Wind power plant fault analysis method and system based on multi-dimensional information association

Abstract

The invention provides a wind power plant fault analysis method and system based on multi-dimensional information association, and relates to the technical field of fault analysis. Field domain configuration information of a target wind power plant is read, a secondary loop topology is established, topology node identification is performed based on data homology and data correlation, a network time protocol is constructed, and the wind power plant fault is analyzed based on the secondary loop topology. Constructing a fault prediction module by combining a fault knowledge graph, determining a transmission timestamp, carrying out data return constraint on the multi-dimensional source data by combining a network time protocol, transmitting the multi-dimensional source data to the fault prediction module, carrying out wind power plant fault prediction, and carrying out fault operation and maintenance management on a target wind power plant based on a fault prediction result. The technical problem that in the prior art, comprehensive analysis on multi-dimensional information of a wind power plant is lacked, and consequently fault prediction accuracy and real-time performance are insufficient is solved. The technical effects of comprehensively and accurately analyzing the multi-dimensional information of the wind power plant and improving the accuracy and real-time performance of fault analysis and prediction are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of fault analysis, and in particular to a wind farm fault analysis method and system based on multi-dimensional information association. Background Art

[0002] As an important part of renewable energy, wind farms are gradually becoming the key to the energy industry. However, during the operation of wind farms, due to their complex electrical and mechanical systems, faults occur frequently. Traditional wind farm fault analysis methods mostly rely on single-dimensional data, such as the operating parameters of a single device or a single type of fault record, and often lack comprehensive analysis of comprehensive information, resulting in low accuracy and real-time performance of fault prediction.

[0003] The existing technology lacks comprehensive analysis of multi-dimensional information of wind farms, resulting in technical problems such as insufficient accuracy and real-time performance of fault prediction. Summary of the invention

[0004] The present application provides a wind farm fault analysis method and system based on multi-dimensional information association, which is used to solve the technical problem in the prior art that there is a lack of comprehensive analysis of wind farm multi-dimensional information, resulting in insufficient accuracy and real-time performance of fault prediction.

[0005] In view of the above problems, the present application provides a wind farm fault analysis method and system based on multi-dimensional information association.

[0006] In a first aspect of the present application, a wind farm fault analysis method based on multi-dimensional information association is provided, the method comprising: reading the field configuration information of the target wind farm and establishing a secondary loop topology; traversing the secondary loop topology, identifying topological nodes based on data homology and data correlation, and constructing a network time protocol, the network time protocol being used to perform timestamp constraints on node data return; based on the secondary loop topology, a fault prediction module is constructed in combination with a fault knowledge graph, the fault prediction module comprising a data fitting layer and a fault decision layer, wherein the fault knowledge graph has data granularity consistency with the multi-dimensional information, the data fitting layer establishes a communication connection with a unified information model, the unified information model is used to perform data de-heterogeneity; determining a transmission timestamp, and performing data return constraints on the multi-dimensional source data in combination with the network time protocol; transmitting the multi-dimensional source data to the fault prediction module, converting and fitting the associated data with the data structure, executing wind farm fault prediction, and determining a fault prediction result; based on the fault prediction result, performing fault operation and maintenance management on the target wind farm.

[0007] A second aspect of the present application provides a wind farm fault analysis system based on multi-dimensional information association, the system comprising: a secondary circuit topology establishment unit, the secondary circuit topology establishment unit being used to read field configuration information of a target wind farm and establish a secondary circuit topology;

[0008] A network time protocol construction unit, the network time protocol construction unit is used to traverse the secondary loop topology, identify topological nodes based on data homology and data correlation, and construct a network time protocol, the network time protocol is used to perform timestamp constraints on node data return; a fault prediction module construction unit, the fault prediction module construction unit is used to construct a fault prediction module based on the secondary loop topology and in combination with a fault knowledge graph, the fault prediction module includes a data fitting layer and a fault decision layer, wherein the fault knowledge graph has data granularity consistency with multidimensional information, the data fitting layer is constructed with a unified information model A communication connection is established, and the unified information model is used to perform data de-heterogeneity; a data return constraint unit, the data return constraint unit is used to determine the transmission timestamp, and in combination with the network time protocol, perform data return constraints on the multi-dimensional source data; a wind farm fault prediction unit, the wind farm fault prediction unit is used to transmit the multi-dimensional source data to the fault prediction module, perform conversion and fitting of associated data and data structure, execute wind farm fault prediction, and determine the fault prediction result; a fault operation and maintenance management unit, the fault operation and maintenance management unit is used to perform fault operation and maintenance management on the target wind farm based on the fault prediction result.

[0009] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0010] The method provided in the embodiment of the present application reads the field configuration information of the target wind farm, establishes the secondary loop topology, traverses the secondary loop topology, identifies the topological nodes based on data homology and data correlation, constructs a network time protocol, and constructs a fault prediction module based on the secondary loop topology in combination with the fault knowledge graph, determines the transmission timestamp, and performs data return constraints on the multi-dimensional source data in combination with the network time protocol, transmits the multi-dimensional source data to the fault prediction module, converts and fits the associated data to the data structure, performs wind farm fault prediction, determines the fault prediction results, and performs fault operation and maintenance management on the target wind farm based on the fault prediction results. The technical effect of comprehensively and accurately analyzing the multi-dimensional information of the wind farm, improving the accuracy and real-time performance of fault analysis and prediction, and improving the operational efficiency and safety of the wind farm is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0012] Figure 1 A schematic diagram of the flow of a wind farm fault analysis method based on multi-dimensional information association provided in this application;

[0013] Figure 2 A schematic diagram of the structure of a wind farm fault analysis system based on multi-dimensional information association provided in this application.

[0014] Explanation of the reference numerals: 11 secondary circuit topology establishing unit, 12 network time protocol building unit, 13 fault prediction module building unit, 14 data return constraint unit, 15 wind farm fault prediction unit, 16 fault operation and maintenance management unit. DETAILED DESCRIPTION

[0015] The present application provides a wind farm fault analysis method and system based on multi-dimensional information association, which is used to solve the technical problem that the prior art lacks comprehensive analysis of wind farm multi-dimensional information, resulting in insufficient accuracy and real-time performance of fault prediction. The application achieves the technical effect of comprehensively and accurately analyzing wind farm multi-dimensional information, improving the accuracy and real-time performance of fault analysis prediction, and improving the operational efficiency and safety of wind farms.

[0016] Below, the technical solutions in the present invention will be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments of the present invention. It should be understood that the present invention is not limited to the example embodiments described herein. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. It should also be noted that, for the convenience of description, only the parts related to the present invention are shown in the accompanying drawings, rather than all of them.

[0017] Embodiment 1, as Figure 1 As shown, the present application provides a wind farm fault analysis method based on multi-dimensional information association, the method comprising:

[0018] Read the field configuration information of the target wind farm and establish the secondary circuit topology.

[0019] Specifically, various configuration information related to the target wind farm is collected through the wind farm's monitoring system, SCADA system, database, etc., to obtain the field configuration information of the target wind farm. The field configuration information refers to a collection of information on various technical and operating parameters related to the wind farm, including but not limited to equipment type, layout location, substation location, line connection relationship, protection equipment, etc.

[0020] The secondary circuit system of a wind farm refers to the electrical system used for control, protection, measurement and signal transmission in the wind farm. Unlike the primary circuit system of a wind farm, which is mainly responsible for the transmission and distribution of electric energy, the secondary circuit system does not directly participate in the transmission of electric energy, but provides necessary auxiliary functions for the operation of the primary circuit. Based on the obtained field configuration information of the target wind farm, the equipment and equipment components in the secondary circuit system are regarded as nodes in the topology graph, and the connections between the equipment are regarded as edges. The secondary circuit topology graph is constructed using graph theory methods, such as adjacency matrix or adjacency list, that is, the secondary circuit topology structure is represented by using an adjacency matrix or adjacency list. The secondary circuit topology graph can cover each key node and equipment more comprehensively, and can clearly show the electrical connection relationship between each device, providing a basis for subsequent fault prediction and operation and maintenance management.

[0021] The secondary loop topology is traversed, topological nodes are identified based on data homology and data correlation, and a network time protocol is constructed. The network time protocol is used to perform timestamp constraints on node data return.

[0022] Specifically, the established secondary circuit topology diagram is traversed to identify all nodes, that is, to identify all devices and device components. According to the homology and correlation of the data, a unique identifier is assigned to each node, and its data type and source are marked. The homology analysis refers to determining the source and flow of data in the secondary circuit topology to ensure the homology of the data, that is, the data comes from the same device or device component. The correlation analysis refers to analyzing the correlation between different data and determining the association between nodes. Construct a network time protocol to constrain the timestamp of data returned between nodes. The timestamp of constraining data returned between nodes refers to unified management and control of time information in the data return process to ensure the timing consistency and accuracy of data transmission between nodes. By effectively constraining the timestamp of data returned between nodes, the timing consistency and accuracy of data during transmission are ensured, providing reliable data support for wind farm fault analysis and prediction.

[0023] Based on the secondary loop topology, a fault prediction module is constructed in combination with a fault knowledge graph. The fault prediction module includes a data fitting layer and a fault decision layer. The fault knowledge graph has data granularity consistency with multidimensional information. A communication connection is established between the data fitting layer and a unified information model. The unified information model is used to perform data de-heterogeneity.

[0024] Specifically, a fault knowledge graph is constructed using historical fault data of wind farms, expert knowledge, etc. The fault knowledge graph is a structured representation of various faults that have occurred in the history of wind farms and their related information, including multi-dimensional information such as fault type, fault cause, fault manifestation, and impact range. The fault knowledge graph has data granularity consistency with the multi-dimensional information. The data granularity consistency ensures that the collected multi-dimensional information is consistent with the fault knowledge graph in terms of data granularity, that is, the degree of data refinement and time resolution match. Furthermore, according to the secondary circuit topology and fault knowledge graph, a fault prediction module is constructed. The fault prediction module includes a data fitting layer and a fault decision layer. The data fitting layer is responsible for fitting the multidimensional data collected in real time with the fault knowledge graph. By analyzing the correlation of multidimensional information, potential fault modes are discovered. The unified information model is used to process multidimensional data from different sources and different formats, and de-heterogeneize it so that the data can be analyzed and processed on the same platform. The fault decision layer performs fault prediction based on the analysis results of the data fitting layer and the fault knowledge graph, which can effectively improve the accuracy and real-time performance of wind farm fault prediction and provide strong support for the operation and maintenance management of wind farms. The fault decision layer is a neural network, which predicts possible future faults by learning the feature representation and mapping relationship of historical fault data.

[0025] The transmission timestamp is determined, and the data return constraints are performed on the multi-dimensional source data in combination with the network time protocol.

[0026] Specifically, the transmission timestamp refers to the time mark added to each piece of data during the data transmission process, recording the specific time points of data generation, transmission and reception. The accuracy of the transmission timestamp directly affects the timing consistency of the data and the accuracy of the analysis. The time of the target wind farm equipment and equipment components is synchronized through the network time protocol to ensure that all equipment is always consistent. Before the data is returned, the multi-dimensional source data is packaged into data packets, and the timestamp of the sending time is embedded in each data packet. This timestamp is obtained by querying the synchronized local clock, ensuring the accuracy and timeliness of the timestamp. At the data receiving end, the received data is checked for timestamp consistency to ensure that the data generation time, transmission time and reception time are within a reasonable range. The timestamp fault tolerance mechanism is used to process data that exceeds the fault tolerance range, realizing the effective data return constraint on the multi-dimensional source data, ensuring the timeliness and accuracy of the data.

[0027] The multi-dimensional source data is transmitted to the fault prediction module, the associated data is converted and fitted with the data structure, the wind farm fault prediction is performed, and the fault prediction result is determined;

[0028] Based on the fault prediction result, fault operation and maintenance management is performed on the target wind farm.

[0029] Specifically, the multi-dimensional source data is used as input data and transmitted to the fault prediction module. The data fitting layer of the fault prediction module is responsible for the conversion and fitting of the associated data and data structure, and uniformly converts data from different sources and formats into a consistent format. Then, the fault decision layer of the fault prediction module performs wind farm fault prediction based on the converted data and generates specific fault prediction results, thereby improving the accuracy and real-time performance of fault analysis and prediction.

[0030] Finally, based on the prediction results output by the module, fault operation and maintenance management is carried out, which includes fault handling, maintenance plan adjustment and preventive measures formulation, so as to ensure the efficient and safe operation of the wind farm, realize accurate prediction and effective management of potential faults in the wind farm, and improve the operational efficiency and safety of the wind farm.

[0031] Furthermore, topological node identification is performed based on data homology and data correlation, including: the data correlation includes spatial correlation and time correlation; a first identifier is determined based on the data homology, a second identifier is determined based on spatial correlation, and a third identifier is determined based on time correlation; based on the first identifier, the second identifier and the third identifier, the secondary loop topology is traversed to perform topological node identification.

[0032] Specifically, data correlation refers to the existence of a collaborative relationship between data, such as two types of data working together to determine a certain fault, etc. Data correlation includes spatial correlation and temporal correlation. The spatial correlation refers to the mutual dependence of data in spatial position or distribution. For example, adjacent wind turbines in the same wind farm may exhibit similar operating conditions or fault modes due to similar environmental conditions (such as wind speed, wind direction, temperature, etc.). The temporal correlation refers to the mutual dependence of data over time, which is used to represent the correlation of the same data set or different data sets at different time points.

[0033] Based on data homology, a first identifier is assigned to node data with homology. The first identifier can reflect the homology of node data and help to quickly identify and distinguish the source of data in subsequent processing. Based on spatial correlation, a second identifier is assigned to nodes with spatial correlation. The second identifier reflects the physical proximity between nodes, which facilitates subsequent spatial analysis and fault location. Based on phase correlation, a third identifier is assigned to data with phase correlation. The third identifier helps to analyze the time-varying laws of data, predict fault behavior, and respond to abnormal situations in a timely manner.

[0034] After determining the first identifier, the second identifier and the third identifier, the secondary circuit topology is traversed based on the first identifier, the second identifier and the third identifier to identify each topological node, which can effectively and accurately identify the nodes of the secondary circuit topology of the wind farm, ensuring that in the fault analysis and prediction process, the homology and correlation of the data can be fully considered, thereby improving the accuracy and reliability of the analysis.

[0035] Furthermore, before constructing the fault prediction module, it includes: calling the operation and maintenance records of the preset time zone of the target wind farm; clustering the operation and maintenance records once based on the fault location, clustering the operation and maintenance records twice based on the fault type, fitting the primary clustering result and the secondary clustering result, and determining N clustering clusters; screening out clusters whose number of clusters is less than a preset number among the N clustering clusters, extracting labels for the entire fault cycle and constructing a fault knowledge graph for the remaining clustering clusters, wherein there is a corresponding relationship between the label group and the clustering cluster.

[0036] Specifically, first, all operation and maintenance records of the target wind farm within a specific preset time zone are called. The operation and maintenance records record in detail the status changes, fault occurrence and resolution of the wind farm equipment during the operation and maintenance process. The characteristic preset time zone is set according to actual needs, such as one year, half a year, three months and other time ranges.

[0037] Next, based on the fault location, the operation and maintenance records are clustered using a clustering algorithm. The fault location refers to the specific equipment or area where the fault occurs, such as a generator, transformer, cable line, etc. The clustering algorithm is, for example, K-means, hierarchical clustering, etc. Based on a clustering, the fault records with the same or similar fault locations in the operation and maintenance records can be grouped together, thereby identifying which locations are high-prone areas for faults.

[0038] After completing the first clustering, the operation and maintenance records are further clustered based on the fault type. The fault type refers to the specific nature or manifestation of the fault, such as mechanical fault, electrical fault, control system fault, environmental impact, etc. Similar to the first clustering, by clustering the fault type, the fault records of the same type can be classified into one category, thereby identifying the patterns of different types of faults.

[0039] Then, the first clustering result is fitted with the second clustering result, and the final N clusters are determined by comprehensively analyzing the relationship between the fault location and the fault type. N is a positive integer, and each cluster represents a group of fault records with similar location and type. The N clusters are screened to remove clusters with less than the preset number of clusters. The removed small clusters cannot accurately reflect the actual fault situation due to insufficient data, so they are removed to ensure that only clusters with sufficient sample numbers are retained to improve the reliability of the analysis. The preset number is a threshold set based on the historical data of the wind farm and the frequency of faults.

[0040] Finally, the remaining clusters are subjected to full-cycle label extraction. Label extraction refers to extracting feature labels related to faults from operation and maintenance records, including key information such as fault location, fault type, fault occurrence time, repair time, and impact range. The extracted labels are associated with clusters in a structured representation. Each label group has a one-to-one correspondence with the corresponding cluster, and a fault knowledge graph is constructed. The fault knowledge graph is a structured knowledge representation method that can clearly display complex information such as the correlation between faults, the causes and consequences of faults, etc. By constructing a fault knowledge graph, effective management and utilization of wind farm operation and maintenance knowledge can be achieved, and the level of intelligent fault handling can be improved.

[0041] Furthermore, a fault knowledge graph is constructed, including: based on the source data granularity of the secondary circuit topology, multi-level data granularity is determined, and a graph hierarchy is constructed; the multi-level data granularity is traversed, and fault labels are preprocessed to determine preprocessed labels, wherein the preprocessing criteria include label decomposition and label elevation; based on the preprocessing labels and the graph hierarchy, the fault knowledge graph is constructed.

[0042] Specifically, the multi-level data granularity is determined according to the degree of refinement of the source data granularity of the secondary circuit topology. The source data has the characteristics of fine granularity, and describes in detail the connection relationship, attributes and status of each component in the circuit. The source data granularity reflects the specific degree of refinement of each node data in the secondary circuit topology. Data granularity refers to the degree of detail and refinement of the data. Multi-level data granularity is to divide the source data according to different levels of refinement, reflecting the characteristics of the data at different levels. For example, it can be divided into equipment level, system level and station level. The hierarchical structure of the fault knowledge graph is constructed according to the determined multi-level data granularity, and each level of the graph level represents a different data granularity.

[0043] Next, the multi-level data granularity is traversed to perform label preprocessing on the fault label. The fault label refers to the characteristic mark related to the fault, including the fault location, fault type, fault occurrence time, repair time, impact range, etc. The fault label is preprocessed by two steps: label decomposition and label elevation. Label decomposition can break down complex or fuzzy fault labels into more specific and easier to understand sub-labels, which helps to refine fault classification and improve the accuracy of the knowledge graph. For example, "fan electrical fault" is decomposed into "fan overload fault" and "fan short circuit fault". Through label elevation, fault labels with similar characteristics or belonging to the same category are merged into higher-level labels to achieve knowledge generalization and aggregation, and reduce label redundancy and complexity. Through label preprocessing, the preprocessed labels are determined to ensure the consistency and accuracy of fault labels at different granularity levels.

[0044] Finally, based on the preprocessed labels and graph levels, the labels are associated with the nodes in the graph level, and these nodes are connected through edges to construct a fault knowledge graph. The fault knowledge graph organizes and associates the preprocessed fault labels according to multi-level data granularity and graph levels to form a structured graph representation. The knowledge graph can comprehensively and accurately reflect the hierarchical relationship and correlation of wind farm faults, provide comprehensive knowledge support for fault analysis and prediction, and thus improve the accuracy and reliability of fault analysis and prediction.

[0045] Furthermore, the data fitting layer establishes a communication connection with the unified information model, including: traversing the multidimensional source data to determine the main data structure, wherein the main data structure is any one of the multidimensional source data; taking the main data structure as the conversion target, combining with the unified information model, traversing the multidimensional source data to perform data isomorphism processing, and determining isomorphic source data; transmitting the isomorphic source data to the data fitting layer.

[0046] Specifically, multidimensional source data refers to various types of data collected from different sensors and devices in a wind farm, such as voltage, current, wind speed, temperature, etc. The multidimensional source data is traversed, and any data structure as a conversion benchmark is selected from the multi-source data, and the data is used as the main data structure for subsequent data conversion and processing. When selecting the main data structure, the representativeness and universality of the data can be considered. Then, the main data structure is used as the conversion target, combined with the unified information model, the multidimensional source data is traversed, and the multidimensional source data from different sources and different formats are converted into the format of the main data structure by means of data cleaning, data conversion and data mapping. The data cleaning includes deduplication, missing value filling, the data conversion includes format unification, unit conversion, and the data mapping includes field matching. After isomorphism processing, the isomorphic source data is determined, that is, all data in the multidimensional source data have been converted into the format of the main data structure, and the isomorphic source data is consistent in structure and format. Finally, the isomorphic source data in a unified format is transmitted to the data fitting layer for data processing and analysis, identification of fault modes, and fault prediction. By establishing a communication connection between the data fitting layer and the unified information model, the heterogeneity problem in data integration is effectively solved, which not only improves the efficiency and accuracy of data processing, but also provides a reliable data basis for wind farm fault analysis and prediction, thereby improving the accuracy and real-time performance of fault prediction.

[0047] Furthermore, the determination of the transmission timestamp includes: determining the data topology type for the multidimensional source data; determining the transmission timestamp based on the data topology type, the transmission timestamp including a time limit dimension and a status dimension, the status dimension including a synchronization timestamp and a stage timestamp; the network time protocol performs backhaul constraints on the multidimensional source data in response to the transmission timestamp.

[0048] Specifically, by analyzing each data in the multidimensional source data, the data topology type is determined, and the data topology type includes data homology and data correlation. Based on the determined data topology type, the dependency and timing requirements of data transmission are clarified, and based on the analysis results, the transmission timestamp is determined. The transmission timestamp includes two dimensions: time limit dimension and state dimension. The time limit dimension refers to the receiving time limit of data transmission, that is, the data is received within the specified time to ensure the real-time and validity of the data. The state dimension includes a synchronization timestamp and a stage timestamp. The synchronization timestamp is used to identify and ensure the consistency of all relevant data at the same time node, and the stage timestamp is used to identify the time point of data in different transmission stages to ensure the accurate recording of data at each stage during the transmission process. The network time protocol is configured on each data acquisition node. The time synchronization mechanism of the network time protocol will respond to the transmission timestamp and strictly constrain the return process of the multidimensional source data. The network time protocol can ensure that all devices participating in data transmission in the network maintain a high degree of time synchronization, thereby avoiding time deviation in the data transmission process and providing a reliable data basis for wind farm fault analysis and prediction.

[0049] Furthermore, after the fault operation and maintenance management of the target wind farm is performed, it includes: based on a preset time period, reading periodic fault decision records; based on the fault decision records, mining new fault data and error fault data, wherein the new fault data and the error fault data meet the preset frequency; based on the new fault data, updating the fault knowledge graph; based on the error fault data, updating and learning the fault decision layer.

[0050] Specifically, a preset time period is set according to the actual target wind farm equipment operating characteristics, fault frequency and maintenance requirements, such as weekly, monthly, three months, etc. Then, all fault decision records within the preset time period are retrieved and read from the wind farm operation database to ensure that the latest fault information can be captured in time. Then, the extracted fault decision records are deeply analyzed, and two types of key data are extracted from the periodic fault decision records using data mining technology: new fault data and error fault data, where the screening of new fault data and error fault data needs to meet the preset frequency, that is, it will be considered as valid data only when it appears more than a certain number of times within the statistical period, so as to ensure the representativeness of the data and the accuracy of the analysis. The new fault data refers to the new fault type or mode that appears within the preset time period, and the error fault data refers to the fault record that is incorrectly classified or processed by the decision layer during the fault decision process. Then, the new fault data is added to the fault knowledge graph, and the fault knowledge graph is updated so that it can more accurately reflect the fault situation and trend of the current wind farm, and then more comprehensively understand the fault phenomenon, and provide a more solid data foundation for subsequent fault identification and diagnosis. At the same time, by analyzing the error fault data, the causes of misjudgment and false alarm in the fault decision process can be identified, and then the fault decision layer can be updated and learned, so that the fault decision layer can gradually improve its accuracy and robustness, and improve the accuracy and reliability of fault prediction. By extracting and analyzing the fault decision records, and then updating the graph and fault decision layer, the fault knowledge graph and fault decision layer can be updated periodically, ensuring the real-time and accuracy of the wind farm fault analysis system, and providing continuous support and optimization for the intelligent operation and maintenance of wind farms.

[0051] Embodiment 2 is based on the same inventive concept as the wind farm fault analysis method based on multi-dimensional information association in the above embodiment. Figure 2 As shown, the present application provides a wind farm fault analysis system based on multi-dimensional information association, wherein the system includes:

[0052] A secondary circuit topology establishing unit 11, wherein the secondary circuit topology establishing unit 11 is used to read the field configuration information of the target wind farm and establish a secondary circuit topology;

[0053] A network time protocol construction unit 12, the network time protocol construction unit 12 is used to traverse the secondary loop topology, identify topological nodes based on data homology and data correlation, and construct a network time protocol, the network time protocol is used to perform timestamp constraints on node data return;

[0054] A fault prediction module construction unit 13, the fault prediction module construction unit 13 is used to construct a fault prediction module based on the secondary circuit topology and in combination with a fault knowledge graph, the fault prediction module includes a data fitting layer and a fault decision layer, wherein the fault knowledge graph and multi-dimensional information have data granularity consistency, the data fitting layer establishes a communication connection with a unified information model, and the unified information model is used to perform data de-heterogeneity;

[0055] A data return constraint unit 14, the data return constraint unit 14 is used to determine a transmission timestamp and perform data return constraints on the multi-dimensional source data in combination with the network time protocol;

[0056] A wind farm fault prediction unit 15, the wind farm fault prediction unit 15 is used to transmit the multi-dimensional source data to the fault prediction module, perform conversion and fitting of associated data and data structure, perform wind farm fault prediction, and determine fault prediction results;

[0057] A fault operation and maintenance management unit 16 is used to perform fault operation and maintenance management on the target wind farm based on the fault prediction result.

[0058] Furthermore, the network time protocol construction unit 12 is used to perform the following steps:

[0059] The data correlation includes spatial correlation and phase correlation; a first identifier is determined based on the data homology, a second identifier is determined based on the spatial correlation, and a third identifier is determined based on the phase correlation; based on the first identifier, the second identifier and the third identifier, the secondary loop topology is traversed to perform topological node identification.

[0060] Furthermore, the fault prediction module construction unit 13 is used to perform the following steps:

[0061] Call the operation and maintenance records of the preset time zone of the target wind farm; cluster the operation and maintenance records once based on the fault location, cluster the operation and maintenance records twice based on the fault type, fit the primary clustering result and the secondary clustering result, and determine N clusters; screen out the clusters whose number of clusters is less than a preset number among the N clusters, extract labels for the entire fault cycle and construct a fault knowledge graph for the remaining clusters, wherein there is a corresponding relationship between the label group and the cluster.

[0062] Furthermore, the fault prediction module construction unit 13 is further configured to perform the following steps:

[0063] Based on the source data granularity of the secondary loop topology, determine the multi-level data granularity and build the graph hierarchy; traverse the multi-level data granularity, perform label preprocessing on the fault labels, and determine the preprocessing labels, wherein the preprocessing standard includes label decomposition and label elevation; based on the preprocessing labels and the graph hierarchy, construct the fault knowledge graph.

[0064] Furthermore, the fault prediction module construction unit 13 is further configured to perform the following steps:

[0065] The multidimensional source data is traversed to determine a main data structure, wherein the main data structure is any one of the multidimensional source data; taking the main data structure as a conversion target, in combination with the unified information model, the multidimensional source data is traversed to perform data isomorphism processing to determine isomorphic source data; and the isomorphic source data is transmitted to the data fitting layer.

[0066] Furthermore, the data transmission constraint unit 14 is used to perform the following steps:

[0067] For the multidimensional source data, determine the data topology type; based on the data topology type, determine the transmission timestamp, the transmission timestamp includes a time limit dimension and a status dimension, the status dimension includes a synchronization timestamp and a stage timestamp; the network time protocol responds to the transmission timestamp to perform backhaul constraints on the multidimensional source data.

[0068] Furthermore, the fault operation and maintenance management unit 16 is used to perform the following steps:

[0069] Based on a preset time period, periodic fault decision records are read; based on the fault decision records, new fault data and error fault data are mined, wherein the new fault data and the error fault data meet a preset frequency; based on the new fault data, the fault knowledge graph is updated; based on the error fault data, the fault decision layer is updated and learned.

[0070] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application should be included in the protection scope of the present application.

[0071] This specification and drawings are merely exemplary illustrations of the present application and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.

Claims

1. A wind farm fault analysis method based on multi-dimensional information association, characterized in that: The method comprises: Read the field configuration information of the target wind farm and establish the secondary circuit topology; Traversing the secondary loop topology, identifying topological nodes based on data homology and data correlation, and constructing a network time protocol, wherein the network time protocol is used to perform timestamp constraints on node data return; Based on the secondary circuit topology, a fault prediction module is constructed in combination with a fault knowledge graph, wherein the fault prediction module includes a data fitting layer and a fault decision layer, wherein the fault knowledge graph and multidimensional information have data granularity consistency, and a communication connection is established between the data fitting layer and a unified information model, and the unified information model is used to perform data de-heterogeneity; Determine the transmission timestamp, and perform data return constraints on the multi-dimensional source data in combination with the network time protocol; The multi-dimensional source data is transmitted to the fault prediction module, the associated data is converted and fitted with the data structure, the wind farm fault prediction is performed, and the fault prediction result is determined; Based on the fault prediction result, fault operation and maintenance management is performed on the target wind farm.

2. The wind farm fault analysis method based on multi-dimensional information association according to claim 1, characterized in that: Topological node identification based on data homology and data correlation, including: The data correlation includes spatial correlation and temporal correlation; Determine a first identifier based on the data homology, determine a second identifier based on spatial correlation, and determine a third identifier based on temporal correlation; Based on the first identifier, the second identifier, and the third identifier, the secondary loop topology is traversed to perform topological node identification.

3. The wind farm fault analysis method based on multi-dimensional information association according to claim 1, characterized in that: Before building the fault prediction module, include: Calling the operation and maintenance records of the target wind farm in a preset time zone; Performing a primary clustering on the operation and maintenance records based on the fault location, performing a secondary clustering on the operation and maintenance records based on the fault type, fitting the primary clustering result and the secondary clustering result, and determining N clusters; The clusters whose number of clusters is less than a preset number in the N clusters are screened out, and labels of the entire fault cycle and fault knowledge graph construction are performed on the remaining clusters, wherein there is a corresponding relationship between the label group and the cluster.

4. The wind farm fault analysis method based on multi-dimensional information association according to claim 3, characterized in that: Construct a fault knowledge graph, including: Based on the source data granularity of the secondary loop topology, determine the multi-level data granularity and build the graph hierarchy; Traversing the multi-level data granularity, performing label preprocessing on the fault label, and determining the preprocessing label, wherein the preprocessing standard includes label decomposition and label up-position; Based on the preprocessing labels and the graph hierarchy, the fault knowledge graph is constructed.

5. The wind farm fault analysis method based on multi-dimensional information association according to claim 1, characterized in that: The data fitting layer establishes a communication connection with the unified information model, including: Traversing the multi-dimensional source data to determine a main data structure, wherein the main data structure is any one of the multi-dimensional source data; Taking the master data structure as the conversion target, combining with the unified information model, traversing the multi-dimensional source data to perform data isomorphism processing, and determining isomorphic source data; The homogeneous source data is transferred to the data fitting layer.

6. The wind farm fault analysis method based on multi-dimensional information association according to claim 1, characterized in that: The determining of the transmission timestamp comprises: For the multidimensional source data, determining a data topology type; Based on the data topology type, determine a transmission timestamp, wherein the transmission timestamp includes a time limit dimension and a state dimension, and the state dimension includes a synchronization timestamp and a phase timestamp; The network time protocol performs a return constraint on the multi-dimensional source data in response to the transmission timestamp.

7. The wind farm fault analysis method based on multi-dimensional information association according to claim 1, characterized in that: After the target wind farm is fault-operated and maintained, the following steps are performed: Read periodic fault decision records based on a preset time period; Based on the fault decision record, mining new fault data and error fault data, wherein the new fault data and the error fault data meet a preset frequency; Based on the newly added fault data, the fault knowledge graph is updated; Based on the error fault data, the fault decision layer is updated and learned.

8. A wind farm fault analysis system based on multi-dimensional information association, characterized in that: The steps for implementing the method according to any one of claims 1 to 7 include: A secondary circuit topology establishing unit, wherein the secondary circuit topology establishing unit is used to read the field configuration information of the target wind farm and establish a secondary circuit topology; A network time protocol construction unit, the network time protocol construction unit is used to traverse the secondary loop topology, identify topological nodes based on data homology and data correlation, and construct a network time protocol, the network time protocol is used to perform timestamp constraints on node data return; A fault prediction module construction unit, the fault prediction module construction unit is used to construct a fault prediction module based on the secondary circuit topology and in combination with a fault knowledge graph, the fault prediction module includes a data fitting layer and a fault decision layer, wherein the fault knowledge graph and multi-dimensional information have data granularity consistency, the data fitting layer establishes a communication connection with a unified information model, and the unified information model is used to perform data de-heterogeneity; A data return constraint unit, the data return constraint unit is used to determine a transmission timestamp and perform data return constraints on the multi-dimensional source data in combination with the network time protocol; A wind farm fault prediction unit, the wind farm fault prediction unit is used to transmit the multi-dimensional source data to the fault prediction module, perform conversion and fitting of associated data and data structure, perform wind farm fault prediction, and determine fault prediction results; A fault operation and maintenance management unit is used to perform fault operation and maintenance management on the target wind farm based on the fault prediction result.

Images