fault penetration node of the iterative fault sample event, and the fault state chain vector corresponding to the energy storage device operation fault data includes a logic vector and a fault penetration node of the energy storage device operation fault data; The extracting, based on the fault state chain vectors associated with each of the X iterative fault sample events, an iterative fault sample event matching the fault state chain vector corresponding to the energy storage device operation fault data from the X iterative fault sample events as a candidate fault sample event, includes: Determine the logical association degree between the energy storage device operation fault data and the X iterative fault sample events respectively based on the logical vector of the energy storage device operation fault data and the logical vectors associated with each of the X iterative fault sample events; Determine the fault permeation node correlation degree between the energy storage device operation fault data and the X iterative fault sample events respectively based on the fault permeation nodes of the energy storage device operation fault data and the fault permeation nodes associated with each of the X iterative fault sample events; Based on the logical association degree and fault penetration node association degree associated with each of the X iterative fault sample events, extracting, from the X iterative fault sample events, an iterative fault sample event that matches the fault state chain vector corresponding to the energy storage device operation fault data as a candidate fault sample event; Wherein, the step of determining the logical correlation between the energy storage device operation fault data and the X iterative fault sample events respectively based on the logical vector of the energy storage device operation fault data and the logical vectors associated with each of the X iterative fault sample events comprises: Parsing the logic vector of the energy storage device operation fault data and the logic vector of each iterative fault sample event to determine the respective logic elements contained therein, wherein the logic elements represent different operation states, component states or logic information related to the operation logic of the energy storage device; For each iterative fault sample event, the logic vector is matched with the logic vector of the energy storage device operation fault data one by one in terms of logic elements. For each logic element, it is determined whether the logic element has the same or similar logical meaning. If it has the same or similar logical meaning, it is further determined whether its logical state is the same. If the logical state is the same, it is marked as a successfully matched logical element. If the logical state is different, it is marked as an unmatched logical element. Thus, the number of successfully matched logical elements and the total number of logical elements in the logic vector of each iterative fault sample event are counted. For each iterative fault sample event, a first ratio value is obtained by dividing the number of successfully matched logical elements by the total number of logical elements, wherein the first ratio value reflects the degree of logical element matching between the iterative fault sample event and the energy storage device operation fault data at the logical vector level; Assign different weights to different logical elements in the logical vector of the iterative fault sample event and the logical vector of the energy storage device operation fault data, and adjust the first proportion value of each iterative fault sample event according to the matching condition of each logical element and the assigned weight to obtain a second proportion value after weight adjustment; The hidden logical dependency relationship existing in the logic vector of the energy storage device operation fault data and the iterative fault sample event is analyzed, and the second proportion value after weight adjustment is increased or decreased according to the hidden logical dependency relationship to obtain a final target logical correlation degree.

4. The method for determining an operating fault of an energy storage device according to claim 3, characterized in that: The determining, based on the fault permeation nodes of the energy storage device operation fault data and the fault permeation nodes associated with the X iterative fault sample events, the fault permeation node association degrees between the energy storage device operation fault data and the X iterative fault sample events respectively includes: Based on the fault permeation node of the energy storage device operation fault data and the fault permeation node corresponding to the a-th iteration fault sample event, determining the permeation time domain correlation between the a-th iteration fault sample event and the energy storage device operation fault data; a is a positive integer not greater than X; Determine a first operation area correlation between the a-th iteration fault example event and the energy storage device operation fault data based on a logical operation area in a fault permeation node of the energy storage device operation fault data and a logical operation area in a fault permeation node corresponding to the a-th iteration fault sample event; Determine a second operation area correlation between the a-th iteration fault sample event and the energy storage device operation fault data based on the non-logical operation area in the fault permeation node of the energy storage device operation fault data and the non-logical operation area in the fault permeation node corresponding to the a-th iteration fault sample event; The penetration time domain correlation degree corresponding to the a-th iterative fault sample event, the first operating area correlation degree, and the second operating area correlation degree are fused and calculated to generate fault penetration node correlation degrees between the energy storage device operation fault data and the X iterative fault sample events.

5. The method for determining operating faults of energy storage equipment according to claim 3, characterized in that: The fault state chain vector corresponding to the iterative fault sample event also includes a trigger state point of the iterative fault sample event; The extracting, from the X iterative fault sample events, an iterative fault sample event matching the fault state chain vector corresponding to the energy storage device operation fault data based on the logical association degree and the fault penetration node association degree associated with each of the X iterative fault sample events as a candidate fault sample event, includes: Determining the correlation of the state points associated with each of the X iterative fault sample events based on the state point coincidence between the trigger state points associated with each of the X iterative fault sample events and the current state point; Performing a fusion calculation on the state point correlation, logic correlation, and fault penetration node correlation corresponding to the a-th iterative fault sample event to generate a first event correlation between the a-th iterative fault sample event and the energy storage device operation fault data; a is a positive integer not greater than X; The iterative fault sample events whose first event correlation degree is greater than the first set correlation degree among the X iterative fault sample events are output as candidate fault sample events that match the fault state chain vector corresponding to the energy storage device operation fault data.

6. The method for determining operating faults of energy storage equipment according to claim 3, characterized in that: The extracting, from the X iterative fault sample events, an iterative fault sample event matching the fault state chain vector corresponding to the energy storage device operation fault data based on the logical association degree and the fault penetration node association degree associated with each of the X iterative fault sample events as a candidate fault sample event, includes: Performing a fusion calculation on the logical correlation and the fault penetration node correlation corresponding to the a-th iteration fault sample event to generate a second event correlation between the a-th iteration fault sample event and the energy storage device operation fault data; a is a positive integer not greater than X; The iterative fault sample events whose second event correlation degree is greater than the second set correlation degree among the X iterative fault sample events are output as candidate fault sample events that match the fault state chain vector corresponding to the energy storage device operation fault data.

7. The method for determining operating faults of energy storage equipment according to claim 1, characterized in that: The method further comprises: Acquire the energy storage device operation update data to be analyzed, and the characteristic attribute label corresponding to the energy storage device operation update data; the characteristic attribute label corresponds to a fault label; Based on the characteristic attribute label, the energy storage device operation update data is decomposed to generate Y iterative fault sample events, and event encoding vector data associated with each of the Y iterative fault sample events is generated; Y is a positive integer; The event encoding vector data associated with each of the Y iterative fault sample events is loaded into a fault feature database corresponding to the feature attribute label.

8. The method for determining operating faults of energy storage equipment according to claim 7, characterized in that: The step of decomposing the energy storage device operation update data based on the characteristic attribute label to generate Y iterative fault sample events, and generating event encoding vector data associated with each of the Y iterative fault sample events, includes: When the characteristic attribute tag is an operating period tag, extracting operating period status data associated with each of the Z components from the energy storage device operation update data; Decomposing the runtime state data associated with each of the Z components to generate runtime state sub-data associated with each of the Z components; Based on the trigger state points associated with the runtime state sub-data of the Z components, adjacent b runtime state sub-data in the runtime state sub-data associated with the Z components are integrated to generate Y integrated runtime state data; b is a positive integer; Perform feature extraction on the Y integrated operation time period status data to generate event coding vector data associated with each of the Y integrated operation time period status data; The Y integrated runtime status data are output as Y iterative fault sample events, and based on the event coding vector data respectively associated with the Y integrated runtime status data, the event coding vector data respectively associated with the Y integrated runtime status data are determined.

9. The method for determining operating faults of energy storage equipment according to claim 7, characterized in that: The step of decomposing the energy storage device operation update data based on the characteristic attribute label to generate Y iterative fault sample events, and generating event encoding vector data associated with each of the Y iterative fault sample events, includes: When the characteristic attribute tag is a device module tag, the energy storage device operation update data is decomposed into unit operation status data associated with each of Y device units; Performing an equivalent logic update on the initial unit state representation information of the c-th device unit to generate updated state representation information of the c-th device unit; c is a positive integer not greater than Y; Generate event coded vector data of initial unit state representation information of the c-th device unit and event coded vector data of updated state representation information of the c-th device unit; Outputting the event coded vector data of the initial unit state representation information and the event coded vector data of the updated state representation information of the c-th device unit as the event coded vector data corresponding to the c-th device unit; The unit operation status data associated with each of the Y equipment units is output as Y iterative fault sample events, and based on the event coding vector data corresponding to the Y equipment units, the event coding vector data associated with each of the Y iterative fault sample events is determined.

10. An operation fault identification system applied to energy storage equipment, characterized in that: The operation fault identification system applied to the energy storage device includes a processor and a machine-readable storage medium, wherein the machine-readable storage medium stores machine executable instructions, and the machine executable instructions are loaded and executed by the processor to implement the operation fault identification method applied to the energy storage device as described in any one of claims 1-9.

Images