Load testing device of wind turbine generator and testing method thereof

Abstract

The invention provides a wind turbine generator load test device and a test method thereof, and the device comprises a load test correlation information obtaining subsystem which is used for obtaining the load test correlation information of a target wind turbine generator; the experience test data calling subsystem is used for calling experience test data of the required test load type set; the load test model training subsystem is used for training a load test model according to the experience test data; and the test subsystem is used for determining a load test scheme and testing according to the load test associated information and the load test model. According to the load testing device of the wind turbine generator and the testing method of the load testing device, a required testing load type set is determined, and the testing load type and empirical data consistent with the required testing load type are selected; a load test model is trained by using empirical data, a test scheme formulating basis extracted according to load test associated information is input into the load test model to obtain a load test scheme, and model training is more accurate and suitable.

Description

Technical Field

[0001] The present invention relates to the technical field of load testing, and particularly relates to a load testing device for a wind turbine and a testing method thereof. Background Art

[0002] All kinds of forces and torques faced by a wind turbine during operation are collectively referred to as loads. These loads mainly originate from the action of wind force and mechanical actions such as the rotation of the wind turbine. The safety and reliability of a wind turbine largely depend on the accurate assessment of the loads it bears. The purpose of load testing is to verify the safety in the design and operation of the wind turbine, and ensure that its structure remains intact under variable environmental and operating conditions. Through such testing, engineers can evaluate various load conditions that the wind turbine may encounter during actual operation, including aerodynamic loads caused by wind force, inertial loads generated by mechanical motion, and additional loads brought by extreme weather events. These test results are crucial for optimizing the design of the wind turbine, improving its durability and performance, and also provide valuable data support for the maintenance of the wind turbine and the prevention of potential failures. Accurate load testing can prevent structural damage, extend the service life of the wind turbine, and ensure its efficient energy conversion during long-term operation.

[0003] The invention patent with the application number: CN202110701216.0 discloses a load testing device and method for an offshore wind turbine. The device includes a wind measurement device, a wave measurement device, a current measurement device, a strain gauge sensor, a data collector, an industrial control computer, and a wireless communication module; the wind measurement device, the wave measurement device, the current measurement device, and the strain gauge sensor are respectively connected to a data collector, and each data collector is connected to the industrial control computer; the industrial control computer is connected to the wireless communication module, and the wireless communication module is used to send the data of the industrial control computer through wireless transmission. In addition to wind, considering waves and currents as factors affecting the loads of the wind turbine can more accurately model the loads received by the wind turbine, accurately evaluate the influence of different factors on the loads of the wind turbine, and thus accurately evaluate the load distribution of the wind turbine during the design stage and optimize the structural design of the wind turbine.

[0004] The above-mentioned prior art is for load modeling in the design stage. However, in addition to the design stage, load testing is also required during the testing and regular maintenance stages after the installation of the wind turbine. The factors affecting the loads of the wind turbine are different in different stages. The prior art models the loads received by the wind turbine for a single stage, and when the subsequent application environment changes, the test accuracy is relatively low, and the test applicable situations are not comprehensive enough.

[0005] In view of this, there is an urgent need for a load testing device for a wind turbine and a testing method thereof to at least solve the above deficiencies. Summary of the Invention

[0006] One of the objectives of the present invention is to provide a load testing device for a wind turbine, determine the set of required test load types, and select the empirical data in the empirical test data where the test load type is the same as the required test load type in the set of required test load types; use the empirical data to train the load testing model corresponding to the set of required test load types, input the basis information for formulating the test plan extracted according to the load testing correlation information into the load testing model to obtain the load testing plan. The empirical data takes into account the influence between different required test load types, and training according to the empirical data improves the training accuracy and suitability of the model.

[0007] A load testing device for a wind turbine provided by an embodiment of the present invention includes:

[0008] A load testing correlation information acquisition subsystem for acquiring the load testing correlation information of the target wind turbine;

[0009] An empirical test data retrieval subsystem for acquiring the set of required test load types and retrieving the empirical test data of the set of required test load types;

[0010] A load testing model training subsystem for training the load testing model corresponding to the set of required test load types according to the empirical test data;

[0011] A testing subsystem for determining the load testing plan of the target wind turbine according to the load testing correlation information and the load testing model and performing corresponding tests.

[0012] Preferably, the load testing correlation information acquisition subsystem acquires the load testing correlation information of the target wind turbine, including:

[0013] Acquire the design parameters, first historical test data, operating environment, and expected load conditions of the target wind turbine;

[0014] Take the design parameters, first historical test data, operating environment, and expected load conditions of the target wind turbine together as the load testing correlation information.

[0015] Preferably, the empirical test data retrieval subsystem acquires the set of required test load types and retrieves the empirical test data of the set of required test load types, including:

[0016] According to the load testing correlation information, determine the second historical test data of the target historical wind turbine with similar information;

[0017] According to the second historical test data, obtain the set of historical test load types of the target historical wind turbine and use it as the set of required test load types;

[0018] Acquire the overload events of the target historical wind turbine;

[0019] Determine the overload type according to the overload event;

[0020] Determine the target load type in the required test load type set excluding the overload type;

[0021] Obtain the first empirical test sub-data of the overload type and the second empirical test sub-data of the target load type according to the second historical test data;

[0022] Use the first empirical test sub-data and the second empirical test sub-data together as the empirical test data.

[0023] Preferably, the empirical test data retrieval subsystem obtains the first empirical test sub-data of the overload type and the second empirical test sub-data of the target load type according to the second historical test data, including:

[0024] Obtain the third historical test data that meets the negative sample extraction conditions before the occurrence of the overload event of the target historical wind turbine;

[0025] Use the second historical test data in the second historical test data except the third historical test data as the fourth historical test data;

[0026] Use the third historical test data corresponding to the overload type as the negative sample test data;

[0027] Use the fourth historical test data corresponding to the overload type as the positive sample test data;

[0028] Mark the non-standard test data features in the negative sample test data and the standard test data features in the positive sample test data;

[0029] Use the marked negative sample test data and positive sample test data as the first empirical test sub-data;

[0030] Use the second historical test data corresponding to the target load type as the second empirical test sub-data;

[0031] Among them, the negative sample extraction conditions include:

[0032] The test time of the third historical test data is before the occurrence time of the overload event and the time interval between the two is less than the preset time interval threshold;

[0033] There is a consistency between the test load type of the third historical test data and the overload type corresponding to the overload event that occurs continuously after the test of the third historical test data.

[0034] Preferably, the empirical test data retrieval subsystem marks the non-standard test data features in the negative sample test data and the standard test data features in the positive sample test data, including:

[0035] Send the negative sample test data to the expert node to obtain the characteristics of non-standard test data annotated by the expert;

[0036] According to the first feature type of the non-standard test data characteristics, mark the standard test data characteristics for the positive sample test data. At the same time, use the remaining test data characteristics of the non-standard test data characteristics in the negative sample test data that have not been annotated as the standard test data characteristics.

[0037] Preferably, the empirical test data retrieval subsystem marks the non-standard test data characteristics in the negative sample test data and the standard test data characteristics in the positive sample test data, and further includes:

[0038] Calculate the data feature similarity between the negative sample test data and the positive sample test data of different second feature types;

[0039] If the data feature similarity is greater than or equal to the preset similarity threshold, mark the corresponding negative sample test data characteristics and positive sample test data characteristics of the second feature type as the standard test data characteristics;

[0040] If the data feature similarity is less than the preset similarity threshold, use the corresponding negative sample test data characteristics of the second feature type as the non-standard test data characteristics and the corresponding positive sample test data characteristics of the second feature type as the standard test data characteristics.

[0041] Preferably, the load test model training subsystem trains the load test model corresponding to the required test load type set according to the empirical test data, including:

[0042] Extract the empirical basis and empirical solution according to the empirical test data;

[0043] Use the empirical basis as the input of the CNN model and the empirical solution as the output of the CNN model to train the load test model.

[0044] Preferably, the test subsystem determines the load test plan for the target wind turbine according to the load test association information and the load test model and conducts the corresponding test, including:

[0045] Extract the basis vector in the load test association information according to the template of the empirical basis extraction;

[0046] Input the basis vector into the load test model to obtain the load test plan output by the model.

[0047] An embodiment of the present invention provides a load test device for a wind turbine, further including:

[0048] Anomaly analysis subsystem, used to determine the abnormal wind turbines with abnormal load tests during the maintenance stage of the target wind turbine, and conduct anomaly attribution according to the abnormal load types of the abnormal wind turbines;

[0049] Among them, the anomaly analysis subsystem determines the abnormal wind turbines with abnormal load tests during the maintenance stage of the target wind turbine, and conducts anomaly attribution according to the abnormal load types of the abnormal wind turbines, including:

[0050] Obtain the wind turbine information of the abnormal wind turbines with the same abnormal load type;

[0051] Conduct principal component analysis according to the wind turbine information to obtain the principal component analysis results associated with the abnormal load type;

[0052] Obtain the anomaly analysis model corresponding to the abnormal load type, and determine the information type of the input feature extraction source for the anomaly analysis model;

[0053] Obtain the association relationship between the principal component analysis results and the information type of the input feature extraction source;

[0054] According to the association relationship, connect to the input feature extraction source information link to obtain the input feature extraction source information and extract the input features, and input the input features into the anomaly analysis model to obtain the anomaly attribution result.

[0055] A load test method for a wind turbine provided by an embodiment of the present invention includes:

[0056] Step 1: Obtain the load test association information of the target wind turbine;

[0057] Step 2: Obtain the set of required test load types, and retrieve the empirical test data of the set of required test load types;

[0058] Step 3: Train the load test model corresponding to the set of required test load types according to the empirical test data;

[0059] Step 4: Determine the load test plan for the target wind turbine according to the load test association information and the load test model, and conduct corresponding tests.

[0060] The beneficial effects of the present invention are:

[0061] The present invention determines the set of required test load types, and selects the empirical data in the empirical test data whose test load types are the same as the required test load types in the set of required test load types; uses the empirical data to train the load test model corresponding to the set of required test load types, inputs the test plan formulation basis information extracted according to the load test association information into the load test model to obtain the load test plan. The empirical data takes into account the influence between different required test load types, and training according to the empirical data improves the training accuracy and suitability of the model.

[0062] Other features and advantages of the present invention will be described in the following specification, and in part will become apparent from the specification, or will be understood by implementing the present invention. The objectives and other advantages of the present invention can be achieved and obtained by the structure specifically pointed out in this application document.

[0063] The technical solution of the present invention will be further described in detail below through the accompanying drawings and embodiments. Description of the Drawings

[0064] The drawings are used to provide a further understanding of the present invention, and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation to the present invention. In the drawings:

[0065] Figure 1 is a schematic diagram of a load test device for a wind turbine in an embodiment of the present invention;

[0066] Figure 2 is a schematic diagram of a load test method for a wind turbine in an embodiment of the present invention. Detailed Embodiments

[0067] The following describes the preferred embodiments of the present invention with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention and are not used to limit the present invention.

[0068] An embodiment of the present invention provides a load test device for a wind turbine, as Figure 1 shown, including:

[0069] A load test associated information acquisition subsystem 1, configured to acquire load test associated information of a target wind turbine; wherein, the target wind turbine is: a wind turbine that needs to perform a load test, such as: a wind turbine in the design stage, a wind turbine in the maintenance stage; the load test associated information is: design parameters of the target wind turbine, historical load data, environmental conditions (such as: wind speed, temperature, etc.), operating conditions, etc.;

[0070] An empirical test data retrieval subsystem 2, configured to acquire a set of required test load types and retrieve empirical test data of the set of required test load types; wherein, the set of required test load types is determined according to the load test associated information (such as: test purpose, design and operating conditions of the target wind turbine), and the required test load types include but are not limited to: aerodynamic load, dynamic load, and maneuvering load, etc.; the empirical test data is: artificial load test data in which the test load type is exactly the same as the required test load type in the set of required test load types, including: equipment used, method steps used, etc.;

[0071] The load test model training subsystem 3 is used to train a load test model corresponding to the required test load type set according to empirical test data; wherein, the load test model is an AI model that intelligently determines a load test plan based on the load test correlation information of the wind turbine;

[0072] The test subsystem 4 is used to determine the load test plan for the target wind turbine according to the load test correlation information and the load test model and conduct corresponding tests; wherein, the load test plan includes: determination of test objectives and conditions, design of test equipment and procedures, execution of actual load tests, establishment of simulation models, acquisition of simulated load data, comparative analysis of test results and simulation results, compilation of test reports and simulation verification reports, and post-test evaluation and optimization;

[0073] Among them, the load test correlation information acquisition subsystem acquires the load test correlation information of the target wind turbine, including:

[0074] Acquire the design parameters, first historical test data, operating environment, and expected load conditions of the target wind turbine; the design parameters include but are not limited to the size of the wind turbine (such as blade length, tower height), material properties, structural layout, power rating, etc.; the first historical test data is the data obtained from previous tests on the target wind turbine; the operating environment is the natural environment and working conditions where the wind turbine is located, including geographical location, climate conditions (such as wind speed, temperature, humidity), topography, etc.; the expected load conditions are various load situations that are expected to be encountered during the operation of the wind turbine according to its design and operating environment, including wind loads, gravity loads, thermal loads caused by temperature changes, possible extreme weather events (such as typhoons, lightning strikes), etc.;

[0075] The design parameters, first historical test data, operating environment, and expected load conditions of the target wind turbine are jointly used as the load test correlation information;

[0076] Among them, the load test model training subsystem trains a load test model corresponding to the required test load type set according to empirical test data, including:

[0077] Extract empirical bases and empirical plans according to empirical test data; wherein, the empirical basis is the characteristic representation in the empirical test data that characterizes the basis for formulating the corresponding historical test plan (empirical plan), such as: the geographical location, climate conditions (wind speed, temperature, humidity), and topography where the target wind turbine is located; the empirical plan is the load test plan corresponding to the empirical basis, such as: if the empirical basis is the wind speed at the geographical location of the target wind turbine, the corresponding empirical plan is the aerodynamic load test plan corresponding to this wind speed;

[0078] Using empirical evidence as the input of the CNN model and empirical solutions as the output of the CNN model, train the load test model;

[0079] Among them, the test subsystem determines the load test plan for the target wind turbine according to the load test association information and the load test model and conducts corresponding tests, including:

[0080] Extract the basis vector in the load test association information according to the empirical evidence extraction template; among them, the empirical evidence extraction template is: a template for the load test association information to extract the basis for formulating the test plan in the load test association information, and the basis vector is: a vector obtained by vectorizing the empirical evidence extracted from the load test association information, and the above vector is the quantization result of the basis for formulating the test plan for the target wind turbine.

[0081] Input the basis vector into the load test model to obtain the load test plan output by the model.

[0082] The working principle and beneficial effects of the above technical solution are:

[0083] The present invention determines the set of required test load types, and selects the empirical data in the empirical test data whose test load types are the same as the required test load types in the set of required test load types; uses the empirical data to train the load test model corresponding to the set of required test load types, and inputs the basis information for formulating the test plan extracted according to the load test association information into the load test model to obtain the load test plan. The empirical data considers the influence between different required test load types, and training according to the empirical data improves the training accuracy and suitability of the model.

[0084] In one embodiment, the empirical test data retrieval subsystem obtains the set of required test load types and retrieves the empirical test data of the set of required test load types, including:

[0085] According to the load test association information, determine the second historical test data of the target historical wind turbine with similar information; among them, the target historical wind turbine is: the wind turbine that has undergone load tests in history; information similarity means that the load test association information of the target historical wind turbine is similar to the load test association information of the target wind turbine; the second historical test data is: the process data of the load test manually carried out according to the load test association information of the target historical wind turbine, including: the equipment and methods used and the load conditions of the target historical wind turbine after the test;

[0086] According to the second historical test data, obtain the historical test load type set of the target historical wind turbine and use it as the set of required test load types; among them, the historical test load type set is obtained by analyzing the second historical test data;

[0087] Obtain the overload events of the target historical wind turbine; wherein, the overload event is an event where the load borne by the target wind turbine exceeds the maximum value specified by the design or safety standard;

[0088] Determine the overload type according to the overload event; wherein, the overload type is the type of load that causes the overload event, for example: additional wind load or other types of load caused by extreme wind speed or operation error;

[0089] Determine the target load type in the required test load type set except the overload type;

[0090] Obtain the first empirical test sub-data of the overload type and the second empirical test sub-data of the target load type according to the second historical test data;

[0091] Use the first empirical test sub-data and the second empirical test sub-data together as the empirical test data;

[0092] Further, the empirical test data retrieval subsystem obtains the first empirical test sub-data of the overload type and the second empirical test sub-data of the target load type according to the second historical test data, including:

[0093] Obtain the third historical test data that meets the negative sample extraction conditions before the overload event of the target historical wind turbine occurs; wherein, the third historical test data is historical test data containing non-standard test operations;

[0094] Use the second historical test data except the third historical test data in the second historical test data as the fourth historical test data;

[0095] Use the third historical test data corresponding to the overload type as the negative sample test data;

[0096] Use the fourth historical test data corresponding to the overload type as the positive sample test data;

[0097] Mark the non-standard test data features in the negative sample test data and the standard test data features in the positive sample test data; wherein, the non-standard test data features are the characteristic values of non-standard test behaviors in the negative sample test data, such as: non-standard conditions, non-standard test operations; the standard test data features are normal and compliant test conditions or operations;

[0098] Use the marked negative sample test data and positive sample test data as the first empirical test sub-data;

[0099] Use the second historical test data corresponding to the target load type as the second empirical test sub-data;

[0100] Among them, the negative sample extraction conditions include:

[0101] The test time of the third historical test data is before the occurrence time of the overload event, and the time interval between the two is less than the preset time interval threshold; wherein, the preset time interval threshold is set manually in advance, and the smaller the time interval, the more likely the test behavior recorded in the third historical test data is to cause the overload event;

[0102] There is a consistency between the test load type of the third historical test data and the overload type corresponding to the overload event that occurs continuously after the test corresponding to the third historical test data;

[0103] Furthermore, the empirical test data retrieval subsystem marks the non-standard test data features in the negative sample test data and the standard test data features in the positive sample test data, including:

[0104] Send the negative sample test data to the expert node to obtain the non-standard test data features marked by the expert; wherein, the expert node is the communication node of the domain expert for the load test of the wind turbine;

[0105] According to the first feature type of the non-standard test data features, mark the standard test data features for the positive sample test data. At the same time, regard the remaining test data features of the non-standard test data features in the negative sample test data that are not marked as the standard test data features; wherein, when marking the standard test data features for the positive sample test data, mark the positive sample test data features corresponding to the first feature type as the standard test data features;

[0106] Furthermore, the empirical test data retrieval subsystem marks the non-standard test data features in the negative sample test data and the standard test data features in the positive sample test data, and also includes:

[0107] Calculate the data feature similarity between the negative sample test data and the positive sample test data of different second feature types; wherein, the second feature type is all data types in the negative sample test data and the positive sample test data, such as: test equipment, test method;

[0108] If the data feature similarity is greater than or equal to the preset similarity threshold, mark the negative sample test data features and the positive sample test data features of the corresponding second feature type as the standard test data features; wherein, the preset similarity threshold is set manually in advance. When the data features are similar, it means that the test behaviors represented by the corresponding test data features in the positive and negative test data are standard;

[0109] If the data feature similarity is less than the preset similarity threshold, regard the negative sample test data features of the corresponding second feature type as the non-standard test data features and the positive sample test data features of the corresponding second feature type as the standard test data features.

[0110] The working principle and beneficial effects of the above technical solution are as follows:

[0111] When determining the set of required test load types, second historical test data similar to the load test-related information can be retrieved, and based on the second historical test data, the set of required test load types is determined. However, not all of the second historical test data is available (for example, the test behaviors recorded in the second historical test data are not standardized). Therefore, an overload event is introduced to obtain the overload type. The target load type excluding the overload type in the set of required test load types is determined, and the second historical test data corresponding to the target load type is standardized test data, and the second historical test data corresponding to the target load type is used as the second empirical test sub-data. The second historical test data corresponding to the overload type contains a part of non-standardized data and a part of standardized data. A negative sample extraction condition is introduced to obtain the third historical test data that meets the negative sample extraction condition before the overload event occurs in the target historical wind turbine, and the third historical test data corresponding to the overload type is used as the negative sample test data, and the fourth historical test data corresponding to the overload type is used as the positive sample test data. The test data features corresponding to the positive sample test data are all standardized test data features, and the corresponding test data features in the negative sample test data contain a part of standardized test data features and a part of non-standardized test data features. Therefore, annotation is required to facilitate the subsequent learning of the neural network. After the annotation is completed, the marked negative sample test data and positive sample test data are used as the first empirical test sub-data.

[0112] Furthermore, the embodiment of the present invention also defines the negative sample extraction condition, which specifically includes:

[0113] Condition 1: The test time of the third historical test data is before the overload event occurrence time and the time interval between the two is less than a preset time interval threshold. The shorter the time interval threshold, the more likely it is that the corresponding third historical test data contains non-standardized test behaviors that cause the corresponding overload event.

[0114] Condition 2: The test load type of the third historical test data is consistent with the overload type corresponding to the overload event that occurs consecutively after the test corresponding to the third historical test data, indicating that the test of the overload type of the overload event that occurs immediately after is actually carried out in the third historical test data that is earlier in time and the overload risk is not detected.

[0115] Using Condition 1 and Condition 2 to define the negative sample extraction condition improves the screening accuracy of the negative sample test data and enhances the extraction efficiency of subsequent non-standardized test data features and standardized test data features.

[0116] Further, mark the non-standard test data features in the negative sample test data and the standard test data features in the positive sample test data, including:

[0117] Method 1: Introduce an expert node. The expert annotates the non-standard test data features, determines the first feature type of the non-standard test data features, and regards the remaining test data features of the non-standard test data features not annotated in the negative sample test data and the positive sample test data features corresponding to the first feature type in the positive sample test data as standard test data features;

[0118] Method 2: Calculate the data feature similarity between the negative sample test data and the positive sample test data. The negative sample test data features with a high similarity deviation are non-standard test data features, and the corresponding deviated positive sample test data features are standard test data features; the negative sample test data features and the positive sample test data features with a data feature similarity greater than or equal to the preset similarity threshold are both standard test data features;

[0119] Introducing both methods for data marking provides sample data for forward learning and reverse learning, improving the comprehensiveness of subsequent learning.

[0120] In one embodiment, the load test device of the wind turbine further includes:

[0121] Anomaly analysis subsystem, used to determine the abnormal wind turbines with abnormal load tests during the maintenance stage of the target wind turbine, and perform anomaly attribution according to the abnormal load types of the abnormal wind turbines; wherein, the abnormal load types are: load types with overload;

[0122] Among them, the anomaly analysis subsystem determines the abnormal wind turbines with abnormal load tests during the maintenance stage of the target wind turbine, and performs anomaly attribution according to the abnormal load types of the abnormal wind turbines, including:

[0123] Obtain the wind turbine information of the abnormal wind turbines with the same abnormal load type; wherein, the wind turbine information includes: the location information, linkage control information, and design information of the abnormal wind turbines;

[0124] Perform principal component analysis based on the wind turbine information to obtain the principal component analysis result associated with the abnormal load type; wherein, the principal component analysis result is: the similar part information of the wind turbine information of the abnormal wind turbines with the same abnormal load type, for example: all are located in area A and all use a certain linkage control module;

[0125] Obtain the anomaly analysis model corresponding to the abnormal load type, and determine the input feature extraction source information type of the anomaly analysis model; wherein, the anomaly analysis model is: an AI model that automatically analyzes the anomaly cause according to the input features corresponding to the abnormal load type;

[0126] Obtain the correlation relationship between the principal component analysis result and the input feature extraction source information type; where the input feature extraction source information type is: the extraction source information type of the input features corresponding to the abnormal load type, such as: climate information; the correlation relationship is, for example: the correlation relationship between area A and climate information;

[0127] According to the correlation relationship, dock with the input feature extraction source information link to obtain the input feature extraction source information and extract the input features, and input the input features into the abnormal analysis model to obtain the abnormal attribution result. Among them, the input feature extraction source information link is: the communication link of the information source of the climate in area A.

[0128] The working principle and beneficial effects of the above technical solution are:

[0129] The load tests set according to experience cannot consider all load situations. Therefore, during the maintenance stage of the target wind turbine, regular monitoring for abnormal analysis is also required;

[0130] Determine the wind turbine information of the abnormal wind turbines of the same abnormal load type, and perform principal component analysis on the wind turbine information to determine the principal component analysis result associated with the abnormal load type; introduce the abnormal analysis model corresponding to the abnormal load type, determine the input data type (input feature extraction source information type) for the model to perform abnormal analysis, obtain the correlation relationship between the principal component analysis result and the input feature extraction source information type, and then, according to the correlation relationship, dock with the input feature extraction source information link. The advantage of such a setting is that, targeted at the principal component analysis result of the wind turbine cluster corresponding to the abnormal load type, it excludes the interference of other irrelevant load abnormal types and other irrelevant wind turbine information, accurately docks with the input feature extraction source information link, reduces the input data volume of the abnormal analysis model, and improves the abnormal analysis efficiency.

[0131] An embodiment of the present invention provides a load test method for a wind turbine, as Figure 2 shown, including:

[0132] Step 1: Obtain the load test correlation information of the target wind turbine;

[0133] Step 2: Obtain the set of required test load types, and retrieve the empirical test data of the set of required test load types;

[0134] Step 3: Train the load test model corresponding to the set of required test load types according to the empirical test data;

[0135] Step 4: Determine the load test plan for the target wind turbine according to the load test correlation information and the load test model and perform corresponding tests.

[0136] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention is also intended to include these modifications and variations.

Claims

1. A load testing device for a wind turbine generator set, characterized in that: include: A load test related information acquisition subsystem is used to acquire load test related information of a target wind turbine; An empirical test data retrieval subsystem is used to obtain a required test load type set and retrieve empirical test data for the required test load type set; A load test model training subsystem is used to train the load test model corresponding to the required test load type set based on empirical test data; The test subsystem is used to determine the load test scheme of the target wind turbine according to the load test association information and the load test model and perform corresponding tests.

2. A load testing device for a wind turbine according to claim 1, characterized in that: The load test related information acquisition subsystem acquires the load test related information of the target wind turbine, including: Obtaining design parameters, first historical test data, operating environment, and expected load conditions of a target wind turbine; The design parameters of the target wind turbine generator set, the first historical test data, the operating environment and the expected load conditions are taken together as load test associated information.

3. A load testing device for a wind turbine generator set according to claim 1, characterized in that: The empirical test data retrieval subsystem obtains the required test load type set and retrieves the empirical test data of the required test load type set, including: Determine, based on the load test association information, second historical test data of a target historical wind turbine generator set with similar information; According to the second historical test data, a historical test load type set of the target historical wind turbine generator set is obtained and used as the required test load type set; Obtain overload events of target historical wind turbines; According to the overload event, determine the overload type; Determine the target load type excluding the overload type in the required test load type set; According to the second historical test data, obtaining first empirical test sub-data of the overload type and second empirical test sub-data of the target load type; The first empirical test sub-data and the second empirical test sub-data are collectively regarded as empirical test data.

4. A load testing device for a wind turbine as claimed in claim 3, characterized in that: The experience test data retrieval subsystem obtains first experience test sub-data of the overload type and second experience test sub-data of the target load type according to the second historical test data, including: Acquire the third historical test data that meets the negative sample extraction condition before the overload event of the target historical wind turbine generator set occurs; using the second historical test data excluding the third historical test data in the second historical test data as fourth historical test data; using the third historical test data corresponding to the overload type as negative sample test data; using fourth historical test data corresponding to the overload type as positive sample test data; Marking irregular test data features in negative sample test data and regular test data features in positive sample test data; The labeled negative sample test data and positive sample test data are used as the first empirical test sub-data; using second historical test data corresponding to the target load type as second empirical test sub-data; Among them, the negative sample extraction conditions include: The test time of the third historical test data is before the time when the overload event occurs and the time interval between the two is less than a preset time interval threshold; The test load type of the third historical test data is consistent with the overload type corresponding to the overload event that occurs subsequently after the test corresponding to the third historical test data.

5. A load testing device for a wind turbine as claimed in claim 4, characterized in that: The empirical test data retrieval subsystem marks the irregular test data features in the negative sample test data and the regular test data features in the positive sample test data, including: Send the negative sample test data to the expert node to obtain the features of the irregular test data annotated by the expert; According to the first feature type of the irregular test data feature, the regular test data feature is marked for the positive sample test data, and at the same time, the remaining test data features of the irregular test data features not marked in the negative sample test data are also used as regular test data features.

6. A load testing device for a wind turbine generator set as claimed in claim 4, characterized in that: The empirical test data retrieval subsystem marks irregular test data features in negative sample test data and regular test data features in positive sample test data, and also includes: Calculate the data feature similarity between the negative sample test data and the positive sample test data of different second feature types; If the data feature similarity is greater than or equal to a preset similarity threshold, the negative sample test data feature and the positive sample test data feature of the corresponding second feature type are marked as standard test data features; If the data feature similarity is less than a preset similarity threshold, the negative sample test data feature of the corresponding second feature type is used as an irregular test data feature, and the positive sample test data feature of the corresponding second feature type is used as a standard test data feature.

7. A load testing device for a wind turbine generator set according to claim 1, characterized in that: The load test model training subsystem trains the load test model corresponding to the required test load type set based on the empirical test data, including: Extract empirical evidence and empirical solutions based on empirical test data; The empirical basis is used as the input of the CNN model, and the empirical solution is used as the output of the CNN model to train the load test model.

8. A load testing device for a wind turbine generator set according to claim 1, characterized in that: The test subsystem determines the load test scheme of the target wind turbine according to the load test association information and load test model and performs corresponding tests, including: Extracting a basis vector from the load test association information based on an empirical basis extraction template; The load test model will be tested based on the vector input to obtain the load test plan of the model output.

9. A load testing device for a wind turbine generator set according to claim 1, characterized in that: Also includes: The abnormality analysis subsystem is used to determine the abnormal wind turbines with abnormal load tests during the maintenance phase of the target wind turbines, and to attribute the abnormalities according to the abnormal load types of the abnormal wind turbines; Among them, the abnormal analysis subsystem determines the abnormal wind turbines with abnormal load test during the maintenance phase of the target wind turbines, and attributes the abnormalities according to the abnormal load types of the abnormal wind turbines, including: Obtain wind turbine information of abnormal wind turbines of the same abnormal load type; Perform principal component analysis based on wind turbine information to obtain principal component analysis results associated with abnormal load types; Obtain an abnormal analysis model corresponding to the abnormal load type, and determine the type of input feature extraction source information of the abnormal analysis model; Obtaining the correlation between the principal component analysis results and the input feature extraction source information type; According to the association relationship, the input feature extraction source information link is connected to obtain the input feature extraction source information and extract the input features, and the input features are input into the anomaly analysis model to obtain the anomaly attribution results.

10. A load testing method for a wind turbine generator set, characterized in that: include: Step 1: Obtain load test related information of the target wind turbine; Step 2: Obtain the required test load type set and retrieve the empirical test data of the required test load type set; Step 3: Train the load test model corresponding to the required test load type set based on the empirical test data; Step 4: Determine the load test plan for the target wind turbine according to the load test association information and the load test model and perform corresponding tests.

Images