Wind driven generator winding temperature data processing method, device and equipment

Abstract

The invention provides a wind driven generator winding temperature data processing method, device and equipment, belongs to the technical field of wind driven generator monitoring, and solves the problems of low accuracy and poor real-time performance of wind driven generator winding temperature prediction. The method comprises the following steps: acquiring current operation data of the wind driven generator; performing conversion processing on the current operation data, and determining a target image matrix; performing noise elimination and feature enhancement processing on the target image matrix to obtain a target feature image; inputting the target feature image into a temperature prediction model to obtain temperature prediction data; and performing early warning on the temperature of the wind driven generator winding according to the temperature prediction data and the dynamic threshold value. According to the scheme, noise is effectively removed, key features are reserved, the stability of temperature prediction is improved, the service life of the wind driven generator is prolonged, and economic losses are reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of wind generator monitoring, in particular to a wind generator winding temperature data processing method, device and equipment. BACKGROUND

[0002] With the development of economy and industry, the limitations of traditional fossil energy are increasingly evident. Its non-renewable nature cannot meet the growing social energy demand, and the environmental pollution problems caused by its use are increasingly serious. In this context, wind energy, as a clean and renewable energy form, has attracted widespread attention. Wind power generation technology is becoming increasingly mature, and wind power is being used on a large scale worldwide. As the key equipment for converting wind energy, the health status monitoring of wind generators has become the focus of the industry. Generators, as the core components of wind turbine generators, directly affect the stability and power generation efficiency of the entire power generation system. Among the many fault types of generators, stator winding faults are relatively common. Currently, various technical means have been developed for wind generator stator winding temperature monitoring. For example, temperature sensors are placed on the generator stator winding to collect temperature data in real time; data acquisition and monitoring control systems are used to transmit and store temperature data; some studies also attempt to use data analysis algorithms to analyze temperature data to assess the operating status of the generator.

[0003] However, the existing technology still has many deficiencies in wind generator winding temperature data processing. First, in the data collection link, due to the complex operating environment of wind generators, temperature sensors are easily affected by electromagnetic interference, vibration and other factors, resulting in noise and errors in the collected temperature data, reducing the accuracy and reliability of the data. Second, in terms of data processing, existing methods focus more on simple data statistics and threshold judgment, lacking deep mining of temperature data features. For example, it is difficult to accurately identify the potential trends and abnormal patterns of stator winding temperature changes, making it difficult to predict possible over-temperature faults in advance. For complex and variable fault causes such as cooling system failure and winding insulation failure, existing methods are difficult to effectively distinguish and diagnose. Third, in terms of data warning, the existing warning mechanism is not flexible and accurate. A fixed temperature threshold is usually used for warning, but in actual operation, the operating conditions of the generator are complex and variable, and the fixed threshold cannot adapt to temperature changes under different operating conditions, easily leading to false positives or false negatives, and failing to provide timely and accurate warning information to maintenance personnel, thereby affecting the timely handling of generator stator winding faults and increasing the risk of secondary damage to the generator, adversely affecting the stable operation and economic benefits of the wind power generation system. Therefore, it is of great practical significance to develop an efficient and accurate wind generator winding temperature data processing method. SUMMARY

[0004] The application provides a wind generator winding temperature data processing method, device and equipment, and solves the problems of low accuracy and poor real-time performance of wind generator winding temperature prediction.

[0005] To solve the above technical problems, the technical scheme of the application is as follows: The application provides a wind generator winding temperature data processing method, device and equipment, and solves the problems of low accuracy and poor real-time performance of wind generator winding temperature prediction. Obtaining current operation data of the wind generator; Converting and processing the current operation data to determine a target image matrix; Performing noise elimination and feature enhancement processing on the target image matrix to obtain a target feature image; Inputting the target feature image into a temperature prediction model to obtain temperature prediction data; the temperature prediction model processes the target feature image through a kernel function to obtain an intermediate result, accumulatively sums and adds a bias term to the intermediate result to obtain the temperature prediction data; According to the temperature prediction data and a dynamic threshold, the temperature of the wind generator winding is prewarned.

[0006] Optionally, obtaining the current operation data of the wind generator comprises: According to a plurality of sensors of the device on the wind generator, the current operation data of the wind generator is obtained, and the current operation data of the wind generator comprises at least one of wind speed, generator active power, generator winding temperature, generator slip ring chamber temperature, pitch angle and motor current.

[0007] Optionally, converting and processing the current operation data to determine a target image matrix comprises: According to the wind speed and the generator active power in the current operation data, buffer boundary data is determined; According to the buffer boundary data, image size data is determined; According to the wind speed and a first scaling coefficient, first position data is determined; According to the generator active power and a second scaling coefficient, second position data is determined; According to the image size data, the first position data and the second position data, a target image matrix is determined, the size of the target image matrix is determined by the image size data, the target image matrix comprises a plurality of pixel points with a preset pixel value, and the positions of the pixel points are determined by the first position data and the second position data.

[0008] Optionally, performing noise elimination and feature enhancement processing on the target image matrix to obtain a target feature image comprises: By traversing the distances between the pixels in the target image matrix, multiple distance data points are obtained; Based on the multiple distance data, determine the average distance; The pixel values ​​of the pixels in the target image matrix are updated based on the mean distance to obtain the target feature image.

[0009] Optionally, the temperature prediction model is trained through the following process: Obtain historical performance data for wind turbines; The historical indicator data is transformed and processed to determine the image matrix; The image matrix is ​​subjected to noise reduction and feature enhancement processing to obtain a feature image; The feature images are subjected to image segmentation and data index extraction to obtain a cleaned dataset; The cleaned dataset is input into a support vector regression model for processing to obtain a temperature prediction model.

[0010] Optionally, the feature images are subjected to image segmentation and data index extraction processing to obtain a cleaned dataset, including: The feature image is binarized to obtain a binarized image; Perform an opening operation on the binarized image to obtain a denoised binarized image; The denoised binarized image is subjected to contour search processing to obtain preliminary mask data; The preliminary mask data is subjected to dilation and padding processing to obtain a preliminary data index; The preliminary data index is clustered to obtain the final data index; The final data index is subjected to data extraction processing to obtain the cleaned dataset.

[0011] Optionally, the cleaned dataset is input into a support vector regression model for processing to obtain a temperature prediction model, including: The training sample data in the cleaned dataset is input into the support vector regression model, which processes the training sample data through a kernel function to obtain intermediate results. The intermediate results are summed and a bias term is added to obtain the predicted data. The prediction data is compared with the training sample data to obtain the prediction error; Based on the prediction error, the weights, biases, and hyperparameters of the support vector regression model are adjusted to obtain the temperature prediction model.

[0012] Optionally, according to the temperature prediction data and the dynamic threshold, a temperature warning of the wind generator winding is performed, including: The temperature prediction data is compared with the dynamic threshold, and when the temperature prediction data is greater than the dynamic threshold, an abnormal temperature warning of the wind generator winding is issued.

[0013] Optionally, the wind generator winding temperature data processing method further includes: The temperature prediction data, the actual temperature data and preset warning temperature data are compared to determine a comparison result. According to the comparison result, the dynamic threshold is updated, wherein, When the temperature prediction data is less than the preset warning temperature data, and the preset warning temperature data is greater than the actual temperature data, a residual error of the temperature prediction data and the actual temperature data is calculated, and the dynamic threshold is updated according to the residual error. When the temperature prediction data is less than the preset warning temperature data, and the preset warning temperature data is less than the actual temperature data, a residual error of the temperature prediction data and the actual temperature data is calculated, and the residual error is accumulated, and the dynamic threshold is updated according to the accumulated residual error. When the temperature prediction data is greater than the preset warning temperature data, and the preset warning temperature data is greater than the actual temperature data, a residual error of the temperature prediction data and the actual temperature data is calculated, and the residual error is accumulated, and the dynamic threshold is updated according to the accumulated residual error.

[0014] The embodiment of the present application also provides a wind generator winding temperature data processing device, including: An acquisition module is configured to acquire current operation data of a wind generator. A processing module is configured to perform conversion processing on the current operation data to determine a target image matrix, perform noise elimination and feature enhancement processing on the target image matrix to obtain a target feature image, input the target feature image into a temperature prediction model to obtain temperature prediction data, and perform processing on the target feature image by the temperature prediction model through a kernel function to obtain an intermediate result, accumulate and sum the intermediate result and add a bias term to obtain the temperature prediction data. A warning module is configured to perform a temperature warning of the wind generator winding according to the temperature prediction data and a dynamic threshold.

[0015] The technical scheme of the present application at least includes the following effects: The scheme of the present application obtains current operation data of the wind driven generator, converts and processes the current operation data to determine a target image matrix, carries out noise elimination and feature enhancement processing on the target image matrix to obtain a target feature image, inputs the target feature image into a temperature prediction model to obtain temperature prediction data, the temperature prediction model processes the target feature image through a kernel function to obtain an intermediate result, accumulates and sums the intermediate result and adds a bias term to obtain the temperature prediction data, and according to the temperature prediction data and a dynamic threshold, the temperature of the wind driven generator winding is warned, noise is effectively removed and key features are retained, the stability of temperature prediction is improved, the service life of the wind driven generator is prolonged, and economic losses are reduced. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 is a flow chart of the wind driven generator winding temperature data processing method provided by the embodiment of the present application; Figure 2 is a pixel point neighborhood schematic diagram provided by the embodiment of the present application; Figure 3 is an image segmentation and cleaning data process schematic diagram provided by the embodiment of the present application; Figure 4 is a data processing process schematic diagram of the temperature prediction model provided by the embodiment of the present application; Figure 5 is a generator winding temperature warning process schematic diagram of the wind driven generator winding temperature data processing method provided by the embodiment of the present application; Figure 6 is a structural diagram of the wind driven generator winding temperature data processing device provided by the embodiment of the present application. DETAILED DESCRIPTION

[0017] Exemplary embodiments of the present application will be described in greater detail below with reference to the accompanying drawings. Although exemplary embodiments of the present application are shown in the drawings, it is understood that the present application can be implemented in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided so that the present application can be more thoroughly understood and the scope of the present application can be accurately conveyed to those skilled in the art.

[0018] As shown in Figure 1 , the embodiment of the present application proposes a wind driven generator winding temperature data processing method, comprising: Step 11, obtaining current operation data of the wind driven generator; Step 12, converting and processing the current operation data to determine a target image matrix; Step 13, carrying out noise elimination and feature enhancement processing on the target image matrix to obtain a target feature image; Step 14, input the target feature image into the temperature prediction model to obtain temperature prediction data; the temperature prediction model processes the target feature image through a kernel function to obtain an intermediate result, accumulates and sums the intermediate result, and adds a bias term to obtain the temperature prediction data; Step 15, according to the temperature prediction data and the dynamic threshold, the temperature of the wind generator winding is prewarned.

[0019] In this embodiment, various types of sensors are installed inside the wind generator, such as temperature sensors, current sensors, voltage sensors, and speed sensors. Among them, the temperature sensor directly measures the winding temperature; the current sensor monitors the winding current; the voltage sensor obtains the voltage across the winding; and the speed sensor measures the speed of the generator rotor. These sensors collect data in real time at a preset sampling frequency and transmit the data to the data acquisition system. At the same time, the control system of the wind generator itself will record some operating parameters, such as the power output and power factor of the generator, which reflect the overall working condition of the generator.

[0020] The collected raw data is cleaned to remove noise and outliers. For example, if a data point collected by the temperature sensor deviates significantly from the normal range, it can be considered an outlier and removed. The method based on standard deviation or the prediction model established based on historical data can be used to identify outliers. The cleaned data is arranged into a two-dimensional matrix according to the different types, where different types of data are arranged as rows or columns of the matrix, and the time series is arranged as another dimension of the matrix; for example, there are n types of data (such as temperature, current, voltage, etc.), and data at m time points are collected, so a n × m data matrix can be constructed. Then the data matrix is converted into an image matrix, that is, each element in the data matrix is mapped to a pixel value of a grayscale image. For example, the normalized data value is multiplied by 255 (for an 8-bit grayscale image) to obtain the corresponding pixel value. In this way, the data matrix is converted into a grayscale image matrix, i.e., a target image matrix.

[0021] After obtaining the target image matrix, an image filtering algorithm is used to eliminate noise in the target image matrix. For example, the average value of the pixels in the neighborhood can be used to replace the value of the center pixel, which is suitable for removing Gaussian noise. The image can also be converted to the frequency domain for processing. The image is converted from the spatial domain to the frequency domain through Fourier transform. In the frequency domain, noise is usually represented as high-frequency components. A low-pass filter can be used to filter out high-frequency noise, and then the image is converted back to the spatial domain through inverse Fourier transform. After removing the noise in the target image matrix, an edge detection algorithm is used to enhance the edge features in the image to facilitate more accurate detection of the edges in the image. After noise elimination and feature enhancement processing, the target feature image obtained can better reflect the temperature-related feature information of the wind turbine generator winding.

[0022] After obtaining the target feature image, the target feature image is input into a temperature prediction model to calculate temperature prediction data. The temperature prediction model uses a kernel function-based method for processing. The role of the kernel function is to map the input target feature image to a high-dimensional feature space, so that the data in the high-dimensional space is more easily linearly classified or regressed. Commonly used kernel functions include linear kernel functions, polynomial kernel functions, and Gaussian radial basis kernel functions. Through kernel function processing, the features in the target feature image are extracted and converted to obtain intermediate results. The intermediate results obtained by kernel function processing are accumulated and summed, and a bias term is added to obtain temperature prediction data. The accumulation and summation process can be regarded as a comprehensive consideration of all feature information, and the bias term is used to adjust the baseline value of the prediction result.

[0023] After obtaining the temperature prediction data, the statistical characteristics of the temperature, such as the mean and standard deviation, are calculated based on the historical temperature data, and then a dynamic threshold is set. For example, the threshold can be set to the mean plus a certain multiple of the standard deviation (such as mean + 3 times standard deviation), and when the temperature prediction data exceeds the threshold, it is considered that there may be a temperature anomaly. At the same time, the threshold is dynamically adjusted in combination with the current operating state of the wind turbine generator (such as power output, speed, etc.). For example, when the generator is in a high-power output state, the heat generation of the winding will increase, and at this time the threshold can be appropriately increased; conversely, when the generator is in a low-power output state, the threshold can be correspondingly reduced.

[0024] According to the comparison result of the temperature prediction data and the dynamic threshold, different warning levels are divided. For example, when the temperature prediction data exceeds the threshold but does not exceed a certain percentage (such as 10%), a yellow warning is issued, prompting the operation and maintenance personnel to pay attention to the winding temperature condition; when the temperature prediction data exceeds the threshold by a certain percentage (such as 10%-20%), an orange warning is issued, suggesting that the operation and maintenance personnel check; when the temperature prediction data exceeds the threshold by a large percentage (such as 20% or more), a red warning is issued, indicating that there may be a serious temperature anomaly, and immediate measures (such as shutdown inspection) need to be taken. Once the warning is triggered, the system will publish the warning information through various ways, such as SMS, email, system interface prompt, etc., to notify the relevant operation and maintenance personnel. The warning information should include the number of the wind turbine, the current temperature prediction data, the dynamic threshold, the warning level and the possible cause analysis, etc., so that the operation and maintenance personnel can understand the situation in time and take appropriate measures.

[0025] The technical scheme effectively removes noise and retains key features by image segmentation and cleaning data, improves the stability of temperature prediction, reduces false positives and false negatives, can monitor winding temperature changes in real time, predict over-temperature anomalies in advance, reduces the risk of shutdown due to temperature out of control, prolongs the service life of the wind turbine, and reduces economic losses.

[0026] In an optional embodiment of the present application, step 11 can include: Step 111, according to the device in the wind turbine, the current running data of the wind turbine is obtained from the plurality of sensors, the current running data of the wind turbine includes at least one of wind speed, generator active power, generator winding temperature, generator slip ring chamber temperature, pitch angle, motor current.

[0027] In this embodiment, the wind speed sensor can be installed on the top of the nacelle or the hub of the wind turbine, away from the shelter of the nacelle, to ensure accurate measurement of natural wind speed; by monitoring the wind speed in real time, the condition of wind resources can be understood, providing a basis for the operation control of the wind turbine. For example, when the wind speed is too low, the wind turbine may not start or the power generation efficiency is low; when the wind speed is too high, measures such as power limiting or shutdown need to be taken to protect the safety of the wind turbine.

[0028] The power sensor can be installed in the power transmission line between the generator and the power grid, directly measuring the active power output by the generator. The generator active power reflects the actual power generation capacity of the wind turbine. By monitoring the active power in real time, the power generation efficiency and performance of the wind turbine can be evaluated, and abnormal conditions in the power generation process can be found in time. For example, if the active power suddenly drops, it may be caused by generator failure, power grid problems or wind speed changes, etc., which need to be investigated and handled in time.

[0029] The generator winding temperature sensor should be installed at the hot spot of the winding to accurately reflect the highest temperature of the winding. It can be achieved by pre-embedding the temperature sensor in the winding. The generator slip ring chamber temperature sensor is installed inside the slip ring chamber, close to the slip ring and carbon brush, to monitor the temperature change of the slip ring chamber. High temperature of the generator winding will cause accelerated aging of the insulation material, reduce the insulation performance of the generator, and even cause short circuit failure. By monitoring the winding temperature in real time, cooling measures such as starting the cooling fan, increasing the flow of cooling liquid, etc. can be taken in time to ensure that the winding temperature is within the safe range. High temperature of the slip ring chamber will affect the normal work of the slip ring and carbon brush, causing the contact resistance to increase and the heat to intensify, and even causing the slip ring and carbon brush to burn out. Monitoring the temperature of the slip ring chamber can timely find the heat dissipation problem of the slip ring chamber and take corresponding measures to improve it, such as cleaning the dust in the slip ring chamber and checking the ventilation system, etc.

[0030] The pitch angle sensor can be installed on the mechanical transmission part of the variable pitch system, such as the variable pitch bearing or the variable pitch drive device, which can measure the pitch angle of the blade in real time. The pitch angle is an important parameter for controlling the power output of the wind turbine. By adjusting the pitch angle, the angle of attack of the wind turbine blade can be changed, so as to control the wind energy absorbed by the blade. At low wind speed, increasing the pitch angle appropriately can increase the lift of the blade and increase the power output of the generator; at high wind speed, reducing the pitch angle can limit the wind energy absorbed by the blade to prevent the generator from overloading. Monitoring the pitch angle can ensure the normal operation of the variable pitch system and realize the power optimization control of the wind turbine.

[0031] The current sensor can be installed at the outgoing end of the motor stator winding to measure the working current of the motor. When installing, attention should be paid to the selection of the range of the sensor to ensure that the current value of the motor under different operating conditions can be accurately measured. The motor current reflects the load condition of the generator. By monitoring the motor current, the overloading, short circuit and other faults of the generator can be found in time. For example, when the motor current suddenly increases, it may be caused by internal short circuit of the generator or sudden increase of the load, which needs to be checked in time to avoid the expansion of the fault.

[0032] In an optional embodiment of the present application, step 12 can include: Step 121, determining buffer boundary data according to the wind speed and generator active power in the current operation data; Step 122, determining image size data according to the buffer boundary data; Step 123, determining first position data according to the wind speed and a first scaling factor; Step 124, determining second position data according to the generator active power and a second scaling factor; In step 125, a target image matrix is determined according to the image size data, the first position data and the second position data, the size of the target image matrix being determined by the image size data, and the target image matrix including a plurality of pixel points with a preset pixel value, the positions of the pixel points being determined by the first position data and the second position data.

[0033] In this embodiment, first, the maximum and minimum values of the wind speed and the generator active power in the current running data are calculated to obtain the value range of the wind speed and the power; on the basis of the value range, a 1% buffer boundary and several additional boundary pixels are added to determine the buffer boundary data, specifically according to: ; ; ; ; The buffer boundary data is determined, wherein, wind buffer is the buffer boundary of the wind speed, wind max is the maximum value of the wind speed, wind min is the minimum value of the wind speed, power buffer is the buffer boundary of the generator active power, power max is the maximum value of the generator active power, power min is the minimum value of the generator active power, wind range is the buffer boundary data of the wind speed, power range is the buffer boundary data of the generator active power.

[0034] After the buffer boundary data is determined, the first scaling coefficient and the second scaling coefficient are used to specify the scaling of the image size to determine the image size data, specifically according to: ; ; The image size data is determined, wherein, M is the width data of the image, delta M is the first scaling coefficient, N is the height data of the image, delta N is the second scaling coefficient.

[0035] After determining the image size data, a two-dimensional image matrix of size (M, N) is created I ( y , x ) according to wind range and power range mapping relationship to (M, N) fills the specified position of the two-dimensional image matrix I ( y , x ) to 255; specifically, according to: M indices =( wind – wind min ) / delta M ; determining first position data, wherein, M indices is the first position data, wind is the wind speed; according to: N indices =( power max – power ) / delta N ; determining second position data, wherein, N indices is the second position data, power is the generator active power; filling the pixel value of the pixel point in the two-dimensional image matrix I ( y , x ) whose first position data is M indices and whose second position data is N indices to 255, thereby determining a target image matrix, that is: I ( y , x )= I ( M indices , N indices )=255.

[0036] In an optional embodiment of the present application, step 13 can include: step 131, traversing the distance between the pixel points in the target image matrix to obtain a plurality of distance data; Step 132: Determine the average distance based on the multiple distance data; Step 133: Update the pixel values ​​of the pixels in the target image matrix according to the mean distance to obtain the target feature image.

[0037] In this embodiment, after determining the target image matrix, starting from any pixel with a pixel value of 255... p In the beginning, such as Figure 2 As shown, the distances to other pixels with a value of 255 in the image are calculated along the 8-neighborhood directions. The average distance in the 8 directions is then used to replace the original value of 255, in order to eliminate noise data and retain the main features. Specifically, according to: ; Determine the set of direction vectors, where, directions It is a set of direction vectors, including 8 direction vectors; For each =255, calculate its value in the above context. The distances in the eight directions mentioned above, i.e. ; in, distance ( y , x () is a set of distances in 8 directions; d k For the first k Distance in each direction, d k The formula for calculation is: ; in, n From pixels ( y , x The number of steps to start in that direction, ) No. k One directional vector; Will After correction by the distance mean, for the current The values ​​are normalized to the range [0, 255] to obtain the target feature image, i.e.: ; ; in, mean () represents the mean function; The pixel value after normalization; I The pixel value of a pixel; max( I ) represents the maximum value of a pixel; min( I () represents the minimum value of a pixel.

[0038] As Figure 3 And Figure 4 In an optional embodiment of the present application, the temperature prediction model is trained by the following process: Step 141, obtaining historical index data of the wind turbine; Step 142, converting the historical index data to determine the image matrix; Step 143, performing noise elimination and feature enhancement processing on the image matrix to obtain the feature image; Step 144, performing image segmentation and data index extraction processing on the feature image to obtain the cleaned data set; Step 145, inputting the cleaned data set into the support vector regression model for processing to obtain the temperature prediction model.

[0039] In this embodiment, first, the historical index data of the wind turbine is obtained according to the method described in step 11; after obtaining the historical index data of the wind turbine, the historical index data of the wind turbine is processed in turn according to the methods described in steps 12 and 13 to obtain the feature image; After obtaining the feature image, image segmentation and data index extraction processing are performed on it to obtain the cleaned data set, the process includes: The feature image is binarized to obtain a binary image; specifically, the binary image of the feature image is obtained by a binarization operation function I feature ( y , x ): using a fixed threshold (e.g. 128) method to convert a grayscale image to a binary image I feature ( y , x ), where: the pixel value is greater than or equal to the threshold value, it is set to 255, otherwise it is 0.

[0040] After obtaining the binary image, an opening operation is performed on it to obtain a denoised binary image; specifically, a 2x2 rectangular kernel is used to perform an opening operation on the binary image I feature ( y , x ) to perform erosion and dilation operations to eliminate noise that is not in the main foreground part, the erosion operation expression is: I eroded ( y , x )= I feature ( y , x ) K , wherein, I eroded ( y , x ) is the binary image after erosion operation, is the erosion operator, K is a 2x2 rectangular kernel used to define the neighborhood range; only the foreground pixels completely covered by the kernel are preserved by erosion operation.

[0041] After obtaining the binary image after erosion operation, inflation operation is performed, and the expression is: I opened ( y , x ) = I eroded ( y , x ) ⊕ K, wherein, I opened ( y , x ) is the binary image after denoising, and ⊕ is the inflation operator; the inflation operation restores the foreground region of the main body eroded.

[0042] After obtaining the binary image after denoising, contour finding processing is performed to obtain preliminary mask data; specifically, the data information of the largest contour is found, and a mask image with the same size as the image but with all pixel values being 0 is created, and the target foreground is drawn on the mask using the obtained contour data information, with a value of 255: the contour coordinates of the largest connected region in the image are found using tools such as OpenCV C = {( x 1, y 1), ( ( x 2, y 2), …}, and then a full 0 matrix mask(y, x) with the same size as the binary image after denoising I opened ( y , x ) is created; the pixels within the contour C are filled with 255 to obtain preliminary mask data.

[0043] After obtaining the preliminary mask data, inflation filling processing is performed to obtain preliminary data index; specifically, to fill the possible target foreground recess, the mask image is inflated twice using a 3x3 rectangular kernel, and then the final target mask is obtained, and this process can be represented as: mask dilated = mask ⊕K' ⊕ K' ; wherein, K' is a 3x3 rectangular kernel used to dilate the foreground region; mask dilated is a preliminary data index used to ensure the foreground region is complete and continuous.

[0044] After obtaining the preliminary data index, it is clustered to obtain a final data index; the final data index is subjected to data extraction processing to obtain a cleaned data set; the specific process includes: after extracting the coordinates of the foreground in the mask, the data index is obtained by inverse mapping according to the above mapping formula from data to image. According to the obtained index, DBSCAN clustering is used to search for similar values nearby, and finally a cleaned data index is obtained, and data cleaning is completed to obtain data:

[0045] wherein, D clean is a cleaned data set; temp is a generator winding temperature; ambientemp is a generator slip ring chamber temperature; pitangle is a pitch angle; current is a motor current.

[0046] The training sample data in the cleaned data set is input into a support vector regression model, the support vector regression model processes the training sample data through a kernel function to obtain an intermediate result; the intermediate result is accumulated and summed and a bias term is added to obtain predicted data; the predicted data is compared with the training sample data to obtain a prediction error; according to the prediction error, the weight data, bias data and hyperparameters of the support vector regression model are adjusted to obtain a temperature prediction model. Specifically: First, the cleaned standardized features (winding temperature, slip ring chamber temperature, pitch angle, current) are input, and the target variable is the temperature change value y t = temp t - temp t-1 .

[0047] Then define the optimization goal of the support vector regression model: minimize the weight vector norm w ∥ 2 and the penalty term of the slack variable C ∑( ξ i + ξ i ), constraint the predicted value in - In-band spacing. Gaussian kernel is used K ( x i , x j )=exp[-∥ x i - x j ∥ 2 / (2 σ 2 )] mapping nonlinear relationship, after converting into dual problem, solving Lagrange multiplier through SMO algorithm, determining support vector and calculating weight w and bias b .

[0048] Finally cross-validation and grid search optimize hyperparameters (C, g, etc.) C , , σ ), finally obtain high-precision temperature change prediction model.

[0049] In an optional embodiment of the application, step 15 can include: Step 151, compare the temperature prediction data with the dynamic threshold, and when the temperature prediction data is greater than the dynamic threshold, issue a temperature abnormality early warning of the wind turbine generator winding.

[0050] In this embodiment, the dynamic threshold can be dynamically adjusted according to the actual running state and historical data of the wind turbine generator. There are many methods to determine the dynamic threshold, for example, based on statistical analysis of historical temperature data, considering the temperature distribution under different seasons and different wind speed intervals, calculating statistical quantities such as mean value and standard deviation of temperature, and then setting an initial threshold combined with a certain safety factor; at the same time, combined with the running parameters of the current wind turbine generator, such as power output size, environmental temperature, etc., the threshold is adjusted in real time. For example, when the generator is in a high-power output state, the winding heat increases, and the dynamic threshold will be correspondingly increased; when the environmental temperature rises, the initial temperature of the winding is higher, and the heat dissipation difficulty increases, and the dynamic threshold will also be appropriately adjusted.

[0051] When the temperature prediction data is greater than the dynamic threshold, that is, the early warning triggering condition is met. For example, if the dynamic threshold is set to 80 DEG C, and the current temperature prediction data is 85 DEG C, the system will determine that the temperature is abnormal, and trigger the early warning mechanism. In order to improve the accuracy and reliability of the early warning, multiple condition combination judgments can also be set. For example, in addition to the temperature prediction data being greater than the dynamic threshold, it is also required that the abnormal state lasts for a certain time, such as 5 minutes continuously, to trigger the early warning. In this way, false alarms caused by instantaneous data fluctuations or interference can be avoided.

[0052] The early warning mode includes local alarm and remote notification. When the local alarm is adopted, an audible and visual alarm device needs to be arranged in the control cabinet of the wind driven generator or the monitoring center. When the early warning is triggered, the alarm device will emit a loud alarm sound and flash a warning light to remind the on-site operation and maintenance personnel. At the same time, a relevant early warning information window will be popped up on the display screen of the monitoring center to display the number of the wind driven generator, the current temperature prediction data, the dynamic threshold value, the early warning time and other detailed information. When the remote notification is adopted, the early warning information can be sent to the relevant operation and maintenance personnel and management personnel through short message, email and instant messaging tool. The short message notification can directly send the number of the wind driven generator and the temperature abnormality to the mobile phone of the operation and maintenance personnel; the email notification can include a more detailed early warning report including the temperature change trend chart, historical data comparison and other information; the instant messaging tool notification can realize real-time communication, and the operation and maintenance personnel can immediately discuss and respond after receiving the notification.

[0053] As shown in FIG. 1, the wind driven generator winding temperature data processing method according to the present application includes the following steps. Figure 5 As shown in FIG. 1, the wind driven generator winding temperature data processing method according to the present application includes the following steps. Step 161, comparing the temperature prediction data, the actual temperature data with the preset early warning temperature data to determine the comparison result. Step 162, updating the dynamic threshold value according to the comparison result, wherein, when the temperature prediction data is less than the preset early warning temperature data, and the preset early warning temperature data is greater than the actual temperature data, calculating the residual error of the temperature prediction data and the actual temperature data, and updating the dynamic threshold value according to the residual error; when the temperature prediction data is less than the preset early warning temperature data, and the preset early warning temperature data is less than the actual temperature data, calculating the residual error of the temperature prediction data and the actual temperature data, accumulating the residual error, and updating the dynamic threshold value according to the accumulated residual error; when the temperature prediction data is greater than the preset early warning temperature data, and the preset early warning temperature data is greater than the actual temperature data, calculating the residual error of the temperature prediction data and the actual temperature data, accumulating the residual error, and updating the dynamic threshold value according to the accumulated residual error.

[0054] In this embodiment, the residual error of the temperature prediction data and the actual temperature data is obtained according to: , wherein, residue is the residual error, f ( y t is the temperature prediction data, y t is the actual temperature data; and according to: , to obtain a dynamic threshold, wherein threshold is the dynamic threshold, mean() is a mean calculation function, and std() is a standard deviation calculation function.

[0055] If the generator winding temperature change value at a certain moment is greater than threshold , it indicates that the winding temperature changes too much in 10 minutes, so a pre-warning is needed. If the actual temperature data exceeds the preset warning temperature, and the temperature prediction data is lower than the preset warning temperature, resulting in no pre-warning, the residual is put into the standard residual to recalculate and update the threshold; if the actual temperature data is lower than the preset warning temperature, and the temperature prediction data exceeds the preset warning temperature, resulting in a warning error, a plurality of groups are accumulated, and then the threshold of the residual feature is recalculated.

[0056] One specific embodiment of the wind generator winding temperature data processing method provided by the embodiment of the application is: Step 1, obtaining current operation data of the wind generator.

[0057] Specifically, historical index data is collected, the time interval is 10 minutes, and the historical index data includes characteristic indexes of wind condition segmentation and characteristic indexes of a temperature fitting model. The structure is as follows: ; , wherein D represents a complete historical time series data set; and are wind speed and generator active power, which are used as characteristic indexes of image segmentation; is generator winding temperature, which is a target monitored by a pre-warning model; is generator slip ring chamber temperature, which represents the environmental temperature of the generator; is a pitch angle, which can be adjusted to affect the output power of the generator, thereby affecting the winding temperature of the generator; is motor current, and when the current increases, the heat generation of the winding will significantly increase, thereby causing the winding temperature to rise.

[0058] Step 2, performing conversion processing on the current operation data to determine a target image matrix.

[0059] Specifically, first, the maximum and minimum values of and are calculated to obtain the value range of wind speed and power; on the basis of the above value range, 1% buffer boundary and additional several boundary pixels are added, 2 boundary pixels are added here, and and are used to specify and scale the image size, and the image scaling coefficients =0.1 and =7 are set to convert the two into image data with wind speed as width and power as heightI y x The relevant calculation formula is: ； ； ； ； ； ； Then, a two-dimensional image matrix of size (M, N) is created According to the mapping relationship of and to (M, N), the specified position of the image is filled with 255. After filling, the distance of the pixel value 255 in the image is calculated along the 8-neighbor direction of the pixel, and the average distance of the 8 directions is taken to replace the original 255 value, so as to eliminate noise data and retain the main features.

[0060] M indices wind wind min delta M N indices power max power delta N Step 3, noise elimination and feature enhancement processing is performed on the target image matrix to obtain a target feature image.

[0061] Specifically, first, according to: ; a set of direction vectors is determined, wherein directions is a set of direction vectors, including 8 direction vectors; For each =255, the distance in the above-mentioned 8 directions is calculated, that is ; wherein,​​​​​​​​​​​​distance y , x ) is a distance set of 8 directions; d k is the distance of the i-th direction, k k The calculation formula of is: d ; wherein, n is the step number along the direction starting from the pixel (i, j), (i, j) is the i-th direction vector; y x corrected by the distance mean value, the value of the current is normalized to [0, 255], so as to obtain the target feature image, that is: k ; ; wherein, mean () is a mean value function; is the pixel value of the pixel point after normalization; I is the pixel value of the pixel point; max() is the maximum value of the pixel point; min() is the minimum value of the pixel point. I I Step 4: input the target feature image into a temperature prediction model to obtain temperature prediction data.

[0062] Specifically, first, the cleaning data is operated using an image segmentation method, and according to the finally obtained image

[0063] , a binary image of the feature image I feature y x is obtained through a binary operation in OpenCV; A 2x2 rectangular kernel is used to perform an open operation on the image I feature y x to eliminate noise that is not in the main foreground part; After eliminating the noise, a contour finding operation in OpenCV is used to find the data information of the largest contour, and a mask image with the same size as the image I is created, all of which are 0, and the obtained contour data information is used to draw the target foreground in the mask, with a value of 255; ​​​​​​​​​​​​​​To fill the possible target foreground concave, a 3x3 rectangular kernel is used twice to dilate the mask image, and then the final target mask is obtained; After extracting the coordinates of the foreground in the mask, the index of the data is obtained by inverse mapping according to the mapping formula from the data to the image. According to the obtained index, DBSCAN clustering is used to search for similar values nearby, and finally the cleaned data index is obtained, and the data after data cleaning is as follows: ; According to The data is used to construct a SVR fitting model for the change value of the generator winding temperature. t represents the time, t -1 represents the last time ; ; The model represents the use t The winding temperature, generator slip ring chamber temperature, pitch angle and motor current at time -1 are used to predict the winding temperature change value at time. t

[0064] The fitting model of the winding temperature change value is obtained The residual characteristics of the model fitting value and the actual value are calculated, and the residual fluctuation threshold is calculated using 3sigma.

[0065] ; ; If the generator winding temperature change value at a certain time is greater than the threshold, it indicates that the winding temperature has changed too much in 10 minutes, so a warning is needed. If no warning is given, the residual is put into the standard residual, and the threshold is recalculated. If the warning is wrong, the threshold of the residual characteristics is recalculated after accumulating 3 groups.

[0066] Step 5, according to the temperature prediction data and the dynamic threshold, the temperature of the wind turbine generator winding is warned.

[0067] When the temperature prediction data is greater than the dynamic threshold, the warning trigger condition is met. For example, if the dynamic threshold is set to 80℃, and the current temperature prediction data is 85℃, the system will determine that the temperature is abnormal, and trigger the warning mechanism. In order to improve the accuracy and reliability of the warning, multiple condition combination judgments can also be set. For example, in addition to the temperature prediction data being greater than the dynamic threshold, the abnormal state also requires a certain time, such as 5 minutes in a row, to trigger the warning. In this way, false alarms caused by instantaneous data fluctuations or interference can be avoided.

[0068] ​Step 6, after completing a prediction, updating the dynamic threshold according to the residual error of the actual temperature data and the temperature prediction data.

[0069] Specifically, the temperature prediction data, the actual temperature data and preset warning temperature data are compared to determine a comparison result; the dynamic threshold is updated according to the comparison result, wherein when the temperature prediction data is less than the preset warning temperature data and the preset warning temperature data is greater than the actual temperature data, a residual error of the temperature prediction data and the actual temperature data is calculated, and the dynamic threshold is updated according to the residual error; when the temperature prediction data is less than the preset warning temperature data and the preset warning temperature data is less than the actual temperature data, a residual error of the temperature prediction data and the actual temperature data is calculated, and the residual error is accumulated, and the dynamic threshold is updated according to the accumulated residual error; when the temperature prediction data is greater than the preset warning temperature data and the preset warning temperature data is greater than the actual temperature data, a residual error of the temperature prediction data and the actual temperature data is calculated, and the residual error is accumulated, and the dynamic threshold is updated according to the accumulated residual error.

[0070] The wind generator winding temperature data processing method provided by the application effectively removes noise and retains key features by image segmentation and cleaning data, improves the stability of temperature prediction, reduces false positives and false negatives, can monitor winding temperature changes in real time, predict over-temperature abnormalities in advance, reduces the risk of shutdown due to temperature loss of control, prolongs the service life of the wind generator, and reduces economic losses.

[0071] As shown in Figure 6 The embodiment of the application further provides a wind generator winding temperature data processing device 60, which comprises: An acquisition module 61 is configured to acquire current operation data of the wind generator. A processing module 62 is configured to perform conversion processing on the current operation data to determine a target image matrix, perform noise elimination and feature enhancement processing on the target image matrix to obtain a target feature image, input the target feature image into a temperature prediction model, perform processing through a kernel function to obtain an intermediate result, and perform accumulation summation and bias addition on the intermediate result to obtain temperature prediction data. A warning module 63 is configured to perform warning on the temperature of the wind generator winding according to the temperature prediction data and a dynamic threshold.

[0072] Optionally, the acquisition module 61 is specifically configured to: According to the plurality of sensors of the wind turbine, current operation data of the wind turbine is acquired, the current operation data of the wind turbine including at least one of wind speed, generator active power, generator winding temperature, generator slip ring chamber temperature, pitch angle, motor current.

[0073] Optionally, the processing module 62 is specifically configured to: According to the wind speed and the generator active power in the current operation data, buffer boundary data is determined; According to the buffer boundary data, image size data is determined; According to the wind speed and the first scaling factor, first position data is determined; According to the generator active power and the second scaling factor, second position data is determined; According to the image size data, the first position data and the second position data, a target image matrix is determined, the size of the target image matrix being determined by the image size data, the target image matrix including a plurality of pixel points with a preset pixel value, the positions of the pixel points being determined by the first position data and the second position data.

[0074] Optionally, the processing module 62 is further specifically configured to: Iterate the distances between the pixel points in the target image matrix to obtain a plurality of distance data; According to the plurality of distance data, a distance mean value is determined; According to the distance mean value, the pixel values of the pixel points in the target image matrix are updated to obtain a target feature image.

[0075] Optionally, the temperature prediction model is trained through the following process: Acquire historical index data of the wind turbine; Convert and process the historical index data to determine an image matrix; Perform noise elimination and feature enhancement processing on the image matrix to obtain a feature image; Perform image segmentation and data index extraction processing on the feature image to obtain a cleaned data set; Input the cleaned data set into a support vector regression model for processing to obtain a temperature prediction model.

[0076] Optionally, the image segmentation and data index extraction processing on the feature image to obtain a cleaned data set comprises: Perform binaryzation processing on the feature image to obtain a binaryzation image; Perform an opening operation on the binaryzation image to obtain a denoised binaryzation image; The denoised binary image is subjected to contour search processing to obtain preliminary mask data; The preliminary mask data is subjected to inflation filling processing to obtain preliminary data index; The preliminary data index is subjected to clustering processing to obtain final data index; The final data index is subjected to data extraction processing to obtain a cleaned data set.

[0077] Optionally, the cleaned data set is input into a support vector regression model for processing to obtain a temperature prediction model, including: The training sample data in the cleaned data set is input into a support vector regression model, and the support vector regression model processes the training sample data through a kernel function to obtain an intermediate result; The intermediate result is accumulated and summed and a bias term is added to obtain prediction data; The prediction data is compared with the training sample data to obtain prediction error; According to the prediction error, the weight data, bias data and hyperparameters of the support vector regression model are adjusted to obtain a temperature prediction model.

[0078] Optionally, the early warning module 63 is specifically configured to: The temperature prediction data is compared with a dynamic threshold, and when the temperature prediction data is greater than the dynamic threshold, an abnormal temperature early warning of the wind turbine winding is issued.

[0079] Optionally, the wind turbine winding temperature data processing device 60 further includes: A determination module 64 is configured to compare the temperature prediction data, the actual temperature data and preset early warning temperature data to determine a comparison result; An update module 65 is configured to update the dynamic threshold according to the comparison result, wherein: When the temperature prediction data is less than the preset early warning temperature data, and the preset early warning temperature data is greater than the actual temperature data, a residual error of the temperature prediction data and the actual temperature data is calculated, and the dynamic threshold is updated according to the residual error; When the temperature prediction data is less than the preset early warning temperature data, and the preset early warning temperature data is less than the actual temperature data, a residual error of the temperature prediction data and the actual temperature data is calculated, and the residual error is accumulated, and the dynamic threshold is updated according to the accumulated residual error; When the temperature prediction data is greater than the preset early warning temperature data, and the preset early warning temperature data is greater than the actual temperature data, a residual of the temperature prediction data and the actual temperature data is calculated, the residual is accumulated, and the dynamic threshold is updated according to the accumulated residual.

[0080] It should be noted that the device corresponds to the above method, and all implementation manners in the method embodiments are applicable to this embodiment and can achieve the same technical effects.

[0081] Those skilled in the art can appreciate that the units and algorithm steps of the examples described in combination with the embodiments disclosed herein can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether the functions are implemented in hardware or software depends on the specific application and design constraints of the technical solutions. A person skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0082] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described system, device and unit can refer to the corresponding processes in the foregoing method embodiments, which will not be repeated here.

[0083] In the embodiments provided by the present application, it should be understood that the disclosed device and method can be implemented by other ways. For example, the device embodiments described above are only schematic, and the division of units is only a logical function division, and there can be another division way in actual implementation, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed mutual couplings or direct couplings or communication connections between different parts can be indirect couplings or communication connections through some interfaces, devices or units, and can be electrical, mechanical or other forms.

[0084] The units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, i.e., they can be located in one place, or can be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the embodiment.

[0085] In addition, each functional unit in each embodiment of the present application can be integrated into a processing unit, or each unit can exist physically, or two or more units can be integrated into one unit.

[0086] If the functions are implemented in the form of software function units and sold or used as independent products, they can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application essentially or the parts that contribute to the prior art or parts of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a ROM, a RAM, a magnetic disk or an optical disk, and various program code storage media.

[0087] In addition, it should be noted that in the device and method of the present application, it is obvious that the components or steps can be decomposed and / or recombined. These decompositions and / or recombination should be considered as equivalent solutions of the present application. Moreover, the steps of performing the above series of processes can naturally be executed in time sequence according to the order of description, but do not necessarily have to be executed in time sequence. Some steps can be executed in parallel or independently of each other. It can be understood by those skilled in the art that all or any steps or components of the method and device of the present application can be implemented in hardware, firmware, software or a combination thereof in any computing device (including processors, storage media, etc.) or network of computing devices, which can be implemented by those skilled in the art using their basic programming skills after reading the description of the present application.

[0088] Therefore, the object of the present application can also be achieved by running a program or a set of programs on any computing device. The computing device can be a commonly known general-purpose device. Therefore, the object of the present application can also be achieved by providing a program product containing program code for implementing the method or device. That is, such a program product also constitutes the present application, and a storage medium storing such a program product also constitutes the present application. Obviously, the storage medium can be any commonly known storage medium or any storage medium developed in the future. It should be noted that in the device and method of the present application, it is obvious that the components or steps can be decomposed and / or recombined. These decompositions and / or recombination should be considered as equivalent solutions of the present application. Moreover, the steps of performing the above series of processes can naturally be executed in time sequence according to the order of description, but do not necessarily have to be executed in time sequence. Some steps can be executed in parallel or independently of each other.

[0089] The above is the preferred embodiment of the present application. It should be noted that for those skilled in the art, without departing from the principles of the present application, a number of improvements and refinements can be made, which should also be considered within the scope of protection of the present application.

Claims

1. A wind generator winding temperature data processing method, characterized by, The method comprises the following steps: obtaining current operation data of a wind turbine; processing the current operation data to determine a target image matrix; performing noise elimination and feature enhancement processing on the target image matrix to obtain a target feature image; inputting the target feature image into a temperature prediction model to obtain temperature prediction data; the temperature prediction model processes the target feature image through a kernel function to obtain an intermediate result, accumulates and sums the intermediate result, and adds a bias term to obtain the temperature prediction data; warning the temperature of the wind turbine winding according to the temperature prediction data and a dynamic threshold.

2. The wind generator winding temperature data processing method of claim 1 wherein, The method comprises the following steps: obtaining current operation data of a wind turbine, comprising:

3. The wind generator winding temperature data processing method of claim 1 wherein, obtaining current operation data of a wind turbine from multiple sensors of the wind turbine, wherein the current operation data of the wind turbine includes at least one of wind speed, generator active power, generator winding temperature, generator slip ring chamber temperature, pitch angle, and motor current. processing the current operation data to determine a target image matrix, comprising: determining buffer boundary data according to the wind speed and the generator active power in the current operation data; determining image size data according to the buffer boundary data; determining first position data according to the wind speed and a first scaling coefficient; determining second position data according to the generator active power and a second scaling coefficient; 4. The wind generator winding temperature data processing method of claim 3 wherein, determining a target image matrix according to the image size data, the first position data, and the second position data, wherein the size of the target image matrix is determined by the image size data, the target image matrix includes multiple pixel points with a preset pixel value, and the positions of the pixel points are determined by the first position data and the second position data. performing noise elimination and feature enhancement processing on the target image matrix to obtain a target feature image, comprising: traversing the distances between the pixel points in the target image matrix to obtain multiple distance data; determining a distance mean value according to the multiple distance data; 5. The wind generator winding temperature data processing method of claim 1, wherein, updating the pixel values of the pixel points in the target image matrix according to the distance mean value to obtain a target feature image. The temperature prediction model is trained through the following process: obtaining historical index data of a wind turbine; processing the historical index data to determine an image matrix; performing noise elimination and feature enhancement processing on the image matrix to obtain a feature image; performing image segmentation and data index extraction processing on the feature image to obtain a cleaned data set; 6. The wind generator winding temperature data processing method of claim 5 wherein, inputting the cleaned data set into a support vector regression model for processing to obtain a temperature prediction model. The method comprises the following steps: performing image segmentation and data index extraction processing on the feature image to obtain a cleaned data set, comprising: performing binaryzation processing on the feature image to obtain a binaryzation image; performing an opening operation on the binaryzation image to obtain a denoised binaryzation image; performing contour finding processing on the denoised binaryzation image to obtain preliminary mask data; performing inflation filling processing on the preliminary mask data to obtain preliminary data indexes; performing clustering processing on the preliminary data indexes to obtain final data indexes; The final data index is subjected to data extraction processing to obtain a cleaned data set.

7. The wind generator winding temperature data processing method of claim 5 wherein, According to the cleaned data set, a support vector regression model is inputted for processing to obtain a temperature prediction model, including: The training sample data in the cleaned data set is inputted into a support vector regression model, and the support vector regression model processes the training sample data through a kernel function to obtain an intermediate result; The intermediate result is accumulated and summed and a bias term is added to obtain predicted data; The predicted data is compared with the training sample data to obtain a prediction error; According to the prediction error, the weight data, bias data and hyperparameters of the support vector regression model are adjusted to obtain a temperature prediction model.

8. The wind generator winding temperature data processing method of claim 1, wherein, According to the temperature prediction data and the dynamic threshold, the temperature of the wind generator winding is warned, including: The temperature prediction data is compared with the dynamic threshold, and when the temperature prediction data is greater than the dynamic threshold, an abnormal temperature warning of the wind generator winding is issued.

9. The wind generator winding temperature data processing method of claim 1, wherein, Further comprising: The temperature prediction data, the actual temperature data and the preset warning temperature data are compared to determine a comparison result; According to the comparison result, the dynamic threshold is updated, wherein When the temperature prediction data is less than the preset warning temperature data, and the preset warning temperature data is greater than the actual temperature data, the residual error of the temperature prediction data and the actual temperature data is calculated, and the dynamic threshold is updated according to the residual error; When the temperature prediction data is less than the preset warning temperature data, and the preset warning temperature data is less than the actual temperature data, the residual error of the temperature prediction data and the actual temperature data is calculated, and the residual error is accumulated, and the dynamic threshold is updated according to the accumulated residual error; When the temperature prediction data is greater than the preset warning temperature data, and the preset warning temperature data is greater than the actual temperature data, the residual error of the temperature prediction data and the actual temperature data is calculated, and the residual error is accumulated, and the dynamic threshold is updated according to the accumulated residual error.

10. A wind generator winding temperature data processing apparatus characterized by, Including: An acquisition module is configured to acquire current operation data of a wind generator; A processing module is configured to convert and process the current operation data to determine a target image matrix; The target image matrix is subjected to noise elimination and feature enhancement processing to obtain a target feature image; the target feature image is inputted into a temperature prediction model to obtain temperature prediction data; the temperature prediction model processes the target feature image through a kernel function to obtain an intermediate result, and the intermediate result is accumulated and summed and a bias term is added to obtain temperature prediction data; A warning module is configured to warn the temperature of the wind generator winding according to the temperature prediction data and the dynamic threshold.

Images