Doubly-fed fan grid-connected fault detection system based on data driving

Abstract

The invention discloses a doubly-fed fan grid-connected fault detection system based on data driving, which belongs to the field of power grid system fault detection and comprises a sensor detection module, an edge calculation module, a data processing module, a central integration module, a fault detection module and a decision module. By introducing a distributed data processing architecture, the problems of real-time response and high-dimensional data processing caused by complexity of a doubly-fed fan system are solved, the influence of frequency fluctuation of a power grid on the system can be remarkably reduced, the pressure and load of the power grid are reduced, and the response speed and flexibility of the system are improved. Through the multi-stage processing and edge computing technology, a fault detection function is realized, faults can be timely and effectively detected, the possibility of fault occurrence is reduced, and the maintenance and operation cost of the system is reduced, so that the reliability and economical efficiency of the overall operation of the wind power plant are improved.

Description

Technical Field

[0001] The present invention relates to the field of power grid system fault detection, and in particular to a data-driven double-fed wind turbine grid-connected fault detection system. Background Art

[0002] To achieve the "dual carbon" goal, China urgently needs to improve the flexibility and stability of wind power grid connection. Doubly-fed wind turbine generators (DFIGs) have become an important choice for wind power development due to their high efficiency and good grid connection characteristics. However, the various fault risks faced by DFIGs during the grid connection process, such as voltage fluctuations and frequency anomalies, pose challenges to the safety of their operation and the stability of the power grid. To this end, it is particularly necessary to improve fault detection and management technologies.

[0003] Traditional fault detection methods mainly rely on analysis based on physical models, such as current and voltage feature analysis and state feedback of control systems. However, these methods may not be able to detect faults quickly and accurately in the face of complex and nonlinear power grid environments, especially for potential hidden faults or complex fault modes, traditional methods have limitations. Summary of the invention

[0004] The purpose of the present invention is to solve the defects in the prior art and to propose a data-driven double-fed wind turbine grid-connected fault detection system.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] In a first aspect, the present invention provides a data-driven double-fed wind turbine grid-connected fault detection system, the system comprising a sensor detection module, an edge computing module, a data processing module, a central integration module, a fault detection module, and a decision module;

[0007] The sensor detection module collects relevant electrical parameters and mechanical parameters of the fan by sensors distributed in different parts of the system;

[0008] The edge computing module is placed close to the sensor for local rapid preliminary processing and fault diagnosis;

[0009] The specific processing method of the edge computing module is as follows:

[0010] Step S1: The edge computing module first performs noise reduction and filtering on the electrical and mechanical parameters collected by the sensor, and removes noise and high-frequency interference in the data through technologies such as wavelet transform or Kalman filtering;

[0011] Step S2: The processed data is lightweight processed by feature extraction and dimensionality reduction algorithms to extract key features, reduce redundant information, and reduce data transmission and computing load;

[0012] Step S3: Based on the feature data after noise reduction and lightweight processing, the system uses a lightweight machine learning model to locate the fault and predict the type, quickly identify the anomaly, and output preliminary fault information;

[0013] The data processing module processes the feature data returned by the edge computing module and further analyzes the system as a whole;

[0014] The central integration module integrates the transmitted data into a global view;

[0015] The fault detection module is used to analyze complex faults based on a global view and transmit fault information to the decision module;

[0016] The decision module is used by operation and maintenance personnel to monitor the system operating status, configuration parameters, and perform operations and adjustments.

[0017] In a possible implementation, the specific steps of fault detection of the fault detection module are as follows:

[0018] Step V1: First, perform preliminary abnormality detection based on the data transmitted to the detection module after processing. Perform preliminary abnormality detection on the processed data through a threshold detection algorithm. For example, mark the instantaneous drop in voltage and the sudden increase in current as abnormal phenomena. Secondly, quickly locate the possible fault area based on the preset fault judgment rules, match the abnormal parameters with the corresponding equipment or system parts, and preliminarily determine the area where the fault occurs.

[0019] Step V2: According to the result of the preliminary positioning, it is passed to the upper-level processing unit, and the historical data comparison module compares it with the historical data to further confirm whether the fault actually occurred. If the fault location is uncertain or suspected to be a multi-point fault, the relevant data is re-checked and a secondary screening is performed to finally confirm the fault location;

[0020] Step V3: After confirming the fault location, perform more detailed feature extraction on the relevant data of the fault area, use a neural network to match the extracted feature quantity with the predefined fault type feature library, and judge the possible fault type based on the combination of different features. For example, abnormal current + abnormal voltage may indicate a short circuit fault; abnormal vibration + temperature may indicate bearing wear or mechanical jamming;

[0021] Step V4: The fault detection module will use a multivariate analysis method to analyze the correlation of multiple abnormal parameters, comprehensively determine the fault type, and combine the historical data of the fan operation, equipment characteristics and fault mode library to determine the severity and impact range of the fault type;

[0022] Step V5: Summarize and standardize the fault location, fault type, severity and other information, convert them into a format that the system can recognize, and transmit them to the central control layer through the system bus, marking the priority of the fault type and possible operation suggestions.

[0023] In a possible implementation, the fault area location algorithm mentioned in step V1 is as follows:

[0024] Step V11: monitor the sensor data through the threshold detection algorithm, preset the normal range of multiple parameters, monitor whether the real-time data exceeds these ranges, and preliminarily diagnose abnormal phenomena. For the grid connection detection of the doubly fed wind turbine system, the electrical parameters monitored here are as follows:

[0025] X(t)={I(t),U(t),P(t)},t=0,1,2,…T

[0026] Where I(t) is the current of the wind turbine at time t, U(t) is the voltage at that time, and P(t) is the power at that time.

[0027] When the monitored electrical parameters exceed the preset threshold range, the system will mark this moment as abnormal. The threshold monitoring rules are as follows:

[0028]

[0029] Step V12: In order to avoid the situation in which a parameter anomaly may cause parameter performance in multiple regions in some complex systems, Bayesian inference is introduced to calculate the probability of fault occurrence based on historical fault data and current detection data. The Bayesian formula is used to detect abnormal data and update the probability of fault region as follows:

[0030]

[0031] In the formula, P(R i ) is the initial probability of failure in each area set according to historical data, where R i represents region i, P(R i |data) is in region R i The probability of observing the current data when a failure occurs.

[0032] In a possible implementation, the fault type analysis algorithm mentioned in step V3 is as follows:

[0033] Step V31: By analyzing the electrical and mechanical signals collected during the operation of the doubly-fed wind turbine, characteristic variables that can represent the system status and potential faults are extracted; electrical signals include current and voltage, etc., which reflect the electrical operation status of the wind turbine. By extracting the time domain and frequency domain characteristics of these signals, electrical faults can be identified;

[0034] In the case of short circuit or overload, the instantaneous current fluctuation is abnormal. The maximum instantaneous value of the current signal is extracted. The formula is as follows:

[0035]

[0036] In the formula, I peak Indicates the maximum value of the current extracted in the entire time series;

[0037] The mean value can reflect the overall current level of the system and is used to capture the current change trend under long-term operation. The formula is as follows:

[0038]

[0039] Where T is the sampling period, I(t) is the current signal;

[0040] In order to extract the frequency components in the current signal, especially the high-frequency components, and to identify the electrical harmonic interference or transient faults in the system, the time domain signal is converted to the frequency domain using Fourier Transform, as follows:

[0041]

[0042] Where I(t) represents the intensity of the current signal at different frequencies f;

[0043] Mechanical signals include vibration and temperature data, which can reflect the mechanical status of the fan;

[0044] Step V32: After feature extraction, the feature quantity is matched with a predetermined fault type feature library. When dealing with complex and nonlinear problems, a neural network is used to match the extracted features with predefined fault types.

[0045] First, the extracted multidimensional features are input into the neural network. Each feature quantity is used as the input node of the network. The activation value of the first layer of neurons is generated by combining the input features through the weight matrix W and the bias term b. The formula is as follows:

[0046] z1=W1·F+b1

[0047] In the formula, F = [I peak ,U drop ,V amp ,T rate ] is the input feature, W1 is the weight matrix, and b1 is the bias;

[0048] Secondly, the activation function ReLU is used to perform nonlinear mapping on the input data, retaining positive values ​​and filtering negative values, helping the network to extract complex features related to the fault type layer by layer. The expression is:

[0049] ReLU(z)=max(0,z)

[0050] Combining them together forms a multi-layer perceptron that can extract features. The specific process is expressed as:

[0051] z=W·ReLU(z)+b

[0052] Step V33: After each layer is activated and weights are updated, the network gradually extracts patterns related to the fault type. Each layer of neurons generates the input of the next layer based on the output of the previous layer until the network reaches the output node of the last layer. The Softmax classifier is used in the last layer to convert the output layer results into the probabilities of different fault types. The expression is:

[0053]

[0054] Where P(y=i|z) represents the probability of fault type i, and zi is the output of the last layer of the neural network; the fault type with the highest probability is selected as the final output based on the output.

[0055] In one possible implementation, the multivariate analysis algorithm mentioned in step V4 is as follows:

[0056] Step V41: At this stage, we have obtained multi-dimensional input data for feature extraction. Through the multi-dimensional classification method of SVM, we further analyze the relationship between these features and determine the fault type. SVM distinguishes different fault categories by constructing an optimal hyperplane, which can map different feature combinations to different fault types. The mathematical description is as follows:

[0057] f(X)=sign(wTX+b)

[0058] Where w is the model weight vector, b is the bias value, and X = [I peak ,U drop ,V amp ,T rate ];

[0059] Step V42: The system adjusts w and b according to the training data to ensure that the input feature data X is correctly classified as a specific fault type, for example: peak and U drop If the fault is abnormal, the system will use SVM to determine that the most likely cause is a short circuit fault; V amp and T rate If both are abnormal, the system may classify it as a mechanical failure, such as bearing wear.

[0060] Step V43: By comparing the abnormality levels between different features, SVM can distinguish the severity of the fault. For a given data point xi, the distance calculation formula from the hyperplane is as follows:

[0061]

[0062] In the formula, w T x i +b represents the projection result of the feature point to the hyperplane, ||w|| is the norm of the weight vector, which is used for normalization; the distance d(x i ) is larger, which means that the data point is farther away from the classification boundary, indicating that the severity of the fault is higher. After the final fault judgment is made through SVM, the system generates a detailed fault report. The report content includes: fault type, fault severity, and potential impact range. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention.

[0064] Figure 1 A block diagram of a data-driven double-fed wind turbine grid-connected fault detection system proposed by the present invention;

[0065] Figure 2 This is a flow chart of a data-driven double-fed wind turbine grid-connected fault detection module proposed by the present invention. DETAILED DESCRIPTION

[0066] In order to better understand the technical solution of the present application, the embodiments of the present application are described in detail below with reference to the accompanying drawings.

[0067] An embodiment of a data-driven doubly-fed wind turbine grid-connected fault detection system applied to electrical fault detection, this embodiment demonstrates the fault detection process of the system when the wind turbine electrical operation is abnormal; first, the sensor detection module collects key electrical parameters such as current, voltage, and power, and the edge computing module reduces noise and filters the collected data, extracts features and performs lightweight processing; then, the system uses a lightweight machine learning model to quickly identify instantaneous abnormal current fluctuations and voltage drops, and locates them as potential short-circuit faults; the fault detection module further analyzes the feature data, combines historical data and feature libraries to confirm the possibility of short-circuit faults, and quantifies the severity of the faults through SVM multivariate analysis; finally, the decision module transmits the detected short-circuit fault location, severity and other information to the operation and maintenance personnel, and provides priority and emergency handling suggestions; it should be clear that the embodiments described are only some embodiments of the present application, and other embodiments not listed are also within the scope of protection of the present application.

[0068] Figure 1 The block diagram of a data-driven double-fed wind turbine grid-connected fault detection system proposed by the present invention is as follows: Figure 1 As shown, the system includes: a sensor detection module, an edge computing module, a data processing module, a central integration module, a fault detection module, and a decision module.

[0069] The sensor detection module collects relevant electrical and mechanical parameters of the fan through sensors distributed in different parts of the system;

[0070] The edge computing module is placed close to the sensor for local and rapid preliminary processing and fault diagnosis;

[0071] The specific processing method of the edge computing module is as follows:

[0072] Step S1: The edge computing module first performs noise reduction and filtering on the electrical and mechanical parameters collected by the sensor, and removes noise and high-frequency interference in the data through technologies such as wavelet transform or Kalman filtering;

[0073] Step S2: The processed data is lightweight processed by feature extraction and dimensionality reduction algorithms to extract key features, reduce redundant information, and reduce data transmission and computing load;

[0074] Step S3: Based on the feature data after noise reduction and lightweight processing, the system uses a lightweight machine learning model to locate the fault and predict the type, quickly identify the anomaly, and output preliminary fault information;

[0075] The data processing module processes the feature data returned by the edge computing module and further analyzes the system as a whole;

[0076] The central integration module integrates the transmitted data into a global view;

[0077] The fault detection module is used to analyze complex faults based on a global view and transmit fault information to the decision module;

[0078] The decision module is used by operation and maintenance personnel to monitor the system operating status, configuration parameters, and perform operations and adjustments.

[0079] Figure 2 This is a flow chart of a data-driven double-fed wind turbine grid-connected fault detection module proposed by the present invention, such as Figure 2As shown in the figure, this module collects the data processed by the integrated module and transmits the data to the fault detection module. First, the threshold detection algorithm is used to identify the abnormal area. Secondly, the probability is updated according to the Bayesian inference through the historical data provided by the system, and finally the fault area is identified. After the fault area is identified, the features are extracted according to the electrical and mechanical signals of the fault area, matched with the fault type, and multivariate in-depth analysis is performed based on the electrical and mechanical multi-dimensional features under the grid-connected operation state to clarify the fault type, location, and depth, and generate a fault report.

[0080] In the embodiment of the present invention, the specific steps of fault detection of the fault detection module are as follows:

[0081] Step V1: First, perform preliminary abnormality detection based on the data transmitted to the detection module after processing. Perform preliminary abnormality detection on the processed data through a threshold detection algorithm. For example, mark the instantaneous drop in voltage and the sudden increase in current as abnormal phenomena. Secondly, quickly locate the possible fault area based on the preset fault judgment rules, match the abnormal parameters with the corresponding equipment or system parts, and preliminarily determine the area where the fault occurs.

[0082] Step V2: According to the result of the preliminary positioning, it is passed to the upper-level processing unit, and the historical data comparison module compares it with the historical data to further confirm whether the fault actually occurred. If the fault location is uncertain or suspected to be a multi-point fault, the relevant data is re-checked and a secondary screening is performed to finally confirm the fault location;

[0083] Step V3: After confirming the fault location, perform more detailed feature extraction on the relevant data of the fault area, use a neural network to match the extracted feature quantity with the predefined fault type feature library, and judge the possible fault type based on the combination of different features. For example, abnormal current + abnormal voltage may indicate a short circuit fault; abnormal vibration + temperature may indicate bearing wear or mechanical jamming;

[0084] Step V4: The fault detection module will use a multivariate analysis method to analyze the correlation of multiple abnormal parameters, comprehensively determine the fault type, and combine the historical data of the fan operation, equipment characteristics and fault mode library to determine the severity and impact range of the fault type;

[0085] Step V5: Summarize and standardize the fault location, fault type, severity and other information, convert them into a format that the system can recognize, and transmit them to the central control layer through the system bus, marking the priority of the fault type and possible operation suggestions.

[0086] In the embodiment of the present invention, the fault area location algorithm mentioned in step V1 is as follows:

[0087] Step V11: monitor the sensor data through the threshold detection algorithm, preset the normal range of multiple parameters, monitor whether the real-time data exceeds these ranges, and preliminarily diagnose abnormal phenomena. For the grid connection detection of the doubly fed wind turbine system, the electrical parameters monitored here are as follows:

[0088] X(t)={I(t),U(t),P(t)},t=0,1,2,…T

[0089] Where I(t) is the current of the wind turbine at time t, U(t) is the voltage at that time, and P(t) is the power at that time.

[0090] When the monitored electrical parameters exceed the preset threshold range, the system will mark this moment as abnormal. The threshold monitoring rules are as follows:

[0091]

[0092] Step V12: In order to avoid the situation in which a parameter anomaly may cause parameter performance in multiple regions in some complex systems, Bayesian inference is introduced to calculate the probability of fault occurrence based on historical fault data and current detection data. The Bayesian formula is used to detect abnormal data and update the probability of fault region as follows:

[0093]

[0094] In the formula, P(R i ) is the initial probability of failure in each area set according to historical data, where R i represents region i, P(R i |data) is in region R i The probability of observing the current data when a failure occurs.

[0095] In the embodiment of the present invention, the fault type analysis algorithm mentioned in step V3 is as follows:

[0096] Step V31: By analyzing the electrical and mechanical signals collected during the operation of the doubly-fed wind turbine, characteristic variables that can represent the system status and potential faults are extracted; electrical signals include current and voltage, etc., which reflect the electrical operation status of the wind turbine. By extracting the time domain and frequency domain characteristics of these signals, electrical faults can be identified;

[0097] In the case of short circuit or overload, the instantaneous current fluctuation is abnormal. The maximum instantaneous value of the current signal is extracted. The formula is as follows:

[0098]

[0099] In the formula, I peak Indicates the maximum value of the current extracted in the entire time series;

[0100] The mean value can reflect the overall current level of the system and is used to capture the current change trend under long-term operation. The formula is as follows:

[0101]

[0102] Where T is the sampling period, I(t) is the current signal;

[0103] In order to extract the frequency components in the current signal, especially the high-frequency components, and to identify the electrical harmonic interference or transient faults in the system, the time domain signal is converted to the frequency domain using Fourier Transform, as follows:

[0104]

[0105] Where I(t) represents the intensity of the current signal at different frequencies f;

[0106] Mechanical signals include vibration and temperature data, which can reflect the mechanical status of the fan;

[0107] Step V32: After feature extraction, the feature quantity is matched with a predetermined fault type feature library. When dealing with complex and nonlinear problems, a neural network is used to match the extracted features with predefined fault types.

[0108] First, the extracted multidimensional features are input into the neural network. Each feature quantity is used as the input node of the network. The activation value of the first layer of neurons is generated by combining the input features through the weight matrix W and the bias term b. The formula is as follows:

[0109] z1=W1·F+b1

[0110] In the formula, F = [I peak ,U drop ,V amp ,T rate ] is the input feature, W1 is the weight matrix, and b1 is the bias;

[0111] Secondly, the activation function ReLU is used to perform nonlinear mapping on the input data, retaining positive values ​​and filtering negative values, helping the network to extract complex features related to the fault type layer by layer. The expression is:

[0112] ReLU(z)=max(0,z)

[0113] Combining them together forms a multi-layer perceptron that can extract features. The specific process is expressed as:

[0114] z=W·ReLU(z)+b

[0115] Step V33: After each layer is activated and weights are updated, the network gradually extracts patterns related to the fault type. Each layer of neurons generates the input of the next layer based on the output of the previous layer until the network reaches the output node of the last layer. The Softmax classifier is used in the last layer to convert the output layer results into the probabilities of different fault types. The expression is:

[0116]

[0117] Where P(y=i|z) represents the probability of fault type i, and zi is the output of the last layer of the neural network; the fault type with the highest probability is selected as the final output based on the output.

[0118] In the embodiment of the present invention, the multivariate analysis algorithm mentioned in step V4 is as follows:

[0119] Step V41: At this stage, we have obtained multi-dimensional input data for feature extraction. Through the multi-dimensional classification method of SVM, we further analyze the relationship between these features and determine the fault type. SVM distinguishes different fault categories by constructing an optimal hyperplane, which can map different feature combinations to different fault types. The mathematical description is as follows:

[0120] f(X)=sign(wTX+b)

[0121] Where w is the model weight vector, b is the bias value, and X = [I peak ,U drop ,V amp ,T rate ];

[0122] Step V42: The system adjusts w and b according to the training data to ensure that the input feature data X is correctly classified as a specific fault type, for example: peak and U drop If the fault is abnormal, the system will use SVM to determine that the most likely cause is a short circuit fault; V amp and T rate If both are abnormal, the system may classify it as a mechanical failure, such as bearing wear.

[0123] Step V43: By comparing the abnormality levels between different features, SVM can distinguish the severity of the fault. For a given data point xi, the distance calculation formula from the hyperplane is as follows:

[0124]

[0125] In the formula, w T x i+b represents the projection result of the feature point to the hyperplane, ||w|| is the norm of the weight vector, which is used for normalization; the distance d(x i ) is larger, which means that the data point is farther away from the classification boundary, indicating that the severity of the fault is higher. After the final fault judgment is made through SVM, the system generates a detailed fault report. The report content includes: fault type, fault severity, and potential impact range.

Claims

1. A data-driven double-fed wind turbine grid-connected fault detection system, characterized in that: It includes a sensor detection module, an edge computing module, a data processing module, a central integration module, a fault detection module, and a decision-making module; The sensor detection module consists of sensors distributed in different parts of the system, which are used to collect relevant electrical parameters and mechanical parameters of the fan; The edge computing module is placed close to the sensor for local and rapid preliminary processing and fault diagnosis; The specific processing method of the edge computing module is as follows: Step S1: The edge computing module first performs noise reduction and filtering on the electrical and mechanical parameters collected by the sensor. Aiming at the common harmonic interference and high-frequency oscillation problems of the grid-connected system, combined with the power electronic characteristics of the doubly-fed wind turbine, an improved Kalman filtering algorithm and multi-scale wavelet transform are used to remove noise; Step S2: The processed data is lightweight processed by feature extraction and dimensionality reduction algorithm to extract key feature quantities; Step S3: Based on the feature data after noise reduction and lightweight processing, the system adopts a lightweight machine learning model and combines the dynamic change trend of the grid-connected parameters to achieve preliminary positioning and type prediction of the fault; The data processing module processes the feature data returned by the edge computing module and further analyzes the system as a whole; The central integration module integrates the transmitted data into a global view; The fault detection module is used to analyze complex faults based on a global view and transmit fault information to the decision module; The decision module is used by operation and maintenance personnel to monitor the system operating status, configuration parameters, and perform operations and adjustments.

2. The data-driven double-fed wind turbine grid-connected fault detection system according to claim 1, characterized in that: The specific steps of fault detection of the fault detection module are as follows: Step V1: Based on the processed data transmitted to the detection module, a dynamic threshold detection algorithm is used to perform preliminary abnormality detection, especially for the fluctuation of grid-connected voltage and current, combined with the preset fault judgment rules, to quickly locate the area where the fault occurs and associate it with the corresponding system parts; Step V2: After the initial location is determined, further confirm whether the fault actually occurred by comparing it with historical data; If the fault location is unclear or there is a suspected multi-point fault, the data will be cross-checked and screened again to finally confirm the fault location and reduce the chance of misjudgment; Step V3: After confirming the fault location, the system performs more detailed feature extraction, combines the power electronics and mechanical characteristics of the doubly fed wind turbine, extracts the feature quantity, and performs classification matching through the neural network intelligent algorithm, matching the fault type with the preset feature library; Step V4: The fault detection module uses a support vector machine to perform multivariate analysis, combines the electrical and mechanical multidimensional characteristics under the grid-connected operation state, determines the fault type and its severity, and generates a comprehensive report to provide accurate fault nature and potential impact range; Step V5: The information on fault type, location, and severity is standardized and transmitted to the central control layer, with the fault type priority and operation suggestions marked to help operation and maintenance personnel make decisions.

3. The data-driven double-fed wind turbine grid-connected fault detection system according to claim 2, characterized in that: The fault location algorithm mentioned in step V1 is as follows: Step V11: Real-time monitoring of grid-connected electrical parameters is performed through a threshold detection algorithm. When the monitored electrical parameters exceed a preset threshold range, the system will mark the current moment as abnormal. Step V12: To avoid misdiagnosis caused by multi-parameter anomalies, Bayesian inference is combined with historical data to calculate the probability of fault occurrence and locate the fault area to improve detection accuracy.

4. The data-driven double-fed wind turbine grid-connected fault detection system according to claim 2, characterized in that: The analysis algorithm of the fault type mentioned in step V3 is as follows: Step V31: extract time domain and frequency domain features from electrical and mechanical signals during the grid-connected process of the doubly-fed wind turbine, analyze the harmonic components in the electrical signals in combination with Fourier transform, and reflect the instantaneous fault characteristics of the system; Step V32: After feature extraction, the extracted multidimensional features are classified through a neural network, and the extracted multidimensional features are input into the neural network. Each feature quantity is used as an input node of the network. The activation value of the first layer of neurons is generated by combining the input features with the weight matrix W and the bias term b. Secondly, the activation function ReLU is used to perform nonlinear mapping on the input data, retaining positive values ​​and filtering negative values, helping the network to extract complex features related to the fault type layer by layer. Combining them together constitutes a multi-layer perceptron that can extract features. Step V33: After each layer is activated and the weights are updated, the network gradually extracts patterns related to the fault type. Each layer of neurons generates the input of the next layer based on the output of the previous layer until the network reaches the output node of the last layer. The Softmax classifier is used in the last layer to convert the output layer results into the probabilities of different fault types. The fault type with the highest probability is selected as the final output based on the output.

5. The data-driven double-fed wind turbine grid-connected fault detection system according to claim 2, characterized in that: The multivariate analysis algorithm proposed in step V4 is as follows: Step V41: Use SVM to build a model and generate an optimal hyperplane through training data. The hyperplane is used to distinguish different types of faults in the feature space, ensuring that different fault categories can be accurately separated in the high-dimensional space. Step V42, SVM maps the input data points into the feature space according to different combinations of features, and classifies the data points according to the optimal hyperplane; In step V43, SVM not only classifies the fault type, but also evaluates the severity of the fault by calculating the distance of the feature data relative to the hyperplane, and generates a detailed fault report based on the SVM classification and evaluation results.

Images