AVAE_SDL-Based Wind Turbine SCADA Fault Early Warning and Location Method, System, Device and Medium

Abstract

The present invention discloses a method, system, device and medium for SCADA fault warning and positioning of wind turbine generators based on AVAE_SDL, including: collecting SCADA data of the gearbox of the wind turbine generator, preprocessing the collected data, and then dividing the preprocessed data into a training set and a test set; constructing an AVAE_SDL model; inputting the training set into the AVAE_SDL model for training, obtaining an optimized model and an anomaly score of the fault degree of the wind turbine generator, and determining a fault threshold; based on the test set and the optimized model, obtaining prediction data of the optimized model and drawing a fault warning diagram; determining the fault location of the gearbox of the wind turbine generator based on the residual error between each input parameter in the SCADA data and the prediction data of the optimized model, and the fault warning diagram. The present invention combines dictionary learning and adversarial variational autoencoders to improve the stability of the warning model; the present invention can significantly improve the accuracy of SCADA fault warning, achieve early warning in advance, and can realize the fault positioning of wind turbine generators.

Description

Technical Field

[0001] The present invention belongs to the technical field of fault detection of wind turbine generators, and relates to a method, system, device and medium for SCADA fault early warning and location of wind turbine generators based on AVAE_SDL. Background Art

[0002] Early fault identification of wind turbine generators is beneficial to avoiding catastrophic consequences and reducing the operation and maintenance costs of wind farms. The monitoring and data acquisition system (SCADA) of wind turbine generators monitors rich parameters, covering all major components of wind turbine generators, and provides process data and status data for the reliable operation of wind turbine generator equipment, such as wind speed, rotation speed, vibration, current, voltage, etc. SCADA data has been widely used in the trend analysis and early fault detection of wind turbine generators. With the development of artificial intelligence technology, applying deep learning to the condition monitoring of wind turbine generators based on SCADA data can effectively improve the early warning effect of faults and the operation reliability of the units. SCADA data is large in quantity and has many parameters. How to effectively mine key information is the key to fault early warning. Since most of the parameters of SCADA are information such as temperature, which reflects faults relatively slowly, it increases the difficulty of operation and maintenance. Implementing early warning and early warning of wind turbine generator faults based on SCADA data has important engineering significance. Summary of the Invention

[0003] The purpose of the present invention is to solve the problems in the prior art that SCADA parameters reflect faults slowly and it is difficult to perform operation and maintenance, and to provide a method, system, device and medium for SCADA fault early warning and location of wind turbine generators based on AVAE_SDL.

[0004] To achieve the above purpose, the present invention adopts the following technical solutions:

[0005] A method for SCADA fault early warning and location of wind turbine generators based on AVAE_SDL includes:

[0006] Collect SCADA data of the gearbox of the wind turbine generator, preprocess the collected data, and then divide the preprocessed data into a training set and a test set;

[0007] Construct an AVAE_SDL model;

[0008] Input the training set into the AVAE_SDL model for training, obtain the optimized model and the anomaly score of the fault degree of the wind turbine generator, and determine the fault threshold;

[0009] Based on the test set and the optimized model, obtain the prediction data of the optimized model and draw a fault early warning diagram;

[0010] Determine the fault location of the wind turbine gearbox based on the residual error between each input parameter in the SCADA data and the predicted data of the optimization model, as well as the fault warning diagram.

[0011] A further improvement of the present invention lies in:

[0012] Furthermore, preprocess the collected data, specifically: data cleaning and normalization preprocessing; the data cleaning eliminates outliers from the collected data; eliminate outliers from the pitch angle and active power, as shown in formula (1):

[0013]

[0014] In the formula, Pow is the active power, v is the generator speed, and θ is the pitch angle;

[0015] The normalization preprocessing is specifically: perform a linear transformation on the cleaned data to map the result to the range of [0, 1]; the normalization formula is:

[0016] X' = (X - X min ) / (X max - X min ) (2)

[0017] Where X is the original data, and Xmax and Xmin are the maximum and minimum values of the original data set respectively.

[0018] Furthermore, input the training set into the AVAE_SDL model for training to obtain the optimized model, specifically: update the encoder / decoder of the AVAE_SDL model, update the dictionary, and update the discriminator; the training set X is input into the variational autoencoder / decoder, the encoder outputs the mean μ and variance σ, the variance is multiplied by random noise and added to the mean to obtain Z I , Z I is input into the decoder to obtain the intermediate latent code Z S and the output result X’, the intermediate latent code Z S is input into the dictionary learning to obtain the sparse coefficient matrix A and the dictionary matrix Dic, and the loss function for updating the encoder / decoder is:

[0019]

[0020] Where is the reconstruction error of the encoder / decoder, is the KL divergence, ||Z S-Dic·A||2 + λA||1 is the loss function in dictionary learning; N is the size of the data, X is the input sample of the model, X’ is the output sample of the model, that is, the reconstructed sample, μ and σ are the output mean and variance of the variational autoencoder VAE, that is, the intermediate variables;

[0021] The loss function for updating the dictionary is:

[0022] L SDL =||Z S -Dic·A||2 + λA||1 (4)

[0023] where, ||·||2 is the second-order norm, ||·||1 is the first-order norm;

[0024] Z I and Z S are jointly input into the discriminator for adversarial discrimination, as shown in formula (5):

[0025]

[0026] The three updates are performed alternately until convergence below the specified error threshold.

[0027] Furthermore, obtain the anomaly score of the fault degree of the wind turbine unit, specifically: use score i to represent the anomaly score of the fault degree of the wind turbine unit, and the calculation method is:

[0028]

[0029] where, the data X and have N rows and n columns, and i represents the row number; score i consists of two parts. The first item is the reconstruction error of the adversarial variational autoencoder, and the other item is the error of sparse dictionary learning; ζ is a constant, set to 0.1; k is the number of parameters of the SCADA data.

[0030] Furthermore, the upper and lower limits of the fault threshold are:

[0031]

[0032] where, μ and ε are the mean and standard deviation of the training dataset fraction, and K is a constant, which is 1.25 times the maximum anomaly score of the training set.

[0033] Furthermore, based on the test set and the optimized model, obtain the prediction data of the optimized model and draw a fault warning diagram, specifically:

[0034] The test set data X is input into the optimized model to obtain the predicted output of the optimized model and the intermediate variable Z of the optimization model S A, Dic, calculate the scores of the test data, and draw the exponentially weighted moving average control chart EWMA, that is, the fault warning chart. When the data in the chart exceeds the fault threshold, it is judged that the wind turbine has a fault and a warning is realized.

[0035] Furthermore, based on the residual error between each input parameter in the SCADA data and the predicted data of the optimization model, and the fault warning chart, determine the fault location of the wind turbine gearbox, specifically:

[0036] Calculate the residual error between the test set data and the predicted data of the optimization model at each time point, and judge whether the residual error exceeds the fault threshold of the fault warning chart. If it exceeds, check the exceeded test set data;

[0037] In the AVAE_SDL model, the residual error is:

[0038]

[0039] Locate according to the residual error error. The SCADA data of the wind turbine contains multiple parameters. The residual error error in formula (8) is the deviation between the actual value and the predicted value of each parameter.

[0040] The SCADA fault warning and location system of the wind turbine based on AVAE_SDL includes:

[0041] A preprocessing module, which collects the SCADA data of the wind turbine gearbox, preprocesses the collected data, and then divides the preprocessed data into a training set and a test set;

[0042] A construction module, which constructs the AVAE_SDL model;

[0043] A training module, which inputs the training set into the AVAE_SDL model for training, obtains the optimization model and the abnormal score of the wind turbine fault degree, and determines the fault threshold;

[0044] An acquisition module, which obtains the predicted data of the optimization model based on the test set and the optimization model, and draws a fault warning chart;

[0045] A determination module, which determines the fault location of the wind turbine gearbox based on the residual error between each input parameter in the SCADA data and the predicted data of the optimization model, and the fault warning chart.

[0046] A terminal device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the above method are implemented.

[0047] A computer-readable storage medium stores a computer program. When the computer program is executed by a processor, the steps of the above method are implemented.

[0048] Compared with the prior art, the present invention has the following beneficial effects:

[0049] In the present invention, the training set is input into the AVAE_SDL model for training to obtain the optimized model and the anomaly score of the fault degree of the wind turbine, and the fault threshold is determined; based on the test set and the optimized model, the prediction data of the optimized model is obtained, and a fault warning graph is drawn; based on the residual error between each input parameter in the SCADA data and the prediction data of the optimized model, and the fault warning graph, the fault location of the wind turbine gearbox is determined. The present invention combines dictionary learning and adversarial variational autoencoders to improve the stability of the warning model; at the same time, the present invention uses the residual error to implement the method of SCADA data fault location; the present invention can significantly improve the SCADA fault warning accuracy rate, realize early warning in advance, and can realize the fault location of the wind turbine generator set. Description of the Drawings

[0050] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention, and therefore should not be regarded as limiting the scope. For those of ordinary skill in the art, other related drawings can be obtained based on these drawings without creative efforts.

[0051] Figure 1 It is a schematic flowchart of a method for SCADA fault warning and location of a wind turbine based on AVAE_SDL of the present invention;

[0052] Figure 2 It is another schematic flowchart of a method for SCADA fault warning and location of a wind turbine based on AVAE_SDL of the present invention;

[0053] Figure 3 It is a structural diagram of the AVAE_SDL model of the present invention;

[0054] Figure 4 It is the effect diagram of the fault warning of unit No. 24;

[0055] Figure 5 It is the effect diagram of the health warning of unit No. 26;

[0056] Figure 6 It is a fault location diagram for the fault of Unit 24;

[0057] Figure 7 It is a schematic structural diagram of the SCADA fault early warning and location system for wind turbines based on AVAE_SDL of the present invention. Specific implementation manners

[0058] To make the objectives, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some but not all of the embodiments of the present invention. Usually, the components of the embodiments of the present invention described and illustrated in the accompanying drawings here can be arranged and designed in various different configurations.

[0059] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed present invention, but merely represents selected embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0060] It should be noted that similar reference numerals and letters denote similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings.

[0061] In the description of the embodiments of the present invention, it should be noted that if terms such as "upper", "lower", "horizontal", "inner", etc. indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, or the orientation or positional relationship in which the product of the invention is usually placed when in use, it is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus cannot be understood as a limitation to the present invention. In addition, terms such as "first", "second", etc. are only used for distinguishing descriptions and cannot be understood as indicating or implying relative importance.

[0062] In addition, if the term "horizontal" appears, it does not mean that the component is required to be absolutely horizontal, but it can be slightly inclined. For example, "horizontal" only means that its direction is more horizontal relative to "vertical", and does not mean that the structure must be completely horizontal, but it can be slightly inclined.

[0063] In the description of the embodiments of the present invention, it should also be noted that unless otherwise clearly specified and limited, if the terms "set", "installed", "connected", and "connected" appear, they should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the communication inside two components. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific situations.

[0064] The following further describes the present invention in detail with reference to the accompanying drawings:

[0065] See Figure 1 and Figure 2 , the present invention discloses a method for SCADA fault warning and location of wind turbines based on AVAE_SDL, including:

[0066] S101, collect the SCADA data of the wind turbine gearbox, preprocess the collected data, and then divide the preprocessed data into a training set and a test set;

[0067] The preprocessing includes data cleaning and normalization preprocessing; the data cleaning is to remove outliers from the collected data; outliers refer to data that deviates from the normal power curve due to reasons such as curtailment of wind power and incorrect wind speed measurement. Since the reasons for the generation of outliers are complex and random, which is not conducive to the training of the warning model, it is necessary to remove them. The pitch angle and active power are used to remove outliers, as shown in formula (1):

[0068]

[0069] In the formula, Pow is the active power, v is the generator speed, and θ is the pitch angle;

[0070] The normalization preprocessing is specifically: perform a linear transformation on the cleaned data to map the result to the range of [0, 1]; the normalization formula is:

[0071] X' = (X - X min ) / (X max - X min ) (2)

[0072] Among them, X is the original data, and Xmax and Xmin are the maximum and minimum values of the original data set respectively.

[0073] S102, construct an AVAE_SDL model.

[0074] See Figure 3, the AVAE_SDL model combines dictionary learning (SDL) and adversarial variational autoencoder (AVAE). SDL is used to extract important features of the latent encoding, and AVAE makes full use of the adversarial nature of the generative adversarial network (GAN) and the posterior learning ability of the variational autoencoder (VAE) to improve the accuracy of anomaly detection.

[0075] S103, Input the training set into the AVAE_SDL model for training, obtain the optimized model and the anomaly score of the fault degree of the wind turbine, and determine the fault threshold.

[0076] Update the encoder / decoder of the AVAE_SDL model, update the dictionary, and update the discriminator; The training set X is input into the variational autoencoder / decoder. The encoder outputs the mean μ and variance σ. Multiply the variance by random noise and add the mean to get Z I , Z I Input into the decoder to obtain the intermediate latent encoding Z S and the output result X’, the intermediate latent encoding Z S is input into dictionary learning to obtain the sparse coefficient matrix A and the dictionary matrix Dic. Update the loss function of the encoder / decoder as:

[0077]

[0078] where, is the reconstruction error of the encoder / decoder, is the KL divergence, ||Z S - Dic·A||2 + λA||1 is the loss function in dictionary learning; N is the size of the data, X is the input sample of the model, X’ is the output sample of the model, that is, the reconstructed sample, μ and σ are respectively the output mean and variance of the variational autoencoder VAE, that is, the intermediate variables;

[0079] Update the loss function of the dictionary as:

[0080] L SDL = ||Z S - Dic·A||2 + λA||1 (4)

[0081] where, ||·||2 is the second-order norm, ||·||1 is the first-order norm;

[0082] Z I and Z S are jointly input into the discriminator for adversarial discrimination, as shown in formula (5):

[0083]

[0084] The three updates are alternated until convergence below the specified error threshold.

[0085] Adopt score i The abnormal score representing the fault degree of the wind turbine generator set, and the calculation method is as follows:

[0086]

[0087] Among them, the data X and There are N rows and n columns, and i represents the row number; score i Consists of two parts. The first item is the reconstruction error of the adversarial variational autoencoder, and the other item is the error of sparse dictionary learning; ζ is a constant, set to 0.1; k is the number of parameters of the SCADA data.

[0088] The upper and lower limits of the fault threshold are:

[0089]

[0090] Among them, μ and ε are the mean and standard deviation of the training dataset fraction, and K is a constant, which is 1.25 times the maximum abnormal score of the training set.

[0091] S104, based on the test set and the optimized model, obtain the prediction data of the optimized model, and draw a fault warning graph.

[0092] The test set data X is input into the optimized model to obtain the prediction output of the optimized model and the intermediate variable Z S of the optimized model, A, Dic, and calculate the score of the test data, and draw an exponentially weighted moving average control chart EWMA, that is, a fault warning graph. When the data in the graph exceeds the fault threshold, it is determined that the wind turbine generator set has a fault and a warning is realized.

[0093] S105, based on the residual error between each input parameter in the SCADA data and the prediction data of the optimized model, and the fault warning graph, determine the fault location of the gearbox of the wind turbine generator set.

[0094] Calculate the residual error between the test set data and the prediction data of the optimized model at each time point, and judge whether the residual error exceeds the fault threshold of the fault warning graph. If it exceeds, check the exceeded test set data;

[0095] In the AVAE_SDL model, the residual error is:

[0096]

[0097] Positioning is realized according to the residual error error. The SCADA data of the wind turbine generator set contains multiple parameters, and the residual error error in formula (8) is the deviation between the actual value and the predicted value of each parameter.

[0098] Embodiment:

[0099] Collect the data of two wind turbines for three years for comparison. The SCADA parameters described in Table 1 are input into both wind turbines. The two wind turbines are Unit 24 and Unit 26 respectively. Preprocess the data of Unit 24 to obtain the training set and the test set, and then input the training set and the test set into the AVAE_SDL model for training in sequence. The upper limit of the fault threshold obtained is 0.25, and the lower limit of the fault threshold is 0.05. As Figure 4 shown, early warning was achieved for Unit 24 before the fault occurred.

[0100] Preprocess the data of Unit 26 to obtain the training set and the test set, and then input the training set and the test set into the AVAE_SDL model for training in sequence. The upper limit of the fault threshold obtained is 0.26, and the lower limit of the fault threshold is 0.07. As Figure 5 shown, Unit 26 has always been within the normal range; at the same time, the positioning result of Unit 24 is as Figure 6 shown. It can be seen that the residual error of the parameters of the high-speed bearing of the gearbox is the largest and exceeds the threshold, and it is determined that there is a fault in the high-speed bearing of the gearbox.

[0101] Table 1 SCADA parameters input into the wind turbine

[0102]

[0103] See Figure 7 , the present invention discloses a wind turbine SCADA fault early warning and positioning system based on AVAE_SDL, including:

[0104] A preprocessing module, which collects the SCADA data of the gearbox of the wind turbine, preprocesses the collected data, and then divides the preprocessed data into a training set and a test set;

[0105] A construction module, which constructs an AVAE_SDL model;

[0106] A training module, which inputs the training set into the AVAE_SDL model for training, obtains the optimized model and the abnormal score of the fault degree of the wind turbine, and determines the fault threshold;

[0107] An acquisition module, which obtains the prediction data of the optimized model based on the test set and the optimized model, and draws a fault early warning diagram;

[0108] A determination module, which determines the fault location of the gearbox of the wind turbine based on the residual error between each input parameter in the SCADA data and the prediction data of the optimized model, and the fault early warning diagram.

[0109] A terminal device provided by an embodiment of the present invention. The terminal device of this embodiment includes: a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps in the above-mentioned method embodiments are implemented. Alternatively, when the processor executes the computer program, the functions of each module / unit in the above-mentioned device embodiments are implemented.

[0110] The computer program can be divided into one or more modules / units, and the one or more modules / units are stored in the memory and executed by the processor to complete the present invention.

[0111] The terminal device can be a computing device such as a desktop computer, a notebook, a palm computer, and a cloud server. The terminal device may include, but is not limited to, a processor and a memory.

[0112] The processor may be a central processing unit (CPU), or may also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.

[0113] The memory can be used to store the computer program and / or module, and the processor realizes various functions of the terminal device by running or executing the computer program and / or module stored in the memory, and by calling the data stored in the memory.

[0114] If the modules / units integrated in the terminal device are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, to implement all or part of the processes in the above-described embodiment methods of the present invention, it can also be completed by a computer program instructing relevant hardware. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, the steps of the above-described various method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in the form of source code, object code, executable file, or some intermediate form, etc. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electrical carrier signal, telecommunication signal, and software distribution medium, etc. It should be noted that the content included in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer-readable medium does not include electrical carrier signals and telecommunication signals.

[0115] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for SCADA fault warning and location of wind turbines based on AVAE_SDL, characterized in that Including: Collect the SCADA data of the wind turbine gearbox, preprocess the collected data, and then divide the preprocessed data into a training set and a test set; Construct an AVAE_SDL model; Input the training set into the AVAE_SDL model for training to obtain an optimized model and an anomaly score of the fault degree of the wind turbine, and determine the fault threshold. The process of inputting the training set into the AVAE_SDL model for training to obtain the optimized model is specifically as follows: Update the encoder / decoder of the AVAE_SDL model, update the dictionary, and update the discriminator; training set X In the input variational autoencoder / decoder, the encoder outputs the mean μ and variance σ , the variance is multiplied by random noise and added to the mean to obtain Z I , Z I Input into the decoder to obtain the intermediate latent code Z S and the output result X’ , the intermediate latent code Z S Is input into dictionary learning to obtain the sparse coefficient matrix A and the dictionary matrix Dic , the loss function for updating the encoder / decoder is: wherein, is the reconstruction error of the encoder / decoder, is the KL divergence, is the loss function in dictionary learning; N is the size of the data, X is the input sample of the model, X’ is the output sample of the model, i.e., the reconstructed sample, μ and σ are the output mean and variance of the variational autoencoder VAE, i.e., the intermediate variables; The loss function for updating the dictionary is: Where, ||·||2 is the second-order norm, and ||·||1 is the first-order norm; Z I And Z S are jointly input into a discriminator for adversarial discrimination as follows: The three updates are carried out alternately until convergence below the specified error threshold; Based on the test set and the optimized model, obtain the prediction data of the optimized model and draw a fault warning diagram; Based on the residual error between each input parameter in the SCADA data and the prediction data of the optimized model, and the fault warning diagram, determine the fault location of the wind turbine gearbox; The process of determining the fault location of the wind turbine gearbox based on the residual error between each input parameter in the SCADA data and the prediction data of the optimized model, and the fault warning diagram is specifically as follows: Calculate the residual error between the test set data and the prediction data of the optimized model at each time point, and judge whether the residual error exceeds the fault threshold of the fault warning diagram. If it exceeds, check the exceeded test set data; In the AVAE_SDL model, the residual error is: According to the residual error error Realize positioning. The SCADA data of the wind turbine contains multiple parameters, and the residual error error is the deviation between the actual value and the predicted value of each parameter.

2. The method for wind turbine SCADA fault early warning and location based on AVAE_SDL according to claim 1, wherein The process of preprocessing the collected data is specifically: data cleaning and normalization preprocessing; the data cleaning eliminates outliers from the collected data; eliminates outliers from the pitch angle and active power, as shown in formula (1): In the formula, Pow is the active power, v is the generator speed, and θ is the pitch angle; The normalization preprocessing is specifically: perform a linear transformation on the cleaned data to map the result to the range of [0, 1]; the normalization formula is: Where, X is the original data, and Xmax and Xmin are the maximum and minimum values of the original data set respectively.

3. The method for SCADA fault early warning and positioning of wind turbine based on AVAE_SDL according to claim 1, wherein The abnormal score for obtaining the fault degree of the wind turbine is specifically: Using score i The abnormal score representing the fault degree of the wind turbine, and the calculation method is: Among them, the data X and have N rows and n columns, i representing the number of rows; score i consists of two parts. The first item is the reconstruction error of the adversarial variational autoencoder, and the other item is the error of sparse dictionary learning; ζ is a constant, set to 0.1; k is the number of parameters of the SCADA data.

4. The method for wind turbine SCADA fault early warning and location based on AVAE_SDL according to claim 1, characterized in that The upper and lower limits of the fault threshold are: wherein, μ and ε are the mean and standard deviation of the training dataset scores, K is a constant and is 1.25 times the maximum anomaly score of the training set.

5. The method for SCADA fault early warning and location of wind turbines based on AVAE_SDL according to claim 1, characterized in that The process of obtaining the prediction data of the optimized model based on the test set and the optimized model and drawing a fault warning diagram is specifically as follows: The test set data X is input into the optimization model to obtain the predicted output of the optimization model and the intermediate variables of the optimization model Z S 、A、Dic , and calculate the score of the test data, draw the exponentially weighted moving average control chart EWMA, that is, the fault warning chart. When the data in the chart exceeds the fault threshold, it is determined that the wind turbine has a fault and early warning is realized.

6. The SCADA fault warning and positioning system for wind turbines based on AVAE_SDL is characterized in that Including: A preprocessing module that collects the SCADA data of the wind turbine gearbox, preprocesses the collected data, and then divides the preprocessed data into a training set and a test set; A construction module that constructs an AVAE_SDL model; A training module that inputs the training set into the AVAE_SDL model for training to obtain an optimized model and an anomaly score of the fault degree of the wind turbine, and determine the fault threshold. The process of inputting the training set into the AVAE_SDL model for training to obtain the optimized model is specifically as follows: Update the encoder / decoder of the AVAE_SDL model, update the dictionary, and update the discriminator; training set X In the input variational autoencoder / decoder, the encoder outputs the mean μ and variance σ , the variance is multiplied by random noise and added to the mean to obtain Z I , Z I The intermediate latent code is obtained by inputting into the decoder Z S and the output result X’ , the intermediate latent code Z S is input into dictionary learning to obtain the sparse coefficient matrix A and the dictionary matrix Dic , the loss function for updating the encoder / decoder is:[[]] Among them, is the reconstruction error of the encoder / decoder, is the KL divergence, is the loss function in dictionary learning; N is the size of the data, X is the input sample of the model, X’ is the output sample of the model, that is, the reconstructed sample, μ and σ are the output mean and variance of the variational autoencoder VAE, that is, the intermediate variables, respectively; The loss function for updating the dictionary is: Where, ||·||2 is the second-order norm, and ||·||1 is the first-order norm; Z I And Z S They are jointly input into a discriminator for adversarial discrimination as follows: The three updates are carried out alternately until convergence below the specified error threshold; An acquisition module, which acquires prediction data of an optimization model based on a test set and the optimization model, and draws a fault warning diagram; A determination module, which determines the fault location of the wind turbine gearbox based on the residual error between each input parameter in the SCADA data and the prediction data of the optimization model, and the fault warning diagram; The step of determining the fault location of the wind turbine gearbox based on the residual error between each input parameter in the SCADA data and the prediction data of the optimization model, and the fault warning diagram is specifically: Calculate the residual error between the test set data and the prediction data of the optimization model at each time point, and determine whether the residual error exceeds the fault threshold of the fault warning diagram. If it exceeds, check the exceeded test set data; In the AVAE_SDL model, the residual error is: According to the residual error error Positioning is achieved. The SCADA data of the wind turbine contains multiple parameters, and the residual error error is the deviation between the actual value and the predicted value of each parameter.

7. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1-5.

8. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the method according to any one of claims 1-5.

Images