Generator bearing temperature abnormity identification method and device
By obtaining generator operation data in real time and inputting an abnormality recognition model for temperature abnormality recognition, the problem of vibration signal analysis method being susceptible to environmental noise interference is solved, and a more accurate generator bearing fault warning is achieved.
Patent Information
- Application Number
- CN202411602926.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-11
- Publication Date
- 2025-02-11
AI Technical Summary
The analysis method based on vibration signals in the prior art is easily disturbed by environmental noise, resulting in distortion of the abnormal recognition results of the generator bearing temperature.
By obtaining the generator's operating data in real time, including winding temperature, bearing temperature, rotation speed, ambient temperature and wind speed, these data are input based on the abnormality recognition model and weighted model of different gradient enhancement decision trees to identify temperature abnormalities.
This method can effectively avoid environmental noise interference, improve the accuracy and reliability of judgment results, and ensure that the early warning results of generator bearing failures are not easily distorted.
Smart Images
Figure CN119416078A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of machine learning technology, and in particular to a method for identifying abnormal generator bearing temperature. Background Art
[0002] Wind power generation is a renewable energy technology that converts wind energy into electrical energy. It is clean, renewable, has a short infrastructure period, and has low operation and maintenance costs. As the core component of wind turbines, generator bearings are able to carry axial and radial loads and reduce friction, which is crucial to the stable operation of generators.
[0003] Generator bearing failure can reduce the power generation efficiency of wind turbines, and even cause the wind turbines to shut down and equipment to be damaged, resulting in serious economic losses. The occurrence of generator bearing failure is usually a gradual process. Considering the difficulty and high cost of installing and maintaining wind turbine equipment, it is of great significance to effectively identify abnormalities before the failure occurs. At present, the wind power industry mainly uses analysis methods based on vibration signals to diagnose generator bearing failures. However, the analysis method based on vibration signals is easily interfered by environmental noise, resulting in distorted abnormal identification results. Summary of the invention
[0004] The embodiment of the present application provides a method and device for identifying abnormal temperature of a generator bearing, thereby solving the technical problem in the prior art that the analysis method based on vibration signals is easily interfered by environmental noise, resulting in distortion of abnormal identification results.
[0005] In a first aspect, an embodiment of the present application provides a method for identifying abnormal temperature of a generator bearing, the method comprising: acquiring operating data of the generator in real time; wherein the operating data comprises the winding temperature, front bearing temperature, rear bearing temperature and rotation speed of the generator, ambient temperature and real-time wind speed; inputting the operating data into a generator bearing temperature abnormality identification model, the generator bearing temperature abnormality identification model comprising a first abnormality identification model and a second abnormality identification model based on different gradient boosting decision trees, and a weighted model; the first abnormality identification model and the second abnormality identification model are configured to receive the operating data and output corresponding first initial prediction data and second initial prediction data; the weighted model is configured to receive the first initial prediction data and the second initial prediction data and output final prediction data; based on the final prediction data, determining whether the temperature of the generator bearing is abnormal; wherein the first abnormality identification model and the second abnormality identification model are obtained by training with a first data set, the first data set comprises the operating data within a historical period of time, and each data in the first data set uses whether the generator bearing temperature is abnormal as a category label; the weighted model is obtained by training with the second data set, the second data set comprises the output results of the first abnormality identification model and the second abnormality identification model and the corresponding category labels using real values.
[0006] In combination with the first aspect, in a possible implementation, before using the first data set to train the first anomaly recognition model and the second anomaly recognition model, the method also includes: performing data cleaning and standardization preprocessing on the first data set; dividing the positive samples and negative samples in the first data set into a training set and a test set according to the same preset ratio.
[0007] In combination with the first aspect, in a possible implementation method, data cleaning is performed on the first data set, including: screening out missing values and outliers in the first data set; filling the missing values with a linear change law using an interpolation method; and eliminating the data corresponding to the outliers and the missing values that cannot be filled.
[0008] In combination with the first aspect, in a possible implementation manner, the preset ratio is 8:2.
[0009] In combination with the first aspect, in a possible implementation, in the process of training the first abnormality recognition model and the second abnormality recognition model, the engine bearing temperature abnormality recognition is defined as a binary classification problem, the loss function of the first abnormality recognition model and the second abnormality recognition model adopts cross entropy loss, and the Sigmoid function is used in the output part of the first abnormality recognition model and the second abnormality recognition model.
[0010] In combination with the first aspect, in a possible implementation, the first anomaly recognition model is established based on LightGBM, and the second anomaly recognition model is established based on XGBoost.
[0011] In combination with the first aspect, in a possible implementation, the weighted model is a back propagation neural network.
[0012] In a second aspect, an embodiment of the present application further provides a device for identifying an abnormal bearing temperature of a generator, the device comprising: an acquisition module for acquiring the operating data of the generator in real time; wherein the operating data comprises the winding temperature, front bearing temperature, rear bearing temperature and rotation speed of the generator, the ambient temperature and the real-time wind speed; a prediction module for inputting the operating data into a generator bearing temperature abnormality recognition model, the generator bearing temperature abnormality recognition model comprising a first abnormality recognition model and a second abnormality recognition model based on different gradient boosting decision trees, and a weighted model; the first abnormality recognition model and the second abnormality recognition model are configured to receive the operating data, and output corresponding first initial prediction data and second initial prediction data; the weighted model is configured to receive the first initial prediction data and the second initial prediction data, and output final prediction data; a judgment module is used to judge whether the temperature of the generator bearing is abnormal based on the final prediction data; wherein the first abnormality recognition model and the second abnormality recognition model are trained using a first data set, the first data set includes the operating data within a historical period of time, and each data in the first data set uses whether the generator bearing temperature is abnormal as a category label; the weighted model is trained using the second data set, and the second data set includes the output results of the first abnormality recognition model and the second abnormality recognition model and the corresponding category labels using real values.
[0013] In a third aspect, an embodiment of the present application provides an electronic device, comprising: one or more processors; and a memory storing computer executable instructions, wherein the computer executable instructions, when executed by the one or more processors, implement a method for identifying abnormal generator bearing temperature as described in the first aspect or any possible implementation of the first aspect.
[0014] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, which stores computer-readable instructions. When the computer-readable instructions are executed by a processor, the method for identifying abnormal generator bearing temperature as described in the first aspect or any possible implementation method of the first aspect is implemented.
[0015] The technical solution provided in the embodiments of the present application has at least the following technical effects: An embodiment of the present application provides a method for identifying abnormal temperature of a generator bearing. The method obtains the operating data of the generator in real time and inputs it into a generator bearing temperature abnormality identification model, and judges whether the temperature of the generator bearing is abnormal based on the final predicted data, thereby being able to warn of a fault of the generator bearing, so that the method is not easily disturbed by environmental noise, thereby ensuring that the judgment result is not easily distorted; and because the generator bearing temperature abnormality identification model includes a first abnormality identification model and a second abnormality identification model based on different gradient boosting decision trees, and a weighted model, the generator bearing temperature abnormality identification model combines the predicted data of the first abnormality identification model and the second abnormality identification model, so that the result determined by the method is more accurate. BRIEF DESCRIPTION OF THE DRAWINGS
[0016] 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 of the present application will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.
[0017] Figure 1 A flow chart of a method for identifying abnormal generator bearing temperature provided in an embodiment of the present application; Figure 2 A schematic diagram of a generator bearing temperature anomaly identification model provided in an embodiment of the present application; Figure 3 A schematic diagram of a generator bearing temperature anomaly recognition model training process provided in an embodiment of the present application; Figure 4 A schematic diagram of a weighted model provided in an embodiment of the present application; Figure 5 A schematic diagram of a generator bearing temperature anomaly identification device provided in an embodiment of the present application; Figure 6 A schematic diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION
[0018] The failure of the generator bearing of a wind turbine will reduce the power generation efficiency of the wind turbine, and even cause the wind turbine to shut down and damage the equipment, resulting in serious economic losses. The occurrence of generator bearing failure is usually a gradual process. Considering the difficulty and high cost of installing and maintaining wind turbine equipment, it is of great significance to effectively identify abnormalities before the failure occurs. At present, the wind power industry mainly uses analysis methods based on vibration signals to diagnose generator bearing failures. However, the analysis method based on vibration signals is easily interfered by environmental noise, resulting in distorted abnormal identification results.
[0019] The embodiment of the present application provides a method and device for identifying abnormal generator bearing temperature, which realizes early warning of generator bearing failure, is not easily interfered by environmental noise, ensures that the judgment result is not easily distorted, and the determined result is more accurate.
[0020] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are 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 are within the scope of protection of this application.
[0021] Below, some terms in the embodiments of the present application are explained to facilitate understanding by those skilled in the art.
[0022] XGBoost (eXtreme Gradient Boosting) is an optimized gradient boosting decision tree. It uses a greedy algorithm to build decision trees. Starting from the first decision tree, it enumerates features on each leaf node and sorts the training samples of the node according to the feature value. It tries to split a leaf node each time, calculates the gain after the split, selects the feature with the largest gain as the split feature, and splits at the best split point of the feature, iterating until the preset conditions are met. In terms of the decision tree column sampling strategy, XGBoost uses a layer-by-layer growth method. Before each node in the same layer is split, a part of the features are randomly sampled to avoid overfitting.
[0023] LightGBM (Light Gradient Boosting Machine) is an optimized gradient boosting decision tree. It uses a histogram-based algorithm to build the decision tree. It uses the bucketing idea to discretize continuous floating-point feature values into k integers, accumulates gradients and counts the number of features according to the buckets where the features are located, and then finds the optimal split point according to the histogram. In terms of decision tree growth strategy, it adopts a leaf-by-leaf growth method with depth limitation. Each time, it finds a leaf with the largest split gain from all current leaves, and then splits it, and repeats it in this way. Compared with the layer-by-layer growth method, when the number of splits is the same, it can reduce more errors and obtain better accuracy.
[0024] Backpropagation Neural Network (BP Neural Network) is a multi-layer feedforward neural network that trains the network through a supervised learning method so that the network output is as close to the expected output as possible. The core of BP Neural Network is the backpropagation algorithm, which calculates the gradient of the loss function with respect to the network parameters and uses these gradients to update the weights and biases of the network, thereby minimizing the loss function. Figure 4 As shown, a BP neural network usually includes an input layer, a hidden layer, and an output layer; the input layer is used to receive input data; there can be multiple hidden layers, each layer consisting of multiple neurons, and the neurons are connected by weights; the output layer is used to generate the final output result.
[0025] The present application embodiment provides a method for identifying abnormal bearing temperature of a generator, such as Figure 1 As shown, the method includes steps S101 to S103.
[0026] S101, acquiring the operation data of the generator in real time, wherein the operation data includes the winding temperature, front bearing temperature, rear bearing temperature and rotation speed, ambient temperature and real-time wind speed of the generator.
[0027] S102, inputting the operating data into a generator bearing temperature anomaly recognition model.
[0028] like Figure 2 As shown, the generator bearing temperature anomaly recognition model 200 includes a first anomaly recognition model 210 and a second anomaly recognition model 220 based on different gradient boosting decision trees, and a weighted model 230; the first anomaly recognition model 210 and the second anomaly recognition model 220 are configured to receive operating data and output corresponding first initial prediction data and second initial prediction data; the weighted model 230 is configured to receive the first initial prediction data and the second initial prediction data, and output final prediction data.
[0029] Among them, the first abnormality recognition model 210 and the second abnormality recognition model 220 are trained using the first data set, the first data set includes operating data within a historical period of time, and each data in the first data set uses whether the generator bearing temperature is abnormal as a category label; the weighted model 230 is trained using the second data set, and the second data set includes the output results of the first abnormality recognition model 210 and the second abnormality recognition model 220 and the corresponding category labels using real values.
[0030] Figure 3The complete training process of the generator bearing temperature anomaly recognition model 200 is shown. Specifically, the operating data of one end of the generator in the historical time is first collected, and each data uses whether the generator bearing temperature is abnormal as a category label to form a first data set; then, the first data set is used to train the first recognition model and the second recognition model to obtain the first initial prediction data and the second initial prediction data respectively, and the first initial prediction data and its corresponding category label using the real value, as well as the second initial prediction data and its corresponding category label using the real value are saved to form a second data set; secondly, the second data set is used to train the weighted model 230; finally, the trained first abnormal recognition model 210, the second abnormal recognition model 220 and the weighted model 230 are combined to form the generator bearing temperature anomaly recognition model 200.
[0031] Continue to refer to Figure 3 Before using the first data set to train the first anomaly recognition model 210 and the second anomaly recognition model 220, the method further includes step S201 and step S202.
[0032] S201, performing data cleaning and standardization preprocessing on the first data set.
[0033] Data cleaning of the first data set can remove missing values and outliers in the first data set, improve the data quality of the first data set, and improve the training efficiency and prediction performance of the first anomaly recognition model 210 and the second anomaly recognition model 220.
[0034] Specifically, when data cleaning is performed on the first data set in step S201, step S301 and step S302 may be used.
[0035] S301, filtering out missing values and outliers in the first data set. For example, using Python's pandas library to identify missing values in the first data set; using the standard deviation method to filter outliers, for each feature data, taking data points that exceed mean ±3 standard deviations as outliers.
[0036] S302. For missing values with a linear variation pattern, use interpolation to fill them.
[0037] Interpolation is a method of estimating the value of unknown data points by using known data points. For example, when executing step S302, the interpolate() function in the pandas library is used in Python to perform linear interpolation.
[0038] S303: Eliminate data corresponding to outliers and unfillable missing values, wherein unfillable missing values refer to missing values that do not have a linear change pattern or other change patterns but are completely random.
[0039] After the data corresponding to the outliers and the unfillable missing values are removed in step 203, the data retained in the first data set can be made more realistic, thereby more accurately training the first anomaly recognition model 210 and the second anomaly recognition model 220.
[0040] Standardization preprocessing can eliminate the dimensional influence between different feature data. The following formula can be used for calculation: in, represents the data before standardization preprocessing, represents the data mean, After the standardized preprocessing is performed by the above calculation formula, the data in the first data set is transformed into a distribution with a mean of 0 and a standard deviation of 1, thereby achieving the purpose of eliminating the dimensional differences between different feature data.
[0041] S202: Divide the positive samples and negative samples in the first data set into a training set and a test set according to the same preset ratio. For example, the preset ratio is 8:2.
[0042] Step S202 ensures that positive samples and negative samples are distributed in the same proportion in the training set and the test set, thereby maintaining the consistency of data distribution; since the test set can more truly reflect the performance of the first anomaly recognition model 210 and the second anomaly recognition model 220 in actual applications, the training set and the test set divided by step S202 can make the prediction results of the trained first anomaly recognition model 210 and the second anomaly recognition model 220 more accurate.
[0043] Exemplarily, the first anomaly recognition model 210 is established based on LightGBM, and the second anomaly recognition model 220 is established based on XGBoost. Both LightGBM and XGBoost are optimization implementations based on gradient-based decision trees. Although the construction strategies of decision trees are different, LightGBM and XGBoost have the same objective function. For a given first data set, the first anomaly recognition model 210 and the second anomaly recognition model 220 use a forward distribution algorithm to learn an additive model containing K trees. The first anomaly recognition model 210 and the second anomaly recognition model 220 are in the form of: ; In the formula, Represents the prediction result, represents the input features, Represents each tree model. First, a classification and regression tree is created as the initial model. A new tree model is added in each iteration. The sample is adjusted by the deviation of the prediction result of the previous step, and the sample is used for the current iteration training. In addition, each iteration only needs to optimize the tree model corresponding to the current iteration step. For the first anomaly recognition model 210 and the second anomaly recognition model 220 obtained after the sth iteration, have: in, represents the i-th sample, It indicates the first s-1 trained sub-models, which have been fixed. Indicates the sub-model corresponding to the current iteration step.
[0044] The objective function consists of two parts: the loss function and the regularization term: ;in, Represents the prediction result, Represents the loss function, which is used to measure the difference between the predicted value and the true value. The cross entropy loss function is often used for calculation in classification tasks. Represents the sum of the loss values of all samples, and the regular term is used to limit the complexity of the model, which is expressed as the sum of the squares of the number of leaf nodes in the tree and the output score W of the leaf node of each tree T: For the s-th iteration model, the first s-1 model prediction results and the regularization term can be regarded as constants for calculation. After simplifying the objective function and expanding the Taylor formula, the objective function can be simplified as follows: ;in, They represent the first and second derivatives of the loss function respectively.
[0045] Define each leaf node The sample set on , the objective function is expressed as: ; In the formula, The optimization problem of the objective function can be transformed into calculating the leaf node output score w that minimizes the objective function. Taking the derivative of w to make the derivative 0 and substituting the result into the objective function, we can get: .
[0046] For example, refer to Figure 4 , the weighted model 230 is a back propagation neural network. The back propagation neural network is a nonlinear network model, and the entire training process can be divided into two parts: forward propagation and back propagation.
[0047] Forward propagation process: After the data in the second data set is loaded into the weighted model 230 from the input layer, it is calculated in the i-th hidden layer neuron node, and the calculation result is processed by the activation function and passed to the next layer node, and finally the prediction result is obtained through the output layer. The calculation process of the neuron node of the back propagation neural network is as follows: Where x represents the input of the weighted model 230, represents the bias term of the i-th layer, represents the activation function used by the weighted model 230, represents the output of the i-th hidden layer.
[0048] After forward propagation, the prediction result of the weighted model 230 is obtained, and the model is optimized by back propagation.
[0049] Back propagation process: Use the loss function to calculate the difference between the predicted result obtained by forward propagation and the true value, and calculate the partial derivative of the loss value corresponding to the weight. For multi-layer neural networks, the chain derivation method is used to calculate the gradient value, and the weight parameters are updated through the gradient descent algorithm until the model converges or reaches the maximum number of iterations set.
[0050] S103. Determine whether the temperature of the generator bearing is abnormal based on the final prediction data.
[0051] When executing step S103, a judgment is made according to a threshold value set in advance. For example, the threshold value is 0.8, when the final predicted data is greater than or equal to 0.8, the temperature of the generator bearing is determined to be abnormal; when the final predicted data is less than 0.8, the temperature of the generator bearing is determined to be normal.
[0052] Furthermore, in the process of training the first abnormality recognition model 210 and the second abnormality recognition model 220, the engine bearing temperature abnormality recognition is defined as a binary classification problem, the loss function of the first abnormality recognition model 210 and the second abnormality recognition model 220 adopts cross entropy loss, and the Sigmoid function is used in the output part of the first abnormality recognition model 210 and the second abnormality recognition model 220.
[0053] Among them, the goal of the binary classification problem is to divide the data into two categories. In the recognition of abnormal engine bearing temperature, the categories are "normal" and "abnormal". The cross entropy loss function is a method to measure the difference between two probability distributions. It can handle probability output well, and when the predicted probability differs greatly from the true label, a larger penalty will be given, which helps the first abnormality recognition model 210 and the second abnormality recognition model 220 to converge to the correct classification boundary faster. The Sigmoid function is a function that maps any real number to the interval (0,1). Using the Sigmoid function in the output part of the binary classification problem can interpret the output results of the first abnormality recognition model 210 and the second abnormality recognition model 220 as probability values, indicating the probability that the data in the first data set belongs to a certain category, which makes the output results more intuitive and easy to interpret.
[0054] When using the first data set to train the first abnormality recognition model 210 and the second abnormality recognition model 220, using a combination of cross entropy loss and Sigmoid function can help the first abnormality recognition model 210 and the second abnormality recognition model 220 to better distinguish between normal and abnormal states. The output results of the first abnormality recognition model 210 and the second abnormality recognition model 220 can be interpreted as the probability of abnormal engine bearing temperature, which provides a basis for further decision-making of the weighted model 230.
[0055] The generator bearing temperature anomaly identification method provided by the embodiment of the present application obtains the operating data of the generator in real time and inputs it into the generator bearing temperature anomaly identification model 200, and judges whether the temperature of the generator bearing is abnormal based on the final predicted data, so as to warn of the fault of the generator bearing, so that the method is not easily disturbed by environmental noise, and ensures that the judgment result is not easily distorted; and because the generator bearing temperature anomaly identification model 200 includes a first anomaly identification model 210 and a second anomaly identification model 220 based on different gradient boosting decision trees, and a weighted model 230, the generator bearing temperature anomaly identification model 200 combines the predicted data of the first anomaly identification model 210 and the second anomaly identification model 220, so that the result determined by the method is more accurate.
[0056] like Figure 5 As shown, the embodiment of the present application further provides a generator bearing temperature anomaly identification device 500 , which includes an acquisition module 501 , a prediction module 502 and a judgment model 503 .
[0057] The acquisition module 501 is used to acquire the operating data of the generator in real time; wherein the operating data includes the winding temperature, front bearing temperature, rear bearing temperature and rotation speed of the generator, ambient temperature and real-time wind speed.
[0058] The prediction module 502 is used to input the operating data into the generator bearing temperature anomaly recognition model, which includes a first anomaly recognition model and a second anomaly recognition model based on different gradient boosting decision trees, and a weighted model; the first anomaly recognition model and the second anomaly recognition model are configured to receive the operating data and output the corresponding first initial prediction data and the second initial prediction data; the weighted model is configured to receive the first initial prediction data and the second initial prediction data and output the final prediction data.
[0059] Exemplarily, the first anomaly recognition model is established based on LightGBM, the second anomaly recognition model is established based on XGBoost, and the weighted model is a back propagation neural network.
[0060] The judgment module 503 is used to judge whether the temperature of the generator bearing is abnormal based on the final prediction data.
[0061] Among them, the first abnormality recognition model and the second abnormality recognition model are trained using the first data set, the first data set includes operating data within a historical period of time, and each data in the first data set uses whether the generator bearing temperature is abnormal as a category label; the weighted model is trained using the second data set, and the second data set includes the output results of the first abnormality recognition model and the second abnormality recognition model and the corresponding category labels using real values.
[0062] The generator bearing temperature anomaly identification device 500 also includes a preprocessing module and a partitioning module. Before using the first data set to train the first anomaly identification model and the second anomaly identification model, the preprocessing module is used to perform data cleaning and standardization preprocessing on the first data set; the partitioning module is used to partition the positive samples and negative samples in the first data set into a training set and a test set according to the same preset ratio.
[0063] Exemplarily, the preset ratio is 8:2.
[0064] When the preprocessing module performs data cleaning on the first data set, it specifically includes: screening out missing values and outliers in the first data set; using interpolation to fill missing values with linear variation rules; and eliminating data corresponding to outliers and missing values that cannot be filled.
[0065] Furthermore, in the process of training the first abnormality recognition model and the second abnormality recognition model, the engine bearing temperature abnormality recognition is defined as a binary classification problem, the loss function of the first abnormality recognition model and the second abnormality recognition model adopts cross entropy loss, and the Sigmoid function is used in the output part of the first abnormality recognition model and the second abnormality recognition model.
[0066] The devices or modules described in the above embodiments can be implemented by computer chips or entities, or by products with certain functions. For the convenience of description, the above devices are described in various modules according to their functions. When implementing the embodiments of the present application, the functions of each module can be implemented in the same or multiple software and / or hardware. Of course, the module that implements a certain function can also be implemented by combining multiple sub-modules or sub-units.
[0067] like Figure 6 As shown, an embodiment of the present application also provides an electronic device 600, which includes one or more processors 601 and a memory 602 storing computer executable instructions; the computer executable instructions implement the above-mentioned generator bearing temperature abnormality identification method when executed by one or more processors 601.
[0068] An embodiment of the present application also provides a computer-readable storage medium, which stores computer-readable instructions. When the computer-readable instructions are executed by a processor, the above-mentioned method for identifying abnormal generator bearing temperature is implemented.
[0069] The above-mentioned storage media include but are not limited to random access memory (RAM), read-only memory (ROM), cache, hard disk drive (HDD) or memory card (Memory Card).
[0070] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other. Each embodiment focuses on the differences from other embodiments.
[0071] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit the present application. Although the present application has been described in detail with reference to the aforementioned embodiments, a person of ordinary skill in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some or all of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the present application.
Claims
1. A method for identifying abnormal temperature of a generator bearing, characterized in that: include: Acquire the operation data of the generator in real time; wherein the operation data includes the winding temperature, front bearing temperature, rear bearing temperature and rotation speed of the generator, ambient temperature and real-time wind speed; Inputting the operating data into a generator bearing temperature anomaly recognition model, the generator bearing temperature anomaly recognition model comprising a first anomaly recognition model and a second anomaly recognition model based on different gradient boosting decision trees, and a weighted model; the first anomaly recognition model and the second anomaly recognition model are configured to receive the operating data and output corresponding first initial prediction data and second initial prediction data; the weighted model is configured to receive the first initial prediction data and the second initial prediction data and output final prediction data; Determining whether the temperature of the generator bearing is abnormal based on the final prediction data; Among them, the first abnormality recognition model and the second abnormality recognition model are trained using a first data set, the first data set includes the operating data within a historical period of time, and each data in the first data set uses whether the generator bearing temperature is abnormal as a category label; the weighted model is trained using the second data set, and the second data set includes the output results of the first abnormality recognition model and the second abnormality recognition model and the corresponding category labels using real values.
2. The method for identifying abnormal generator bearing temperature according to claim 1, characterized in that: Before using the first data set to train the first anomaly recognition model and the second anomaly recognition model, the method further includes: Performing data cleaning and standardization preprocessing on the first data set; The positive samples and negative samples in the first data set are divided into a training set and a test set according to the same preset ratio.
3. The method for identifying abnormal generator bearing temperature according to claim 2, characterized in that: The performing data cleaning on the first data set includes: Screening out missing values and outliers in the first data set; For the missing values with linear variation rules, interpolation method is used to fill them; Eliminate the data corresponding to the outliers and the missing values that cannot be filled.
4. The method for identifying abnormal generator bearing temperature according to claim 2, characterized in that: The preset ratio is 8:
2.
5. The method for identifying abnormal generator bearing temperature according to claim 1, characterized in that: In the process of training the first abnormality recognition model and the second abnormality recognition model, the engine bearing temperature abnormality recognition is defined as a binary classification problem, the loss function of the first abnormality recognition model and the second abnormality recognition model adopts cross entropy loss, and the Sigmoid function is used in the output part of the first abnormality recognition model and the second abnormality recognition model.
6. The method for identifying abnormal generator bearing temperature according to claim 1, characterized in that: The first anomaly recognition model is established based on LightGBM, and the second anomaly recognition model is established based on XGBoost.
7. The method for identifying abnormal generator bearing temperature according to claim 1, characterized in that: The weighted model is a back propagation neural network.
8. A device for identifying abnormal temperature of a generator bearing, characterized in that: include: An acquisition module, used for acquiring the operation data of the generator in real time; wherein the operation data includes the winding temperature, front bearing temperature, rear bearing temperature and rotation speed of the generator, ambient temperature and real-time wind speed; A prediction module, configured to input the operating data into a generator bearing temperature anomaly recognition model, wherein the generator bearing temperature anomaly recognition model includes a first anomaly recognition model and a second anomaly recognition model based on different gradient boosting decision trees, and a weighted model; the first anomaly recognition model and the second anomaly recognition model are configured to receive the operating data and output corresponding first initial prediction data and second initial prediction data; the weighted model is configured to receive the first initial prediction data and the second initial prediction data and output final prediction data; A judgment module, used for judging whether the temperature of the generator bearing is abnormal based on the final prediction data; Among them, the first abnormality recognition model and the second abnormality recognition model are trained using a first data set, the first data set includes the operating data within a historical period of time, and each data in the first data set uses whether the generator bearing temperature is abnormal as a category label; the weighted model is trained using the second data set, and the second data set includes the output results of the first abnormality recognition model and the second abnormality recognition model and the corresponding category labels using real values.
9. An electronic device, characterized in that: include: one or more processors; as well as A memory storing computer executable instructions, wherein the computer executable instructions, when executed by the one or more processors, implement the method for identifying abnormal temperature of a generator bearing according to any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-readable instructions, and when the computer-readable instructions are executed by a processor, the method for identifying abnormal temperature of a generator bearing according to any one of claims 1 to 7 is implemented.