Wind turbine generator fault detection method, device and equipment and readable storage medium

Abstract

The invention discloses a wind turbine generator fault detection method, device and equipment and a readable storage medium, and relates to the technical field of wind turbine generator fault detection.The wind turbine generator fault detection method comprises the steps that multi-source operation data of a wind turbine generator are obtained, and the multi-source operation data at least comprise vibration data, temperature data and electrical data; determining a multi-dimensional feature vector of the wind turbine generator according to the multi-source operation data; and determining a fault diagnosis result of the wind turbine generator according to the multi-dimensional feature vector and a preset fault diagnosis model. The method and the device aim at improving the accuracy of fault detection of the wind turbine generator.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of wind turbine fault detection, and in particular to a wind turbine fault detection method, device, equipment and readable storage medium. BACKGROUND

[0002] With the development of wind power industry towards large-scale and intelligent direction, 10MW land-based doubly-fed wind turbine has become the main model of land-based wind power due to its large single capacity, high power generation efficiency and low cost per kilowatt. However, the large-scale of wind turbine also brings problems such as the increase of structural complexity, the deterioration of load condition and the enhancement of system coupling, which leads to a significant increase in the failure risk of key components such as transmission chain.

[0003] Traditional fault detection methods mainly rely on manual inspection, threshold alarm and simple statistical analysis. When facing multi-source heterogeneous operating data (such as vibration, temperature and electrical data) of super-capacity units, there are problems such as low data utilization, insufficient feature extraction and poor working condition adaptability, which leads to poor accuracy of wind turbine fault detection, making it difficult to achieve accurate identification and early warning of early and weak faults, and unable to meet the urgent needs of modern wind farm intelligent operation for high reliability and low operation cost.

[0004] Therefore, how to improve the accuracy of wind turbine fault detection is a technical problem to be solved at present. SUMMARY

[0005] The main purpose of the present application is to provide a wind turbine fault detection method, device, equipment and readable storage medium, which aims to improve the accuracy of wind turbine fault detection.

[0006] To achieve the above purpose, the present application provides a wind turbine fault detection method, which comprises: obtaining multi-source operating data of a wind turbine, the multi-source operating data at least comprising vibration data, temperature data and electrical data; determining a multi-dimensional feature vector of the wind turbine according to the multi-source operating data; determining a fault diagnosis result of the wind turbine according to the multi-dimensional feature vector and a preset fault diagnosis model.

[0007] In an embodiment, the step of determining the multi-dimensional feature vector of the wind turbine according to the multi-source operating data comprises: performing feature extraction according to the vibration data to obtain vibration features of the wind turbine; performing feature extraction according to the temperature data to obtain temperature features of the wind turbine; performing feature extraction according to the electrical data to obtain electrical features of the wind turbine; The vibration feature, the temperature feature, and the electrical feature are weighted to obtain a multi-dimensional feature vector of the wind turbine.

[0008] In an embodiment, the fault diagnosis model comprises a fault type database storing a standard fault feature vector corresponding to each wind turbine fault type, and the step of determining the fault diagnosis result of the wind turbine according to the multi-dimensional feature vector and a preset fault diagnosis model comprises: The multi-dimensional feature vector is model trained according to the fault diagnosis model to obtain a fault probability corresponding to each wind turbine fault type; The wind turbine fault type corresponding to the probability value with the largest value in all the fault probabilities is determined as the predicted fault type of the wind turbine; The fault diagnosis result of the wind turbine is determined according to the predicted fault type.

[0009] In an embodiment, the step of determining the fault diagnosis result of the wind turbine according to the predicted fault type comprises: The probability value with the largest value is determined as a first confidence of the predicted fault type; A standard fault feature vector corresponding to the predicted fault type is obtained from the fault type database as a target fault feature vector, and a second confidence of the predicted fault type is determined according to the multi-dimensional feature vector and the target fault feature vector; A comprehensive confidence of the predicted fault type is determined according to the first confidence and the second confidence, and the fault diagnosis result of the wind turbine is determined according to the comprehensive confidence.

[0010] In an embodiment, the step of determining the fault diagnosis result of the wind turbine according to the comprehensive confidence comprises: It is detected whether the comprehensive confidence exceeds a preset confidence threshold; If the comprehensive confidence exceeds the confidence threshold, the predicted fault type is determined as the fault diagnosis result of the wind turbine; If the comprehensive confidence does not exceed the confidence threshold, the fault diagnosis model is updated, the updated fault diagnosis model is taken as a next preset fault diagnosis model, and the step of determining the fault diagnosis result of the wind turbine according to the multi-dimensional feature vector and a preset fault diagnosis model is executed again.

[0011] In an embodiment, the step of updating the fault diagnosis model according to the fault diagnosis model comprises: An incremental training sample of the fault diagnosis model is acquired, and the incremental training sample is composed of the multi-dimensional feature vector when the fault diagnosis model is triggered to update; A model loss parameter is determined according to the incremental training sample and historical training samples of the fault diagnosis model, and the fault diagnosis model is updated according to the model loss parameter to obtain an updated fault diagnosis model.

[0012] In an embodiment, the step of determining the second confidence of the predicted fault type according to the multi-dimensional feature vector and the target fault feature vector comprises: An algorithm processing is performed on the multi-dimensional feature vector and the target fault feature vector based on a fault diagnosis algorithm of the fault diagnosis model to obtain a cosine similarity and an Euclidean distance between the multi-dimensional feature vector and the target fault feature vector; The Euclidean distance is mapped to a Gaussian kernel similarity through a preset Gaussian kernel function, and a weighted processing is performed on the cosine similarity and the Gaussian kernel similarity to obtain the second confidence of the predicted fault type.

[0013] In addition, the present application also provides a wind turbine fault detection device, which comprises: An acquisition module is configured to acquire multi-source operation data of a wind turbine, wherein the multi-source operation data at least comprises vibration data, temperature data and electrical data; A feature extraction module is configured to determine a multi-dimensional feature vector of the wind turbine according to the multi-source operation data; A fault diagnosis module is configured to determine a fault diagnosis result of the wind turbine according to the multi-dimensional feature vector and a preset fault diagnosis model.

[0014] In addition, the present application also provides an electronic device, which comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the computer program is executed by the processor to implement the steps of the wind turbine fault detection method as described above.

[0015] In addition, the present application also provides a computer readable storage medium, wherein the computer readable storage medium stores a computer program, and the computer program is executed by a processor to implement the steps of the wind turbine fault detection method as described above.

[0016] The wind turbine fault detection method provided in the application obtains multi-source operation data with strict time sequence alignment and strong internal physical correlation by synchronously collecting vibration data, temperature data and electrical data of the wind turbine, so as to provide comprehensive and synchronous data basis for subsequent feature extraction; next, the multi-source operation data can be accurately obtained according to the multi-source operation data, so as to overcome the defects of insufficient feature representation ability of traditional single threshold or simple statistical model under variable speed and variable load conditions; finally, the multi-dimensional feature vector is input into the preset fault diagnosis model, so that the accurate discrimination of the wind turbine fault detection can be realized by processing the multi-dimensional feature vector through the preset fault diagnosis model, and the fault diagnosis result of the wind turbine can be quickly and accurately output, thereby significantly improving the accuracy of the wind turbine fault detection. BRIEF DESCRIPTION OF DRAWINGS

[0017] The accompanying drawings incorporated in and forming a part of the specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the application.

[0018] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the accompanying drawings required to be used in the embodiments or prior art description will be briefly introduced. Obviously, those skilled in the art can obtain other drawings from these drawings without creative labor.

[0019] Figure 1 is a flowchart of the first embodiment of the wind turbine fault detection method of the present application; Figure 2 is a flowchart of the second embodiment of the wind turbine fault detection method of the present application; Figure 3 is a flowchart of the wind turbine fault detection method related to the embodiment of the present application; Figure 4 is a structural diagram of the wind turbine fault detection device related to the embodiment of the present application; Figure 5 is a structural diagram of the electronic device related to the embodiment of the present application.

[0020] The implementation of the present application, functional characteristics and advantages will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION

[0021] With reference to the drawings, the technical solutions in the embodiments of the present application will be clearly and completely described below. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative effort are within the scope of protection of the present application.

[0022] It should be noted that if the embodiments of the present application involve directionality indications (such as up, down, left, right, front, back, etc.), the directionality indications are only used to explain the relative positional relationship, movement condition, etc. between components in a certain specific posture (as shown in the drawings), and if the specific posture changes, the directionality indications also change accordingly.

[0023] In addition, if the embodiments of the present application involve descriptions such as “first”, “second”, etc., the descriptions of “first”, “second”, etc. are only for description purposes, and cannot be understood as indicating or implying the relative importance of the indicated technical features or implicitly indicating the number of the indicated technical features. Therefore, the features limited by “first” and “second” can explicitly or implicitly include at least one of the features. In addition, the technical solutions of the various embodiments can be combined with each other, but it must be based on the fact that a person of ordinary skill in the art can realize it, and when the combination of technical solutions contradicts each other or cannot be realized, it should be considered that the combination of technical solutions does not exist and is not within the scope of protection claimed by the present application.

[0024] The exemplary embodiments will be described in detail herein below with reference to the accompanying drawings. Unless otherwise indicated, the same numbers on different drawings represent the same or similar elements. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application.

[0025] With the development of wind power industry towards large-scale and intelligentization, 10MW onshore doubly-fed wind turbine has become the main model of onshore wind power due to its large single capacity, high power generation efficiency, and low cost per kilowatt. However, the large-scale of wind turbine also brings problems such as increased structural complexity, deteriorated load conditions, and enhanced system coupling, which leads to a significant increase in the failure risk of key components such as the transmission chain.

[0026] Traditional fault detection methods mainly rely on manual inspection, threshold alarm and simple statistical analysis. When facing multi-source heterogeneous operating data (such as vibration, temperature and electrical data) of super-capacity units, there are problems such as low data utilization, insufficient feature extraction, poor working condition adaptability, etc., which leads to poor accuracy of wind turbine fault detection, making it difficult to achieve accurate identification and early warning of early weak faults, and unable to meet the urgent needs of modern wind farm intelligent operation for high reliability and low operation cost.

[0027] Therefore, improving the accuracy of wind turbine fault detection is a pressing technical problem that needs to be solved.

[0028] To address the shortcomings of the aforementioned wind turbine fault detection schemes, this application proposes a wind turbine fault detection method. The solution of this application's embodiment is as follows: The wind turbine fault detection method established in this application simultaneously collects vibration data, temperature data, and electrical data from the wind turbine to obtain multi-source operating data with strict time alignment and strong inherent physical correlation, providing a comprehensive and synchronous data foundation for subsequent feature extraction. Next, based on the multi-source operating data, multi-source operating data can be accurately obtained, thereby overcoming the shortcomings of traditional single-threshold or simple statistical models in terms of insufficient feature representation capabilities under varying speed and load conditions. Finally, the multi-dimensional feature vectors are input into a preset fault diagnosis model, enabling accurate identification of wind turbine faults through the processing of multi-dimensional feature vectors by the preset fault diagnosis model. This allows for rapid and accurate output of wind turbine fault diagnosis results, significantly improving the accuracy of wind turbine fault detection.

[0029] It should be noted that the executing entity in this embodiment can be a computing service device with data processing, network communication, and program execution functions, such as a tablet computer, personal computer, or mobile phone, or a device or electronic device capable of performing the above functions. The following description uses an electronic device as an example to illustrate this embodiment and the subsequent embodiments.

[0030] This application provides a method for detecting faults in wind turbine generators, referring to... Figure 1 As shown, Figure 1 This is a flowchart illustrating the first embodiment of the wind turbine fault detection method of this application. The wind turbine fault detection method provided in this application includes steps S10 to S30.

[0031] Step S10: Obtain multi-source operating data of the wind turbine, wherein the multi-source operating data includes at least vibration data, temperature data and electrical data.

[0032] In this embodiment, a sensor network and data acquisition system are used to synchronously collect multi-source heterogeneous operating data (i.e., multi-source operating data) of key components of the wind turbine (such as the main shaft, gearbox, generator, etc.), including vibration data, temperature data, and electrical data (such as current, voltage, power, etc.). This fundamentally changes the limitation of traditional fault detection methods that rely on a single data source, and realizes comprehensive and multi-physical field synchronous perception of the wind turbine's operating status, so as to provide a comprehensive and synchronous data foundation for subsequent feature extraction.

[0033] It should be noted that the multi-source operation data refers to a multi-element time series data set with different attributes and sampling frequencies collected from different subsystems and different physical quantity measuring points of the wind turbine.

[0034] In specific embodiments, for a 1.5 MW doubly-fed asynchronous wind turbine, vibration data (sampling frequency 10 kHz) is obtained from an acceleration sensor installed on the high-speed shaft of the gearbox, temperature data (sampling frequency 1 Hz) is obtained from PT100 temperature sensors embedded in the bearings of the gearbox and the generator winding, and electrical data such as three-phase current, voltage and active power (sampling frequency 1 Hz) is read in real time from the main control system of the unit; Next, after aligning the three types of data by time stamp, they jointly constitute the multi-source operation data of the wind turbine in a specific time window (such as the past 10 minutes).

[0035] Step S20: determining a multi-dimensional feature vector of the wind turbine according to the multi-source operation data.

[0036] In this embodiment, after data preprocessing according to the multi-source operation data, the multi-source operation data after data preprocessing is subjected to feature extraction to obtain a multi-dimensional feature vector that can comprehensively reflect the current operation health status of the wind turbine, thereby providing standardized and information-condensed high-quality input data for subsequent fault diagnosis models.

[0037] It should be noted that the multi-dimensional feature vector is a fixed-length numerical array formed by feature extraction and fusion of multi-source data such as vibration, temperature, and electrical data.

[0038] In specific embodiments, after data preprocessing of the multi-source operation data of the wind turbine in a specific time window, since the multi-source operation data at least includes vibration data, temperature data and electrical data, 8 vibration features such as vibration effective value, vibration peak value, vibration peak value interval, vibration kurtosis, margin factor, 1 times rotation frequency amplitude, 2 times rotation frequency amplitude and 3 times rotation frequency amplitude are extracted from the vibration data; 2 temperature features such as temperature mean value and temperature rise rate in a specific time window are extracted from the temperature data; 3 electrical features such as three-phase current, voltage and active power are extracted from the electrical data, and the vibration features, temperature features and electrical features are subjected to weighted processing to obtain a multi-dimensional feature vector that can comprehensively reflect the current operation health status of the wind turbine.

[0039] Step S30: determining a fault diagnosis result of the wind turbine according to the multi-dimensional feature vector and a pre-set fault diagnosis model.

[0040] In this embodiment, the multi-dimensional feature vector capable of comprehensively reflecting the current operation health status of the wind turbine is output to the pre-trained fault diagnosis model, so that the accurate discrimination of the wind turbine fault detection can be realized through the processing of the multi-dimensional feature vector by the pre-set fault diagnosis model, and then the fault diagnosis result of the wind turbine can be quickly and accurately output, thereby significantly improving the accuracy of the wind turbine fault detection.

[0041] It should be noted that the pre-set fault diagnosis model is a wind turbine fault output model trained based on a machine learning algorithm (such as support vector machine SVM, random forest, deep learning network).

[0042] In summary, the wind turbine fault detection method provided in the present application synchronously collects vibration data, temperature data and electrical data of the wind turbine to obtain multi-source operation data with strictly aligned time sequence and strong internal physical correlation, thereby providing a comprehensive and synchronous data basis for subsequent feature extraction. Next, the multi-source operation data can be accurately obtained according to the multi-source operation data, thereby overcoming the defects of insufficient feature representation ability of traditional single threshold or simple statistical model under variable speed and variable load conditions. Finally, the multi-dimensional feature vector is input into the pre-set fault diagnosis model, so that the accurate discrimination of the wind turbine fault detection can be realized through the processing of the multi-dimensional feature vector by the pre-set fault diagnosis model, and then the fault diagnosis result of the wind turbine can be quickly and accurately output, thereby significantly improving the accuracy of the wind turbine fault detection.

[0043] Further, based on the first embodiment of the wind turbine fault detection method of the present application, the second embodiment of the wind turbine fault detection method of the present application is proposed.

[0044] Further, in some other possible embodiments, the step S20 of determining the multi-dimensional feature vector of the wind turbine according to the multi-source operation data can further include steps S201 to S204.

[0045] Step S201: performing feature extraction according to the vibration data to obtain vibration features of the wind turbine.

[0046] In this embodiment, feature extraction is performed according to high-frequency vibration data, and vibration features representing damage and deterioration states of mechanical components (such as bearings and gears) in the wind turbine can be extracted. Specifically, time domain statistical analysis and frequency domain transformation analysis are performed on the vibration data, so that the non-stationary and complex vibration waveform of the vibration data itself can be converted into a set of vibration features with clear physical meaning (such as wear degree, impact strength and misalignment degree).

[0047] It should be noted that the vibration features can include vibration time domain features and vibration frequency domain features of the vibration data, the vibration time domain features can include vibration effective value, vibration peak value, vibration peak value interval, vibration kurtosis and margin factor, and the vibration frequency domain features can include 1 times rotation frequency amplitude, 2 times rotation frequency amplitude and 3 times rotation frequency amplitude. Among them, the vibration effective value is used to reflect the average energy of the vibration of the wind turbine; the vibration peak value is used to reflect the maximum instantaneous impact of the wind turbine; the vibration peak value interval is used to reflect the overall range of the vibration amplitude of the wind turbine; the vibration kurtosis is used to reflect the sharpness of the vibration signal distribution of the wind turbine; the margin factor is used to reflect the ratio of the vibration peak value to the vibration root amplitude, which is sensitive to surface damage type faults; the 1 times rotation frequency amplitude mainly reflects the unbalance or misalignment fault characteristics of the rotor in the wind turbine; the 2 times rotation frequency amplitude mainly reflects the misalignment fault of the wind turbine; the 3 times rotation frequency amplitude mainly reflects the possible faults related to other mechanical looseness or nonlinear vibration in the wind turbine.

[0048] Step S202: performing feature extraction according to the temperature data to obtain temperature features of the wind turbine.

[0049] In the embodiment, the temperature features reflecting the thermal state change of the wind turbine can be accurately extracted by performing feature extraction according to the temperature data of the wind turbine.

[0050] It should be noted that the temperature features are a set of numerical values for quantitatively describing the current level, stability and change trend of the temperature data. The temperature features usually include temperature mean value and temperature rise rate in a specific time window.

[0051] Step S203: performing feature extraction according to the electrical data to obtain electrical features of the wind turbine.

[0052] In the embodiment, the electrical features reflecting the health status of the electrical system of the wind turbine, such as the motor body, power electronic converter and grid connection, are extracted from the electrical data.

[0053] It should be noted that the electrical features are a set of numerical values for quantitatively describing the three-phase balance, waveform quality, power characteristics and energy conversion efficiency of the electrical system of the wind turbine. Common electrical features include current unbalance degree reflecting motor or grid asymmetry, total harmonic distortion rate reflecting waveform distortion, power factor reflecting reactive power situation, and standard deviation of power or current.

[0054] Step S204: performing weighted processing according to the vibration features, the temperature features and the electrical features to obtain a multi-dimensional feature vector of the wind turbine.

[0055] In the embodiment, the weighted vibration coefficient of the vibration feature, the weighted temperature coefficient of the temperature feature, and the weighted electrical coefficient of the electrical feature are determined, and the product between the vibration feature and the weighted vibration coefficient, the product between the temperature feature and the weighted temperature coefficient, and the product between the electrical feature and the weighted electrical coefficient are summed to obtain a multi-dimensional feature vector capable of comprehensively reflecting the current operation health status of the wind turbine, thereby providing standardized and information-condensed high-quality input data for the subsequent fault diagnosis model.

[0056] It should be noted that the weighted vibration coefficient, the weighted temperature coefficient, and the weighted electrical coefficient can be defined according to application requirements, and the present application does not make any limitation here.

[0057] Further, in some possible embodiments, the fault diagnosis model comprises a fault type database storing standard fault feature vectors corresponding to respective wind turbine fault types, and the step S30 of determining the fault diagnosis result of the wind turbine according to the multi-dimensional feature vector and the preset fault diagnosis model can further comprise the following implementation steps S301 to S303.

[0058] The step S301 comprises: performing model training on the multi-dimensional feature vector according to the fault diagnosis model to obtain a fault probability corresponding to each wind turbine fault type.

[0059] In the embodiment, after the multi-dimensional feature vector is input into the preset fault diagnosis model, a fault probability corresponding to each wind turbine fault type is calculated according to a softmax function of the fault diagnosis model, thereby providing an accurate and reliable fault probability value for subsequent prediction of the fault type of the wind turbine.

[0060] It should be noted that the wind turbine fault type can be a gearbox fault type, a main shaft and bearing fault type, a coupling fault type, a generator fault type, and a power conversion system fault type of the wind turbine.

[0061] The step S302 comprises: determining, as a predicted fault type of the wind turbine, a wind turbine fault type corresponding to a probability value with the largest value among all the fault probabilities.

[0062] In the embodiment, from all the fault probabilities output by the fault diagnosis model, a wind turbine fault type corresponding to a probability value with the largest value is selected as the predicted fault type of the wind turbine, thereby realizing conversion from a continuous probability space to a discrete fault label and providing an explicit and unambiguous preliminary diagnosis conclusion for the operation and maintenance personnel.

[0063] The step S303 comprises: determining the fault diagnosis result of the wind turbine according to the predicted fault type.

[0064] In the embodiment, the accuracy of the wind turbine fault detection is realized according to the predicted fault type, and then the fault diagnosis result of the wind turbine can be quickly and accurately obtained, and the accuracy of the wind turbine fault detection is significantly improved.

[0065] Further, in other possible embodiments, the step S303 of determining the fault diagnosis result of the wind turbine according to the predicted fault type can further include the following steps S3031 to S3033.

[0066] The step S3031 is to determine the probability value with the largest value as the first confidence of the predicted fault type.

[0067] In the embodiment, the probability value with the largest value is determined as the first confidence of the predicted fault type, so that the complex and multi-dimensional classification uncertainty in the model is condensed into a standardized and interpretable scalar score, and direct evidence from the internal output of the fault diagnosis model is provided for the subsequent judgment of the reliability of the diagnosis result.

[0068] It should be noted that the first confidence essentially reflects that, under the current multi-dimensional feature vector, the fault diagnosis model considers that the predicted fault type has a relative probability advantage over other wind turbine fault types.

[0069] The step S3032 is to obtain a standard fault feature vector corresponding to the predicted fault type from the fault type database as a target fault feature vector, and determine a second confidence of the predicted fault type according to the multi-dimensional feature vector and the target fault feature vector.

[0070] In the embodiment, the standard fault feature vector corresponding to the predicted fault type is obtained from the fault type database as the target fault feature vector; then, the cosine similarity and the Euclidean distance between the multi-dimensional feature vector and the target fault feature vector are calculated, and the Euclidean distance is mapped to the Gaussian kernel similarity by using the Gaussian kernel function; next, the cosine similarity and the Gaussian kernel similarity are weighted and fused to generate the second confidence, which effectively solves the problem of overconfidence or misjudgment of abnormal patterns that may exist when relying solely on the internal confidence of the model, and significantly improves the accuracy of the wind turbine fault detection.

[0071] The step S3033 is to determine a comprehensive confidence of the predicted fault type according to the first confidence and the second confidence, and determine the fault diagnosis result of the wind turbine according to the comprehensive confidence.

[0072] In the embodiment, the first weighting coefficient corresponding to the first confidence degree and the second weighting coefficient corresponding to the second confidence degree are determined; then, the product between the first confidence degree and the first weighting coefficient is superimposed with the product between the second confidence degree and the second weighting coefficient, and the comprehensive confidence degree of the predicted fault type can be determined; finally, the fault diagnosis result of the wind turbine can be accurately obtained according to the comprehensive confidence degree, thereby significantly reducing the risk of false alarm caused by accidental misjudgment of the fault diagnosis model or atypical features.

[0073] Further, in some possible embodiments, with reference to Figure 3 The step S3033 of determining the fault diagnosis result of the wind turbine according to the comprehensive confidence degree can further include the following implementation steps A10 to A30.

[0074] Step A10: detecting whether the comprehensive confidence degree exceeds a preset confidence degree threshold.

[0075] In the embodiment, the comprehensive confidence degree is compared with a preset confidence degree threshold representing an acceptable minimum reliability standard, thereby establishing a quality control standard for the fault diagnosis result at the system level, and effectively preventing the interference of false alarms of low reliability prediction on operation and maintenance decisions.

[0076] It should be noted that the confidence degree threshold can be customized according to application requirements, and the present application does not make any limitation here.

[0077] Step A20: if the comprehensive confidence degree exceeds the confidence degree threshold, determining the predicted fault type as the fault diagnosis result of the wind turbine.

[0078] In the embodiment, when the comprehensive confidence degree exceeds the preset confidence degree threshold, the predicted fault type is determined as the fault diagnosis result of the wind turbine, which ensures that the wind turbine can make a quick, decisive and reliable response to typical and explicit fault types of the wind turbine, meets the core requirements of timeliness and certainty of fault diagnosis, and is the basis of efficient and automated operation and maintenance.

[0079] Step A30: if the comprehensive confidence degree does not exceed the confidence degree threshold, performing model updating according to the fault diagnosis model, taking the updated fault diagnosis model as the next preset fault diagnosis model, and returning to execute the step of determining the fault diagnosis result of the wind turbine according to the multi-dimensional feature vector and the preset fault diagnosis model.

[0080] In the embodiment, if the comprehensive confidence does not exceed the preset confidence threshold, model updating is performed according to the fault diagnosis model, and the updated fault diagnosis model is taken as a next preset fault diagnosis model, and the step of determining the fault diagnosis result of the wind turbine according to the multi-dimensional feature vector and the preset fault diagnosis model is returned to be executed, so that a one-off and static diagnosis process can be changed into a dynamic intelligent system with self-improvement and continuous adaptation capability, and the accuracy and reliability of the fault diagnosis model in a long-term complex and changeable actual operation environment can be maintained and continuously improved.

[0081] Further, in other possible embodiments, the step A30 of performing model updating according to the fault diagnosis model can further include the following implementation steps A301 to A302.

[0082] The step A301 of obtaining an incremental training sample of the fault diagnosis model, the incremental training sample being composed of the multi-dimensional feature vector triggering the fault diagnosis model updating.

[0083] In the embodiment, when the fault diagnosis model updating is triggered, the multi-dimensional feature vector input into the preset fault diagnosis model is determined, and an incremental training sample is constructed according to the multi-dimensional feature vector and the wind turbine operation condition corresponding to the multi-dimensional feature vector, so as to provide a training sample for incremental learning for subsequent model updating.

[0084] The step A302 of determining a model loss parameter according to the incremental training sample and a historical training sample of the fault diagnosis model, and performing model updating on the fault diagnosis model according to the model loss parameter to obtain an updated fault diagnosis model.

[0085] In the embodiment, a sampling reserved sample set is extracted from the historical training sample of the preset fault diagnosis model, the reserved sample set including a representative sample set of all wind turbine fault types and a difficult example sample in which the fault diagnosis model performs poorly in historical training; next, the fault diagnosis model is triggered to perform model optimization training based on the incremental training sample and the reserved sample set to obtain a model loss parameter of the fault diagnosis model; finally, the fault diagnosis model is updated according to the model loss parameter, and the updated fault diagnosis model can be accurately obtained.

[0086] It should be noted that the expression corresponding to the model loss parameter is:

[0087] wherein, the model loss parameter can be understood as a model loss value that needs to be minimized in model optimization training. refers to the classification loss when the model optimization training is performed based on the incremental training sample and the reserved sample set, and the classification loss is used to measure the good or bad degree of the current fault diagnosis model in classifying the current training sample integrated with the incremental training sample and the reserved sample set; represents a balance coefficient of the fault diagnosis model, which can be customized according to application requirements, and the balance coefficient is used to adjust the weight or influence strength of the item in the model loss parameter; represents a balance coefficient of the fault diagnosis model, which can be customized according to application requirements, and the balance coefficient is used to adjust the weight or influence strength of the item in the model loss parameter; represents a regularization loss, which is not directly used to measure classification errors, but to impose constraints or penalties on the changes of the model loss parameter itself to prevent catastrophic forgetting. That is, in incremental learning, the fault diagnosis model is prevented from sharply forgetting the judgment ability of the previously learned and still effective old fault mode when learning new fault types or new data.

[0088] Further, in some feasible embodiments, the above step S3032 of determining the second confidence of the predicted fault type according to the multi-dimensional feature vector and the target fault feature vector can further include the following implementation steps B10 to B20.

[0089] Step B10: performing algorithm processing on the multi-dimensional feature vector and the target fault feature vector based on the fault diagnosis algorithm of the fault diagnosis model to obtain the cosine similarity and the Euclidean distance between the multi-dimensional feature vector and the target fault feature vector.

[0090] In the embodiment, the multi-dimensional feature vector and the target fault feature vector are processed by the fault diagnosis algorithm of the fault diagnosis model to obtain the cosine similarity and the Euclidean distance between the multi-dimensional feature vector and the target fault feature vector. The cosine similarity focuses on analyzing the consistency in direction and form between the two feature vectors (i.e., between the multi-dimensional feature vector and the target fault feature vector), and the greater the value of the cosine similarity indicates that the proportion of the feature composition of the fault mode is more similar. The Euclidean distance directly measures the absolute geometric distance between the two feature vectors in the feature space, and the smaller the value of the Euclidean distance indicates that the feature values are closer as a whole, thereby providing an accurate and reliable data basis for subsequent comprehensive confidence calculation.

[0091] In specific embodiments, the above step B10 of performing algorithm processing on the multi-dimensional feature vector and the target fault feature vector based on the fault diagnosis algorithm of the fault diagnosis model to obtain the cosine similarity and the Euclidean distance between the multi-dimensional feature vector and the target fault feature vector can further include steps B101 to B102.

[0092] Step B101: obtaining two n-dimensional feature vectors, one of the two n-dimensional feature vectors being the multi-dimensional feature vector The other n-dimensional eigenvector among the two n-dimensional eigenvectors is the target fault eigenvector. .

[0093] Step B102: The fault diagnosis algorithm includes a cosine similarity algorithm, which calculates the cosine similarity between the multi-dimensional feature vector and the target fault feature vector. The expression for this cosine similarity algorithm is as follows:

[0094]

[0095]

[0096] in, This represents the inner product between the multi-dimensional feature vector and the target fault feature vector; Representing multi-dimensional feature vectors The first in 1 eigenvector; Represents the target fault feature vector The first in 1 eigenvector; Representing multi-dimensional feature vectors The Euclidean norm; Represents the target fault feature vector The Euclidean norm; It represents the cosine similarity between the multi-dimensional feature vector and the target fault feature vector.

[0097] Step B102: The fault diagnosis algorithm includes a distance calculation algorithm, which calculates the Euclidean distance between the multi-dimensional feature vector and the target fault feature vector. The expression for this distance calculation algorithm is as follows:

[0098]

[0099] in, This represents the sum of squared differences between the multi-dimensional feature vector and the target fault feature vector. This represents the Euclidean distance between the multi-dimensional feature vector and the target fault feature vector.

[0100] Step B20: Map the Euclidean distance to Gaussian kernel similarity using a preset Gaussian kernel function, and perform weighted processing based on the cosine similarity and the Gaussian kernel similarity to obtain the second confidence level of the predicted fault type.

[0101] In the embodiment, firstly, the Euclidean distance is nonlinearly transformed by using a Gaussian kernel function, and the Euclidean distance representing the absolute spatial proximity is smoothly and monotonously mapped into a normalized Gaussian kernel similarity, which can effectively suppress the distance calculation sensitivity caused by the slight fluctuation of the eigenvalue or the outlier, and enhance the stability of the measurement. Subsequently, the cosine similarity reflecting the consistency of the vector direction is weighted and fused with the Gaussian kernel similarity reflecting the absolute position proximity, so as to comprehensively combine the two complementary geometric perspectives of the morphological similarity and the numerical proximity, and finally generate a second confidence that can more comprehensively and stably represent the matching degree between the current sample feature and the target fault mode, thereby significantly improving the reliability of the fault diagnosis and the decision reliability.

[0102] In specific embodiments, the step B20 of mapping the Euclidean distance into a Gaussian kernel similarity by using a preset Gaussian kernel function can further include the following implementation steps B201 to B202.

[0103] Step B201: setting a kernel parameter The kernel parameter may be set according to experience in advance , or calculated based on the standard deviation of the Euclidean distance between historical samples .

[0104] Step B202: calculating the Gaussian kernel similarity of the Euclidean distance mapping based on the preset Gaussian kernel function, and the expression of the Gaussian kernel function is as follows:

[0105] It should be noted that represents the Gaussian kernel similarity, represents the kernel parameter, and represents the Euclidean distance.

[0106] In summary, the wind turbine fault detection method provided in the present application synchronously collects the vibration data, temperature data and electrical data of the wind turbine to obtain multi-source operation data with strictly aligned time sequence and strong internal physical correlation, thereby providing a comprehensive and synchronous data basis for subsequent feature extraction. Next, the multi-source operation data can be accurately obtained from the multi-source operation data, thereby overcoming the defects of the traditional single threshold or simple statistical model in the insufficient feature representation ability under variable speed and variable load conditions. Finally, the multi-dimensional feature vector is input into the preset fault diagnosis model, so that the multi-dimensional feature vector can be processed by the preset fault diagnosis model to accurately identify the wind turbine fault detection, and then the fault diagnosis result of the wind turbine can be quickly and accurately output, thereby significantly improving the accuracy of the wind turbine fault detection.

[0107] Further, the application also provides a wind turbine fault detection device, the wind turbine fault detection device comprises: An acquisition module H01 is configured to acquire multi-source operation data of a wind turbine, wherein the multi-source operation data comprises at least vibration data, temperature data and electrical data; A feature extraction module H02 is configured to determine a multi-dimensional feature vector of the wind turbine according to the multi-source operation data; A fault diagnosis module H03 is configured to determine a fault diagnosis result of the wind turbine according to the multi-dimensional feature vector and a preset fault diagnosis model.

[0108] Further, the application also provides an electronic device. Please refer to Figure 5 , Figure 5 is a structural schematic diagram of an electronic device related to the embodiment of the application. The electronic device of the embodiment of the application can be a device running a wind turbine fault detection method locally.

[0109] The electronic device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, and the computer program is used to implement the steps of the wind turbine fault detection method described above when executed by the processor.

[0110] As shown in Figure 5 , the electronic device of the embodiment of the application can comprise a processor 1001 such as a CPU, a communication bus 1002, a user interface 1003, a network interface 1004 and a memory 1005. The communication bus 1002 is used to realize the connection and communication among these components. The user interface 1003 can comprise a display screen (Display) and an input unit such as a keyboard (Keyboard). Optionally, the user interface 1003 can further comprise a standard wired interface and a wireless interface. The network interface 1004 can optionally comprise a standard wired interface and a wireless interface (such as a Wi-Fi interface).

[0111] The memory 1005 is arranged on the main body of the electronic device, and the memory 1005 stores a program which is executed by the processor 1001 to realize corresponding operations. The memory 1005 is also used to store parameters for use by the electronic device. The memory 1005 can be a high-speed RAM memory or a stable memory (non-volatile memory) such as a disk memory. Optionally, the memory 1005 can also be a storage device independent of the aforementioned processor 1001.

[0112] Those skilled in the art can understand that Figure 5 the structure of the electronic device shown in the above description does not constitute a limitation on the electronic device, and can comprise more or fewer components than those shown in the diagram, or combine certain components, or different component arrangements.

[0113] As shown in Figure 5 , the memory 1005 as a storage medium can include an operating system, a network communication module, a user interface module, and a wind turbine fault detection program.

[0114] In Figure 5 the electronic device, the processor 1001 can be used to call the control program stored in the memory 1005, and execute the steps of the wind turbine fault detection method as described above.

[0115] In addition, the present application also provides a computer readable storage medium, the computer readable storage medium has a computer program stored thereon, the computer program is executed by the processor to realize the steps of the wind turbine fault detection method as described above.

[0116] It should be noted that in this paper, the term "includes", "contains" or any other variant thereof is intended to cover non-exclusive inclusion, so that the process, method, article or system including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or includes elements inherent to such process, method, article or system. Without more limitations, the element defined by the statement "includes a" does not exclude the presence of another identical element in the process, method, article or system including the element.

[0117] The above-mentioned application embodiment serial number is only for description, not representing the pros and cons of the embodiment.

[0118] Through the description of the above embodiments, those skilled in the art can clearly understand that the above-mentioned embodiment method can be realized by means of software and necessary general hardware platform, of course, it can also be realized by hardware, but in many cases, the former is a better embodiment. Based on such understanding, the technical solutions of the present application can be embodied in the form of software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, including a plurality of instructions for making an electronic device (which can be a mobile phone, computer, server, or network device, etc.) execute the method described in each embodiment of the present application.

[0119] The above is only the preferred embodiment of the present application, and does not limit the patent scope of the present application, and any equivalent structure or equivalent process transformation made by using the contents of the specification and drawings, or directly or indirectly applied to other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A wind turbine generator fault detection method characterized by, The wind turbine fault detection method comprises: Obtaining multi-source operation data of the wind turbine, wherein the multi-source operation data at least comprises vibration data, temperature data and electrical data; Determining a multi-dimensional feature vector of the wind turbine according to the multi-source operation data; Determining a fault diagnosis result of the wind turbine according to the multi-dimensional feature vector and a preset fault diagnosis model.

2. The wind turbine generator fault detection method of claim 1, wherein, The step of determining the multi-dimensional feature vector of the wind turbine according to the multi-source operation data comprises: Extracting features according to the vibration data to obtain vibration features of the wind turbine; Extracting features according to the temperature data to obtain temperature features of the wind turbine; Extracting features according to the electrical data to obtain electrical features of the wind turbine; Performing weighted processing on the vibration features, the temperature features and the electrical features to obtain the multi-dimensional feature vector of the wind turbine.

3. The wind turbine generator fault detection method of claim 1, wherein, The fault diagnosis model comprises a fault type database, wherein the fault type database stores standard fault feature vectors corresponding to each wind turbine fault type, and the step of determining the fault diagnosis result of the wind turbine according to the multi-dimensional feature vector and the preset fault diagnosis model comprises: Performing model training on the multi-dimensional feature vector according to the fault diagnosis model to obtain fault probabilities corresponding to each wind turbine fault type; Determining a wind turbine fault type corresponding to a probability value with the largest value among all the fault probabilities as the predicted fault type of the wind turbine; Determining the fault diagnosis result of the wind turbine according to the predicted fault type.

4. The wind turbine generator fault detection method of claim 3 wherein, The step of determining the fault diagnosis result of the wind turbine according to the predicted fault type comprises: Determining the probability value with the largest value as a first confidence degree of the predicted fault type; Obtaining a standard fault feature vector corresponding to the predicted fault type from the fault type database as a target fault feature vector, and determining a second confidence degree of the predicted fault type according to the multi-dimensional feature vector and the target fault feature vector; Determining a comprehensive confidence degree of the predicted fault type according to the first confidence degree and the second confidence degree, and determining the fault diagnosis result of the wind turbine according to the comprehensive confidence degree.

5. The wind turbine generator fault detection method of claim 4 wherein, The step of determining the fault diagnosis result of the wind turbine according to the comprehensive confidence degree comprises: Detecting whether the comprehensive confidence degree exceeds a preset confidence threshold; If the comprehensive confidence degree exceeds the confidence threshold, determining the predicted fault type as the fault diagnosis result of the wind turbine; If the comprehensive confidence degree does not exceed the confidence threshold, performing model updating according to the fault diagnosis model, taking the updated fault diagnosis model as a next preset fault diagnosis model, and returning to execute the step of determining the fault diagnosis result of the wind turbine according to the multi-dimensional feature vector and the preset fault diagnosis model.

6. The wind turbine generator fault detection method of claim 5 wherein, The step of performing model updating according to the fault diagnosis model comprises: An incremental training sample of the fault diagnosis model is acquired, and the incremental training sample is composed of the multi-dimensional feature vector when the fault diagnosis model is triggered to update; A model loss parameter is determined according to the incremental training sample and historical training samples of the fault diagnosis model, and the fault diagnosis model is updated according to the model loss parameter to obtain an updated fault diagnosis model.

7. The wind turbine generator fault detection method of claim 4 wherein, The step of determining the second confidence of the predicted fault type according to the multi-dimensional feature vector and the target fault feature vector comprises: An algorithm processing is performed on the multi-dimensional feature vector and the target fault feature vector based on a fault diagnosis algorithm of the fault diagnosis model to obtain a cosine similarity and an Euclidean distance between the multi-dimensional feature vector and the target fault feature vector; The Euclidean distance is mapped to a Gaussian kernel similarity through a preset Gaussian kernel function, and a weighted processing is performed on the cosine similarity and the Gaussian kernel similarity to obtain the second confidence of the predicted fault type.

8. A wind turbine generator fault detection apparatus characterized by, The wind turbine fault detection device comprises: An acquisition module is configured to acquire multi-source operation data of a wind turbine, wherein the multi-source operation data at least comprises vibration data, temperature data and electrical data; A feature extraction module is configured to determine a multi-dimensional feature vector of the wind turbine according to the multi-source operation data; A fault diagnosis module is configured to determine a fault diagnosis result of the wind turbine according to the multi-dimensional feature vector and a preset fault diagnosis model.

9. An electronic device, comprising: The electronic device comprises a memory, a processor and a computer program stored on the memory and executable on the processor, and the computer program is executed by the processor to implement the steps of the wind turbine fault detection method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a computer program, and the computer program is executed by the processor to implement the steps of the wind turbine fault detection method according to any one of claims 1 to 7.

Images