Wind turbine generator gearbox anomaly detection method, device, equipment and medium

Abstract

The invention relates to the technical field of wind power generation equipment detection, and discloses a wind turbine generator gearbox anomaly detection method, device and equipment and a medium, and the method comprises the steps: extracting the operation state data of a wind turbine generator gearbox from a data collection and monitoring control system of a wind turbine generator, and carrying out the preprocessing of the operation state data of the gearbox; inputting the preprocessed operation state data of the gearbox into a pre-trained deep neural network model, and outputting an anomaly detection score; calculating a gearbox fault threshold value corresponding to the preprocessed operation state data of the gearbox by using an exponential weighted moving average algorithm; according to the method, the operation state of the gear box is monitored through the data acquisition and monitoring control system, and the gear box anomaly detection is comprehensively performed by using the deep neural network model and the exponential weighted moving average algorithm. And the gearbox abnormity detection efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of wind power generation equipment detection, and in particular to a method, device, equipment and medium for detecting abnormality of a gear box of a wind turbine set. Background Art

[0002] At present, the wind power industry is developing rapidly. Wind power generation is an indispensable part of the renewable energy field, and a large number of wind turbines are connected to the grid and put into operation. During the operation of wind turbines, the gearbox is one of the important components of the wind turbine transmission chain. Its health status directly affects the power generation efficiency and production safety of the wind turbine. Therefore, abnormal detection of the gearbox can effectively ensure the stable operation of the entire wind turbine and effectively avoid accidents such as gear breakage.

[0003] In the prior art, workers conduct on-site inspections of gearboxes during the operation of wind turbines. During the actual inspection, the wind turbines need to be stopped for inspection, and abnormalities such as abnormal noise and abnormal temperature in the gearbox need to be inspected, and the gearbox needs to be recorded. However, this method relies on manual inspections, which results in a large consumption of manpower and material resources. Summary of the invention

[0004] In view of this, the present invention provides a method, device, equipment and medium for detecting abnormality of a wind turbine gearbox, so as to solve the problem that abnormality detection of the gearbox is achieved by manual inspection in the prior art, which relies on manual inspection and consumes a lot of manpower and material resources.

[0005] In a first aspect, the present invention provides a method for detecting abnormality in a wind turbine gearbox, the method comprising:

[0006] Extracting the operating status data of the gearbox of the wind turbine from the data acquisition and monitoring control system of the wind turbine, and preprocessing the operating status data of the gearbox;

[0007] The pre-processed gearbox operating status data is input into a pre-trained deep neural network model and an anomaly detection score is output;

[0008] Calculating a gearbox fault threshold corresponding to the preprocessed gearbox operating status data using an exponentially weighted moving average algorithm;

[0009] Abnormality detection is performed on the gearbox using the abnormality detection score and the fault threshold of the gearbox.

[0010] The present invention monitors the operating status of the gearbox of the wind turbine set through the data acquisition and monitoring control system of the wind turbine set, pre-processes the operating status data of the gearbox to facilitate analysis of the operating status data of the gearbox, and inputs the pre-processed operating status data of the gearbox into a pre-trained deep neural network model to monitor the gearbox abnormality using the deep neural network model, and calculates the gearbox fault threshold using an exponentially weighted moving average algorithm, thereby combining the abnormality detection score and the gearbox fault threshold to achieve gearbox abnormality detection, replace manual inspection of gearbox abnormalities, save manpower and material resources, improve the efficiency of gearbox abnormality detection, and facilitate staff to repair faults in a timely manner.

[0011] In an optional implementation, the preprocessing of the operating status data of the gearbox includes:

[0012] Extracting power data points from the operating status data of the gearbox, and filtering out power-limited data points from the power data points according to a pre-acquired active power threshold;

[0013] Using the quartile method to screen out outliers from the operating status data of the gearbox;

[0014] The operating status data of the gearbox is subjected to data normalization processing.

[0015] The present invention pre-processes the operating status data of the gearbox, screens out power-limited data points, eliminates the influence of power-limited data points on data analysis, uses the quartile method to screen out outliers, and improves the anti-interference performance of the data. At the same time, screening out power-limited data points and screening out outliers can also reduce the amount of data, improve data processing efficiency, normalize the data, reduce the differences between data, and enhance the stability of the data.

[0016] In an optional implementation, the deep neural network model is trained as follows:

[0017] A generator is constructed using a deep deconvolutional network, a discriminator is constructed using a deep convolutional network, and a deep neural network model is constructed using the generator and the discriminator;

[0018] The normal operating status data of the wind turbine gearbox is obtained, and the deep neural network model is trained using the normal operating status data of the gearbox as a training data set.

[0019] The present invention utilizes the generator and the discriminator to construct a deep neural network model, combining the advantages of a deep deconvolution network and a deep convolution network, while utilizing the unsupervised characteristics of the deep neural network model to reduce the requirements for the training data set.

[0020] In an optional implementation, the method of constructing a generator using a deep deconvolutional network includes:

[0021] Gaussian noise is added, and the noise vector corresponding to the Gaussian noise is added to the generator constructed using the deep deconvolution network.

[0022] The present invention prevents overfitting of the deep neural network model, improves the generalization ability of the deep neural network model, and effectively improves the robustness of the deep neural network model by adding Gaussian noise to the generator constructed using the deep deconvolution network.

[0023] In an optional embodiment, the step of inputting the preprocessed gearbox operating status data into a pre-trained deep neural network model and outputting anomaly detection scores comprises:

[0024] Input the preprocessed gearbox operating status data into a pre-trained deep neural network model to obtain reconstructed output data;

[0025] The difference between the reconstructed output data and the pre-acquired target data is determined as an abnormality detection score.

[0026] The present invention inputs the preprocessed gearbox operating status data into a pre-trained deep neural network model to utilize the unsupervised characteristics of the deep neural network model. There is no need to mark the abnormal data, but to obtain reconstructed output data. The anomaly detection score is determined according to the difference between the reconstructed output data and the pre-acquired target data, so as to facilitate preliminary detection of gearbox faults.

[0027] In an optional implementation, the gearbox fault threshold is calculated according to the following formula:

[0028]

[0029] Among them, UCL(t) is the upper threshold limit, LCL(t) is the lower threshold limit, μ is the mean of the gearbox operating status data score, ε is the standard deviation of the gearbox operating status data score, K is a constant, λ is the weight of the historical score, and t is the time index.

[0030] The present invention calculates the upper and lower limits of the gearbox fault threshold by using an exponentially weighted moving average algorithm. The exponentially weighted moving average algorithm can smooth data and reduce interference from other data, making it easier to perform gearbox fault detection based on the upper and lower limits of the gearbox fault threshold, thereby improving fault detection efficiency.

[0031] In a second aspect, the present invention provides a wind turbine gearbox abnormality detection device, the device comprising:

[0032] A preprocessing module, used to extract the operating status data of the gearbox of the wind turbine from the data acquisition and monitoring control system of the wind turbine, and preprocess the operating status data of the gearbox;

[0033] An output module, used to input the pre-processed gearbox operating status data into a pre-trained deep neural network model and output anomaly detection scores;

[0034] A calculation module, used for calculating a gearbox fault threshold corresponding to the preprocessed gearbox operating status data by using an exponentially weighted moving average algorithm;

[0035] The detection module is used to perform abnormality detection on the gearbox by using the abnormality detection score and the fault threshold of the gearbox.

[0036] In a third aspect, the present invention provides a computer device, comprising: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the wind turbine gearbox abnormality detection method of the first aspect or any corresponding embodiment thereof by executing the computer instructions.

[0037] In a fourth aspect, the present invention provides a computer-readable storage medium having computer instructions stored thereon, the computer instructions being used to enable a computer to execute the wind turbine gearbox abnormality detection method of the first aspect or any corresponding embodiment thereof.

[0038] In a fifth aspect, the present invention provides a computer program product, comprising computer instructions for causing a computer to execute the wind turbine gearbox abnormality detection method of the first aspect or any corresponding embodiment thereof. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0040] Figure 1 is a schematic flow chart of a method for detecting abnormality in a wind turbine gearbox according to an embodiment of the present invention;

[0041] Figure 2 is a flow chart of another method for detecting abnormality of a wind turbine gearbox according to an embodiment of the present invention;

[0042] Figure 3is a schematic diagram of a deep neural network structure according to an embodiment of the present invention;

[0043] Figure 4 is an application schematic diagram of a method for detecting abnormality in a wind turbine gearbox according to an embodiment of the present invention;

[0044] Figure 5 is a structural block diagram of a wind turbine gearbox abnormality detection device according to an embodiment of the present invention;

[0045] Figure 6 It is a schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0046] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0047] According to an embodiment of the present invention, an embodiment of a method for detecting abnormality in a gearbox of a wind turbine is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0048] In this embodiment, a method for detecting abnormality of a wind turbine gearbox is provided, which can be used in a mobile terminal. Figure 1 is a flow chart of a method for detecting abnormality of a wind turbine gearbox according to an embodiment of the present invention. Figure 1 As shown, the process includes the following steps:

[0049] Step S101 : extracting the operating status data of the gearbox of the wind turbine from the data acquisition and monitoring control system of the wind turbine, and preprocessing the operating status data of the gearbox.

[0050] In an embodiment of the present invention, a supervisory control and data acquisition (SCADA) system of a wind turbine generator set monitors the operating status of a wind turbine generator set gearbox in real time to obtain operating status data of the wind turbine generator set gearbox, and pre-processes the operating status data of the gearbox obtained by monitoring the SCADA system to reduce the amount of data processing.

[0051] Step S102: input the preprocessed gearbox operating status data into a pre-trained deep neural network model, and output an anomaly detection score.

[0052] In an embodiment of the present invention, a deep neural network (DNN) model is used to detect the abnormal operating state of the gearbox. The deep neural network model needs to be trained in advance. By utilizing the unsupervised characteristics of the deep neural network model, the preprocessed gearbox operating state data is input into the pre-trained deep neural network model, and an abnormality detection score is output to detect the abnormal operating state of the gearbox.

[0053] Step S103, using an exponentially weighted moving average algorithm to calculate a gearbox fault threshold corresponding to the preprocessed gearbox operating status data.

[0054] In an embodiment of the present invention, when calculating the gearbox fault threshold, an exponentially weighted moving average (EWMA) algorithm is used to perform data smoothing and result visualization processing. According to the EWMA algorithm, combined with the information of previous subgroups or observations, an EWMA control chart can be customized to detect the offset of the data size, so as to use the EWMA control chart to analyze the data change trend. The gearbox fault analysis is performed on the pre-processed gearbox operating status data using the exponentially weighted moving average algorithm, and the obtained fault threshold is determined as the gearbox fault threshold. In actual applications, the fault threshold can also be adjusted in combination with the safety factor or fault standard to make the fault threshold more suitable for the actual application scenario of the actual gearbox.

[0055] Specifically, the gearbox fault threshold is calculated according to the following formula:

[0056]

[0057] Among them, UCL(t) is the upper threshold limit, LCL(t) is the lower threshold limit, μ is the mean of the gearbox operating status data score, ε is the standard deviation of the gearbox operating status data score, K is a constant, λ is the weight of the historical score, and t is the time index.

[0058] By using the exponentially weighted moving average algorithm to calculate the upper and lower limits of the gearbox fault threshold, the exponentially weighted moving average algorithm can smooth the data and reduce interference from other data, making it easier to detect gearbox faults based on the upper and lower limits of the gearbox fault threshold and improving fault detection efficiency.

[0059] Step S104: perform abnormality detection on the gearbox using the abnormality detection score and the fault threshold of the gearbox.

[0060] In the embodiment of the present invention, the abnormality detection score and the fault threshold of the gearbox are used for statistical scoring, and the statistical score exceeding the fault threshold for multiple consecutive times is used as an alarm condition for gearbox abnormality detection.

[0061] Specifically, the following formula is used to perform statistical scoring on the anomaly detection score and the gearbox fault threshold:

[0062] Se t =λ·score t +(1-λ)·Se t-1 (2)

[0063] Among them, Se t is the statistical score, λ is the weight of the historical score, which is generally set to 0.3, and score t is the anomaly detection score at time t, where t is the time index.

[0064] The wind turbine gearbox abnormality detection method provided in the present embodiment monitors the operating status of the wind turbine gearbox through the data acquisition and monitoring control system of the wind turbine, pre-processes the operating status data of the gearbox to facilitate analysis of the operating status data of the gearbox, and inputs the pre-processed operating status data of the gearbox into a pre-trained deep neural network model to monitor the gearbox abnormality using the deep neural network model, and calculates the gearbox fault threshold using an exponentially weighted moving average algorithm, thereby combining the abnormality detection score and the gearbox fault threshold to achieve gearbox abnormality detection, replace manual inspection of gearbox abnormalities, save manpower and material resources, improve the efficiency of gearbox abnormality detection, and facilitate staff to repair faults in a timely manner.

[0065] In this embodiment, a method for detecting abnormality of a wind turbine gearbox is provided, which can be used in the above-mentioned mobile terminal. Figure 2 is a flow chart of a method for detecting abnormality of a wind turbine gearbox according to an embodiment of the present invention. Figure 2 As shown, the process includes the following steps:

[0066] Step S201: extracting the operating status data of the gearbox of the wind turbine from the data acquisition and monitoring control system of the wind turbine, and preprocessing the operating status data of the gearbox.

[0067] Specifically, the above step S201 includes:

[0068] Step S2011: extract power data points from the operating status data of the gearbox, and filter out limited power data points from the power data points according to a pre-acquired active power threshold.

[0069] Step S2012: using the quartile method to screen out outliers from the gearbox operating status data.

[0070] Step S2013, performing data normalization processing on the operating status data of the gearbox.

[0071] In an embodiment of the present invention, the operating status data of the gearbox is preprocessed, and the wind power capture characteristics of the unit are clarified according to the basic principle of wind energy capture of the wind turbine. The power of the unit can be adjusted by adjusting the pitch angle and the tip speed ratio, and the variable pitch wind turbine generally starts and stops the unit and captures wind energy by adjusting the pitch angle. Therefore, controlling the change of the pitch angle can basically obtain the required active power, determine the active power threshold, and use this feature to delete the abandoned wind data. Among them, the abandoned wind data refers to the power data that is forced to be abandoned during the wind power generation process due to equipment maintenance, wind speed exceeding the operating range of the wind turbine, etc., which causes the wind power to be unable to be fully utilized or fully connected to the power grid, and is forced to be abandoned.

[0072] Specifically, power data points are extracted from the operating status data of the gearbox, the extracted power data points are compared with the active power threshold obtained in advance, and the power data points exceeding the active power threshold are screened out.

[0073] The quartile method is used to screen out outliers from the gearbox operating status data. First, the gearbox operating status data is sorted, and then the quartiles and interquartile range are calculated. The discrete value boundaries are defined, including the upper boundary and the lower boundary. The sorted data is traversed, and the data falling outside the upper boundary and the lower boundary are marked as discrete values. The discrete values ​​are screened out to complete the outlier screening of the gearbox operating status data.

[0074] The gearbox operating status data is normalized and compressed to [0, 1] or other ranges in proportion to eliminate the impact of excessive differences in the numerical range, thereby facilitating subsequent data processing to ensure the regularity of the data.

[0075] The gearbox operating status data is preprocessed to filter out power-limited data points to eliminate the impact of power-limited data points on data analysis. The quartile method is used to filter out outliers to improve the data's anti-interference performance. At the same time, filtering out power-limited data points and outliers can also reduce the amount of data, improve data processing efficiency, normalize the data, reduce differences between data, and enhance data stability.

[0076] For details, please see Figure 1 Step S101 of the illustrated embodiment will not be described in detail here.

[0077] Step S202: input the preprocessed gearbox operating status data into a pre-trained deep neural network model, and output an anomaly detection score.

[0078] Specifically, the deep neural network model is trained according to the following steps:

[0079] In step Sa, a generator is constructed using a deep deconvolutional network, a discriminator is constructed using a deep convolutional network, and a deep neural network model is constructed using the generator and the discriminator.

[0080] Step Sb: acquiring the normal operating status data of the wind turbine gearbox, and using the normal operating status data of the gearbox as a training data set to train the deep neural network model.

[0081] In an embodiment of the present invention, a generator is constructed using deep deconvolutional neural networks (DNNs). Specifically, a noise vector is used as an input of the deep deconvolutional network, and the generator is gradually constructed using deconvolution layers and upsampling strategies.

[0082] The discriminator is constructed using deep convolutional neural networks (CNNs). Specifically, the received image or other forms of data are used as the input layer, the convolution layer extracts features from the input data, and the fully connected layer outputs a scalar score or probability to represent the probability that the input data is a real sample. The discriminator is constructed through the input layer, the convolution layer and the fully connected layer.

[0083] The generator and discriminator are used to build a deep neural network model based on generative adversarial networks (GANs). The generator and discriminator are integrated, and generator training and discriminator training are alternately performed to build a deep neural network model.

[0084] Specifically, the deep neural network can be divided into input layer, hidden layer and output layer according to the position of different layers, such as Figure 3 As shown in the figure, the first layer is the input layer, the last layer is the output layer, the middle layer is the hidden layer, and each layer is a fully connected layer, that is, any neuron in the i-th layer must be connected to any neuron in the i+1-th layer, and there are continuous weights between neurons. Each neuron in the neural network can be regarded as a simple linear function. After the data input to each neuron is calculated, it will be affected by the activation function and then input to the next neural network layer for further training.

[0085] In order to prevent overfitting in the deep neural network model, the present invention introduces the dropout method ( Figure 3 The gray neurons in the figure are the neurons filtered out by dropout. In this neural network, dropout will traverse each layer of the network and set the probability of eliminating nodes in the neural network. In each layer of the network, the probability of keeping and eliminating the set nodes is 0.1, that is, 10% of the neurons in each layer do not participate in training, and then the next layer is trained.

[0086] Specific loss functions and cost functions are selected for different data situations. In the back propagation process, according to the chain rule, we can reversely recurse from the output layer to the first layer to obtain all partial derivatives about the parameters W and b, and then apply the gradient descent method to continuously update the parameters. When the loss no longer changes during training, it proves that the network has been trained.

[0087] When training a deep neural network, the more practical mean square error is used as the loss function to measure the loss between the output calculated by the training sample and the actual training sample output. That is, for each sample, the expected minimization formula is:

[0088]

[0089] where a L and y are vectors with the same feature dimension and output dimension. In terms of the selection of activation function, since the Sigmoid activation function can map all data into numbers between (0, 1) and can express the activated and unactivated states of neurons, it works better when the feature difference is relatively complex. In the back propagation process, according to the chain rule of derivation, we can reversely recurse from the output layer to the first layer to obtain all partial derivatives of the parameters W and b, and then apply the gradient descent method to continuously update the parameters. When the training loss no longer continues to change, it proves that the network has been trained.

[0090] The operating status data of the wind turbine gearbox under normal conditions is obtained. By utilizing the unsupervised characteristics of the deep neural network and selecting specific parameters, the deep neural network model is trained using the gearbox operating status data under normal conditions as the training data set.

[0091] The deep neural network model is constructed using the generator and the discriminator, which combines the advantages of deep deconvolutional networks and deep convolutional networks. At the same time, the unsupervised characteristics of the deep neural network model are utilized to reduce the requirements for the training data set.

[0092] Specifically, in the above step Sa, constructing a generator using a deep deconvolutional network includes:

[0093] Step Sa1, adding Gaussian noise, and adding the noise vector corresponding to the Gaussian noise to the generator constructed using the deep deconvolution network.

[0094] In an embodiment of the present invention, when constructing a generator, Gaussian noise is added to enhance the diversity and randomness of generated samples, which can be achieved by superimposing a Gaussian distributed random noise vector on the input layer of the generator.

[0095] By adding Gaussian noise to the generator constructed using a deep deconvolutional network, the deep neural network model is prevented from overfitting, the generalization ability of the deep neural network model is improved, and the robustness of the deep neural network model is effectively improved.

[0096] Specifically, the above step S202 includes:

[0097] Step S2021, input the preprocessed gearbox operating status data into a pre-trained deep neural network model to obtain reconstructed output data.

[0098] Step S2022: Determine the difference between the reconstructed output data and the pre-acquired target data as an anomaly detection score.

[0099] In an embodiment of the present invention, the preprocessed operating status data of the gearbox is input into a pre-trained deep neural network model for data reconstruction. The purpose of data reconstruction is to make the output data as close as possible to the target data, that is, to obtain reconstructed output data. The reconstructed output data is compared with the pre-acquired target data to calculate the difference between the reconstructed output data and the target data, and the difference is determined as the anomaly detection score.

[0100] Specifically, the anomaly detection score is calculated according to the following formula: i :

[0101]

[0102] Among them, score i is the anomaly detection score, i=1, 2, 3...N, X i,k For the target data, Reconstruct output data.

[0103] Specifically, the loss function of the deep neural network model is determined according to the correspondence between the reconstructed output data and the target data. The loss function measures the difference between the reconstructed output data of the deep neural network model and the target data. The deep neural network model is updated according to the difference using an optimizer and back-propagated. The training converged model is tested using test data to evaluate the generalization ability of the deep neural network model.

[0104] Step S203: Calculate the gearbox fault threshold corresponding to the preprocessed gearbox operating status data using an exponentially weighted moving average algorithm.

[0105] For details, please see Figure 1 Step S103 of the illustrated embodiment will not be described in detail here.

[0106] Step S204: perform abnormality detection on the gearbox using the abnormality detection score and the fault threshold of the gearbox.

[0107] For details, please see Figure 1 Step S104 of the illustrated embodiment will not be described in detail here.

[0108] The wind turbine gearbox abnormality detection method provided in this embodiment inputs the preprocessed gearbox operating status data into a pre-trained deep neural network model to utilize the unsupervised characteristics of the deep neural network model. There is no need to mark the abnormal data, but to obtain reconstructed output data. The abnormality detection score is determined based on the difference between the reconstructed output data and the pre-acquired target data to facilitate preliminary detection of gearbox faults.

[0109] As a specific application example of the embodiment of the present invention, Figure 4 As shown in the figure, the SCADA data in the healthy period, that is, the data without abnormality in the wind turbine gearbox, is used as the data source to preprocess the SCADA data, including screening out power-limited data points, screening outliers, and data normalization. Then the deep neural network DNN is trained, Gaussian noise is added, the generator is constructed by the deep deconvolution network, the discriminator is constructed by the deep convolution network, and the deep neural network model is constructed. The data is smoothed by the EWMA algorithm, and the adaptive threshold is set to complete the construction of the entire model for wind turbine gearbox abnormality detection. The SCADA data to be tested is obtained, the SCADA data to be tested is preprocessed, and input into the pre-trained deep neural network DNN. The data is smoothed by the EWMA algorithm, and the results are visualized to complete the abnormality detection of the wind turbine gearbox.

[0110] In this embodiment, a wind turbine gearbox abnormality detection device is also provided, which is used to implement the above-mentioned embodiments and preferred implementation modes, and will not be repeated hereafter. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.

[0111] This embodiment provides a wind turbine gearbox abnormality detection device, such as Figure 5 As shown, including:

[0112] The preprocessing module 501 is used to extract the operating status data of the gearbox of the wind turbine from the data acquisition and monitoring control system of the wind turbine and to preprocess the operating status data of the gearbox;

[0113] An output module 502 is used to input the pre-processed gearbox operating status data into a pre-trained deep neural network model and output an anomaly detection score;

[0114] A calculation module 503, used to calculate a gearbox fault threshold corresponding to the preprocessed gearbox operating status data using an exponentially weighted moving average algorithm;

[0115] The detection module 504 is used to perform abnormality detection on the gearbox by using the abnormality detection score and the fault threshold of the gearbox.

[0116] In some optional implementations, the preprocessing module 501 includes:

[0117] The first screening unit is used to extract power data points from the operating status data of the gearbox, and screen out limited power data points in the power data points according to a pre-acquired active power threshold.

[0118] The second screening unit is used to screen out outliers from the gearbox operating status data using the quartile method.

[0119] The normalization processing unit is used to perform data normalization processing on the operating status data of the gearbox.

[0120] In some optional implementations, the output module 502 includes:

[0121] The data acquisition unit is used to input the preprocessed gearbox operating status data into a pre-trained deep neural network model to obtain reconstructed output data.

[0122] The determination unit is used to determine the difference between the reconstructed output data and the pre-acquired target data as an abnormality detection score.

[0123] The further functional description of each of the above modules and units is the same as that of the above corresponding embodiments and will not be repeated here.

[0124] The wind turbine gearbox abnormality detection device in this embodiment is presented in the form of a functional unit, where the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that executes one or more software or fixed programs, and / or other devices that can provide the above functions.

[0125] The embodiment of the present invention also provides a computer device having the above Figure 5 The wind turbine gearbox abnormality detection device shown.

[0126] See also Figure 6 , Figure 6 is a schematic diagram of the structure of a computer device provided by an optional embodiment of the present invention, such as Figure 6 As shown, the computer device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. Various components are connected to each other using different buses for communication, and can be installed on a common mainboard or installed in other ways as needed. The processor can process the instructions executed in the computer device, including instructions stored in or on the memory to display the graphical information of the GUI on an external input / output device (such as, a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Similarly, multiple computer devices can be connected, and each device provides some necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Figure 6 A processor 10 is taken as an example.

[0127] The processor 10 may be a central processing unit, a network processor or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be a dedicated integrated circuit, a programmable logic device or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic or any combination thereof.

[0128] The memory 20 stores instructions executable by at least one processor 10, so that at least one processor 10 executes the method shown in the above embodiment.

[0129] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created according to the use of the computer device, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some optional embodiments, the memory 20 may optionally include a memory remotely arranged relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0130] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid state drive; the memory 20 may also include a combination of the above types of memory.

[0131] The computer device also includes an input device 30 and an output device 40. The processor 10, the memory 20, the input device 30 and the output device 40 may be connected via a bus or other means. Figure 6 The example of connecting through bus is taken in the following.

[0132] The input device 30 can receive input digital or character information and generate key signal input related to user settings and function control of the computer device, such as a touch screen, etc. The output device 40 can include a display device, etc.

[0133] The embodiment of the present invention also provides a computer-readable storage medium. The method according to the embodiment of the present invention can be implemented in hardware, firmware, or can be implemented as a computer code that can be recorded in a storage medium, or can be implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and will be stored in a local storage medium through a network download, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state hard disk, etc.; further, the storage medium can also include a combination of the above types of memories. It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor, or hardware, the method shown in the above embodiment is implemented.

[0134] A part of the present invention may be applied as a computer program product, such as a computer program instruction, which, when executed by a computer, can call or provide the method and / or technical solution according to the present invention through the operation of the computer. Those skilled in the art should understand that the existence of the computer program instruction in a computer-readable medium includes, but is not limited to, a source file, an executable file, an installation package file, etc., and accordingly, the way in which the computer program instruction is executed by the computer includes, but is not limited to: the computer directly executes the instruction, or the computer compiles the instruction and then executes the corresponding compiled program, or the computer reads and executes the instruction, or the computer reads and installs the instruction and then executes the corresponding installed program. Here, the computer-readable medium may be any available computer-readable storage medium or communication medium accessible to the computer.

[0135] Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A method for detecting abnormality of a wind turbine gearbox, characterized in that: The method comprises: Extracting the operating status data of the gearbox of the wind turbine from the data acquisition and monitoring control system of the wind turbine, and preprocessing the operating status data of the gearbox; Input the preprocessed gearbox operating status data into a pre-trained deep neural network model and output anomaly detection scores; Calculating a gearbox fault threshold corresponding to the preprocessed gearbox operating status data using an exponentially weighted moving average algorithm; Abnormality detection is performed on the gearbox using the abnormality detection score and the fault threshold of the gearbox.

2. The method according to claim 1, characterized in that The preprocessing of the operating status data of the gearbox includes: Extracting power data points from the operating status data of the gearbox, and filtering out power-limited data points from the power data points according to a pre-acquired active power threshold; Using the quartile method to screen out outliers from the operating status data of the gearbox; The operating status data of the gearbox is subjected to data normalization processing.

3. The method according to claim 1, characterized in that The deep neural network model is trained as follows: A generator is constructed using a deep deconvolutional network, a discriminator is constructed using a deep convolutional network, and a deep neural network model is constructed using the generator and the discriminator; Acquire the normal operating status data of the wind turbine gearbox, and use the normal operating status data of the gearbox as a training data set to train the deep neural network model.

4. The method according to claim 3, characterized in that The generator is constructed by using a deep deconvolution network, comprising: Gaussian noise is added, and the noise vector corresponding to the Gaussian noise is added to the generator constructed using the deep deconvolution network.

5. The method according to claim 1, characterized in that The pre-processed gearbox operating status data is input into a pre-trained deep neural network model to output anomaly detection scores, including: Input the preprocessed gearbox operating status data into a pre-trained deep neural network model to obtain reconstructed output data; The difference between the reconstructed output data and the pre-acquired target data is determined as an abnormality detection score.

6. The method according to claim 1, characterized in that The gearbox fault threshold is calculated according to the following formula: Among them, UCL(t) is the upper threshold limit, LCL(t) is the lower threshold limit, μ is the mean of the gearbox operating status data score, ε is the standard deviation of the gearbox operating status data score, K is a constant, λ is the weight of the historical score, and t is the time index.

7. A wind turbine gearbox abnormality detection device, characterized in that: The device comprises: A preprocessing module, used to extract the operating status data of the gearbox of the wind turbine from the data acquisition and monitoring control system of the wind turbine, and preprocess the operating status data of the gearbox; An output module, used to input the pre-processed gearbox operating status data into a pre-trained deep neural network model and output anomaly detection scores; A calculation module, used for calculating a gearbox fault threshold corresponding to the preprocessed gearbox operating status data by using an exponentially weighted moving average algorithm; The detection module is used to perform abnormality detection on the gearbox by using the abnormality detection score and the fault threshold of the gearbox.

8. A computer device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the wind turbine gearbox abnormality detection method according to any one of claims 1 to 6 by executing the computer instructions.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the wind turbine gearbox abnormality detection method according to any one of claims 1 to 6.

10. A computer program product, characterized in that The method comprises computer instructions, wherein the computer instructions are used to enable a computer to execute the wind turbine gearbox abnormality detection method according to any one of claims 1 to 6.

Images