Fan gearbox abnormity monitoring method and system based on multi-source data fusion

Abstract

The invention provides a fan gear box abnormity monitoring method and system based on multi-source data fusion, and the method comprises the steps: carrying out the feature extraction of the multi-mode data of a fan gear box, and obtaining a plurality of feature data of the fan gear box; performing importance evaluation on each piece of feature data, and selecting a plurality of pieces of target feature data to form a feature data set according to a sequence of importance from high to low; inputting the feature data set into a pre-constructed fan gear box anomaly identification model for anomaly prediction to obtain an anomaly level of the fan gear box; and executing the early warning strategy corresponding to the abnormality level according to the abnormality level of the fan gear box, so that the abnormality of the fan gear box is monitored by using the feature data with relatively high importance in the multi-modal data, and the efficiency and the accuracy of the abnormality monitoring result of the fan gear box are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of wind power generation equipment condition monitoring, and in particular to a wind turbine gearbox abnormality monitoring method and system based on multi-source data fusion. Background Art

[0002] Gearboxes are key components in wind turbine transmission systems, transmitting force and load, and are widely used in mechanical engineering. Gearboxes often operate in complex and harsh environments, making them prone to various anomalies that can affect wind turbine operation. Therefore, timely and accurate diagnosis of gearbox anomalies is crucial.

[0003] Typically, when monitoring gearbox abnormalities, a single type of sensor is often used to obtain data for analysis. However, this method of data acquisition is incomplete and easily affected by the environment, resulting in low accuracy in wind turbine gearbox abnormality monitoring results. Summary of the Invention

[0004] The present invention provides a method and system for monitoring abnormalities of a wind turbine gearbox based on multi-source data fusion, which are used to solve the defect of low accuracy of abnormality monitoring results of a wind turbine gearbox in the prior art.

[0005] In one aspect, the present invention provides a method for monitoring abnormalities of a wind turbine gearbox based on multi-source data fusion, which comprises: Perform feature extraction on the multimodal data of the wind turbine gearbox to obtain multiple feature data of the wind turbine gearbox; Evaluate the importance of each feature data, and select multiple target feature data in descending order of importance to form a feature data set; Inputting the feature data set into a pre-built wind turbine gearbox anomaly recognition model to perform anomaly prediction and obtain an anomaly level of the wind turbine gearbox; Execute the warning strategy corresponding to the abnormality level.

[0006] According to a wind turbine gearbox abnormality monitoring method based on multi-source data fusion provided by the present invention, before inputting the feature data set into a pre-built wind gearbox abnormality recognition model for abnormality prediction and obtaining the abnormality level of the wind gearbox, the method further includes: Determining similarities between the plurality of feature data and historical feature data of each historical case in the historical case library, and selecting a historical case with the greatest similarity, wherein the historical case with the greatest similarity includes a plurality of historical selected feature data; If the plurality of target feature data are different from the plurality of historically selected feature data, extracting N distinguishing feature data; Arbitrating the N distinguishing feature data to obtain arbitration feature data; The feature data set is updated based on the arbitration feature data.

[0007] According to a wind turbine gearbox abnormality monitoring method based on multi-source data fusion provided by the present invention, the N distinguishing feature data are arbitrated to obtain arbitration feature data, including: After removing one distinguishing feature data from the plurality of feature data each time, the remaining feature data is used to train the random forest model to obtain the performance of the random forest model at that time; After traversing all distinguishing feature data, n distinguishing feature data are selected as the arbitration feature data in order of performance from low to high, where n is less than or equal to N.

[0008] According to a wind turbine gearbox abnormality monitoring method based on multi-source data fusion provided by the present invention, the N distinguishing feature data are arbitrated to obtain arbitration feature data, including: Determining an interaction value between each distinguishing feature data and other distinguishing feature data to obtain a comprehensive interaction value of each distinguishing feature data; In descending order of comprehensive interaction values, m distinguishing feature data are selected as the arbitration feature data, where m is less than or equal to N.

[0009] According to a wind turbine gearbox abnormality monitoring method based on multi-source data fusion provided by the present invention, determining the interaction value between each distinguishing feature data and other distinguishing feature data to obtain the comprehensive interaction value of each distinguishing feature data includes: Pairing the N distinguishing feature data in pairs to obtain a plurality of distinguishing feature data pairs; For each pair of distinguishing feature data, identifying an interaction value between the first distinguishing feature data and the second distinguishing feature data in each pair of distinguishing feature data according to a preset interactivity identification rule; Based on the interaction value between the first distinguishing feature data and the second distinguishing feature data in each distinguishing feature data pair, the interaction values ​​between each distinguishing feature data and other distinguishing feature data are determined and added to obtain a comprehensive interaction value of each distinguishing feature data.

[0010] According to a wind turbine gearbox abnormality monitoring method based on multi-source data fusion provided by the present invention, for each pair of distinguishing feature data, according to a preset interactive identification rule, an interaction value between the first distinguishing feature data and the second distinguishing feature data in each pair of distinguishing feature data is identified, including: After removing the first distinguishing feature data from the plurality of feature data, training a random forest model using the remaining feature data to obtain performance of the first random forest model; After removing the second distinguishing feature data from the plurality of feature data, training a random forest model using the remaining feature data to obtain performance of a second random forest model; After simultaneously removing the first distinguishing feature data and the second distinguishing feature data from the plurality of feature data, training a random forest model using the remaining feature data to obtain performance of a third random forest model; Based on the performance of the first random forest model, the performance of the second random forest model, the performance of the third random forest model, and the performance of the fourth random forest model, an interaction value between the first distinguishing feature data and the second distinguishing feature data in each pair of distinguishing feature data is obtained; wherein, the performance of the fourth random forest model is a model obtained by training the random forest model using the multiple feature data.

[0011] According to a wind turbine gearbox abnormality monitoring method based on multi-source data fusion provided by the present invention, based on the performance of the first random forest model, the performance of the second random forest model, the performance of the third random forest model, and the performance of the fourth random forest model, an interaction value between the first distinguishing feature data and the second distinguishing feature data in each distinguishing feature data pair is obtained, including: Calculating a sum of the performance of the first random forest model and the performance of the second random forest model; Calculating a difference between the sum value and the performance of the third random forest model and the performance of the fourth random forest model; According to the difference, an interaction value between the first distinguishing feature data and the second distinguishing feature data in each distinguishing feature data pair is obtained.

[0012] On the other hand, the present invention also provides a wind turbine gearbox abnormality monitoring system based on multi-source data fusion, which includes: An extraction module is used to extract features from the multimodal data of the wind turbine gearbox to obtain multiple feature data of the wind turbine gearbox; The screening module is used to evaluate the importance of each feature data and select multiple target feature data in descending order of importance to form a feature data set; an identification module, configured to input the feature data set into a pre-built wind turbine gearbox anomaly identification model to perform anomaly prediction and obtain an anomaly level of the wind turbine gearbox; The execution module is used to execute the warning strategy corresponding to the abnormal level.

[0013] On the other hand, the present invention also provides an electronic device, which includes a memory, a processor, and a computer program stored in the memory and runnable on the processor. When the processor executes the program, it implements any of the above-mentioned methods for monitoring abnormalities of wind gearboxes based on multi-source data fusion.

[0014] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the above-described methods for monitoring abnormalities of a wind turbine gearbox based on multi-source data fusion.

[0015] On the other hand, the present invention further provides a computer program product, comprising a computer program, which, when executed by a processor, implements any of the above-described methods for monitoring abnormalities of a wind turbine gearbox based on multi-source data fusion.

[0016] The wind turbine gearbox abnormality monitoring method and system based on multi-source data fusion provided by the present invention perform feature extraction on the multi-modal data of the wind turbine gearbox to obtain multiple feature data of the wind turbine gearbox; perform importance evaluation on each feature data, and select multiple target feature data in descending order of importance to form a feature data set; input the feature data set into a pre-built wind turbine gearbox abnormality recognition model to perform abnormality prediction to obtain the abnormality level of the wind turbine gearbox; execute the early warning strategy corresponding to the abnormality level, realize the use of feature data with higher importance in the multi-modal data to monitor the abnormality of the wind turbine gearbox, and improve the efficiency and accuracy of the abnormality monitoring results of the wind turbine gearbox. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the present invention or the prior art, a brief introduction is given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0018] Figure 1 1 is a flow chart of a method for monitoring abnormalities of a wind turbine gearbox based on multi-source data fusion provided by an embodiment of the present invention; Figure 2 is a schematic diagram of a process for constructing a feature dataset provided by an embodiment of the present invention; Figure 3 1 is a schematic structural diagram of a wind turbine gearbox abnormality monitoring system based on multi-source data fusion provided by an embodiment of the present invention; Figure 4 It is a structural diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0019] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0020] Figure 1 It is a flow chart of a method for monitoring abnormalities of a wind turbine gearbox based on multi-source data fusion provided by an embodiment of the present invention.

[0021] like Figure 1 As shown, the execution subject of the wind turbine gearbox abnormality monitoring method based on multi-source data fusion provided by the embodiment of the present invention can be an electronic device, and the method mainly includes the following steps: 101. Perform feature extraction on multimodal data of the wind turbine gearbox to obtain multiple feature data of the wind turbine gearbox; In a specific implementation, various sensors and other data acquisition devices can be used to acquire multimodal data of the wind turbine gearbox. For example, this multimodal data may include vibration parameters, temperature parameters, and oil parameters of the wind turbine gearbox. Alignment and fusion operations are then performed on the multimodal data to extract features, thereby obtaining multiple feature data of the wind turbine gearbox.

[0022] 102. Evaluate the importance of each feature data, and select multiple target feature data in descending order of importance to form a feature data set; In a specific implementation process, after feature extraction is performed on the obtained multimodal data, high-dimensional feature data is usually obtained. Although high-dimensional feature data can more accurately detect anomalies, the high dimensionality of the dataset increases the complexity of calculation and analysis. Therefore, in order to balance accuracy and computational complexity, each feature data can be evaluated for importance, and multiple target feature data can be selected in descending order of importance to form a feature dataset. The specific number of target feature data selected can be set according to actual needs. The method of evaluating the importance of each feature data can be implemented using the random forest feature selection method.

[0023] 103. Input the feature data set into a pre-built wind turbine gearbox anomaly recognition model to perform anomaly prediction and obtain an anomaly level of the wind turbine gearbox; In a specific implementation process, a fan gearbox anomaly recognition model can be pre-trained based on historical data. In this way, after the importance of each feature data is evaluated and multiple target feature data are screened out to form a feature data set, the feature data set can be input into the pre-built fan gearbox anomaly recognition model for anomaly prediction to obtain the anomaly level of the fan gearbox.

[0024] For example, multiple target feature data in the feature data set may include temperature change rate, vibration kurtosis, oil contamination, etc. The fan gearbox abnormality recognition model can identify the abnormality of the fan gearbox based on the temperature change rate, vibration kurtosis, oil contamination, etc., and classify them into levels to obtain the abnormality level of the fan gearbox.

[0025] For example, if the temperature change rate ΔT is greater than 0.1°C / min for three consecutive windows, the vibration kurtosis is greater than 4.0, and the oil particles are greater than 50 / 100ml, the fan gearbox abnormality level can be determined to be level 1. If the temperature change rate is greater than 0.2°C / min for two consecutive windows, the vibration kurtosis is greater than 4.5, and the oil particles are greater than 100 / 100ml, the fan gearbox abnormality level can be determined to be level 2. If the temperature change rate is greater than 0.3°C / min in the current window, the vibration kurtosis is greater than 5.0, and the oil particles are greater than 200 / 100ml, the fan gearbox abnormality level can be determined to be level 3.

[0026] It should be noted that, in this embodiment, a new data buffer pool can be added (retaining data from the last 30 days) to incrementally learn the fan gearbox anomaly recognition model, adjust the parameters of the fan gearbox anomaly recognition model, and the decision threshold for abnormality level classification in the fan gearbox anomaly recognition model can be adaptively adjusted. For example, it can dynamically change according to seasonal operating conditions, such as focusing on the oil viscosity index in winter, and the weight of the index and the decision threshold corresponding to the index can be dynamically adjusted.

[0027] 104. Execute the warning strategy corresponding to the abnormality level.

[0028] In a specific implementation, a relationship between different abnormality levels and early warning strategies can be pre-set. Thus, after obtaining the abnormality level of the wind gearbox, the early warning strategy corresponding to the abnormality level of the wind gearbox can be obtained through this relationship. For example, if the abnormality level of the wind gearbox is level one, a first prompt message can be sent to strengthen monitoring. If the abnormality level of the wind gearbox is level two, a second prompt message can be sent to prepare for shutdown and inspection. If the abnormality level of the wind gearbox is level three, a third prompt message can be sent and the wind gearbox can be shut down immediately.

[0029] The wind gearbox abnormality monitoring method based on multi-source data fusion of this embodiment performs feature extraction on the multimodal data of the wind gearbox to obtain multiple feature data of the wind gearbox; performs importance evaluation on each feature data, and selects multiple target feature data in descending order of importance to form a feature data set; inputs the feature data set into a pre-built wind gearbox abnormality recognition model to perform abnormality prediction to obtain the abnormality level of the wind gearbox; executes the early warning strategy corresponding to the abnormality level, realizes the use of feature data with higher importance in the multimodal data to monitor the abnormality of the wind gearbox, and improves the efficiency and accuracy of the abnormality monitoring results of the wind gearbox.

[0030] In a specific implementation process, in order to more accurately screen the feature data, the present invention also provides the following technical solutions: Figure 2 FIG. 1 is a flow chart of constructing a feature data set according to an embodiment of the present invention. Figure 2 As shown, the method is implemented before step 103 of "inputting the feature data set into a pre-built wind turbine gearbox anomaly recognition model to perform anomaly prediction and obtain an anomaly level of the wind turbine gearbox", which may include the following steps: 201. Determine the similarity between multiple feature data of the wind turbine gearbox and historical feature data of each historical case in the historical case library, and select the historical case with the maximum similarity; In a specific implementation, a similarity algorithm such as cosine similarity can be used to calculate the similarity between multiple feature data of the wind turbine gearbox and the historical feature data of each historical case, and the historical case with the greatest similarity is selected to obtain a historical operating condition similar to the current operating condition of the wind turbine gearbox. The historical case with the greatest similarity includes multiple historical selected feature data.

[0031] 202. If the plurality of target feature data are different from the plurality of historically selected feature data, extract N distinguishing feature data; In a specific implementation process, after preliminarily selecting multiple target feature data using the random forest feature selection method, semantic recognition, comparison and other operations can be performed on the multiple target feature data and multiple historically selected feature data to detect whether the multiple target feature data are the same as the multiple historically selected feature data. If the two are different, N distinguishing feature data can be extracted.

[0032] 203. Arbitrate the N distinguishing feature data to obtain arbitration feature data; After obtaining a plurality of target characteristic data, the N distinguishing characteristic data may be arbitrated to obtain arbitration characteristic data, that is, at least one distinguishing characteristic data is selected from the N distinguishing characteristic data.

[0033] In a specific implementation process, this step can be implemented as follows: The first method is: after removing one distinguishing feature data from the multiple feature data each time, the remaining feature data is used to train the random forest model to obtain the performance of the random forest model at that time; after traversing all the distinguishing feature data, n distinguishing feature data are selected in order of performance from low to high as the arbitration feature data, where n is less than or equal to N.

[0034] That is to say, for each distinguishing feature data, it can be removed and the remaining feature data can be used to train the random forest model, and the performance of the random forest model trained at that time can be verified to obtain the performance of the random forest model when a certain distinguishing feature data is not included. If the performance of the random forest model at that time is low, it means that the removed distinguishing feature data has a greater impact on the model and needs to be retained. Conversely, if the performance of the random forest model at that time is high, it means that the removed distinguishing feature data has a smaller impact on the model and can be not retained. Therefore, n distinguishing feature data can be selected in order of performance from low to high as the arbitration feature data, and n can be set according to actual needs.

[0035] The second type: Although the performance of the random forest model trained at that time is higher after the individual distinguishing feature data is removed, the importance of the distinguishing feature data will be determined to be lower. However, in actual applications, when the distinguishing feature data is combined with other distinguishing feature data, it will have a greater impact on the performance of the random forest model. Therefore, in order to more accurately screen the distinguishing feature data, this embodiment can determine the interaction value between each distinguishing feature data and other distinguishing feature data to obtain the comprehensive interaction value of each distinguishing feature data; and select m distinguishing feature data as the arbitration feature data in descending order of the comprehensive interaction value, where m is less than or equal to N.

[0036] That is to say, if the interaction value between a distinguishing feature data and other distinguishing feature data is larger, its importance is higher, and it needs to be retained; otherwise, if the interaction value between a distinguishing feature data and other distinguishing feature data is larger, its importance is lower, and it can be not retained.

[0037] In a specific implementation process, the second method can be implemented as follows: a1. Pairing the N distinguishing feature data in pairs to obtain multiple distinguishing feature data pairs; For each pair of distinguishing feature data, the two distinguishing feature data included therein can be recorded as the first distinguishing feature data and the second distinguishing feature data.

[0038] b1. for each pair of distinguishing feature data, identifying the interaction value between the first distinguishing feature data and the second distinguishing feature data in each pair of distinguishing feature data according to a preset interactivity identification rule; In a specific implementation process, this step can be implemented as follows: (1) after removing the first distinguishing feature data from the plurality of feature data, training a random forest model using the remaining feature data to obtain the performance of the first random forest model; (2) after removing the second distinguishing feature data from the plurality of feature data, training a random forest model using the remaining feature data to obtain the performance of a second random forest model; (3) after simultaneously removing the first distinguishing feature data and the second distinguishing feature data from the plurality of feature data, training a random forest model using the remaining feature data to obtain the performance of a third random forest model; (4) obtaining an interaction value between the first distinguishing feature data and the second distinguishing feature data in each distinguishing feature data pair based on the performance of the first random forest model, the performance of the second random forest model, the performance of the third random forest model, and the performance of the fourth random forest model; In a specific implementation process, the performance of the fourth random forest model is a model obtained by training the random forest model using the multiple feature data, that is, a model obtained through training without removing any distinguishing feature data.

[0039] In a specific implementation process, the sum of the performance of the first random forest model and the performance of the second random forest model can be calculated; the difference between the sum and the performance of the third random forest model and the performance of the fourth random forest model can be calculated; based on the difference, the interaction value between the first distinguishing feature data and the second distinguishing feature data in each distinguishing feature data pair can be obtained, and the corresponding calculation formula is as shown in Formula (1): (1) in, represents the interaction value between the first distinguishing feature data and the second distinguishing feature data in each distinguishing feature data pair, represents the performance of the first random forest model after removing the first distinguishing feature data, represents the performance of the second random forest model after removing the second distinguishing feature data, represents the performance of the third random forest model after removing the first distinguishing feature data and the second distinguishing feature data, Represents the performance of the fourth random forest model based on all feature data. Indicates the contribution of the first distinguishing feature data and the second distinguishing feature data to the performance of the random forest model, Indicates the contribution of the first distinguishing feature data to the performance of the random forest model, Indicates the contribution of the second distinguishing feature data to the performance of the random forest model.

[0040] After simplifying the above formula, we can get formula (2): (2) in, It means that there is an interactive characteristic between the first distinguishing feature data and the second distinguishing feature data, and the larger the value, the stronger the interactive characteristic. The interactive value between the first distinguishing feature data and the second distinguishing feature data can be calculated according to Actual value, Indicates that there is no interaction between the first distinguishing feature data and the second distinguishing feature data. In this case, the interaction value between the first distinguishing feature data and the second distinguishing feature data can be The value is adjusted to 0. In this way, when the comprehensive interaction value is calculated later, any non-existent interactions are calculated as 0, which will not affect the calculation of the comprehensive interaction value.

[0041] c1. Based on the interaction value between the first distinguishing feature data and the second distinguishing feature data in each distinguishing feature data pair, determine the interaction value between each distinguishing feature data and other distinguishing feature data and add them together to obtain a comprehensive interaction value of each distinguishing feature data.

[0042] In a specific implementation process, when determining the interaction value between the first distinguishing feature data and the second distinguishing feature data in each pair of distinguishing feature data, the interaction value between any distinguishing feature data and other distinguishing feature data can also be determined. Then, all the interaction values ​​between any distinguishing feature data and other distinguishing feature data are added together to obtain the comprehensive interaction value of any distinguishing feature data. After traversing all the distinguishing feature data, the comprehensive interaction value of each distinguishing feature data is obtained.

[0043] d1. Select m distinguishing feature data as the arbitration feature data in descending order of comprehensive interaction values.

[0044] In a specific implementation, the higher the comprehensive interaction value of a particular distinguishing feature data, the more important the distinguishing feature data is. Therefore, m distinguishing feature data may be selected as the arbitration feature data in descending order of comprehensive interaction value. Here, m is less than or equal to N and can be set based on actual needs.

[0045] 204. Update the feature data set based on the arbitration feature data.

[0046] In a specific implementation process, the same feature data from multiple target feature data and multiple historically selected feature data may be extracted and merged with the obtained arbitration feature data to update the feature data set.

[0047] The wind turbine gearbox abnormality monitoring method based on multi-source data fusion of this embodiment, after selecting multiple target feature data using the random forest feature selection method to form a feature data set, can adjust the feature data by combining multiple historical selected feature data from historical cases corresponding to working conditions similar to the current working conditions of the wind turbine gearbox. In this way, the random forest feature selection method and historical cases can be combined to more accurately obtain the required feature data to form a feature data set, and then the wind turbine gearbox abnormality recognition model can be used to perform abnormality prediction to further improve the recognition accuracy.

[0048] Based on the same general inventive concept, the present invention also protects a wind gearbox abnormality monitoring system based on multi-source data fusion. The wind gearbox abnormality monitoring system based on multi-source data fusion provided by the present invention is described below. The wind gearbox abnormality monitoring system based on multi-source data fusion described below and the wind gearbox abnormality monitoring method based on multi-source data fusion described above can be referenced to each other.

[0049] Figure 3 FIG is a structural diagram of a wind turbine gearbox abnormality monitoring system based on multi-source data fusion provided by an embodiment of the present invention, such as Figure 3 As shown, the wind turbine gearbox abnormality monitoring system based on multi-source data fusion of this embodiment includes an extraction module 31 , a screening module 32 , an identification module 33 and an execution module 34 .

[0050] The extraction module 31 is used to extract features from the multimodal data of the wind turbine gearbox to obtain multiple feature data of the wind turbine gearbox; The screening module 32 is used to evaluate the importance of each feature data and select multiple target feature data in descending order of importance to form a feature data set; The recognition module 33 is used to input the characteristic data set into a pre-built wind turbine gearbox abnormality recognition model to perform abnormality prediction and obtain the abnormality level of the wind turbine gearbox; The execution module 34 is used to execute the warning strategy corresponding to the abnormality level.

[0051] Figure 4is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. The wind turbine gearbox anomaly monitoring system based on multi-source data fusion may include: a processor 410, a communications interface 420, a memory 430, and a communication bus 440. The processor 410, the communications interface 420, and the memory 430 communicate with each other via the communication bus 440. The processor 410 may call the logic instructions in the memory 430 to execute the wind turbine gearbox anomaly monitoring method based on multi-source data fusion.

[0052] Furthermore, the logic instructions in the aforementioned memory 430 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product, stored in a storage medium, includes instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0053] On the other hand, the present invention also provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the wind gearbox abnormality monitoring method based on multi-source data fusion provided by the above methods.

[0054] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the wind turbine gearbox abnormality monitoring method based on multi-source data fusion provided by the above methods.

[0055] It should be noted that the relevant information that may be involved in the various embodiments of this application are all strictly in accordance with the requirements of laws and regulations, follow the principles of legality, legitimacy and necessity, and are based on the reasonable purposes of business scenarios to process information that users actively provide during the use of products / services or generated due to the use of products / services, as well as information obtained with user authorization.

[0056] The user personal information processed by this application will vary depending on the specific product / service scenario and must be based on the specific scenario in which the user uses the product / service. This may involve the user's account information, device information, or other related information. This application will treat the relevant information and its processing with a high degree of diligence.

[0057] This application attaches great importance to the security of relevant information and has taken reasonable and feasible security protection measures that comply with industry standards to protect relevant information and prevent unauthorized access, public disclosure, use, modification, damage or loss of relevant information.

[0058] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0059] Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented using software plus a necessary general-purpose hardware platform, or of course, hardware. Based on this understanding, the essence of the above technical solution, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, or an optical disk, and includes a number of instructions for causing a computer device (such as a personal computer, server, or network device) to execute the methods described in each embodiment or certain portions of the embodiments.

[0060] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A wind turbine gearbox abnormality monitoring method based on multi-source data fusion, characterized in that: include: Perform feature extraction on the multimodal data of the wind turbine gearbox to obtain multiple feature data of the wind turbine gearbox; Evaluate the importance of each feature data, and select multiple target feature data in descending order of importance to form a feature data set; Inputting the feature data set into a pre-built wind turbine gearbox anomaly recognition model to perform anomaly prediction and obtain an anomaly level of the wind turbine gearbox; Execute the warning strategy corresponding to the abnormality level.

2. The wind turbine gearbox abnormality monitoring method based on multi-source data fusion according to claim 1 is characterized in that: Before inputting the feature data set into a pre-built wind turbine gearbox anomaly recognition model for anomaly prediction and obtaining the anomaly level of the wind turbine gearbox, the method further includes: Determining similarities between the plurality of feature data and historical feature data of each historical case in the historical case library, and selecting a historical case with the greatest similarity, wherein the historical case with the greatest similarity includes a plurality of historical selected feature data; If the plurality of target feature data are different from the plurality of historically selected feature data, extracting N distinguishing feature data; Arbitrating the N distinguishing feature data to obtain arbitration feature data; The feature data set is updated based on the arbitration feature data.

3. The wind turbine gearbox abnormality monitoring method based on multi-source data fusion according to claim 2 is characterized in that: Arbitrating the N distinguishing feature data to obtain arbitration feature data includes: After removing one distinguishing feature data from the plurality of feature data each time, the remaining feature data is used to train the random forest model to obtain the performance of the random forest model at that time; After traversing all distinguishing feature data, n distinguishing feature data are selected as the arbitration feature data in order of performance from low to high, where n is less than or equal to N.

4. The wind turbine gearbox abnormality monitoring method based on multi-source data fusion according to claim 2 is characterized in that: Arbitrating the N distinguishing feature data to obtain arbitration feature data includes: Determining an interaction value between each distinguishing feature data and other distinguishing feature data to obtain a comprehensive interaction value of each distinguishing feature data; In descending order of comprehensive interaction values, m distinguishing feature data are selected as the arbitration feature data, where m is less than or equal to N.

5. The wind turbine gearbox abnormality monitoring method based on multi-source data fusion according to claim 4 is characterized in that: Determining an interaction value between each distinguishing feature data and other distinguishing feature data to obtain a comprehensive interaction value of each distinguishing feature data includes: Pairing the N distinguishing feature data in pairs to obtain a plurality of distinguishing feature data pairs; For each pair of distinguishing feature data, identifying an interaction value between the first distinguishing feature data and the second distinguishing feature data in each pair of distinguishing feature data according to a preset interactivity identification rule; Based on the interaction value between the first distinguishing feature data and the second distinguishing feature data in each distinguishing feature data pair, the interaction values ​​between each distinguishing feature data and other distinguishing feature data are determined and added to obtain a comprehensive interaction value of each distinguishing feature data.

6. The wind turbine gearbox abnormality monitoring method based on multi-source data fusion according to claim 5 is characterized in that: For each pair of distinguishing feature data, identifying the interaction value between the first distinguishing feature data and the second distinguishing feature data in each pair of distinguishing feature data according to a preset interactivity identification rule, including: After removing the first distinguishing feature data from the plurality of feature data, training a random forest model using the remaining feature data to obtain performance of the first random forest model; After removing the second distinguishing feature data from the plurality of feature data, training a random forest model using the remaining feature data to obtain performance of a second random forest model; After simultaneously removing the first distinguishing feature data and the second distinguishing feature data from the plurality of feature data, training a random forest model using the remaining feature data to obtain performance of a third random forest model; Based on the performance of the first random forest model, the performance of the second random forest model, the performance of the third random forest model, and the performance of the fourth random forest model, an interaction value between the first distinguishing feature data and the second distinguishing feature data in each pair of distinguishing feature data is obtained; wherein, the performance of the fourth random forest model is a model obtained by training the random forest model using the multiple feature data.

7. The method for monitoring abnormality of a wind turbine gearbox based on multi-source data fusion according to claim 6, characterized in that: Obtaining an interaction value between the first distinguishing feature data and the second distinguishing feature data in each distinguishing feature data pair based on the performance of the first random forest model, the performance of the second random forest model, the performance of the third random forest model, and the performance of the fourth random forest model, comprising: Calculating a sum of the performance of the first random forest model and the performance of the second random forest model; Calculating a difference between the sum value and the performance of the third random forest model and the performance of the fourth random forest model; According to the difference, an interaction value between the first distinguishing feature data and the second distinguishing feature data in each distinguishing feature data pair is obtained.

8. A wind turbine gearbox abnormality monitoring system based on multi-source data fusion, characterized in that: include: An extraction module is used to extract features from the multimodal data of the wind turbine gearbox to obtain multiple feature data of the wind turbine gearbox; The screening module is used to evaluate the importance of each feature data and select multiple target feature data in descending order of importance to form a feature data set; an identification module, configured to input the feature data set into a pre-built wind turbine gearbox anomaly identification model to perform anomaly prediction and obtain an anomaly level of the wind turbine gearbox; The execution module is used to execute the warning strategy corresponding to the abnormal level.

9. An electronic device, characterized in that: The invention comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method for monitoring abnormalities of a wind turbine gearbox based on multi-source data fusion as claimed in any one of claims 1 to 7 is implemented.

10. A non-transitory computer-readable storage medium, characterized in that A computer program is stored thereon, and when the computer program is executed by a processor, the wind turbine gearbox abnormality monitoring method based on multi-source data fusion as described in any one of claims 1 to 7 is implemented.

Images