A method and system for constructing a wind turbine knowledge graph

Abstract

The application discloses a wind power generator knowledge graph construction method and system, which investigates meteorological and humidity conditions of a wind power generator when a fault entity appears, and analyzes the correlation between the conditions and the occurrence of the entity. The method can determine which specific conditions are associated with the occurrence of the entity, and calculate the probability of the occurrence of the entity under different conditions. In this way, the application can effectively screen and analyze multidimensional data.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of data processing, and particularly relates to a wind turbine knowledge graph construction method and system. BACKGROUND

[0002] In power system management, constructing a knowledge graph is a key process aimed at improving the efficiency and reliability of the system. However, the current traditional method of constructing a power system fault knowledge graph mainly relies on the operating parameters of power equipment itself, lacking comprehensive analysis from multiple fields. For example, the current method of constructing a wind turbine knowledge graph only relies on operating data, fault logs, and maintenance records of wind turbines. This method ignores the influence of external factors such as weather conditions, geographical features, and regional load changes on power system faults. Therefore, the current wind turbine knowledge graph construction method only relies on internal operating parameters and fails to effectively integrate data from multiple fields such as meteorology and geography, resulting in an incomplete and inaccurate knowledge graph in describing and predicting power system faults. The constructed knowledge graph may not fully reflect all potential factors leading to faults, thereby affecting the effectiveness of fault diagnosis and preventive measures. SUMMARY

[0003] To solve the problems existing in the prior art, the present application proposes a wind turbine knowledge graph construction method and system.

[0004] The technical solution of the present application is as follows:

[0005] A wind turbine knowledge graph construction method, comprising:

[0006] Obtaining time series data of multiple wind turbines under multiple conditions and multiple fault types that have occurred;

[0007] Respectively taking each fault type as a fault entity, dividing the wind turbines under each fault entity into target class wind turbines that have appeared the fault entity and normal class wind turbines that have not appeared the fault entity; according to the correlation index of the target class wind turbines and the normal class wind turbines under each fault entity, judging whether there is a specific correlation between each fault entity and any condition, and obtaining the conditions that have a specific correlation with each fault entity;

[0008] For each entity fault, the time series data of each wind turbine in the target wind turbine under each condition within a preset time period after each fault is obtained, the influence correlation between each target wind turbine and any condition for each fault entity is judged, the fault entities and conditions that have influence correlation for each target wind turbine are obtained, and the ratio of the number of times of faults of the fault entities and conditions that have influence correlation for each target wind turbine to a preset number of times of faults before the current detection is obtained, and meanwhile, the influence combination of each target wind turbine in the target wind turbine under each fault entity is formed based on the conditions that have influence correlation with each fault entity.

[0009] Based on the ratio of the number of times of faults of the fault entities and conditions that have influence correlation for each target wind turbine to a preset number of times of faults before the current detection, the first probability and the second probability of the influence combination of each target wind turbine in the target wind turbine under each fault entity are obtained, and the influence combination is screened according to the first probability and the second probability, and the updated influence combination of each target wind turbine in the target wind turbine under each fault entity is obtained.

[0010] Based on the updated probability of the updated influence combination and the updated probability of the conditions that have influence correlation with each fault entity, the wind turbine knowledge graph is constructed.

[0011] Further, the specific steps of judging whether there is a specific correlation between each fault entity and any condition and obtaining the conditions that have a specific correlation with each fault entity according to the group correlation index of the target wind turbine and the normal wind turbine under each fault entity include:

[0012] Optionally, one fault entity and one condition are selected, denoted as fault entity A and the a-th condition, and any two target wind turbines in the target wind turbine under the fault entity A are grouped to obtain the absolute value of the difference of the mean value of the time series data of each group under the a-th condition, denoted as the first difference of the a-th condition of each group.

[0013] According to the obtaining method of the first difference of the a-th condition of each group in the target wind turbine, the first difference of the a-th condition of each group in the normal wind turbine is obtained.

[0014] Based on the first difference of the a-th condition of each group in the target wind turbine and the first difference of the a-th condition of each group in the normal wind turbine, the correlation index of the a-th condition is calculated by using the following formula:

[0015]

[0016] In the formula, G aCY is the correlation index of the a-th condition a CY' is the mean of the first difference value of the a-th condition of all groups in the target wind turbine a CY' is the mean of the first difference value of the a-th condition of all groups in the normal wind turbine, norm[] is a linear normalization function, and || is an absolute value;

[0017] If the correlation index of the a-th condition is greater than a preset threshold, it is determined that the a-th condition has a specific correlation with the fault entity A;

[0018] Selecting other conditions, repeating the above steps, and determining whether the other conditions have a specific correlation with the fault entity A;

[0019] Selecting other fault entities, repeating the above steps, and determining whether each fault entity has a specific correlation with any condition, thereby obtaining the conditions that each fault entity has a specific correlation with.

[0020] Further, for each entity fault, the time series data of each condition of each wind turbine in the target wind turbine within a preset time period after each fault is obtained, the influence correlation between each target wind turbine and any condition for each fault entity is determined, and the fault entities and conditions that have an influence correlation with each target wind turbine are obtained, and the specific steps of obtaining the ratio of the number of times of occurrence of a fault to a preset number of times of occurrence of a fault between the fault entities and conditions that have an influence correlation with each target wind turbine in a preset number of times of occurrence of a fault before the current detection include:

[0021] Optionally, a fault entity and a condition are selected, denoted as fault entity A and a-th condition;

[0022] For the fault entity A, the mean of the time series data of the a-th condition of each target wind turbine within a preset time period after each fault is denoted as the first mean of each target wind turbine after each fault;

[0023] The mean of the time series data of each target wind turbine within a preset time period after each fault is denoted as the second mean;

[0024] In the target wind turbine, the absolute value of the difference between the first mean and the second mean of each target wind turbine after each fault is calculated; when the absolute value of the difference is greater than one-fourth of the second mean, it is determined that the a-th condition of each target wind turbine has an influence correlation with the A-th fault entity when each fault occurs;

[0025] In each target wind turbine in a preset number of times of occurrence of a fault before the current monitoring, the ratio of the number of times of occurrence of a fault between the a-th condition and the A-th fault entity to a preset number of times of occurrence of a fault is obtained.

[0026] selecting other conditions, repeating the above steps, judging the influence correlation between each target wind turbine and any condition of the fault entity A, and obtaining the conditions of the influence correlation between each target wind turbine and the fault entity A, and obtaining the ratio of the number of failures of the conditions of the influence correlation between each target wind turbine and the fault entity A in the preset number of failures before the current detection to the preset number of failures;

[0027] selecting other fault entities, repeating the above steps, judging the influence correlation between each target wind turbine and any condition of each fault entity, and obtaining the conditions of the influence correlation between each target wind turbine and the fault entity, and obtaining the ratio of the number of failures of the conditions of the influence correlation between each target wind turbine and the fault entity in the preset number of failures before the current detection to the preset number of failures.

[0028] Further, the specific steps of forming the influence combination of each target wind turbine in the target wind turbine class under each fault entity based on the conditions of the influence correlation between each fault entity include:

[0029] selecting one fault entity, denoted as fault entity A;

[0030] forming an influence combination of all conditions of the influence correlation between each target wind turbine and the A-th fault entity;

[0031] selecting other fault entities, repeating the above steps, and forming the influence combination of each target wind turbine in the target wind turbine class under each fault entity.

[0032] Further, the specific steps of obtaining the first probability and the second probability of the influence combination of each target wind turbine in the target wind turbine class under each fault entity based on the ratio of the number of failures of each condition in the influence combination of each target wind turbine in the target wind turbine class under the fault entity A in the preset number of failures before the current detection to the preset number of failures include:

[0033] selecting one fault entity, denoted as fault entity A;

[0034] denoting the average of the ratio of the number of failures of each condition in the influence combination of each target wind turbine in the target wind turbine class under the fault entity A in the preset number of failures before the current detection to the preset number of failures as the first probability of the influence combination of each target wind turbine in the target wind turbine class under the fault entity A;

[0035] denoting the average of the first probability of the influence combination of all target wind turbines in the target wind turbine class under the fault entity A as the second probability;

[0036] selecting another fault entity, repeating the above steps to obtain the first probability and the second probability of the influence combination of each target wind turbine in the target class of wind turbines under each fault entity.

[0037] Further, the specific step of screening the influence combination according to the first probability and the second probability to obtain the updated influence combination of each target wind turbine in the target class of wind turbines under each fault entity comprises:

[0038] Optionally, one fault entity is selected and recorded as fault entity A;

[0039] When the first probability of the influence combination of each target wind turbine in the target class of wind turbines under the fault entity A is greater than the second probability, the influence combination of each target wind turbine is recorded as the updated influence combination.

[0040] Selecting another fault entity, repeating the above steps to obtain the updated influence combination of each target wind turbine in the target class of wind turbines under each fault entity.

[0041] Further, the specific step of constructing the wind turbine knowledge graph based on the updated probability of the updated influence combination and the updated probability of the condition associated with each fault entity comprises:

[0042] Optionally, one fault entity is selected and recorded as fault entity A;

[0043] The average value of the ratio of the number of failures of each condition associated with the fault entity A to the preset number of failures before the current detection is calculated, and recorded as the updated probability of each condition associated with the fault entity A;

[0044] The average value of the association index of all conditions in each updated influence combination is recorded as the association index of the updated influence combination;

[0045] According to the updated probability of each condition associated with the fault entity A and the association index of the updated influence combination, the power knowledge graph of multi-dimensional data is constructed, and the specific method is as follows:

[0046] The attribute vector of the node constructed by the fault entity A in the knowledge graph is [z, x, c, v, b, n], wherein z is a combination of all conditions associated with the fault entity A, x is a combination of the association indexes of all conditions associated with the fault entity A; c is a combination of the updated probabilities of all conditions associated with the fault entity A; v is a combination of all updated influence combinations; b is a combination of the association indexes of all updated influence combinations; n is a combination of the first probabilities of all updated influence combinations;

[0047] In all the existing fault types of all wind power generators, a power knowledge graph is constructed according to attribute vectors corresponding to all fault types, taking each fault type as a node.

[0048] A wind power generator knowledge graph construction system comprises a multi-dimensional data acquisition module, a first data processing module, a second data processing module, a third data processing module and a knowledge graph construction module.

[0049] The multi-dimensional data acquisition module is configured to acquire time series data of multiple wind power generators under multiple conditions and multiple existing fault types.

[0050] The first data processing module is configured to divide, under each fault entity, a target class of wind power generators and a normal class of wind power generators that do not appear the fault entity, and determine whether there is a specific association between each fault entity and any condition according to the correlation index of the target class of wind power generators and the normal class of wind power generators under each fault entity, and acquire conditions that have a specific association with each fault entity.

[0051] The second data processing module is configured to acquire, for each entity fault, time series data of each wind power generator in the target class of wind power generators under each condition within a preset time period after each fault, determine the influence association between each target wind power generator and any condition for each fault entity, and acquire fault entities and conditions that have an influence association with each target wind power generator, and obtain a ratio of the number of faults to a preset value of the fault entity and condition that have an influence association with each target wind power generator in the target class of wind power generators under each fault entity in a preset number of times of occurrence of faults before the current detection, and form an influence combination of each target wind power generator in the target class of wind power generators under each fault entity based on the conditions that have an influence association with each fault entity.

[0052] The third data processing module is configured to acquire a first probability and a second probability of the influence combination of each target wind power generator in the target class of wind power generators under each fault entity based on the ratio of the number of faults to a preset value of the fault entity and condition that have an influence association with each target wind power generator in the target class of wind power generators under each fault entity in a preset number of times of occurrence of faults before the current detection, and screen the influence combination according to the first probability and the second probability to acquire an updated influence combination of each target wind power generator in the target class of wind power generators under each fault entity.

[0053] The knowledge graph construction module is configured to complete construction of the wind power generator knowledge graph based on an update probability of the updated influence combination and an update probability of the conditions that have an influence association with each fault entity.

[0054] An electronic device comprising a memory storing a computer program and a processor configured to invoke and run the computer program stored in the memory to perform the method according to any one of the preceding method claims.

[0055] A computer readable storage medium storing a computer program configured to perform the steps of the method according to any one of the preceding method claims when executed by a processor.

[0056] Compared with the prior art, the present application has the following beneficial effects:

[0057] The present application proposes a wind turbine knowledge graph construction method and system. The method can investigate the weather and humidity conditions when the fault entity occurs in detail, and analyze the correlation between these conditions (including weather conditions, temperature conditions, humidity conditions, geographical and topographical conditions, and regional load change conditions) and the occurrence of the fault entity. The method can determine which specific conditions are associated with the occurrence of the fault entity, and calculate the probability of the occurrence of the fault entity under different conditions. In this way, the present application can effectively screen and analyze multi-dimensional data. In addition, the present application also integrates the results into the knowledge graph, thereby enriching the content of the constructed wind turbine knowledge graph and optimizing its structure. This integration helps to improve the completeness, accuracy and real-time performance of the wind turbine knowledge graph, significantly improving the efficiency of power system management and maintenance. BRIEF DESCRIPTION OF DRAWINGS

[0058] Figure 1 A wind turbine knowledge graph construction method flowchart in an embodiment. DETAILED DESCRIPTION

[0059] The present application will be further illustrated below in conjunction with the accompanying drawings and specific embodiments. It should be understood that these embodiments are only used to illustrate the present application and not to limit the scope of the present application. After reading the present application, those skilled in the art can make various modifications to the present application, and all such modifications fall within the scope defined by the appended claims.

[0060] Embodiment one:

[0061] A wind turbine knowledge graph construction method of the present application, as shown in Figure 1 , comprising:

[0062] S1, obtaining time series data of a plurality of wind power generators under a plurality of conditions and a plurality of fault types that have occurred; the conditions can be meteorological conditions (precipitation, thunderstorm activity), temperature conditions, humidity conditions, geographical terrain conditions (altitude, terrain features, landform), and regional load variation conditions, etc.; the relevant data of meteorological conditions, temperature conditions, and humidity conditions in the region where the target wind power station is located can be obtained by accessing a meteorological website; the relevant data of geographical terrain conditions in the region where the target wind power station is located can be obtained by accessing a geographic information system (GIS) service provider. The time series data collected in combination with a plurality of conditions are cleaned to remove erroneous, duplicate or incomplete records, and time series data under a plurality of conditions are obtained.

[0063] S2, each fault type is taken as a fault entity, and the wind power generator is divided into a target class of wind power generators representing the fault entity and a normal class of wind power generators not representing the fault entity under each fault entity; according to the correlation index of the target class of wind power generators and the normal class of wind power generators under each fault entity, it is judged whether there is a specific correlation between each fault entity and any condition, and the conditions that have a specific correlation with each fault entity are obtained.

[0064] S3, for each entity fault, the time series data of each wind power generator in the target class of wind power generators under each condition within a preset time period after each fault is obtained, the influence correlation between each target wind power generator and any condition for each fault entity is judged, the fault entity and the condition that have an influence correlation corresponding to each target wind power generator are obtained, and the ratio of the number of faults of the fault entity and the condition that have an influence correlation corresponding to each target wind power generator to a preset number of times of faults in a preset number of times of faults before the current detection is obtained, and the influence combination of each target wind power generator in the target class of wind power generators under each fault entity is formed based on the conditions that have an influence correlation with each fault entity.

[0065] S4, based on the ratio of the number of faults of the fault entity and the condition that have an influence correlation corresponding to each target wind power generator to a preset number of times of faults in a preset number of times of faults before the current detection, the first probability and the second probability of the influence combination of each target wind power generator in the target class of wind power generators under each fault entity are obtained; the influence combination is screened according to the first probability and the second probability, and the updated influence combination of each target wind power generator in the target class of wind power generators under each fault entity is obtained.

[0066] S5, based on the update probability of the updated influence combination and the update probability of the condition that has an influence correlation with each fault entity, the construction of the wind power generator knowledge graph is completed.

[0067] Embodiment two:

[0068] The further optional design based on the embodiment one is that, in the example, whether there is a specific association between each fault entity and any condition is judged according to the group association index of the target type wind turbine and the normal type wind turbine under each fault entity, and the specific steps of obtaining the condition which has a specific association with each fault entity include:

[0069] An optional fault entity and a condition are denoted as fault entity A and the a th condition, in the target type wind turbine under the fault entity A, any two target wind turbines form a group, the absolute value of the difference of the mean value of the a th condition time sequence data of each group is obtained, and is denoted as the first difference value of the a th condition of each group;

[0070] According to the obtaining method of the first difference value of the a th condition of each group in the target type wind turbine, the first difference value of the a th condition of each group in the normal type wind turbine is obtained;

[0071] Based on the first difference value of the a th condition of each group in the target type wind turbine and the first difference value of the a th condition of each group in the normal type wind turbine, the association index of the a th condition is calculated by using the following formula:

[0072]

[0073] In the formula, G a is the association index of the a th condition, CY a is the mean value of the first difference value of the a th condition of all groups in the target type wind turbine, CY' a is the mean value of the first difference value of the a th condition of all groups in the normal type wind turbine, norm[] is a linear normalization function, and || is an absolute value;

[0074] If the association index of the a th condition is greater than a preset threshold, it is determined that the a th condition has a specific association with the fault entity A;

[0075] Other conditions are selected, and the above steps are repeated to determine whether the other conditions have a specific association with the fault entity A;

[0076] Other fault entities are selected, and the above steps are repeated to determine whether there is a specific association between each fault entity and any condition, so as to obtain the condition which has a specific association with each fault entity.

[0077] The acquisition process of the condition which has a specific association with the fault entity is further described below by taking an example:

[0078] Firstly, a specific fault of a target wind turbine is determined, the specific fault is taken as an entity, and an analysis target A1 is obtained;

[0079] Using the data samples of the target wind turbine, the differences of meteorological, geographical conditions i are calculated respectively by comparing the data samples j with and without the analysis target A1 Where i = 1 when the analysis target A1 exists, and i = 0 otherwise;

[0080] It should be noted that when obtaining the differences of meteorological, geographical conditions, especially the meteorological differences, the meteorological data when the wind turbine fails is selected, and for the data samples without the analysis target, the mean value of the regional meteorological data is used as the corresponding meteorological, geographical condition value.

[0081] The mean values of the differences of conditions i under the two conditions are calculated respectively That is And

[0082] When the mean value of the condition difference in the data sample with the analysis target A1 is greater, it means that the relationship score of the current condition and the equipment failure is smaller, and vice versa. The mean value of the condition difference in the data sample without the analysis target A1 is greater, it means that the relationship score of the current condition and the equipment failure is greater. ;

[0083]

[0084] According to different analysis targets A n and different conditions i, the relationship score of condition i and analysis target is repeatedly obtained , that is, the relationship score of condition i and different equipment failures (analysis target A n ) of the target wind turbine. The relationship score between analysis target A1 and different conditions can also be repeatedly obtained.

[0085] A threshold is set to screen the relationship score of the above condition and the specific failure. When the relationship score of condition i and analysis target , it is considered that condition i and analysis target A n have a relationship (specific relationship), and the associated condition of the specific failure is identified. Select any analysis target A n , and evaluate the risk of condition i to the analysis target A n for the wind turbine data samples in the same region.

[0086] Example Three:

[0087] The further optional design based on the embodiment two of the present application is that, in the present example, for each entity fault, the time sequence data of each wind turbine in each target wind turbine under each condition within a preset time period after each fault is acquired, the influence correlation between each target wind turbine and any condition for each fault entity is judged, the fault entity and the condition existing the influence correlation corresponding to each target wind turbine are acquired, and the specific steps of obtaining the ratio of the number of faults of the fault entity and the condition existing the influence correlation corresponding to each target wind turbine in the preset number of times of faults before the current detection to the preset number include:

[0088] Optionally, one fault entity and one condition are recorded as fault entity A and the a-th condition;

[0089] For the fault entity A, the mean value of the time sequence data of each target wind turbine under the a-th condition within a preset time period after each fault is recorded as the first mean value of each target wind turbine after each fault;

[0090] The mean value of the time sequence data of each target wind turbine within a preset time period after each fault is recorded as the second mean value;

[0091] In the target wind turbine, the absolute value of the difference between the first mean value and the second mean value of each target wind turbine after each fault is calculated; when the absolute value of the difference is greater than one fourth of the second mean value, it is determined that the a-th condition and the A-th fault entity exist the influence correlation when each target wind turbine occurs the fault;

[0092] In each target wind turbine in the preset number of times of faults before the current detection, the ratio of the number of faults existing the influence correlation between the a-th condition and the A-th fault entity to the preset number (which can be recorded as the probability of the occurrence of the fault caused by each condition of each target wind turbine in the target wind turbine, simply referred to as the probability of the condition);

[0093] Other conditions are selected, and the above steps are repeated to judge the influence correlation between each target wind turbine and any condition for the fault entity A, and the condition existing the influence correlation corresponding to each target wind turbine for the fault entity A is acquired, and the ratio of the number of faults of the condition existing the influence correlation corresponding to each target wind turbine for the fault entity A in the preset number of times of faults before the current detection to the preset number is obtained;

[0094] Select other fault entities, repeat the above steps, determine the influence correlation between each target wind turbine and each fault entity and any condition, and obtain the fault entities and conditions that have influence correlation with each target wind turbine, and obtain the ratio of the number of times of failure of the fault entities and conditions that have influence correlation with each target wind turbine to the preset number of times of failure in the current detection.

[0095] The above process is further illustrated by an example as follows:

[0096] The time sequence data of any wind turbine is analyzed, and the time sequence data difference of each condition at different times (it should be noted that the time sequence data difference obtained at this time only includes the conditions that have a relationship with the analysis target A n n , i.e. conditions with specific correlation) is compared, and the probability of each condition leading to failure is evaluated.

[0097] First, search the historical data samples of the current wind turbine to determine the corresponding data (time stamp, meteorological and geographical conditions, etc.) of the analysis target A1 appearing at different times;

[0098] Calculate the data mean H i of condition i in time sequence, and obtain the data difference value when each failure (analysis target A1) occurs; The smaller the data difference value, the smaller the correlation between the time stamp t and the condition i, and vice versa, the greater the possibility caused by condition i;

[0099] Set a threshold value when the data difference value when each failure (analysis target A1) occurs , it is considered that condition i has a greater impact on the occurrence of this failure. Here, the conditions with greater impact can be considered as conditions that have influence correlation with the fault entity, and the conditions with smaller impact can be considered as conditions that do not have influence correlation with the fault entity;

[0100] Statistically count the number of times condition i has influence on analysis target A1, and combine the total number of times analysis target A1 occurs in the current wind turbine to obtain the probability of condition i leading to failure

[0101] Repeat the above steps to obtain the probability of different conditions leading to failure (analysis target A1)

[0102] Example Four:

[0103] ​The further optional design based on the embodiment three of the present application is that the specific steps of forming the influence combination of each target wind turbine in the target type wind turbine under each fault entity based on the conditions associated with the influence of each fault entity in the present example include:

[0104] Optionally, one fault entity is selected and recorded as fault entity A;

[0105] All conditions associated with the influence of the A-th fault entity in each target wind turbine are combined to form an influence combination;

[0106] Other fault entities are selected and the above steps are repeated to form the influence combination of each target wind turbine in the target type wind turbine under each fault entity.

[0107] Embodiment five:

[0108] The further optional design based on the embodiment four of the present application is that the specific steps of obtaining the first probability and the second probability of the influence combination of each target wind turbine in the target type wind turbine under each fault entity based on the ratio of the number of times of failure to the preset number of times of failure of the conditions associated with the influence of each target wind turbine before the current detection in the preset number of times of failure before the current detection in the present example include:

[0109] Optionally, one fault entity is selected and recorded as fault entity A;

[0110] The average value of the ratio of the number of times of failure to the preset number of times of failure of each condition in the influence combination of each target wind turbine in the target type wind turbine under the fault entity A (i.e. the probability of each condition) before the current detection in the preset number of times of failure before the current detection is recorded as the first probability of the influence combination of each target wind turbine in the target type wind turbine under the fault entity A;

[0111] The average value of the first probability of the influence combination of all target wind turbines in the target type wind turbine under the fault entity A is recorded as the second probability;

[0112] Other fault entities are selected and the above steps are repeated to obtain the first probability and the second probability of the influence combination of each target wind turbine in the target type wind turbine under each fault entity.

[0113] Embodiment six:

[0114] The further optional design based on the embodiment five of the present application is that the specific steps of screening the influence combination according to the first probability and the second probability to obtain the updated influence combination of each target wind turbine in the target type wind turbine under each fault entity in the present example include:

[0115] Optionally, one fault entity is selected and recorded as fault entity A;

[0116] When the first probability of the influence combination of each target wind turbine in the target class of wind turbines of the fault entity A is greater than the second probability, the influence combination of each target wind turbine is recorded as an updated influence combination.

[0117] Other fault entities are selected, and the above steps are repeated to obtain an updated influence combination of each target wind turbine in the target class of wind turbines of each fault entity.

[0118] Embodiment seven:

[0119] The application further comprises the following optional design based on embodiment six: In this embodiment, the specific steps for constructing the wind turbine knowledge graph based on the updated probability of the updated influence combination and the updated probability of the condition associated with each fault entity include:

[0120] An optional fault entity is recorded as fault entity A;

[0121] The average value of the ratio of the number of failures of each condition associated with the fault entity A to the preset number of failures before the current detection (i.e., the probability of each condition) is calculated, and the average value is recorded as the updated probability of each condition associated with the fault entity A.

[0122] The average value of the association index of all conditions in each updated influence combination is recorded as the association index of the updated influence combination.

[0123] According to the updated probability of each condition associated with the fault entity A and the association index of the updated influence combination, the power knowledge graph of the multi-dimensional data is constructed, and the specific method is as follows:

[0124] The attribute vector of the node constructed by the fault entity A in the knowledge graph is [z, x, c, v, b, n], wherein z is a combination of all conditions associated with the fault entity A, x is a combination of the association indexes of all conditions associated with the fault entity A; c is a combination of the updated probabilities of all conditions associated with the fault entity A; v is a combination of all updated influence combinations; b is a combination of the association indexes of all updated influence combinations; and n is a combination of the first probabilities of all updated influence combinations.

[0125] Among all the fault types of all wind turbines, each fault type is taken as a node, and the power knowledge graph is constructed according to the attribute vector corresponding to each fault type.

[0126] Embodiment eight:

[0127] The wind turbine knowledge graph construction system of the application comprises a multidimensional data acquisition module, a first data processing module, a second data processing module, a third data processing module and a knowledge graph construction module.

[0128] The multidimensional data acquisition module is configured to acquire time series data of a plurality of wind turbines under a plurality of conditions and a plurality of fault types that have occurred.

[0129] The first data processing module is configured to divide, under each fault entity, a target class of wind turbines and a normal class of wind turbines that do not appear the fault entity, and determine whether there is a specific association between each fault entity and any condition according to the correlation index of the target class of wind turbines and the normal class of wind turbines under each fault entity, and acquire conditions that have a specific association with each fault entity.

[0130] The second data processing module is configured to acquire, for each entity fault, time series data of each wind turbine in the target class of wind turbines under each condition within a preset time period after each fault, determine the influence association between each target wind turbine and any condition for each fault entity, and acquire fault entities and conditions that have an influence association with each target wind turbine, and obtain a ratio of the number of faults of the fault entities and conditions that have an influence association with each target wind turbine in the target class of wind turbines to a preset number of times of faults before the current detection, and form an influence combination of each target wind turbine in the target class of wind turbines under each fault entity based on the conditions that have an influence association with each fault entity.

[0131] The third data processing module is configured to acquire a first probability and a second probability of the influence combination of each target wind turbine in the target class of wind turbines under each fault entity based on the ratio of the number of faults of the fault entities and conditions that have an influence association with each target wind turbine in the target class of wind turbines to the preset number of times of faults before the current detection, and screen the influence combination according to the first probability and the second probability to acquire an updated influence combination of each target wind turbine in the target class of wind turbines under each fault entity.

[0132] The knowledge graph construction module is configured to complete construction of the wind turbine knowledge graph based on an update probability of the updated influence combination and an update probability of the conditions that have an influence association with each fault entity.

[0133] Embodiment Nine

[0134] An electronic device comprises a memory and a processor, the memory stores a computer program, and the processor is configured to call and run the computer program stored in the memory to execute the method according to any one of the above embodiments.

[0135] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the method of any one of the above.

[0136] Application embodiment:

[0137] The wind turbine knowledge graph construction method of the application comprises the following specific steps:

[0138] Step S001: In a predetermined time period, obtain multiple condition time series data and the types of faults that have occurred for each wind turbine in different regions.

[0139] In a predetermined time period, obtain multiple condition time series data for each wind turbine in different regions. The conditions include meteorological conditions, temperature conditions, humidity conditions, etc.

[0140] In a predetermined time period, obtain the types of faults that have occurred for each wind turbine in different regions.

[0141] In this embodiment, the condition time series data is obtained from reports of meteorological temperature observation stations, environmental humidity monitoring information systems, etc.

[0142] It should be noted that this embodiment takes the collection interval of each condition time series data as 4h and the collection duration as 5 days as an example for description.

[0143] Each condition time series data obtained is classified and standardized to ensure consistency in format and facilitate subsequent data comparison and analysis.

[0144] Step S002: Select any type of fault as the A-th fault entity, and divide all wind turbines into target class wind turbines and normal class wind turbines according to whether the A-th fault entity exists; and determine whether each condition is associated with the A-th fault entity according to the condition time series data of the target class wind turbines and the normal class wind turbines.

[0145] The fault conditions and probabilities of wind turbines under different meteorological and geographical conditions differ significantly. These differences often result from the combined effects of multiple complex factors, such as climate characteristics, topographic changes, and soil properties. By comparing and analyzing the fault conditions and probabilities of wind turbines in different regions under different meteorological and geographical conditions, the degree of association between specific conditions and equipment faults can be obtained, and the relationships between meteorological information and geographical information and entities in the knowledge graph can be identified and defined. For example, local equipment faults of wind turbines may be associated with specific weather conditions, or certain geographical features may affect the operating efficiency of equipment in wind turbines.

[0146] In all the existing fault types of all wind turbines, any one of the fault types is recorded as the A-th fault entity. The fault type can be blade damage, gearbox failure, etc.

[0147] According to whether the A-th fault entity exists, all the wind turbines in different areas are divided into two categories. The wind turbines in which the A-th fault entity exists are recorded as target category wind turbines, and the wind turbines in which the A-th fault entity does not exist are recorded as normal category wind turbines.

[0148] In the target category wind turbines, any two target wind turbines form a group, and the absolute value of the difference between the average of the a-th conditional time series data of each group is obtained, which is recorded as the first difference of the a-th condition of each group.

[0149] According to the obtaining method of the first difference of the a-th condition of each group in the target category wind turbines, the first difference of the a-th condition of each group in the normal category wind turbines is obtained.

[0150] This embodiment quantifies the relationship strength between the condition and the fault, and gives a calculation method of a condition correlation index as follows:

[0151]

[0152] In the formula, G a is the correlation index of the a-th condition, CY a is the average of the first difference of the a-th condition of all groups in the target category wind turbines, CY' a is the average of the first difference of the a-th condition of all groups in the normal category wind turbines, norm[] is a linear normalization function, and || is an absolute value.

[0153] In the above formula, G a The value is closer to 1, indicating that the relationship between the a-th condition and the A-th fault entity is stronger; otherwise, the value is closer to 0, indicating that the relationship is weaker.

[0154] If the correlation index of the a-th condition is greater than a preset threshold, it is determined that the a-th condition and the A-th fault entity have a specific correlation.

[0155] In this embodiment, the preset threshold is 0.6, which is taken as an example for description. In other embodiments, other values can be set, and this embodiment is not limited.

[0156] Similarly, it is determined whether other conditions and the A-th fault entity have a specific correlation.

[0157] It should be noted that the correlation degree analysis of this embodiment is only used to reveal the relationship between meteorological and geographical conditions and equipment faults, and cannot directly indicate the influence degree of a specific condition on equipment faults.

[0158] Step S003: obtaining each conditional time series data of each target wind turbine in the target class wind turbine within a preset time period after each failure; and obtaining the probability of each condition of each target wind turbine in the target class wind turbine according to each conditional time series data of each target wind turbine in the target class wind turbine within a preset time period after each failure.

[0159] In the target class wind turbine, the a-th conditional time series data of the L-th target wind turbine within a preset time period after each failure is obtained. In this embodiment, the preset value is 6, which is taken as an example. It should be noted that the a-th conditional time series data at the time of failure is only obtained within one week after the failure.

[0160] The mean value of the a-th conditional time series data of the L-th target wind turbine within a preset time period after each failure is denoted as the first mean value of the L-th target wind turbine at the time of each failure.

[0161] Further, the mean value of the time series data of the L-th target wind turbine within a preset time period after each failure is denoted as the second mean value.

[0162] In the target class wind turbine, the absolute value of the difference between the first mean value and the second mean value of the L-th target wind turbine at the time of each failure is calculated. When the absolute value of the difference is greater than one fourth of the second mean value, it is determined that there is an influence correlation between the a-th condition and the A-th failure entity of the L-th target wind turbine at the time of each failure.

[0163] It should be noted that the smaller the above-mentioned absolute value of the difference, the smaller the correlation between the failure and the condition at a certain moment, and vice versa.

[0164] In the target class wind turbine, the ratio of the number of failures corresponding to the influence correlation between the a-th condition and the A-th failure entity to the preset value within a preset time period after each failure of the L-th target wind turbine is denoted as the probability of each condition of each target wind turbine in the target class wind turbine.

[0165] Similarly, the probability of each condition of the L-th target wind turbine in the target class wind turbine is obtained.

[0166] Step S004: obtaining the first probability of the influence combination of each target wind turbine in the target class wind turbine according to whether there is a specific correlation between each condition and the A-th failure entity; obtaining the second probability according to the first probability of the influence combination of all target wind turbines in the target class wind turbine; and screening out the updated influence combination from the influence combination according to the first probability and the second probability of the influence combination of each target wind turbine in the target class wind turbine.

[0167] In analyzing the probability of failure occurrence, the failure may occur due to the joint influence of multiple conditions, or it may be caused by a single condition, so when the two conditions are not completely mutually exclusive and not completely independent, the two conditions should be combined into one condition in the process of building the knowledge graph.

[0168] Similarly, determine whether there is an influence association between each condition and the A th fault entity. All conditions with influence associations are combined into an influence combination, denoted as the influence combination of the L th target wind turbine in the target class wind turbine.

[0169] Obtain the mean value of the probability of each condition in the influence combination, denoted as the first probability of the L th target wind turbine influence combination in the target class wind turbine.

[0170] Similarly, obtain the first probability of each target wind turbine influence combination in the target class wind turbine.

[0171] The mean value of the first probability of all target wind turbine influence combinations in the target class wind turbine is denoted as the second probability.

[0172] When the first probability of the L th target wind turbine influence combination in the target class wind turbine is greater than the second probability, determine that each condition in the L th target wind turbine influence combination belongs to a joint influence condition, and the influence combination of the L th target wind turbine is denoted as an updated influence combination, which is divided into the same node in the knowledge graph.

[0173] Thus, some target wind turbines in the target class wind turbine have updated influence combinations.

[0174] Step S005: According to the probability of each condition of each wind turbine and the first probability of the updated influence combination, the power knowledge graph construction of multi-dimensional data is completed.

[0175] In the fault analysis of wind turbines, it is extremely important to deeply understand the probability of failure occurrence under specific conditions or condition combinations. This method can provide more comprehensive data analysis, so as to more accurately predict and prevent potential failures.

[0176] Obtain the probability of each condition of all wind turbines and the first probability of the influence combination.

[0177] It should be noted that the greater the mean value of the probability, the greater the risk caused by the corresponding condition, and the higher the risk assessment score.

[0178] According to the conditions and condition combinations in the above analysis process, according to the relationship between the conditions and the failure and the probability caused, add the corresponding attribute vector to the node in the known knowledge graph.

[0179] Calculate the mean of the probability of each condition associated with the A-th fault entity, denoted as the updated probability of each condition associated with the A-th fault entity.

[0180] Combine the mean of the association index of all conditions in each updated impact combination, denoted as the association index of the updated impact combination.

[0181] For the A-th fault entity, if it has a relationship with any one condition and condition combination, the attribute vector of the node in the knowledge graph is [z, x, c, v, b, n], where z is a combination of all conditions associated with the A-th fault entity, x is a combination of the association index of all conditions associated with the A-th fault entity; c is a combination of the updated probability of all conditions associated with the A-th fault entity; v is a combination of all updated impact combinations; b is a combination of the association index of all updated impact combinations; n is a combination of the first probability of all updated impact combinations.

[0182] Similarly, the attribute vector corresponding to each fault type is obtained, and in all fault types of all wind power generators, each fault type is taken as a node, and the power knowledge graph is constructed according to the attribute vector corresponding to each fault type. The construction of the knowledge graph is a known technology, which will not be described in detail here.

[0183] At this point, the construction of the power knowledge graph of multi-dimensional data is completed.

[0184] It should be noted that for the update of the knowledge graph, first of all, it depends on the continuous monitoring and collection of data sources and environmental changes to determine the update requirement. The running state of the wind power generator and the environmental change are continuously monitored, and new data collected from external data sources such as meteorology and geography are regularly checked and updated. The obtained data is cleaned, and the entity target in the new data is identified. According to the above method, the relationship between entities is extracted from the new data, and the attribute information of the entity is updated or added. Finally, the quality of the data and knowledge is checked to ensure that the updated knowledge graph is accurate.

[0185] At this point, the present application is completed.

[0186] The above is only a specific embodiment of the present application, but the protection scope of the present application is not limited thereto, and any changes or replacements easily thought of by those skilled in the art within the technical scope disclosed by the present application should be covered within the protection scope of the present application.

Claims

1. A method for constructing a knowledge graph of a wind turbine generator, characterized in that, include: Acquire time-series data of multiple wind turbines under various conditions and identify multiple fault types that have occurred; Each fault type is treated as a fault entity. Under each fault entity, wind turbines are divided into target wind turbines that have fault entities and normal wind turbines that do not have fault entities. Based on the correlation index of target wind turbines and normal wind turbines under each fault entity, it is determined whether there is a specific correlation between each fault entity and any condition, and the conditions that have a specific correlation with each fault entity are obtained. For each entity failure, time-series data of each wind turbine in the target type of wind turbine is obtained under each condition within a preset time period after each failure. The influence association between each target wind turbine and each failure entity and any condition is determined. The failure entities and conditions with influence associations for each target wind turbine are obtained. The ratio of the number of failures of the failure entities and conditions with influence associations for each target wind turbine to the preset value of failures before the current detection is obtained. At the same time, the influence combination of each target wind turbine in the target type of wind turbine is formed based on the conditions with influence associations with each failure entity. Based on the ratio of the number of faults occurring to a preset value in the number of faults occurring before the current detection, which is associated with the existence of each target wind turbine, the first probability and the second probability of the influence combination of each target wind turbine in the target type wind turbine under each fault entity are obtained; the influence combination is filtered according to the first probability and the second probability to obtain the updated influence combination of each target wind turbine in the target type wind turbine of each fault entity. The wind turbine knowledge graph is constructed based on the update probability of the updated impact combination and the update probability of the condition that has an impact association with each fault entity.

2. The method for constructing a knowledge graph of a wind turbine generator according to claim 1, characterized in that, The specific steps for determining whether there is a specific association between each fault entity and any condition based on the inter-group correlation index of the target type wind turbine and normal type wind turbine under each fault entity, and obtaining the conditions that have a specific association with each fault entity, include: Choose any fault entity and a condition, denoted as fault entity A and the a-th condition. Among the target wind turbines under fault entity A, group any two target wind turbines together and obtain the absolute value of the difference between the mean of the time series data of the a-th condition in each group, denoted as the first difference of the a-th condition in each group. Based on the method of obtaining the first difference of the first condition of each group in the target type of wind turbine, the first difference of the first condition of each group in the normal type of wind turbine is obtained. Based on the first difference of the a-th condition in each group of target-type wind turbines and the first difference of the a-th condition in each group of normal-type wind turbines, the correlation index of the a-th condition is calculated using the following formula. In the formula, G a CY is the correlation index for the a-th condition. a Let CY' be the mean of the first differences of the a-th condition in all groups of the target class of wind turbines. a Let be the mean of the first difference of the a-th condition in all groups of normal wind turbine generators, norm[] is the linear normalization function, and || is the absolute value; If the correlation index of the a-th condition is greater than the preset threshold, it is determined that there is a specific correlation between the a-th condition and the faulty entity A. Select other conditions, repeat the above steps, and determine whether there is a specific relationship between the other conditions and the faulty entity A; Select other faulty entities and repeat the above steps to determine whether there is a specific association between each faulty entity and any condition, thereby obtaining the conditions under which each faulty entity has a specific association.

3. The method for constructing a knowledge graph of a wind turbine generator according to claim 2, characterized in that, The specific steps for obtaining time-series data of each wind turbine in the target wind turbine class under each condition within a preset time period after each fault for each entity fault, determining the influence association between each target wind turbine on each fault entity and any condition, obtaining the fault entities and conditions with influence associations corresponding to each target wind turbine, and obtaining the ratio of the number of faults with influence associations corresponding to each target wind turbine to a preset value in the number of faults that occurred before the current detection, for each entity fault, include: Choose one faulty entity and one condition, denoted as faulty entity A and the a-th condition; For faulty entity A, the mean of the a-th conditional time series data of each target wind turbine within a preset time period after each fault is recorded as the first mean of each target wind turbine after each fault. The average of the time-series data of each target wind turbine within a preset time period after each failure is recorded as the second average. In the target type of wind turbine, calculate the absolute value of the difference between the first mean and the second mean after each failure of each target wind turbine; when the absolute value of the difference is greater than one-quarter of the second mean, it is determined that there is an influence correlation between the a-th condition and the A-th failure entity of each target wind turbine at each failure. In each target wind turbine, the ratio of the number of failures corresponding to the influence association between the a-th condition and the A-th fault entity to the preset value in the number of failures before the current monitoring; Select other conditions, repeat the above steps, determine the influence association between each target wind turbine on the fault entity A and any condition, obtain the conditions that have an influence association with the fault entity A for each target wind turbine, and obtain the ratio of the number of times the conditions that have an influence association with the fault entity A for each target wind turbine occur to the preset value of faults before the current detection. Select other fault entities, repeat the above steps, determine the influence association between each target wind turbine and each fault entity and any condition, obtain the fault entities and conditions with influence association for each target wind turbine, and obtain the ratio of the number of faults with influence association for each target wind turbine to the preset value of faults occurring before the current detection.

4. The method for constructing a knowledge graph of a wind turbine generator according to claim 3, characterized in that, The specific steps for forming the influence combination of each target wind turbine in the target type of wind turbine under each fault entity based on the condition that there is an influence association with each fault entity include: Choose any faulty entity and denote it as faulty entity A; Group all conditions in each target wind turbine that have an impact association with the Ath fault entity into an impact combination; Select other faulty entities and repeat the above steps to form the combination of influences of each target wind turbine in the target type of wind turbine under each faulty entity.

5. The method for constructing a knowledge graph of a wind turbine generator according to claim 4, characterized in that, The specific steps for obtaining the first and second probabilities of the influence combination of each target wind turbine under each fault entity, based on the ratio of the number of faults occurring in the preset number of faults before the current detection to the preset value, according to the correlation between the fault entity and the condition associated with the existence of each target wind turbine, include: Choose any faulty entity and denote it as faulty entity A; The average of the ratios of the number of failures occurring to the preset value in the influence combination of each condition of each target wind turbine in the target wind turbine category under the fault entity A is denoted as the first probability of the influence combination of each target wind turbine in the target wind turbine category under the fault entity A. The average of the first probability of the combination of the effects of all target wind turbines in the target type wind turbines under the fault entity A is denoted as the second probability. Select other faulty entities and repeat the above steps to obtain the first probability and the second probability of the influence combination of each target wind turbine in the target type wind turbine under each faulty entity.

6. The method for constructing a knowledge graph of a wind turbine generator according to claim 5, characterized in that, The specific steps for filtering the influence combinations based on the first probability and the second probability to obtain the updated influence combination for each target wind turbine in each fault entity target class wind turbine include: Choose any faulty entity and denote it as faulty entity A; When the first probability of the influence combination of each target wind turbine in the target type of wind turbine under the fault entity A is greater than the second probability, the influence combination of each target wind turbine is recorded as the updated influence combination. Select other faulty entities and repeat the above steps to obtain the updated impact combination of each target wind turbine in the target class of each faulty entity.

7. The method for constructing a knowledge graph of a wind turbine generator according to claim 6, characterized in that, The specific steps for constructing the wind turbine knowledge graph based on the update probability of the updated influence combination and the update probability of the condition that has an influence association with each fault entity include: Choose any faulty entity and denote it as faulty entity A; Calculate the average of the ratios of the number of failures to the preset value in the number of failures that occur before the current detection for each condition that is associated with the faulty entity A for all target wind turbines. This is recorded as the update probability of each condition that is associated with the faulty entity A. The mean of the correlation indices of all conditions in each update-affected combination is denoted as the correlation index of the update-affected combination. Based on the update probability of each condition that has an impact association with the faulty entity A, and the association index of the update impact combination, the power knowledge graph of the multidimensional data is constructed. The specific method is as follows: The attribute vector of the node constructed by the faulty entity A in the knowledge graph is [z,x,c,v,b,n], where z is the combination of all conditions that have an impact relationship with the faulty entity A, x is the combination of the association indices of all conditions that have an impact relationship with the faulty entity A, c is the combination of the update probabilities of all conditions that have an impact relationship with the faulty entity A, v is the combination of all update impact combinations, b is the combination of the association indices of all update impact combinations, and n is the combination of the first probabilities of all update impact combinations. Among all the fault types that have occurred in all wind turbines, a power knowledge graph is constructed with each fault type as a node, based on the attribute vectors corresponding to all fault types.

8. A knowledge graph construction system for wind turbine generators, characterized in that, It includes a multi-dimensional data acquisition module, a first data processing module, a second data processing module, a third data processing module, and a knowledge graph construction module; The multi-dimensional data acquisition module is used to acquire time-series data of multiple wind turbines under multiple conditions and multiple fault types that have occurred. The first data processing module is used to treat each type of fault as a fault entity, and under each fault entity, divide the wind turbine into target wind turbines that have fault entities and normal wind turbines that have not fault entities; based on the correlation index of the target wind turbines and normal wind turbines under each fault entity, determine whether there is a specific correlation between each fault entity and any condition, and obtain the conditions that have a specific correlation with each fault entity. The second data processing module is used to acquire time-series data of each wind turbine in the target type of wind turbine under each condition within a preset time period after each fault for each entity fault, determine the influence association between each target wind turbine on each fault entity and any condition, and acquire the fault entity and condition with influence association corresponding to each target wind turbine, and obtain the ratio of the number of faults with influence association corresponding to each target wind turbine to the preset value of faults occurring before the current detection, and at the same time form the influence combination of each target wind turbine in the target type of wind turbine under each fault entity based on the condition with influence association with each fault entity. The third data processing module is used to obtain a first probability and a second probability of the influence combination of each target wind turbine under each fault entity based on the ratio of the number of faults occurring in the preset number of faults before the current detection to the preset value of the fault entity and conditions associated with the existence of each target wind turbine; and to filter the influence combination according to the first probability and the second probability to obtain the updated influence combination of each target wind turbine in the target wind turbine of each fault entity. The knowledge graph construction module is used to construct the wind turbine knowledge graph based on the update probability of the update influence combination and the update probability of the condition that has an influence association with each fault entity.

9. An electronic device, characterized in that, The electronic device includes a memory and a processor, the memory storing a computer program, and the processor being configured to invoke and run the computer program stored in the memory to perform the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1 to 7 above.

Images