A wind turbine fault detection method and system

Abstract

The application discloses a kind of wind generator fault detection method and system, it is related to wind generator fault detection field, including, the discrete quantity and continuous quantity of wind turbine in SCADA are acquired, the first correlation degree between two discrete quantities is determined, if the first correlation degree meets first preset requirement, then determine that there is migration correlation between two discrete quantities;Second correlation degree between two continuous quantities is determined, if the second correlation degree meets second preset requirement, then determine that there is migration correlation between two continuous quantities;Source quantity and target quantity in the two discrete quantities and two continuous quantities that migration correlation exists are distinguished, and source quantity and target quantity correspond;According to source quantity, establish classification model, according to the predicted quantity of classification model prediction source quantity and target quantity in a period of time in the future, according to predicted quantity, determine the fault condition of wind turbine.The application is by being substituted into discrete quantity and continuous quantity by migration learning, reduce the required computing resources of different system quantity, improve analysis effect.

Description

Technical Field

[0001] This application relates to the field of wind turbine generator fault detection technology, and more specifically, to a wind turbine generator fault detection method and system. Background Technology

[0002] A wind turbine is an electrical device that converts wind energy into mechanical work, which drives a rotor to rotate and ultimately outputs alternating current (AC). A wind turbine generally consists of a wind turbine rotor, a generator (including its components), a directional control unit (tail fin), a tower, a speed-limiting safety mechanism, and an energy storage device. The working principle of a wind turbine is relatively simple: the wind turbine rotates under the influence of wind, converting the kinetic energy of the wind into the mechanical energy of the turbine shaft. The generator, driven by the turbine shaft, rotates to generate electricity. Broadly speaking, wind energy is also solar energy, so a wind turbine can also be described as a thermal energy generator that uses the sun as a heat source and the atmosphere as a working medium.

[0003] Wind turbine systems are quite complex. When a wind turbine malfunctions, it often involves the interaction of various parameters within the system, making analysis time-consuming and labor-intensive. Current technologies for analyzing wind turbine faults require analyzing all parameters of each system individually due to their different parameter characteristics, resulting in high computational resource consumption and poor analysis quality.

[0004] Therefore, how to reduce computing resources and improve analysis results is a technical problem that needs to be solved. Summary of the Invention

[0005] This invention provides a method for detecting faults in wind turbine generators, addressing the problems of high computational resource consumption and poor analysis results in existing technologies. This method is applied to a SCADA system and includes:

[0006] Obtain the discrete and continuous quantities of the fan in SCADA, determine the first correlation degree between each pair of discrete quantities, and if the first correlation degree meets the first preset requirement, then determine that there is a migration correlation between each pair of discrete quantities.

[0007] Determine the second degree of correlation between each pair of continuous quantities. If the second degree of correlation meets the second preset requirement, then it is determined that there is a migration correlation between the pair of continuous quantities.

[0008] Distinguish between source and target quantities in pairwise discrete and pairwise continuous quantities that have migration correlations, and ensure that the source and target quantities correspond to each other;

[0009] A classification model is established based on the source quantity. The predicted quantities of the source quantity and the target quantity are then used to predict the future quantities over a period of time. The failure status of the wind turbine is determined based on the predicted quantities.

[0010] In some embodiments of this application, the source quantity includes a source discrete quantity and a source continuous quantity, and the target quantity includes a target discrete quantity and a target continuous quantity, wherein the source discrete quantity and the target discrete quantity correspond to each other, and the source continuous quantity and the target continuous quantity correspond to each other.

[0011] In some embodiments of this application, the first correlation degree meets a first preset requirement, including:

[0012] The first degree of relevance includes support and confidence.

[0013] If the support exceeds the first threshold and the confidence exceeds the second threshold, then a migration association is determined to exist between the pairwise discrete quantities.

[0014] In some embodiments of this application, the method further includes;

[0015] If the support does not exceed the first threshold and the confidence exceeds the second threshold, then it is determined that there is no migration association between the pairwise discrete quantities.

[0016] If the support exceeds the first threshold and the confidence does not exceed the second threshold, then it is determined that there is no migration association between the pairwise discrete quantities.

[0017] If the support does not exceed the first threshold and the confidence does not exceed the second threshold, then it is determined that there is no migration association between the pairwise discrete quantities.

[0018] In some embodiments of this application, determining a second degree of correlation between pairwise continuous quantities includes:

[0019] The second correlation degree is obtained based on the first matrix, the second matrix, and the weight coefficient matrix.

[0020] In some embodiments of this application, the method further includes:

[0021] The degree of opposition is adjusted based on the degree of opposition coefficient, and the weight coefficient matrix is ​​adjusted based on the degree of opposition coefficient.

[0022] In some embodiments of this application, the second correlation degree meets a second preset requirement, including:

[0023] If the second correlation degree exceeds the third threshold, then a migration correlation is determined to exist between the pairwise continuous quantities;

[0024] If the second correlation does not exceed the third threshold, then it is determined that there is no migration correlation between the pairwise continuous quantities.

[0025] In some embodiments of this application, the source and target quantities in pairwise discrete quantities and pairwise continuous quantities that have migration correlations are distinguished, including:

[0026] In the pairwise discrete quantities with transfer correlation, the one with more labeled training samples is taken as the source discrete quantity, and the other with fewer labeled training samples is taken as the target discrete quantity.

[0027] In the pairwise continuous quantities with transfer association, the one with more labeled training samples is taken as the source continuous quantity, and the other with fewer labeled training samples is taken as the target continuous quantity.

[0028] In some embodiments of this application, a classification model is established based on the source quantity, and the predicted quantities of the source quantity and the target quantity over a future period are predicted based on the classification model, including:

[0029] Establish a discrete quantity classification model based on the source discrete quantity, and predict the discrete quantity in the future period based on the discrete quantity classification model.

[0030] A continuous quantity classification model is established based on the source continuous quantity, and the predicted continuous quantity is predicted within a future period based on the continuous quantity classification model.

[0031] Correspondingly, this application also provides a wind turbine fault detection system, applied in a SCADA system, the system comprising:

[0032] The first correlation module is used to obtain the discrete and continuous quantities of the wind turbine in SCADA, determine the first correlation degree between each pair of discrete quantities, and if the first correlation degree meets the first preset requirement, then it is determined that there is a migration correlation between each pair of discrete quantities.

[0033] The second association module is used to determine the second association degree between pairs of continuous quantities. If the second association degree meets the second preset requirements, then it is determined that there is a migration association between the pairs of continuous quantities.

[0034] The differentiation module is used to distinguish the source and target quantities among pairwise discrete quantities and pairwise continuous quantities that have migration relationships, and the source and target quantities correspond to each other;

[0035] The prediction module is used to build a classification model based on the source quantity, predict the source quantity and target quantity for a future period of time based on the classification model, and determine the failure status of the wind turbine based on the predicted quantity.

[0036] By applying the above technical solutions, discrete and continuous quantities of the wind turbine in SCADA are obtained. The first correlation degree between each pair of discrete quantities is determined. If the first correlation degree meets a first preset requirement, a transfer correlation is determined between the pair of discrete quantities. Similarly, the second correlation degree between each pair of continuous quantities is determined. If the second correlation degree meets a second preset requirement, a transfer correlation is determined between the pair of continuous quantities. The source and target quantities in the pair of discrete and continuous quantities with transfer correlations are distinguished, and the source and target quantities correspond. A classification model is established based on the source quantities. The predicted quantities of the source and target quantities over a future period are predicted based on the classification model. The fault status of the wind turbine is determined based on the predicted quantities. This application reduces the computational resources required for different system quantities and improves the analysis effect by incorporating transfer learning into discrete and continuous quantities. Attached Figure Description

[0037] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0038] Figure 1 A flowchart illustrating a wind turbine fault detection method according to an embodiment of the present invention is shown.

[0039] Figure 2 A schematic diagram of the structure of a wind turbine fault detection system proposed in an embodiment of the present invention is shown. Detailed Implementation

[0040] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0041] This application provides a wind turbine fault detection method applied to a SCADA system. Besides the main equipment, the wind turbine generators, wind farms also contain many other devices, such as weather masts, noise monitoring stations, main substations, and SCADA systems. Typically, each wind farm has a control room within its main substation. The SCADA system provides communication links between the control room and other equipment in the wind farm, thereby providing information about the substation, weather masts, and wind turbine generators. Since the SCADA system plays a crucial role in the operation and maintenance of the distributed control system of a wind farm, it generally monitors two types of parameters: discrete and continuous.

[0042] like Figure 1 As shown, the method includes the following steps:

[0043] Step S101: Obtain the discrete and continuous quantities of the wind turbine in SCADA, determine the first correlation degree between each pair of discrete quantities, and if the first correlation degree meets the first preset requirement, then determine that there is a migration correlation between each pair of discrete quantities.

[0044] In this embodiment, SCADA parameters are categorized into discrete and continuous parameters based on their type. Discrete parameters are mainly reflected in four systems: the yaw system, the hydraulic system, the generator system, and the cooling system. Discrete parameters can intuitively represent the operating state of a system, expressed as either 0 or 1, such as the yaw brake switch status or the high / low temperature of lubricating oil. Continuous parameters are mainly reflected in six aspects: angle, pressure, temperature, speed, vibration, and electrical factors. Continuous parameters can reflect the development trend of a parameter over a period of time, expressed as a continuous and slow change over time, such as the generator speed or the gearbox temperature. Before utilizing transfer learning, it is necessary to determine whether there is a correlation between discrete and continuous parameters. Only after a correlation is established can transfer learning be applied to subsequent steps. All possible combinations are performed on each pair of discrete parameters, and the same applies to continuous parameters.

[0045] In some embodiments of this application, the source quantity includes a source discrete quantity and a source continuous quantity, and the target quantity includes a target discrete quantity and a target continuous quantity, wherein the source discrete quantity and the target discrete quantity correspond to each other, and the source continuous quantity and the target continuous quantity correspond to each other.

[0046] In this embodiment, transfer learning is a novel machine learning method that uses existing knowledge to solve problems in different but related domains. It relaxes two fundamental assumptions of traditional machine learning, aiming to transfer knowledge learned from a labeled source domain to the target domain when the source and target domain data have different data distributions. This addresses learning problems where the target domain has only a small amount of labeled sample data or even none at all. In this approach, the source domain corresponds to the source discrete quantity and the source continuous quantity, while the target domain corresponds to the target discrete quantity and the target continuous quantity. Applying transfer learning between different discrete quantities or different continuous quantities reduces the waste of computational resources.

[0047] To ensure the adaptability of the correlation degree between different discrete quantities, in some embodiments of this application, the first correlation degree meets a first preset requirement, including: the first correlation degree includes support and confidence; if the support exceeds a first threshold and the confidence exceeds a second threshold, then it is determined that there is a migration correlation between the pairwise discrete quantities. In some embodiments of this application, the method further includes: if the support does not exceed the first threshold and the confidence exceeds the second threshold, then it is determined that there is no migration correlation between the pairwise discrete quantities; if the support exceeds the first threshold and the confidence does not exceed the second threshold, then it is determined that there is no migration correlation between the pairwise discrete quantities; if the support does not exceed the first threshold and the confidence does not exceed the second threshold, then it is determined that there is no migration correlation between the pairwise discrete quantities.

[0048] In this embodiment, the first correlation degree is the correlation degree between pairwise discrete quantities obtained using association rules. The specific calculation process is a conventional technique in this field and will not be elaborated here. Support and confidence reflect the effectiveness and certainty of this association rule, respectively. Support characterizes the importance or probability of occurrence of the association rule in the parameter database; that is, the higher the support, the higher the correlation. Confidence characterizes the credibility of the association rule; that is, the higher the confidence, the higher the credibility. Only when both confidence and support meet the requirements is a transferable association considered to exist between the two.

[0049] Step S102: Determine the second correlation degree between pairs of continuous quantities. If the second correlation degree meets the second preset requirement, then it is determined that there is a migration correlation between pairs of continuous quantities.

[0050] In some embodiments of this application, determining the second correlation degree between pairwise continuous quantities includes: obtaining the second correlation degree based on the first matrix, the second matrix, and the weight coefficient matrix.

[0051] In this embodiment, set pair analysis is chosen to calculate the correlation between two sets based on the property of continuous quantities (continuous change over time). Set pair analysis involves defining two sets A and B as a set pair H = (A, B). Under the context of a problem Q, the expansion analysis of set pair H yields P characteristics. X of these characteristics are shared by A and B, Z characteristics are mutually exclusive, and the remaining Y = P × Z characteristics are neither shared by nor mutually exclusive between the two sets. The multivariate correlation degree expression is established as follows:

[0052] μ=a+b1i1+b2i2+…+b l-2 i l-2 +cj

[0053] μ is the degree of connection between sets A and B, ranging from -1 to 1. a, b, and c are the degree of identity, difference, and opposition of the two sets under problem Q, respectively, where a = X / P, b = Y / P, and c = Z / P. i is the degree of difference coefficient, which is specified to take a value in the range of -1 to 1 depending on different situations, and j is the degree of opposition coefficient.

[0054] Let matrix R = [a, b1…b2] l-2 [C] is the similarity and difference evaluation matrix of the evaluation indicators, and matrix E = [1, i1...i...]. l-2 ,j] T

[0055] R is the first matrix, and E is the second matrix.

[0056]

[0057] W m R mThese are the constant weight coefficient matrix and the similarity / dissimilarity inverse evaluation matrix, respectively.

[0058] To improve calculation accuracy, in some embodiments of this application, the method further includes: correcting the degree of opposition coefficient based on the degree of opposition, and correcting the weight coefficient matrix based on the degree of opposition coefficient.

[0059] In this embodiment, since different degrees of opposition will affect the weighting coefficient, it needs to be corrected.

[0060] The degree of opposition is adjusted based on the degree of opposition, specifically as follows:

[0061] A preset opposition array C0(C1, C2, C3, C4) is provided, where C1, C2, C3, and C4 are preset values, and C1 < C2 < C3 < C4.

[0062] A preset opposition correction coefficient array J0(J1, J2, J3, J4) is provided, where J1, J2, J3, and J4 are all preset values, and 0.8 < J1 < J2 < J3 < J4 < 1.2.

[0063] Based on the relationship between the degree of opposition and the preset degree of opposition, determine the degree of opposition correction coefficient and make corrections accordingly;

[0064] If c < C1, then the first preset opposition degree correction coefficient J1 is used as the opposition degree correction coefficient, and the opposition degree correction coefficient is corrected. The corrected opposition degree coefficient is j*J1.

[0065] If C1≤c<C2, then the second preset opposition degree correction coefficient J2 is used as the opposition degree correction coefficient to correct the opposition degree correction coefficient. The corrected opposition degree correction coefficient is j*J2.

[0066] If C2≤c<C3, then the third preset opposition degree correction coefficient J3 is used as the opposition degree correction coefficient to correct the opposition degree correction coefficient. The corrected opposition degree correction coefficient is j*J3.

[0067] If C3≤c<C4, then the fourth preset opposition degree correction coefficient J4 is used as the opposition degree correction coefficient to correct the opposition degree correction coefficient. The corrected opposition degree coefficient is j*J4.

[0068] The weighting coefficient matrix is ​​adjusted based on the degree of opposition coefficient, specifically as follows:

[0069] Set the modified opposition coefficient to D, and preset the modified opposition coefficient array D0(D1, D2, D3, D4), where D1, D2, D3, and D4 are all preset values, and D1 < D2 < D3 < D4.

[0070] A preset weight coefficient correction value array Q0(Q1, Q2, Q3, Q4) is provided, where Q1, Q2, Q3, and Q4 are all preset values, and 0.8 < Q1 < Q2 < Q3 < Q4 < 1.2.

[0071] Based on the relationship between the corrected opposition coefficient and the preset opposition coefficient, the weight coefficient correction value is determined, and the weights are corrected accordingly.

[0072] If D < D1, determine the first preset weight coefficient correction value Q1 as the weight coefficient correction value, and the corrected weight coefficient is W. m *Q1;

[0073] If D1 ≤ D < D2, determine the second preset weight coefficient correction value Q2 as the weight coefficient correction value, and the corrected weight coefficient is W. m *Q2;

[0074] If D2 ≤ D < D3, determine the third preset weight coefficient correction value Q3 as the weight coefficient correction value, and the corrected weight coefficient is W. m *Q3;

[0075] If D3 ≤ D < D4, determine the fourth preset weight coefficient correction value Q4 as the weight coefficient correction value, and the corrected weight coefficient is W. m *Q4.

[0076] In some embodiments of this application, the second correlation degree meets a second preset requirement, including:

[0077] If the second correlation degree exceeds the third threshold, then a migration correlation is determined to exist between the pairwise continuous quantities;

[0078] If the second correlation does not exceed the third threshold, then it is determined that there is no migration correlation between the pairwise continuous quantities.

[0079] Step S103: Distinguish between source and target quantities in pairwise discrete and pairwise continuous quantities that have migration correlation, and ensure that the source and target quantities correspond to each other.

[0080] In this embodiment, the source and target quantities are determined based on the labeled training samples, which is equivalent to applying the learning method for the source quantity to the target quantity. The quantity with a larger number of effective labeled training samples is used as the source quantity, and the quantity with a smaller number is used as the target quantity, thereby improving the effectiveness of transfer learning.

[0081] In some embodiments of this application, the source and target quantities in pairwise discrete quantities and pairwise continuous quantities that have migration correlations are distinguished, including:

[0082] In the pairwise discrete quantities with transfer correlation, the one with more labeled training samples is taken as the source discrete quantity, and the other with fewer labeled training samples is taken as the target discrete quantity.

[0083] In the pairwise continuous quantities with transfer association, the one with more labeled training samples is taken as the source continuous quantity, and the other with fewer labeled training samples is taken as the target continuous quantity.

[0084] Step S104: Establish a classification model based on the source quantity, predict the predicted quantities of the source quantity and target quantity over a future period based on the classification model, and determine the fault status of the wind turbine based on the predicted quantities.

[0085] In this embodiment, a reliable classification model is established to predict the target domain data using a large number of labeled training samples or source domain data. The source domain data and the target domain data may not have the same data distribution.

[0086] In some embodiments of this application, a classification model is established based on the source quantity, and the predicted quantities of the source quantity and the target quantity over a future period are predicted based on the classification model, including:

[0087] Establish a discrete quantity classification model based on the source discrete quantity, and predict the discrete quantity in the future period based on the discrete quantity classification model.

[0088] A continuous quantity classification model is established based on the source continuous quantity, and the predicted continuous quantity is predicted within a future period based on the continuous quantity classification model.

[0089] By applying the above technical solutions, discrete and continuous quantities of the wind turbine in SCADA are obtained. The first correlation degree between each pair of discrete quantities is determined. If the first correlation degree meets a first preset requirement, a transfer correlation is determined between the pair of discrete quantities. Similarly, the second correlation degree between each pair of continuous quantities is determined. If the second correlation degree meets a second preset requirement, a transfer correlation is determined between the pair of continuous quantities. The source and target quantities in the pair of discrete and continuous quantities with transfer correlations are distinguished, and the source and target quantities correspond. A classification model is established based on the source quantities. The predicted quantities of the source and target quantities over a future period are predicted based on the classification model. The fault status of the wind turbine is determined based on the predicted quantities. This application reduces the computational resources required for different system quantities and improves the analysis effect by incorporating transfer learning into discrete and continuous quantities.

[0090] Through the above description of the embodiments, those skilled in the art can clearly understand that the present invention can be implemented in hardware or by means of software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solution of the present invention can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, external hard drive, etc.) and includes several instructions to cause a computer device (such as a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0091] Correspondingly, this application also provides a wind turbine fault detection system, applied in a SCADA system, such as... Figure 2 As shown, the system includes:

[0092] The first correlation module 201 is used to obtain the discrete and continuous quantities of the wind turbine in SCADA, determine the first correlation degree between each pair of discrete quantities, and if the first correlation degree meets the first preset requirement, then it is determined that there is a migration correlation between each pair of discrete quantities.

[0093] The second association module 202 is used to determine the second association degree between pairs of continuous quantities. If the second association degree meets the second preset requirements, then it is determined that there is a migration association between pairs of continuous quantities.

[0094] The differentiation module 203 is used to distinguish the source quantity and the target quantity among pairwise discrete quantities and pairwise continuous quantities that have migration correlation, and the source quantity and the target quantity correspond to each other;

[0095] The prediction module 204 is used to establish a classification model based on the source quantity, predict the predicted quantities of the source quantity and the target quantity over a future period based on the classification model, and determine the failure status of the wind turbine based on the predicted quantities.

[0096] In some embodiments of this application, the source quantity includes a source discrete quantity and a source continuous quantity, and the target quantity includes a target discrete quantity and a target continuous quantity, wherein the source discrete quantity and the target discrete quantity correspond to each other, and the source continuous quantity and the target continuous quantity correspond to each other.

[0097] In some embodiments of this application, the first association module 201 is used for:

[0098] The first degree of relevance includes support and confidence.

[0099] If the support exceeds the first threshold and the confidence exceeds the second threshold, then a migration association is determined to exist between the pairwise discrete quantities.

[0100] In some embodiments of this application, the first association module 201 is used for;

[0101] If the support does not exceed the first threshold and the confidence exceeds the second threshold, then it is determined that there is no migration association between the pairwise discrete quantities.

[0102] If the support exceeds the first threshold and the confidence does not exceed the second threshold, then it is determined that there is no migration association between the pairwise discrete quantities.

[0103] If the support does not exceed the first threshold and the confidence does not exceed the second threshold, then it is determined that there is no migration association between the pairwise discrete quantities.

[0104] In some embodiments of this application, the second association module 202 is used for:

[0105] The second correlation degree is obtained based on the first matrix, the second matrix, and the weight coefficient matrix.

[0106] In some embodiments of this application, the second association module 202 is used for:

[0107] The degree of opposition is adjusted based on the degree of opposition coefficient, and the weight coefficient matrix is ​​adjusted based on the degree of opposition coefficient.

[0108] In some embodiments of this application, the second association module 202 is used for:

[0109] If the second correlation degree exceeds the third threshold, then a migration correlation is determined to exist between the pairwise continuous quantities;

[0110] If the second correlation does not exceed the third threshold, then it is determined that there is no migration correlation between the pairwise continuous quantities.

[0111] In some embodiments of this application, the distinguishing module 203 is used for:

[0112] In the pairwise discrete quantities with transfer correlation, the one with more labeled training samples is taken as the source discrete quantity, and the other with fewer labeled training samples is taken as the target discrete quantity.

[0113] In the pairwise continuous quantities with transfer association, the one with more labeled training samples is taken as the source continuous quantity, and the other with fewer labeled training samples is taken as the target continuous quantity.

[0114] In some embodiments of this application, the prediction module 204 is used for:

[0115] Establish a discrete quantity classification model based on the source discrete quantity, and predict the discrete quantity in the future period based on the discrete quantity classification model.

[0116] A continuous quantity classification model is established based on the source continuous quantity, and the predicted continuous quantity is predicted within a future period based on the continuous quantity classification model.

[0117] Those skilled in the art will understand that the modules in the system of the implementation scenario can be distributed throughout the system of the implementation scenario as described, or they can be modified to reside in one or more systems different from this implementation scenario. The modules of the above-mentioned implementation scenario can be merged into one module, or they can be further divided into multiple sub-modules.

[0118] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for detecting faults in wind turbine generators, applied in a SCADA system, characterized in that, The method includes: Obtain the discrete and continuous quantities of the fan in SCADA, determine the first correlation degree between each pair of discrete quantities, and if the first correlation degree meets the first preset requirement, then determine that there is a migration correlation between each pair of discrete quantities. Determine the second degree of correlation between each pair of continuous quantities. If the second degree of correlation meets the second preset requirement, then it is determined that there is a migration correlation between the pair of continuous quantities. Distinguish between source and target quantities in pairwise discrete and pairwise continuous quantities that have migration correlations, and ensure that the source and target quantities correspond to each other; A classification model is established based on the source quantity; the predicted quantities of the source quantity and the target quantity are predicted based on the classification model over a period of time; and the failure status of the wind turbine is determined based on the predicted quantities. Determining the second degree of correlation between pairwise continuous quantities includes: The second correlation degree is obtained based on the first matrix, the second matrix, and the weight coefficient matrix; The method further includes: The degree of opposition is adjusted based on the degree of opposition coefficient, and the weight coefficient matrix is ​​adjusted based on the degree of opposition coefficient.

2. The method as described in claim 1, characterized in that, The source quantity includes source discrete quantity and source continuous quantity, and the target quantity includes target discrete quantity and target continuous quantity. The source discrete quantity and the target discrete quantity correspond to each other, and the source continuous quantity and the target continuous quantity correspond to each other.

3. The method as described in claim 1, characterized in that, The first degree of correlation meets the first preset requirements, including: The first degree of relevance includes support and confidence. If the support exceeds the first threshold and the confidence exceeds the second threshold, then a migration association is determined to exist between the pairwise discrete quantities.

4. The method as described in claim 3, characterized in that, The method further includes; If the support does not exceed the first threshold and the confidence exceeds the second threshold, then it is determined that there is no migration association between the pairwise discrete quantities. If the support exceeds the first threshold and the confidence does not exceed the second threshold, then it is determined that there is no migration association between the pairwise discrete quantities. If the support does not exceed the first threshold and the confidence does not exceed the second threshold, then it is determined that there is no migration association between the pairwise discrete quantities.

5. The method as described in claim 1, characterized in that, The second correlation degree meets the second preset requirements, including: If the second correlation degree exceeds the third threshold, then a migration correlation is determined to exist between the pairwise continuous quantities; If the second correlation does not exceed the third threshold, then it is determined that there is no migration correlation between the pairwise continuous quantities.

6. The method as described in claim 2, characterized in that, Distinguish between source and target quantities in pairwise discrete and pairwise continuous quantities that exhibit migration correlation, including: In the pairwise discrete quantities with transfer correlation, the one with more labeled training samples is taken as the source discrete quantity, and the other with fewer labeled training samples is taken as the target discrete quantity. In the pairwise continuous quantities with transfer association, the one with more labeled training samples is taken as the source continuous quantity, and the other with fewer labeled training samples is taken as the target continuous quantity.

7. The method as described in claim 6, characterized in that, A classification model is built based on the source quantity, and the predicted quantities of the source quantity and the target quantity are predicted over a future period based on the classification model, including: Establish a discrete quantity classification model based on the source discrete quantity, and predict the discrete quantity in the future period based on the discrete quantity classification model. A continuous quantity classification model is established based on the source continuous quantity, and the predicted continuous quantity is predicted within a future period based on the continuous quantity classification model.

8. A wind turbine fault detection system, applied in a SCADA system, characterized in that, The system includes: The first correlation module is used to obtain the discrete and continuous quantities of the wind turbine in SCADA, determine the first correlation degree between each pair of discrete quantities, and if the first correlation degree meets the first preset requirement, then it is determined that there is a migration correlation between each pair of discrete quantities. The second association module is used to determine the second association degree between pairs of continuous quantities. If the second association degree meets the second preset requirements, then it is determined that there is a migration association between the pairs of continuous quantities. The differentiation module is used to distinguish the source and target quantities among pairwise discrete quantities and pairwise continuous quantities that have migration relationships, and the source and target quantities correspond to each other; The prediction module is used to build a classification model based on the source quantity, predict the source quantity and target quantity for a future period of time based on the classification model, and determine the failure status of the wind turbine based on the predicted quantity. The second association module determines the second degree of association between pairwise continuous quantities, including: The second correlation degree is obtained based on the first matrix, the second matrix, and the weight coefficient matrix; The system also includes: The degree of opposition is adjusted based on the degree of opposition coefficient, and the weight coefficient matrix is ​​adjusted based on the degree of opposition coefficient.

Images