Fault detection method and device for wind turbine generator, edge equipment and storage medium

Abstract

The embodiment of the invention discloses a fault detection method and device for a wind turbine generator, edge equipment and a storage medium, and relates to the technical field of wind power generation. The method comprises the following steps: acquiring current vibration data of a generator bearing of a wind turbine generator, and calculating a root mean square of the current vibration data to obtain an original root mean square; normalizing the original root mean square to obtain a standard root mean square; normalizing the standard root mean square by using the cloud server to obtain a target root mean square; and performing fault detection on the wind turbine generator according to the target root mean square. In the technical scheme provided by the embodiment of the invention, the normalizing degree of the target root mean square is improved through normalizing the original root mean square, so that the applicability of the 3-sigma rule is improved, the false alarm rate of fault detection is reduced, the accuracy and efficiency of fault detection are improved, and the operation efficiency and reliability of the wind turbine generator are improved.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of wind power generation technology, and in particular to a fault detection method, device, edge device and storage medium for a wind turbine set. Background Art

[0002] Due to the complex operating conditions of wind turbines, a condition monitoring system (CMS) of wind turbines is usually used to collect vibration data of generator bearings and calculate the root mean square (RMS) of the vibration data. The RMS value is then used to perform fault detection on the generator bearings of the wind turbine, that is, an alarm is triggered when the RMS value is not within the threshold range.

[0003] In the prior art, the threshold range of the generator bearing is determined using the three-sigma rule, and the wind turbine fault detection is performed based on the threshold range. However, the set of RMS values ​​calculated using vibration data does not belong to the standard normal distribution. In this case, directly using the 3-sigma rule to determine the threshold range will result in a high false alarm rate for fault detection. Summary of the invention

[0004] The embodiments of the present application provide a fault detection method, apparatus, edge device and storage medium for a wind turbine group, which realize the fault detection function of the wind turbine group to solve the problem in the prior art that the false alarm rate of fault detection is high due to directly using the 3-sigma rule to determine the threshold range when the RMS set does not belong to the standard normal distribution.

[0005] In a first aspect, an embodiment of the present application provides a fault detection method for a wind turbine generator set, which is applied to a side device, and the method includes:

[0006] Acquire the current vibration data of the generator bearing of the wind turbine generator set, and calculate the root mean square of the current vibration data to obtain the original root mean square, where the current vibration data is obtained by sampling the generator bearing at the current moment using a sensor with a preset sampling frequency;

[0007] The original root mean square is normalized to obtain the standard root mean square;

[0008] Use the cloud server to normalize the standard RMS to obtain the target RMS;

[0009] Fault detection of wind turbines is performed based on the target RMS.

[0010] In an embodiment of the present application, the edge device can obtain the current vibration data of the generator bearing of the wind turbine, and calculate the root mean square of the current vibration data to obtain the original root mean square, and then normalize the original root mean square to obtain the standard root mean square, and use the cloud server to normalize the standard root mean square to obtain the target root mean square, and then perform fault detection on the wind turbine according to the target root mean square, thereby realizing the fault detection function of the wind turbine. In the above technical scheme, by normalizing the original root mean square, the normalization degree of the target root mean square is improved, so that the target root mean square set is more in line with the standard normal distribution, thereby improving the applicability of the 3-sigma rule and reducing the false alarm rate of fault detection, thereby avoiding the problem of a high false alarm rate of fault detection caused by directly using the 3-sigma rule to determine the threshold range when the RMS set does not belong to the standard normal distribution, and improving the accuracy of fault detection; through the cloud-edge collaboration of edge devices and cloud servers, the efficiency of fault detection is improved, thereby avoiding the expansion of faults and the increase of downtime, thereby improving the operating efficiency and reliability of wind turbines, and reducing operation and maintenance costs.

[0011] In a second aspect, an embodiment of the present application provides a fault detection device for a wind turbine generator set, which is applied to a side device, and the device includes:

[0012] A calculation module is used to obtain the current vibration data of the generator bearing of the wind turbine generator set, and calculate the root mean square of the current vibration data to obtain the original root mean square. The current vibration data is obtained by sampling the generator bearing at the current moment using a sensor with a preset sampling frequency;

[0013] The first processing module is used to perform normalization processing on the original root mean square to obtain a standard root mean square;

[0014] The second processing module is used to perform normalization processing on the standard root mean square using the cloud server to obtain the target root mean square;

[0015] The fault detection module is used to perform fault detection on the wind turbine generator set according to the target root mean square.

[0016] In a third aspect, an embodiment of the present application provides an edge device, the edge device comprising:

[0017] at least one processor; and a memory communicatively coupled to the at least one processor;

[0018] The memory stores a computer program that can be executed by at least one processor, and the computer program is executed by at least one processor so that the at least one processor can execute the fault detection method for a wind turbine set according to any embodiment of the present application.

[0019] In a fourth aspect, an embodiment of the present application provides a wind turbine fault detection system, the system comprising a wind turbine, a cloud server, and an edge device for executing the wind turbine fault detection method of any embodiment of the present application.

[0020] In a fifth aspect, an embodiment of the present application provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, a fault detection method for a wind turbine set as in any embodiment of the present application is implemented.

[0021] The description of the second, third, fourth and fifth aspects of the present application can refer to the detailed description of the first aspect; and the beneficial effects of the second, third, fourth and fifth aspects can refer to the beneficial effect analysis of the first aspect, which will not be repeated here.

[0022] In this application, the name of the above-mentioned wind turbine fault detection device does not limit the device or functional module itself. In actual implementation, these devices or functional modules may appear with other names. As long as the functions of each device or functional module are similar to those of this application, they belong to the scope of the claims of this application and their equivalent technologies.

[0023] These and other aspects of the present application will become more apparent from the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0025] Figure 1 It is a flow chart of a fault detection method for a wind turbine generator set provided in an embodiment of the present application;

[0026] Figure 2 is a structural schematic diagram of a fault detection system for a wind turbine generator set provided in an embodiment of the present application;

[0027] Figure 3 is another flow chart of the fault detection method for a wind turbine generator set provided in an embodiment of the present application;

[0028] Figure 4 is a structural schematic diagram of a fault detection device for a wind turbine generator set provided in an embodiment of the present application;

[0029] Figure 5 It is a structural diagram of the edge device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0030] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of this application.

[0031] It should be noted that the terms "first", "second", "target", and "original" in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including", "having", and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products, or devices.

[0032] Figure 1 It is a flow chart of the fault detection method for a wind turbine set provided in an embodiment of the present application. This embodiment can be applied to scenarios where fault detection of a wind turbine set is required. The fault detection method for a wind turbine set provided in this embodiment can be executed by a fault detection device for a wind turbine set provided in an embodiment of the present application, and the device can be implemented by software and / or hardware. In a specific embodiment, the fault detection device for the wind turbine set can be integrated in the fault detection system of the wind turbine set, and specifically can be integrated in the edge device of the fault detection system of the wind turbine set, and the edge device can be a computer, etc.

[0033] In a specific embodiment, the structure of the fault detection system of the wind turbine generator set can be as follows: Figure 2 As shown, Figure 2 The wind turbine fault detection system shown includes a wind turbine, an edge device and a cloud server; the edge device is a local device of the wind turbine, which can obtain various data of the wind turbine, and one wind turbine corresponds to one edge device; the cloud server can communicate with each edge device, that is, it can obtain data in the edge device and send data to the edge device.

[0034] Combine the following Figure 2The fault detection system of the wind turbine set shown in the figure is used to illustrate a fault detection method of a wind turbine set provided in an embodiment of the present application. The execution subject of the method can be an edge device of the fault detection system of the wind turbine set. Figure 1 The fault detection method of the wind turbine generator set in this embodiment includes but is not limited to the following steps:

[0035] S110, obtaining current vibration data of a generator bearing of the wind turbine generator set, and calculating a root mean square of the current vibration data to obtain an original root mean square.

[0036] Among them, generator bearings are an important part of the generator in a wind turbine. They are responsible for supporting rotating parts (such as generator rotors) and allowing them to rotate smoothly. The performance of generator bearings directly affects the operating efficiency and life of the generator. In wind turbines, due to the harsh working environment (such as high-speed rotation, vibration, and large temperature changes), generator bearings are easily damaged. Therefore, monitoring the status of generator bearings is crucial to preventing failures and extending the life of wind turbines.

[0037] The current vibration data is obtained by sampling the generator bearing at the current moment using a sensor with a preset sampling frequency; optionally, the current vibration data is X t ={X t1 ,X t2 ,...,X tn}, wherein t represents the current moment, and n represents the number of samples obtained by sampling the generator bearing at the current moment using a sensor with a preset sampling frequency; illustratively, the sensor may be a vibration sensor or an acceleration sensor, and the preset sampling frequency may be 12800 Hz or 25600 Hz, etc.

[0038] The root mean square is an effective way to express vibration amplitude, taking into account the average energy of the vibration signal over the entire time period. The original root mean square is the result of the root mean square calculation of the current vibration data collected by the sensor, reflecting the vibration level of the generator bearing at the current moment.

[0039] Specifically, suitable sensors, such as acceleration sensors or vibration sensors, can be selected in advance according to the characteristics and monitoring requirements of the generator bearings of the wind turbine set. Then, the sampling frequency of the sensor can be set to a preset sampling frequency according to the operating speed and vibration characteristics of the generator bearings. The selected sensor can be installed at an appropriate position near the generator bearings, such as on the bearing seat, bearing cover or casing near the bearing, to ensure that the sensor can accurately capture the vibration signal of the generator bearing. At this time, the sensor can communicate with the edge device, that is, the edge device can control the sensor.

[0040] When it is necessary to perform fault detection on a wind turbine within a preset time period, for each moment within the preset time period, the edge device can select any moment as the current moment and start the sensor to sample the vibration data of the generator bearing at the current moment using the sensor with a preset sampling frequency, and then pre-process the sampled vibration data, such as removing noise and filtering, to obtain the current vibration data, and then calculate the root mean square of the current vibration data according to the root mean square calculation formula to obtain the original root mean square, that is, the calculation formula of the original root mean square can be:

[0041]

[0042] Among them, RMS t Represents the original RMS at the current time (i.e. time t).

[0043] After obtaining the original root mean square at the current moment, the generator bearings at other moments within the preset time period can be sampled using a sensor with a preset sampling frequency to obtain vibration data at other moments, and the root mean square of the vibration data at other moments can be calculated respectively to obtain the original root mean square set corresponding to the preset time period. At this time, the original root mean square set does not conform to the standard normal distribution.

[0044] S120, normalize the original root mean square to obtain a standardized root mean square.

[0045] The standard root mean square is the result obtained after normalizing the original root mean square. At this time, the standard root mean square set corresponding to the preset time period is close to the standard normal distribution.

[0046] Specifically, after obtaining the original root mean square, the original root mean square can be normalized to obtain a standard root mean square. For example, the original root mean square at the current moment can be standardized using the mean-variance algorithm and the law of large numbers to obtain the standard root mean square at the current moment, and then the aforementioned steps are repeated to obtain the standard root mean square at other moments in the preset time period, thereby obtaining a standard root mean square set corresponding to the preset time period, so that the standard root mean square set is close to the standard normal distribution.

[0047] S130. Using a cloud server, normalize the standard root mean square to obtain a target root mean square.

[0048] The target root mean square is the result obtained after further normalization of the standard root mean square. At this time, the target root mean square set corresponding to the preset time period is closer to the standard normal distribution.

[0049] Specifically, after obtaining the standard root mean square, the cloud server can be used to normalize the standard root mean square to obtain the target root mean square. For example, the standard root mean square set corresponding to the preset time period can be sent to the cloud server, so that the cloud server uses a deep learning method to normalize the standard root mean square set to obtain the target root mean square set corresponding to the preset time period; or, the cloud server pre-trains and optimizes the neural network model based on the standard root mean square of the wind turbine before the preset time period, so that the neural network model can improve the normalization degree of the data, and sends the trained neural network model to the edge device, and then the edge device uses the trained neural network model to normalize the standard root mean square set to obtain the target root mean square set; at this time, compared with the standard root mean square set, the target root mean square set is closer to the standard normal distribution, and the degree of normalization is improved.

[0050] S140. Perform fault detection on the wind turbine generator set according to the target RMS value.

[0051] Specifically, after obtaining the target RMS, the edge device can obtain the fault threshold range corresponding to the preset time period from the cloud server. That is, since the target RMS set after normalization is approximately consistent with the standard normal distribution, the cloud server can determine the fault threshold range of the wind turbine based on the 3-sigma rule, that is, the fault threshold range is [μ-3σ,μ+3σ], where μ represents the mean of the target RMS set and σ represents the standard deviation of the target RMS set. The fault threshold range is used to detect whether the generator bearing of the wind turbine is faulty.

[0052] Then, the edge device can perform fault detection on the wind turbine according to the target root mean square and the fault threshold range, that is, for each target root mean square in the target root mean square set, it can be determined whether the target root mean square is within the fault threshold range. If the target root mean square is within the fault threshold range, it indicates that the vibration level of the generator bearing of the wind turbine at a certain moment is within the safe range, and it can be determined that the wind turbine is in a normal state at a certain moment; if the target root mean square is not within the fault threshold range, it indicates that the vibration level of the generator bearing of the wind turbine at a certain moment exceeds the safe range, and it can be determined that the wind turbine is in a fault state at a certain moment, and generate fault warning information, and then display the fault warning information. The fault warning information at this time may include information such as fault type, severity, occurrence time and occurrence location, so as to remind the operation and maintenance personnel to take timely measures for inspection and maintenance to avoid further expansion of the fault.

[0053] Optionally, S110 to S140 may be repeatedly executed to perform fault detection on wind turbines in other time periods.

[0054] According to the technical solution of the embodiment of the present application, the edge device can obtain the current vibration data of the generator bearing of the wind turbine set, and calculate the root mean square of the current vibration data to obtain the original root mean square, and then normalize the original root mean square to obtain the standard root mean square, and use the cloud server to normalize the standard root mean square to obtain the target root mean square, and then perform fault detection on the wind turbine set according to the target root mean square, thereby realizing the fault detection function of the wind turbine set. In the above technical scheme, by normalizing the original root mean square, the normalization degree of the target root mean square is improved, so that the target root mean square set is more in line with the standard normal distribution, thereby improving the applicability of the 3-sigma rule and reducing the false alarm rate of fault detection, thereby avoiding the problem of a high false alarm rate of fault detection caused by directly using the 3-sigma rule to determine the threshold range when the RMS set does not belong to the standard normal distribution, and improving the accuracy of fault detection; through the cloud-edge collaboration of edge devices and cloud servers, the efficiency of fault detection is improved, thereby avoiding the expansion of faults and the increase of downtime, thereby improving the operating efficiency and reliability of wind turbines, and reducing operation and maintenance costs.

[0055] The following further describes a wind turbine fault detection method provided by an embodiment of the present application. Figure 3 This is another flow chart of the fault detection method for a wind turbine provided in an embodiment of the present application. The embodiment of the present application is optimized based on the above embodiments.

[0056] See also Figure 3 The method of this embodiment includes but is not limited to the following steps:

[0057] S310, obtaining current vibration data of a generator bearing of a wind turbine generator set, and calculating a root mean square of the current vibration data to obtain an original root mean square.

[0058] The implementation process and technical principle of S310 are the same as those of S110, and reference may be made to the description of S110 above, which will not be repeated here.

[0059] S320, using a mean-variance algorithm to normalize the original root mean square to obtain an intermediate root mean square.

[0060] Among them, the intermediate root mean square is the result obtained after the original root mean square is normalized using the mean-variance algorithm. At this time, the intermediate root mean square eliminates the dimension and scale differences of the original root mean square, and the intermediate root mean square set corresponding to the preset time period is close to the standard normal distribution.

[0061] Specifically, after obtaining the original root mean square set corresponding to the preset time period, for each moment in the preset time period, any moment can be selected as the current moment, and then the original root mean square of the current moment is normalized using the mean-variance algorithm to obtain the intermediate root mean square of the current moment, and then other moments are selected as the current moment, and the above steps are repeated to obtain the intermediate root mean square of each moment, that is, the intermediate root mean square set corresponding to the preset time period. At this time, compared with the original root mean square set, the intermediate root mean square set is closer to the standard normal distribution.

[0062] Furthermore, the original RMS is normalized using the mean-variance algorithm to obtain the intermediate RMS, including Sa1-Sa3:

[0063] Sa1. Obtain historical vibration data of the wind turbine generator set, and calculate the root mean square of the historical vibration data to obtain the historical root mean square.

[0064] Among them, the historical vibration data is obtained by sampling the generator bearings at historical moments using a sensor with a preset sampling frequency; the historical moment is the moment before the current moment, and the historical moment at this time can be multiple moments. For example, when the current moment is 13:00 on September 1, the historical moment can be 13:00 every day in the past three months.

[0065] The historical RMS is the result obtained by performing RMS calculation on the historical vibration data collected by the sensor at the same historical moment, which reflects the vibration level of the generator bearing at a specific historical moment.

[0066] Specifically, multiple preset historical moments can be obtained, and then for each of the multiple historical moments, the historical vibration data corresponding to each historical moment can be obtained from the buffer of the edge device, that is, after the edge device uses the sensor to collect the vibration data of each moment in real time, the vibration data can be stored in the buffer for subsequent use; then the root mean square calculation formula is used to calculate the root mean square of the historical vibration data corresponding to each historical moment, and the historical root mean square corresponding to each historical moment is obtained, thereby obtaining multiple historical root mean squares.

[0067] For example, it is known that the current time is 13:00 on September 1, historical time 1 is 13:00 on August 31, historical time 2 is 13:00 on August 30, and historical time 3 is 13:00 on August 29. At this time, the historical vibration data 1 corresponding to the historical time 1, the historical vibration data 2 corresponding to the historical time 2, and the historical vibration data 3 corresponding to the historical time 3 are respectively obtained, and then the root mean square of the historical vibration data 1 is calculated to obtain the historical root mean square 1 corresponding to the historical time 1, the root mean square of the historical vibration data 2 is calculated to obtain the historical root mean square 2 corresponding to the historical time 2, and the root mean square of the historical vibration data 3 is calculated to obtain the historical root mean square 3 corresponding to the historical time 3.

[0068] Optionally, in another implementation, the edge device may directly obtain the historical RMS corresponding to each historical moment from the cache area.

[0069] Sa2. Calculate the mean of the historical RMS to obtain the historical mean, and calculate the mean square deviation of the historical RMS to obtain the historical mean square deviation.

[0070] The historical mean is the average of multiple historical RMS values. The historical mean square deviation is the standard deviation of multiple historical RMS values, which is used to measure the degree of deviation between the historical RMS values ​​and the historical mean.

[0071] Specifically, after obtaining the historical RMSs corresponding to multiple historical moments, the mean calculation formula can be used to calculate the mean of the multiple historical RMSs to obtain the historical mean, and then the mean square deviation calculation formula can be used to calculate the mean square deviation of the multiple historical RMSs to obtain the historical mean square deviation.

[0072] Sa3. Determine the intermediate root mean square based on the original root mean square, historical mean and historical root mean square deviation.

[0073] Specifically, after obtaining the historical mean and historical mean square error, the intermediate root mean square can be determined according to the original root mean square, the historical mean and the historical mean square error. That is, the difference between the original root mean square and the historical mean at the current moment can be calculated, and the ratio between the difference and the historical mean square error can be calculated. Then, the ratio is determined as the intermediate root mean square at the current moment. That is, the calculation formula of the intermediate root mean square can be:

[0074]

[0075] in, represents the middle RMS at the current moment, represents the historical mean, σ RMS represents the historical mean square error.

[0076] In an embodiment of the present application, the original root mean square is normalized by using the mean-variance algorithm and the law of large numbers, thereby improving the normalization degree of the intermediate root mean square, thereby making the intermediate root mean square set closer to the standard normal distribution, providing data support for the subsequent determination of the standard root mean square, thereby reducing the false alarm rate of fault detection and improving the accuracy of fault detection.

[0077] S330, using the inverse error function to normalize the intermediate root mean square to obtain a standard root mean square.

[0078] Among them, the inverse error function is the inverse function of the error function (Error Function, referred to as erf); the error function is the cumulative distribution function of the standard normal distribution, and its mathematical expression is:

[0079]

[0080] Among them, z represents the input value of the error function; π represents the ratio of pi; u represents the integral variable, which changes from -z to z during the integration process. It has no direct physical or statistical meaning and is only an intermediate variable used for integration operations. The inverse error function can be expressed as erf -1 (z).

[0081] Specifically, after obtaining the intermediate root mean square set, for each moment in the preset time period, any moment can be selected as the current moment, and then the intermediate root mean square of the current moment is normalized using the inverse error function to obtain the standard root mean square of the current moment, and then other moments are selected as the current moment, and the above steps are repeated to obtain the standard root mean square of each moment, that is, the standard root mean square set corresponding to the preset time period. At this time, compared with the intermediate root mean square set, the standard root mean square set is closer to the standard normal distribution. After that, S340 or S350 can be executed.

[0082] Furthermore, the intermediate RMS is normalized using the inverse error function to obtain the standard RMS, including Sb1-Sb2:

[0083] Sb1. Map the intermediate RMS to a preset interval to obtain a candidate RMS.

[0084] The preset interval is a predefined range of values ​​used to map the intermediate root mean square to a specific interval; illustratively, since the inverse error function is used to convert data between -1 and 1 back to a real number, the preset interval can be (-1,1).

[0085] The candidate root mean square is a result obtained by mapping the intermediate root mean square to a preset interval, that is, the candidate root mean square is within the preset interval.

[0086] Specifically, the intermediate root mean square at the current moment can be mapped to a preset interval to obtain a candidate root mean square at the current moment, so that the candidate root mean square at the current moment is within the preset interval, for example, the candidate root mean square at the current moment is within (-1,1); that is, the calculation formula of the candidate root mean square can be:

[0087]

[0088] Among them, α t Represents the candidate RMS at the current moment.

[0089] Sb2. Use the inverse error function to perform nonlinear transformation on the candidate RMS to obtain the standard RMS.

[0090] Specifically, after obtaining the candidate root mean square at the current moment, the inverse error function value corresponding to the candidate root mean square at the current moment can be determined based on the inverse error function, that is, the inverse error function value of the inverse error function can be determined by using a table lookup method. Specifically, the inverse error function value of each data within a preset interval is calculated in advance, and the data and its corresponding inverse error function value are stored in a preset relationship table, and then the preset relationship table is queried based on the candidate root mean square at the current moment to obtain the inverse error function value corresponding to the candidate root mean square at the current moment.

[0091] Afterwards, the product between the inverse error function value and the set value can be calculated to scale the inverse error function value to obtain the standard root mean square at the current moment. Therefore, the calculation formula of the standard root mean square can be:

[0092]

[0093] in, represents the standard root mean square at the current moment; K represents a set value, for example, K can be

[0094] Optionally, the standard RMS set corresponding to the wind turbine in the preset time period can be expressed as Here, p represents the number of each moment in the preset time period, and each moment is separated by 1 second.

[0095] In an embodiment of the present application, the intermediate root mean square is normalized by using an inverse error function, thereby improving the degree of normalization of the standard root mean square. That is, compared with the intermediate root mean square set, the standard root mean square set is closer to the standard normal distribution, which provides data support for the subsequent determination of the target root mean square and facilitates subsequent data analysis or model prediction, thereby reducing the false alarm rate of fault detection and improving the accuracy of fault detection.

[0096] S340, sending the standard root mean square to the cloud server, so that the cloud server uses a pre-trained neural network model to normalize the standard root mean square to obtain the target root mean square.

[0097] Among them, the neural network model is a model pre-trained by the cloud server based on the vibration data of the wind turbines in the same wind farm, and is used to normalize the root mean square of the same wind farm to make the data more consistent with the standard normal distribution.

[0098] Specifically, after obtaining the standard RMS set corresponding to the preset time period, the edge device can send the standard RMS set to the cloud server; then, the cloud server can receive the standard RMS set sent by the edge device, and then input the standard RMS set into a pre-trained neural network model to normalize the standard RMS set using the neural network model, and determine the output value of the neural network model as the target RMS set, at which time the target RMS set is closer to the strong assumption of the standard normal distribution and is a dimensionless value. After that, S370 can be executed.

[0099] Optionally, the neural network model can be trained in the following manner: using sample data as a training data set, inputting the training data set into an initial neural network model to obtain a training output, and using the difference between the training output and a standard normal distribution as a loss to optimize the initial neural network model to obtain a neural network model; that is, the training of the neural network model is unsupervised training, and the loss function used is a negative log-likelihood loss function, which is used to calculate the difference between the training output and the standard normal distribution.

[0100] The sample data may include the standard root mean square of the historical vibration data of each wind turbine in the preset wind farm; the preset wind farm is a specific wind farm set in advance, and a wind farm may include multiple wind turbines and multiple edge devices, and one wind turbine corresponds to one edge device; the standard root mean square of the historical vibration data is the data obtained after the root mean square of the historical vibration data is normalized using S320 and S330. Optionally, the training data set can be expressed as φ={θ1,θ2,...,θ m}, where θ m represents the standard RMS set of the mth wind turbine generator set before a preset time period, and m represents the number of wind turbine generator sets included in the preset wind farm.

[0101] The initial neural network model is a neural network structure that has been constructed but has not been optimized by any training data before the training process begins. Exemplarily, the initial neural network model may be a probabilistic neural network (PNN).

[0102] Specifically, the cloud server can obtain the standard root mean square of the historical vibration data of each wind turbine from each edge device in the preset wind farm, and form these standard root mean squares into a training data set, and then input the training data set into the initial neural network model to obtain the training output, and use the negative log-likelihood loss function to calculate the difference between the training output and the standard normal distribution to obtain the loss value, and then use the back propagation algorithm and the gradient descent method to optimize the initial neural network model, aiming to adjust the weights and biases in the initial neural network model to minimize the loss value and obtain the final neural network model; that is, the training process of the neural network model is to transform the training data set φ={θ1,θ2,...,θ m} is fitted with the standard normal distribution N(0,1), which can be expressed as:

[0103]

[0104] in, represents the training output; F(·) represents the activation function, which can be the Sigmoid activation function; b i represents the bias of the i-th layer; l i represents the weight of the i-th layer; g represents the number of layers of the neural network, which is in the range of [2,15].

[0105] It should be noted that the neural network model obtained by training based on the standard root mean square of historical vibration data of each wind turbine in a preset wind farm can normalize the root mean square of each wind turbine in the preset wind farm, that is, different wind turbines in the same wind farm can be detected using the same fault threshold range, reducing the multi-scale problem of different wind turbines.

[0106] In an embodiment of the present application, the neural network model is unsupervisedly trained based on the standard root mean square of the historical vibration data of each wind turbine in a preset wind farm, and the negative log-likelihood loss function is selected as the training target, which improves the normalization degree of the neural network model, and then improves the normalization degree of the target root mean square, thereby providing strong support for subsequent fault diagnosis.

[0107] S350, obtaining a pre-trained neural network model from a cloud server.

[0108] Specifically, after the cloud server completes the training of the neural network model, it can send the trained neural network model to each edge device in the preset wind farm; thereafter, each edge device in the preset wind farm can receive the neural network model sent by the cloud server.

[0109] S360. Use a neural network model to normalize the standard root mean square to obtain a target root mean square.

[0110] Specifically, after obtaining the standard root mean square set and neural network model corresponding to the preset time period, for each moment in the preset time period, the standard root mean square of the current moment can be input into the pre-trained neural network model, so as to use the neural network model to normalize the standard root mean square of the current moment, and determine the output value of the neural network model as the target root mean square of the current moment, so as to obtain the target root mean square set corresponding to the preset time period. At this time, compared with the standard root mean square set, the target root mean square set is closer to the strong assumption of the standard normal distribution and is a dimensionless value. S370 can be executed afterwards.

[0111] It should be noted that S340, S350 and S360 are parallel schemes, and any one of the schemes can be selected for execution according to actual conditions; there is no obvious order of execution of S340, S350 and S360, and the execution order can be determined according to actual conditions, and this embodiment does not make specific limitations on this.

[0112] S370. Perform fault detection on the wind turbine generator set according to the target root mean square.

[0113] The implementation process and technical principle of S370 are the same as those of S140, and reference may be made to the above description of S140, which will not be repeated here.

[0114] The technical solution of the embodiment of the present application can obtain the current vibration data of the generator bearing of the wind turbine generator set, and calculate the root mean square of the current vibration data to obtain the original root mean square, and then use the mean-variance algorithm to normalize the original root mean square to obtain an intermediate root mean square, and use the inverse error function to normalize the intermediate root mean square to obtain a standard root mean square. The normalization degree of the standard root mean square is improved by the double normalization processing of the mean-variance algorithm and the inverse error function, so that the standard root mean square set is closer to the standard normal distribution, which provides data support for the subsequent determination of the target root mean square; then, the standard root mean square is sent to the cloud server, so that the cloud server uses a pre-trained neural network model to normalize the standard root mean square to obtain the target root mean square. Alternatively, a pre-trained neural network model is obtained from a cloud server, and the standard root mean square is normalized using the neural network model to obtain the target root mean square, which is then normalized for the third time using the neural network model to further improve the normalization degree of the target root mean square; after three normalization processes in sequence, the normalization degree of the target root mean square is improved, making the target root mean square set closer to the standard normal distribution, thereby improving the applicability of the 3-sigma rule, reducing the false alarm rate of fault detection, and improving the accuracy of fault detection; by training the neural network model on a cloud server and performing statistical calculations on edge devices, the advantages of cloud-edge collaboration are fully utilized, the problem of insufficient computing power of edge devices is avoided, and the efficiency of fault detection is thereby improved.

[0115] Figure 4 is a schematic diagram of a structure of a fault detection device for a wind turbine generator set provided in an embodiment of the present application, with reference to Figure 4 , the fault detection device of the wind turbine generator set may include:

[0116] The calculation module 410 is used to obtain the current vibration data of the generator bearing of the wind turbine generator set, and calculate the root mean square of the current vibration data to obtain the original root mean square, where the current vibration data is obtained by sampling the generator bearing at the current moment using a sensor with a preset sampling frequency;

[0117] The first processing module 420 is used to perform normalization processing on the original root mean square to obtain a standard root mean square;

[0118] The second processing module 430 is used to perform normalization processing on the standard root mean square using the cloud server to obtain a target root mean square;

[0119] The fault detection module 440 is used to perform fault detection on the wind turbine generator system according to the target RMS value.

[0120] In one embodiment, the second processing module 430 is specifically used for:

[0121] The standard root mean square is sent to the cloud server, so that the cloud server uses the pre-trained neural network model to normalize the standard root mean square to obtain the target root mean square.

[0122] In one embodiment, the second processing module 430 is further used for:

[0123] Get the pre-trained neural network model from the cloud server;

[0124] The standard RMS is normalized using a neural network model to obtain the target RMS.

[0125] In one embodiment, the first processing module 420 is specifically used for:

[0126] The original RMS is normalized using the mean-variance algorithm to obtain the intermediate RMS;

[0127] The intermediate RMS is normalized using the inverse error function to obtain the standard RMS.

[0128] In one embodiment, the first processing module 420 uses a mean-variance algorithm to normalize the original root mean square to obtain an intermediate root mean square, including:

[0129] Acquire historical vibration data of the wind turbine generator set, and calculate the root mean square of the historical vibration data to obtain the historical root mean square, where the historical vibration data is obtained by sampling the generator bearing at a historical moment using a sensor with a preset sampling frequency;

[0130] Calculate the mean of the historical RMS to obtain the historical mean, and calculate the mean square deviation of the historical RMS to obtain the historical mean square deviation;

[0131] Determine the intermediate RMS based on the raw RMS, historical mean, and historical RMS deviation.

[0132] In one embodiment, the first processing module 420 uses an inverse error function to normalize the intermediate root mean square to obtain a standard root mean square, including:

[0133] Mapping the intermediate RMS to a preset interval to obtain a candidate RMS;

[0134] The candidate RMS is transformed nonlinearly using the inverse error function to obtain the standard RMS.

[0135] In one embodiment, the neural network model in the second processing module 430 is trained in the following manner:

[0136] The sample data is used as a training data set, and the training data set is input into an initial neural network model to obtain a training output. The difference between the training output and a standard normal distribution is used as a loss to optimize the initial neural network model to obtain a neural network model; the sample data includes the standard root mean square of historical vibration data of each wind turbine in a preset wind farm.

[0137] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional modules is used as an example for illustration. In actual applications, the above-mentioned functions can be assigned to different functional modules as needed, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. The specific working process of the functional modules described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0138] The fault detection device for a wind turbine set provided in this embodiment can be applied to the fault detection method for a wind turbine set provided in any of the above embodiments, and has corresponding functions and beneficial effects.

[0139] Figure 5 It is a structural diagram of the edge device provided in an embodiment of the present application. Figure 5 A block diagram of an exemplary edge device 11 suitable for implementing the embodiments of the present application is shown. Figure 5 The edge device 11 shown is merely an example and should not bring any limitation to the function and scope of use of this embodiment.

[0140] like Figure 5As shown, the edge device 11 is in the form of a general-purpose computing electronic device. The components of the edge device 11 may include, but are not limited to: one or more processors or processing units 16, a system memory 28, and a bus 18 connecting different system components (including the system memory 28 and the processing unit 16).

[0141] Bus 18 represents one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, a processor or a local bus using any of a variety of bus architectures. By way of example, these architectures include, but are not limited to, an Industry Standard Architecture (ISA) bus, a Micro Channel Architecture (MAC) bus, an Enhanced ISA bus, a Video Electronics Standards Association (VESA) local bus, and a Peripheral Component Interconnect (PCI) bus.

[0142] The edge device 11 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by the edge device 11, including volatile and non-volatile media, removable and non-removable media.

[0143] The system memory 28 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) 30 and / or cache memory 32. The edge device 11 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, the storage system 34 may be used to read and write non-removable, non-volatile magnetic media ( Figure 5 not shown, usually called a "hard drive"). Although Figure 5 Not shown, a disk drive for reading and writing to a removable non-volatile disk (e.g., a "floppy disk"), and an optical drive for reading and writing to a removable non-volatile optical disk (e.g., a CD-ROM, a DVD-ROM, or other optical medium) may be provided. In these cases, each drive may be connected to the bus 18 via one or more data medium interfaces. The system memory 28 may include at least one program product having a set (e.g., at least one) of program modules that are configured to perform the functions of the various embodiments of the present application.

[0144] A program / utility 40 having a set (at least one) of program modules 42 may be stored, for example, in system memory 28, such program modules 42 including, but not limited to, an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment. Program modules 42 generally perform the functions and / or methods of the embodiments described herein.

[0145] The edge device 11 may also communicate with one or more external devices 14 (e.g., keyboards, pointing devices, displays 24, etc.), one or more devices that enable a user to interact with the edge device 11, and / or any device that enables the edge device 11 to communicate with one or more other computing devices (e.g., a network card, a modem, etc.). Such communication may be performed through an input / output (I / O) interface 22. Furthermore, the edge device 11 may also communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) through a network adapter 20.

[0146] like Figure 5 As shown, the network adapter 20 communicates with other modules of the edge device 11 via the bus 18. It should be understood that although Figure 5 Not shown, other hardware and / or software modules may be used in conjunction with the edge device 11, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0147] The processing unit 16 executes various functional applications and page displays by running the programs stored in the system memory 28, for example, implementing a fault detection method for a wind turbine set provided in any embodiment of the present application.

[0148] An embodiment of the present application provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, a fault detection method for a wind turbine set provided in any embodiment of the present application is implemented.

[0149] The computer storage medium of this embodiment can adopt any combination of one or more computer-readable media. Computer-readable media can be computer-readable signal media or computer-readable storage media. Computer-readable storage media can be, for example, but not limited to: electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or devices, or any combination of the above. More specific examples (non-exhaustive list) of computer-readable storage media include: electrical connections with one or more wires, portable computer disks, hard disks, random access memories (RAM), read-only memories (ROM), erasable programmable read-only memories (EPROM or flash memory), optical fibers, portable compact disk read-only memories (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the above. In this document, computer-readable storage media can be any tangible medium containing or storing programs, which can be used by instruction execution systems, devices or devices or used in combination with them.

[0150] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, which carry computer-readable program code. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than a computer-readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0151] The program code embodied on the computer readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

[0152] Computer program code for performing the operation of the present application can be written in one or more programming languages ​​or combinations thereof, including object-oriented programming languages, such as Java, Smalltalk, C++, and conventional procedural programming languages, such as "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as an independent software package, partially on the user's computer, partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computer (e.g., using an Internet service provider to connect through the Internet).

[0153] It should be understood by those skilled in the art that the above modules or steps of the present application can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, optionally, they can be implemented by a program code executable by a computer device, so that they can be stored in a storage device and executed by the computing device, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present application is not limited to any specific combination of hardware and software.

[0154] In addition, the acquisition, storage, use, and processing of data in the technical solution of this application comply with the relevant provisions of national laws and regulations.

[0155] Note that the above are only preferred embodiments of the present application and the technical principles used. Those skilled in the art will understand that the present application is not limited to the specific embodiments herein, and that various obvious changes, readjustments and substitutions can be made by those skilled in the art without departing from the scope of protection of the present application. Therefore, although the present application is described in more detail through the above embodiments, the present application is not limited to the above embodiments, and may include more other equivalent embodiments without departing from the inventive concept of the present application, and the scope of the present application is determined by the scope of the appended claims.

Claims

1. A fault detection method for a wind turbine generator set, It is characterized in that Applied to edge devices, the method includes: Acquire current vibration data of a generator bearing of a wind turbine generator set, and calculate a root mean square of the current vibration data to obtain an original root mean square, wherein the current vibration data is obtained by sampling the generator bearing at the current moment using a sensor with a preset sampling frequency; Normalizing the original root mean square to obtain a standardized root mean square; Using a cloud server to normalize the standard root mean square to obtain a target root mean square; Fault detection is performed on the wind turbine generator set according to the target root mean square.

2. The fault detection method for a wind turbine generator set according to claim 1, It is characterized in that The method of normalizing the standard root mean square by using a cloud server to obtain a target root mean square includes: The standard root mean square is sent to the cloud server, so that the cloud server uses a pre-trained neural network model to normalize the standard root mean square to obtain the target root mean square.

3. The fault detection method for a wind turbine generator set according to claim 1, It is characterized in that The method of normalizing the standard root mean square by using a cloud server to obtain a target root mean square includes: Obtaining a pre-trained neural network model from the cloud server; The standard root mean square is normalized using the neural network model to obtain the target root mean square.

4. The fault detection method for a wind turbine generator set according to claim 1, It is characterized in that The normalizing the original root mean square to obtain a standard root mean square includes: Normalizing the original root mean square using a mean-variance algorithm to obtain an intermediate root mean square; The intermediate root mean square is normalized using an inverse error function to obtain the standard root mean square.

5. The fault detection method for a wind turbine generator set according to claim 4, It is characterized in that The method of normalizing the original root mean square by using the mean-variance algorithm to obtain an intermediate root mean square comprises: Acquire historical vibration data of the wind turbine generator set, and calculate the root mean square of the historical vibration data to obtain a historical root mean square, wherein the historical vibration data is obtained by sampling the generator bearing at a historical moment using a sensor with a preset sampling frequency; Calculating the mean of the historical root mean square to obtain the historical mean, and calculating the mean square deviation of the historical root mean square to obtain the historical mean square deviation; The intermediate root mean square is determined according to the original root mean square, the historical mean and the historical root mean square deviation.

6. The fault detection method for a wind turbine generator set according to claim 4, It is characterized in that The method of normalizing the intermediate root mean square by using an inverse error function to obtain the standard root mean square comprises: Mapping the intermediate root mean square into a preset interval to obtain a candidate root mean square; The candidate RMS is nonlinearly transformed using an inverse error function to obtain the standard RMS.

7. The fault detection method for a wind turbine generator set according to claim 2 or 3, It is characterized in that The neural network model is trained in the following way: The sample data is used as a training data set, and the training data set is input into an initial neural network model to obtain a training output. The difference between the training output and a standard normal distribution is used as a loss to optimize the initial neural network model to obtain the neural network model; the sample data includes the standard root mean square of historical vibration data of each wind turbine in a preset wind farm.

8. A fault detection device for a wind turbine generator set, It is characterized in that Applied to edge equipment, the device includes: A calculation module, used to obtain current vibration data of a generator bearing of a wind turbine generator set, and calculate a root mean square of the current vibration data to obtain an original root mean square, wherein the current vibration data is obtained by sampling the generator bearing at the current moment using a sensor with a preset sampling frequency; A first processing module is used to perform normalization processing on the original root mean square to obtain a standard root mean square; A second processing module is used to perform normalization processing on the standard root mean square by using a cloud server to obtain a target root mean square; A fault detection module is used to perform fault detection on the wind turbine generator set according to the target root mean square.

9. An edge device, It is characterized in that The edge device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the fault detection method for a wind turbine set as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, It is characterized in that When the program is executed by a processor, the fault detection method for a wind turbine set as claimed in any one of claims 1 to 7 is implemented.

Images