A filter board temperature monitoring method, device, equipment and medium

CN118836127BActive Publication Date: 2024-11-29HUADIAN ELECTRIC POWER SCI INST CO LTD
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202411327230.9
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-09-20
Publication Date
2024-11-29
Estimated Expiration
2044-09-20

AI Technical Summary

Technical Problem

The prior art is difficult to effectively monitor the temperature of the filter plate in the wind turbine, resulting in insufficient accuracy and effectiveness of fault warning.

Method used

The variable screening method based on Pearson's correlation coefficient method is used to obtain the relevant variables in the wind turbine state monitoring data, and a multivariate filter plate temperature model based on deep neural network is constructed, and the difference between the model predicted value and the actual temperature value is judged. Fault warning.

Benefits of technology

It improves the accuracy and effectiveness of filter plate temperature monitoring, realizes the fault warning of wind turbines, and enhances the overall performance of the monitoring system.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118836127B_ABST
    Figure CN118836127B_ABST
Patent Text Reader

Abstract

The present application discloses a filter plate temperature monitoring method, device, equipment and medium, which relates to the field of wind turbine status monitoring, including: obtaining the correlation coefficient between each variable in the data to be processed and the filter plate temperature based on the Pearson correlation coefficient method, and performing variable screening according to each correlation coefficient and preset variable screening conditions, determining a training set and a test set through the obtained screened data and preset data processing rules, and using the training set and the test set to respectively perform model training and model testing on the constructed initial multivariable filter plate temperature model, in the process of monitoring the filter plate temperature of the wind turbine based on the obtained target multivariable filter plate temperature model, judging whether the preset warning triggering conditions are met through the current warning threshold information and the local residual sequence, and completing the corresponding fault warning operation. The filter plate temperature is monitored based on multiple variables to achieve fault warning of the wind turbine and improve monitoring accuracy.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of wind turbine status monitoring, and in particular to a filter board temperature monitoring method, device, equipment and medium. Background Art

[0002] The filter board is an important component. Whether it operates normally will determine whether the wind turbine can operate normally. Not only that, the temperature of the filter board can also reflect the operating status of the filter board.

[0003] At present, the research on data-driven fault warning methods for specific parameters mainly relies on learning the historical state information of variables, and realizing abnormal state warning by predicting the future trend of variables. In fact, wind turbines are complex systems with multiple components tightly coupled, and there are close connections between the internal components or variables. Considering the historical information of a certain variable alone is not enough to mine the most accurate historical information of the variable.

[0004] As can be seen from the above, how to monitor the temperature of the filter board based on multiple variables to achieve fault warning of the wind turbine is a problem that needs to be solved urgently. Summary of the invention

[0005] In view of this, the purpose of the present invention is to provide a filter board temperature monitoring method, device, equipment and medium, which can effectively monitor the filter board temperature based on multiple variables, and thus effectively realize the fault warning of the wind turbine generator, and improve the accuracy and effectiveness of monitoring. The specific scheme is as follows:

[0006] In a first aspect, the present application provides a filter board temperature monitoring method, comprising:

[0007] Obtaining the correlation coefficient between each variable in the data to be processed and the temperature of the filter plate based on the Pearson correlation coefficient method, and performing corresponding variable screening according to each of the correlation coefficients and the preset variable screening conditions to obtain corresponding screened data; wherein the data to be processed is the status monitoring data corresponding to the wind turbine;

[0008] Determine the corresponding training set and test set by the screened data and the preset data processing rules, and use the training set and the test set to perform corresponding model training and model testing on the constructed initial multivariable filter board temperature model, so as to obtain the corresponding target multivariable filter board temperature model; the initial multivariable filter board temperature model is a model constructed based on a deep neural network;

[0009] In the process of monitoring the filter board temperature of the wind turbine based on the target multivariable filter board temperature model, the current warning threshold information and the local residual sequence are used to determine whether the preset warning trigger conditions are currently met to complete the corresponding fault warning operation; wherein, the residual sequence includes a number of values ​​for characterizing the difference between the model prediction value and the corresponding actual temperature value.

[0010] Optionally, the method of obtaining the correlation coefficient between each variable in the data to be processed and the filter plate temperature based on the Pearson correlation coefficient method, and performing corresponding variable screening according to each of the correlation coefficients and preset variable screening conditions, includes:

[0011] Acquire status monitoring data of the wind turbine to obtain the data to be processed;

[0012] Calculate the correlation coefficient between each variable in the data to be processed and the filter plate temperature based on the Pearson correlation coefficient method;

[0013] Determine whether each of the correlation coefficients is greater than a corresponding preset coefficient threshold to obtain a corresponding determination result;

[0014] For the judgment result indicating that it is greater than the preset coefficient threshold, the corresponding variable is determined as the target input variable to complete the corresponding variable screening operation and obtain the corresponding screened data.

[0015] Optionally, using the training set and the test set to perform corresponding model training and model testing on the constructed initial multivariable filter board temperature model respectively includes:

[0016] Using the training set to perform corresponding model training on the constructed initial multivariable filter plate temperature model until a preset model training termination condition is met, thereby obtaining a corresponding trained model;

[0017] In the process of model testing the trained model based on the test set, the corresponding model performance evaluation operation is performed through the root mean square error calculation formula, the predicted value to be evaluated output by the trained model and the corresponding actual temperature value; wherein, the predicted value to be evaluated is the temperature prediction value obtained after the trained model processes the target input variable in any sample in the training set.

[0018] Optionally, determining corresponding training sets and test sets through the screened data and preset data processing rules includes:

[0019] After cleaning the screened data, randomly shuffling the cleaned data to obtain corresponding shuffled data;

[0020] Based on the Min-Max Scaling function, the shuffled data is normalized accordingly to obtain normalized data;

[0021] The normalized data is divided into data sets based on a preset data set division ratio to obtain corresponding training sets and test sets.

[0022] Optionally, before using the training set and the test set to perform corresponding model training and model testing on the constructed initial multivariable filter board temperature model to obtain the corresponding target multivariable filter board temperature model, the method further includes:

[0023] A deep neural network with a preset number of layers is established, and the deep neural network is configured with the corresponding number of nodes, activation function, loss function, optimizer, and callback function to complete the corresponding model building operation and obtain the corresponding initial multivariable filter board temperature model.

[0024] Optionally, in the process of monitoring the filter board temperature of the wind turbine based on the target multivariable filter board temperature model, judging whether a preset warning trigger condition is currently met through the current warning threshold information and the local residual sequence to complete the corresponding fault warning operation, including:

[0025] In the process of monitoring the filter plate temperature of the wind turbine based on the target multivariable filter plate temperature model, determining the difference between the current model predicted value and the corresponding actual temperature value, and saving the difference in a local residual sequence;

[0026] Determine current warning threshold information and offset statistics using an exponentially weighted moving average control chart generated based on the residual sequence;

[0027] By comparing the offset statistics and the warning threshold information, it is determined whether the preset warning triggering condition is currently met to complete the corresponding fault warning operation.

[0028] In a second aspect, the present application provides a filter board temperature monitoring device, comprising:

[0029] A variable screening module, used for obtaining the correlation coefficient between each variable in the data to be processed and the temperature of the filter plate based on the Pearson correlation coefficient method, and performing corresponding variable screening according to each of the correlation coefficients and preset variable screening conditions to obtain corresponding screened data; wherein the data to be processed is the status monitoring data corresponding to the wind turbine;

[0030] A target model acquisition module is used to determine a corresponding training set and a test set through the screened data and preset data processing rules, and use the training set and the test set to perform corresponding model training and model testing on the constructed initial multivariable filter board temperature model, so as to obtain a corresponding target multivariable filter board temperature model; the initial multivariable filter board temperature model is a model constructed based on a deep neural network;

[0031] A fault warning module is used to determine whether the preset warning trigger conditions are currently met through the current warning threshold information and the local residual sequence during the process of monitoring the filter board temperature of the wind turbine based on the target multivariable filter board temperature model, so as to complete the corresponding fault warning operation; wherein the residual sequence includes a number of values ​​used to characterize the difference between the model prediction value and the corresponding actual temperature value.

[0032] Optionally, the variable screening module includes:

[0033] A data acquisition unit, used for acquiring the status monitoring data of the wind turbine to obtain the data to be processed;

[0034] A correlation calculation unit, used for calculating the correlation coefficient between each variable in the data to be processed and the filter plate temperature based on the Pearson correlation coefficient method;

[0035] A coefficient judgment unit, used to judge whether each of the correlation coefficients is greater than a corresponding preset coefficient threshold value, so as to obtain a corresponding judgment result;

[0036] The variable screening completion unit is used to determine the corresponding variable as the target input variable for the judgment result indicating that it is greater than the preset coefficient threshold, so as to complete the corresponding variable screening operation and obtain the corresponding screened data.

[0037] In a third aspect, the present application provides an electronic device, including:

[0038] Memory, used to store computer programs;

[0039] The processor is used to execute the computer program to implement the aforementioned filter board temperature monitoring method.

[0040] In a fourth aspect, the present application provides a computer-readable storage medium for storing a computer program, wherein the computer program implements the aforementioned filter board temperature monitoring method when executed by a processor.

[0041] The present application first obtains the correlation coefficient between each variable in the data to be processed and the filter board temperature based on the Pearson correlation coefficient method, and performs corresponding variable screening according to each of the correlation coefficients and the preset variable screening conditions to obtain the corresponding screened data; wherein, the data to be processed is the status monitoring data corresponding to the wind turbine; the corresponding training set and test set are determined by the screened data and the preset data processing rules, and the constructed initial multivariable filter board temperature model is respectively trained and tested by the training set and the test set to obtain the corresponding target multivariable filter board temperature model; the initial multivariable filter board temperature model is a model constructed based on a deep neural network; in the process of monitoring the filter board temperature of the wind turbine based on the target multivariable filter board temperature model, the current warning threshold information and the local residual sequence are used to determine whether the preset warning trigger condition is currently met to complete the corresponding fault warning operation; wherein, the residual sequence includes a number of values ​​used to characterize the difference between the model prediction value and the corresponding actual temperature value. As can be seen from the above, before establishing the target multivariable filter board temperature model, the present application first screens the various variables to be processed based on the Pearson correlation coefficient method to obtain the training set and test set corresponding to the initial multivariable filter board temperature model for training and testing, and performs corresponding model training and model testing on the initial multivariable filter board temperature model through the training set and the test set to obtain the corresponding target multivariable filter board temperature model, and then when using the multivariable filter board temperature model to monitor the filter board temperature, it is determined whether to perform a fault warning for the wind turbine based on the current local residual sequence. In this way, it is possible to effectively monitor the filter board temperature based on multiple variables, and then effectively realize the fault warning of the wind turbine, thereby improving the accuracy and effectiveness of monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0043] Figure 1 This is a flow chart of a filter board temperature monitoring method disclosed in this application;

[0044] Figure 2 A schematic diagram of a condition monitoring variable of a wind turbine provided in this application;

[0045] Figure 3 A flow chart of a specific filter board temperature monitoring method disclosed in this application;

[0046] Figure 4 A specific filter board temperature abnormality warning diagram provided for this application;

[0047] Figure 5 This is a structural flow chart of a filter board temperature monitoring device disclosed in this application;

[0048] Figure 6 This is a structural diagram of an electronic device disclosed in this application. DETAILED DESCRIPTION

[0049] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0050] In the power system, monitoring the temperature of the filter board can effectively realize fault warning of the wind turbine. At present, the research on specific parameters based on data-driven fault warning methods mainly relies on the learning of historical state information of variables, and realizes abnormal state warning by predicting the future trend of variables. In fact, as a complex system with multiple tightly coupled components, the wind turbine set has a close relationship between the internal components or variables. Considering the historical information of a certain variable alone is not enough to mine the most accurate historical information of the variable. To this end, the present application provides a filter board temperature monitoring method, which can monitor the filter board temperature based on multiple variables to realize fault warning of wind turbines.

[0051] See also Figure 1 As shown, an embodiment of the present invention discloses a filter board temperature monitoring method, comprising:

[0052] Step S11, based on the Pearson correlation coefficient method, obtain the correlation coefficient between each variable in the data to be processed and the filter board temperature, and perform corresponding variable screening according to each of the correlation coefficients and preset variable screening conditions to obtain corresponding screened data; wherein the data to be processed is status monitoring data corresponding to the wind turbine.

[0053] In this embodiment, the wind turbine is a complex system with multiple components tightly coupled. There is a very close relationship between the internal components and variables. Considering the historical information of a certain variable alone is not enough to mine the most accurate historical information of the variable. The influence of multiple variables on the temperature of the filter board should be considered at the same time. The data to be processed in this embodiment can be SCADA data obtained from the Supervisory Control And Data Acquisition (SCADA) system, such as power, wind speed, generator speed, pitch motor torque and various temperatures, etc. Figure 2 If all variables are used for multivariable filter board modeling, it is obviously a waste of resources. In order to improve efficiency and save costs, it is necessary to perform variable screening operations on the variables in the above SCADA data to obtain the training set and test set corresponding to the initial multivariable filter board temperature model for training and testing.

[0054] It can be understood that when performing variable screening operations on the above variables, the correlation coefficient between the above variables and the filter board temperature can be calculated by the Pearson correlation coefficient method. If the correlation coefficient between the variables in the data to be processed and the filter board temperature meets the preset variable screening conditions, the variables in the data to be processed are subjected to variable screening processing to obtain the corresponding screened data. Specifically, the correlation coefficient between each variable in the data to be processed and the filter board temperature is obtained based on the Pearson correlation coefficient method, and the corresponding variable screening is performed according to each of the correlation coefficients and the preset variable screening conditions, including: obtaining the state monitoring data of the wind turbine to obtain the data to be processed; calculating the correlation coefficient between each variable in the data to be processed and the filter board temperature based on the Pearson correlation coefficient method; respectively judging whether each of the correlation coefficients is greater than the corresponding preset coefficient threshold value to obtain the corresponding judgment result; for the judgment result indicating that it is greater than the preset coefficient threshold value, determining the corresponding variable as the target input variable to complete the corresponding variable screening operation and obtain the corresponding screened data.

[0055] In a first specific implementation manner, the Pearson correlation coefficient method is used to calculate the correlation between each variable in the data to be processed and the filter board temperature, and the calculation formula is as follows:

[0056] ;

[0057] Where n is the number of samples; X is any variable in the data to be processed; Y is the target variable (filter board temperature); , is the sample mean; r X,Yis the correlation between variable X and variable Y. It should be pointed out that the range of the Pearson correlation coefficient is [-1,1]. The larger the absolute value of the Pearson correlation coefficient, the stronger the correlation; the closer the value of the coefficient is to 0, the smaller the correlation. Specifically, the preset coefficient threshold can be set to 0.6. If the correlation coefficient between each variable in the data to be processed and the filter board temperature calculated based on the Pearson correlation coefficient method is greater than 0.6, then the variables greater than 0.6 are preliminarily selected as the target input variables for filter board temperature modeling; if the correlation coefficient between each variable in the data to be processed and the filter board temperature calculated based on the Pearson correlation coefficient method is less than or equal to 0.6, then the variables less than or equal to 0.6 and the variable data related to the variables are deleted from the data to be processed. It should be pointed out that the method for calculating the correlation between each variable in the data to be processed and the filter board temperature is not limited to the Pearson correlation coefficient method.

[0058] In a second specific implementation, the Spearman's rank correlation coefficient method (Spearman's rank correlation coefficient) can be used to measure the rank correlation between each variable in the data to be processed and the filter board temperature. The value range of the Spearman's rank correlation coefficient is also [-1,1], where 1 indicates a complete positive correlation and -1 indicates no correlation at all.

[0059] In a third specific implementation, the Kendall's coefficient of concordance method can be used to calculate the degree of correlation between each variable in the data to be processed and the filter plate temperature, which is similar to the Spearman rank correlation coefficient method, but the Kendall's coefficient of concordance method provides more classifications and can be applied to different situations. The value range of the Kendall's coefficient of concordance is also [-1,1], where the closer the absolute value is to 1, the stronger the correlation is, and the closer the absolute value is to 0, the weaker the correlation is.

[0060] Furthermore, the target input variables determined based on the above-mentioned Pearson correlation coefficient method include: NCC300 temperature, NCC320 temperature, machine-side inductor temperature, grid-side inductor temperature, machine-side semiconductor temperature, grid-side semiconductor temperature, ambient temperature, maximum fault current, gearbox bearing temperature, gearbox oil temperature, driving direction generator bearing temperature, non-driving direction generator bearing temperature, maximum generator winding temperature, pitch motor 1 temperature, pitch motor 2 temperature, pitch motor 3 temperature, yaw inverter temperature, cabin temperature, cabin battery temperature, blade 1 angle, blade 2 angle, blade 3 angle, blade 1 maximum angle, blade 2 maximum angle, blade 3 maximum angle, generator average speed, generator maximum speed, control generator average speed, etc. 29 variables.

[0061] Step S12, determining the corresponding training set and test set through the screened data and preset data processing rules, and using the training set and the test set to perform corresponding model training and model testing on the constructed initial multivariable filter board temperature model, so as to obtain the corresponding target multivariable filter board temperature model; the initial multivariable filter board temperature model is a model constructed based on a deep neural network.

[0062] In this embodiment, before training the initial multivariable filter board temperature model, it is necessary to first obtain a training set and a test set for training and testing the initial multivariable filter board temperature model, pre-process the screened data, and determine the corresponding training set and test set through preset data processing rules. Specifically, the corresponding training set and test set are determined through the screened data and preset data processing rules, including: after cleaning the screened data, randomly shuffling the cleaned data to obtain the corresponding shuffled data; normalizing the shuffled data based on the Min-Max Scaling function to obtain normalized data; and dividing the normalized data based on the preset data set division ratio to obtain the corresponding training set and test set.

[0063] It is understandable that after obtaining the filtered data, the filtered data is cleaned, that is, the filtered data is rechecked to find and correct the information in the filtered data. The duplicate filtered data can be deleted, but in order to prevent the wrong data from being deleted, a backup can be made while deleting the duplicate filtered data; then the cleaned data is randomly shuffled, and the obtained shuffled data is normalized accordingly using the Min-Max Scaling function, that is, the shuffled data is scaled so that the shuffled data is mapped to a specific interval, such as [-1,1]. In the specific application process, the method of normalizing the shuffled data is not limited to the Min-Max Scaling function, as long as the data of different scales or ranges can be unified to the same scale range. The obtained normalized data is divided according to the preset data set division ratio. The preset data set division ratio can be adjusted accordingly according to the training scheme of the initial multivariate filter plate temperature model, which is not specifically limited here.

[0064] It should be pointed out that the initial multivariable filter board temperature model is a model constructed based on a deep neural network. Before training and testing the initial multivariable filter board temperature model, a deep neural network that meets the requirements is first established. Specifically, before using the training set and the test set to perform corresponding model training and model testing on the constructed initial multivariable filter board temperature model to obtain the corresponding target multivariable filter board temperature model, it also includes: establishing a deep neural network with a preset number of layers, and performing corresponding node number configuration, activation function configuration, loss function configuration, optimizer configuration, and callback function configuration on the deep neural network to complete the corresponding model construction operation and obtain the corresponding initial multivariable filter board temperature model.

[0065] Furthermore, Keras (an artificial neural network library) or other methods can be used to establish a deep neural network for the establishment of a target multivariable filter board temperature model. First, a deep neural network with a preset number of layers is established, and the corresponding number of nodes of the deep neural network is configured, and the configuration of the number of nodes is in a one-to-one correspondence with the preset number of layers. In a specific implementation, a four-layer deep neural network is established, and the number of nodes of the four layers is set to 25, 5, 2, and 1 respectively, and the corresponding activation functions are configured as ReLU (Linear rectification function, linear rectification function) function, ReLU function, ReLU function, Sigmoid (an S-shaped growth curve) function, MSE (mean_squared_error, mean square error) is configured as the model loss function, Adam (Adaptive Moment Estimation, a deep learning optimization algorithm) optimizer is configured, and EarlyStopping (a callback function) is introduced to avoid overfitting. Among them, the activation function configuration, loss function configuration, optimizer configuration, and callback function configuration can be adjusted accordingly based on actual conditions.

[0066] It is further necessary to understand that after the initial multivariable filter board temperature model is constructed based on the deep neural network, the initial multivariable filter board temperature model needs to be subjected to corresponding model training and model testing. Specifically, the use of the training set and the test set to respectively perform corresponding model training and model testing on the constructed initial multivariable filter board temperature model includes: using the training set to perform corresponding model training on the constructed initial multivariable filter board temperature model until the preset model training termination condition is met, and a corresponding trained model is obtained; in the process of model testing the trained model based on the test set, a corresponding model performance evaluation operation is performed through the root mean square error calculation formula, the predicted value to be evaluated output by the trained model, and the corresponding actual temperature value; wherein, the predicted value to be evaluated is the temperature prediction value obtained after the trained model processes the target input variable in any sample in the training set.

[0067] In this embodiment, before the training set performs corresponding model training on the constructed initial multivariable filter board temperature model, a preset model training termination condition can be set. If the model training reaches the preset model training termination condition, the training is stopped to obtain the corresponding trained model.

[0068] In a specific implementation, before the training set performs corresponding model training on the initial multivariable filter board temperature model, the preset number of training rounds is customized according to actual conditions. If the preset number of training rounds is 3, then after all samples in the training set undergo 3 rounds of model training in the constructed initial multivariable filter board temperature model, the model training is terminated to obtain the corresponding trained model.

[0069] In another specific implementation, the change of training loss during training is observed through the model loss function, and the termination of training is controlled according to the change of loss. If the training loss keeps decreasing, the model training is stopped, thereby preventing the model from overfitting.

[0070] It can be further understood that after obtaining the above-mentioned trained model, the trained model can be evaluated using RMSE (Root Mean Squared Error) or other methods that can reflect the error between the predicted value and the actual value. The RMSE calculation formula is as follows:

[0071] ;

[0072] Where n is the number of test set samples; and are respectively the model predicted value (that is, the predicted value to be evaluated) and the actual measured value (that is, the actual temperature value) of the ith input sample in the test set.

[0073] Step S13, in the process of monitoring the filter board temperature of the wind turbine based on the target multivariable filter board temperature model, determine whether the preset warning trigger conditions are currently met through the current warning threshold information and the local residual sequence to complete the corresponding fault warning operation; wherein, the residual sequence includes a number of values ​​for characterizing the difference between the model prediction value and the corresponding actual temperature value.

[0074] In this embodiment, the residual sequence includes a number of differences between model predicted values ​​and corresponding actual temperature values. In the process of monitoring the filter board temperature of the wind turbine based on the target multivariable filter board temperature model, when the difference between the model predicted value and the actual measured value of the filter board temperature gradually increases over a period of time, that is, when the preset early warning trigger condition is met, it indicates that the wind turbine has a fault, and the fault early warning operation of the wind turbine is immediately triggered; if the difference between the model predicted value and the actual measured value of the filter board temperature over a period of time is small, that is, when the preset early warning trigger condition is not met, it indicates that the state of the wind turbine is normal, and there is no need to trigger the fault early warning operation of the wind turbine.

[0075] It can be understood that the Exponentially Weighted Moving-Average (EWMA) control chart can detect abnormalities of wind turbines by monitoring the residual sequence. Specifically, a difference sequence consisting of several differences between the model prediction value and the corresponding actual temperature value is collected over a period of time. The data in the difference sequence should be continuous and collected in chronological order to ensure the timeliness and relevance of the data in the difference sequence. The calculation of the EWMA statistic involves assigning different weights to each data point. Generally, newer data points will obtain higher weights, while older data points will obtain fewer weights. The newness and weight of the data point are positively correlated. Then, the EWMA statistic of each data point is calculated according to the EWMA model, and the control limit is calculated based on the EWMA statistic. The formation of the control limit is based on the sensitivity to process changes to ensure timely detection of abnormal changes in the process. The control line of the EWMA chart can also be determined according to demand to ensure the sensitivity and accuracy of the EWMA chart. In the process of using the EWMA chart for monitoring, if the sample point exceeds the control line, it may indicate that the wind turbine is abnormal, and an alarm can be triggered in time.

[0076] As can be seen from the above, before establishing the target multivariable filter board temperature model, the present application first screens the various variables to be processed based on the Pearson correlation coefficient method to obtain the training set and test set corresponding to the initial multivariable filter board temperature model for training and testing, and performs corresponding model training and model testing on the initial multivariable filter board temperature model through the training set and the test set to obtain the corresponding target multivariable filter board temperature model, and then when using the multivariable filter board temperature model to monitor the filter board temperature, it is determined whether to perform a fault warning for the wind turbine based on the current local residual sequence. In this way, it is possible to effectively monitor the filter board temperature based on multiple variables, and then effectively realize the fault warning of the wind turbine, thereby improving the accuracy and effectiveness of monitoring.

[0077] It can be seen from the above embodiments that the present application implements fault warning of wind turbines based on an exponentially weighted moving average control diagram. Therefore, the process of implementing fault warning based on an exponentially weighted moving average control diagram is described.

[0078] See also Figure 3 As shown, an embodiment of the present invention discloses a filter board temperature monitoring method, comprising:

[0079] Step S21, based on the Pearson correlation coefficient method, obtain the correlation coefficient between each variable in the data to be processed and the filter board temperature, and perform corresponding variable screening according to each of the correlation coefficients and preset variable screening conditions to obtain corresponding screened data; wherein the data to be processed is status monitoring data corresponding to the wind turbine.

[0080] Step S22, determining the corresponding training set and test set through the screened data and preset data processing rules, and using the training set and the test set to perform corresponding model training and model testing on the constructed initial multivariable filter board temperature model, so as to obtain the corresponding target multivariable filter board temperature model; the initial multivariable filter board temperature model is a model constructed based on a deep neural network.

[0081] Step S23: in the process of monitoring the filter board temperature of the wind turbine based on the target multivariable filter board temperature model, determine the difference between the current model predicted value and the corresponding actual temperature value, and save the difference in a local residual sequence.

[0082] In this embodiment, the filter board temperature of the wind turbine is detected by using the trained target multivariable filter board temperature model, the model prediction value at the current moment is determined based on the target multivariable filter board temperature model, and the difference between the model prediction value and the actual temperature value corresponding to the current moment is calculated, that is, the residual between the model prediction value and the actual temperature value at the current moment, and by determining the residual between the model prediction value and the actual temperature value corresponding to the current moment over a period of time, a residual sequence composed of a plurality of residuals over a period of time is obtained, and by monitoring the residual sequence, the fault information of the wind turbine can be obtained in advance.

[0083] Step S24: Determine the current warning threshold information and the offset statistic using the exponentially weighted moving average control chart generated based on the residual sequence.

[0084] In this embodiment, a corresponding exponentially weighted moving average control chart can be generated based on the residual sequence within a period of time. The exponentially weighted moving average control chart can realize abnormal detection of wind turbines by monitoring the residual sequence. The exponentially weighted moving average control chart can be used to calculate the offset statistics at the current moment. The calculation method is as follows:

[0085] ;

[0086] in, is the current moment in the exponentially weighted moving average control chart The offset statistic of t-1 is the offset statistic at time t-1 in the exponentially weighted moving average control chart; For the current moment The residual of is a smoothing parameter. In a specific implementation , here The value of can be adjusted accordingly according to the actual situation. It should be noted that the central value z of the exponentially weighted moving average control chart is 0 is the mean of the residuals of all samples in the model test set, that is, the exponentially weighted moving average control chart is based on the exponentially weighted moving average of the sample means of all historical data. The control limits of EWMA are defined as follows:

[0087] ;

[0088] ;

[0089] in, is the lower control limit; is the upper control limit; It is a constant that determines the alarm sensitivity, which is generally 3, but can also be determined based on actual conditions. is the estimated variance of EWMA, and the calculation formula of the estimated variance is as follows:

[0090] ;

[0091] In this embodiment, is the variance of the residuals of the samples in the model test set.

[0092] Step S25: By comparing the offset statistics and the warning threshold information, it is determined whether the preset warning triggering condition is currently met to complete the corresponding fault warning operation.

[0093] In this embodiment, a continuously increasing difference over a period of time can usually indicate an abnormality of the equipment, and setting an alarm threshold can obtain fault information in advance so as to make a maintenance plan. By comparing the temperature prediction value and the actual temperature value of the filter board in the exponentially weighted moving average control diagram generated based on the residual sequence, if the offset statistics of the temperature prediction value and the actual temperature value gradually increase over a period of time, it indicates that the residual sequence at the current moment reaches the preset alarm threshold, thus triggering the fault warning operation of the wind turbine.

[0094] In a specific implementation, the SCADA data of wind turbine No. 52 of a certain wind farm is used for variable screening and multivariable filter board temperature modeling. The fault case data occurring at 13:11:32 on 2019.02.17 is used for the condition monitoring and early warning research of the wind turbine. The predicted and actual temperature values ​​of the filter board temperature during this period and the EWMA early warning effect are shown in Figure 2. Figure 4 As shown in the figure, in a period of time far from the occurrence of the fault, the model prediction value of the filter board temperature is consistent with the actual temperature value, and the residual is small; as the fault time gets closer, the degree of consistency with the actual temperature value gradually deteriorates, and the residual gradually increases, which warns the occurrence of this fault in advance. The above residual sequence is sent to the EWMA model to accurately obtain the fault warning time. EWMA is very sensitive to the change of residuals. The alarm index of EWMA reaches the upper threshold at 32 points, and the warning amount is 28, which means that the alarm is triggered 4 hours and 40 minutes in advance.

[0095] Among them, the specific implementation process of the above steps S21 and S22 can refer to the corresponding content in the above embodiments, and will not be repeated here.

[0096] It can be seen from the above that a residual sequence is obtained based on the difference between the temperature prediction value and the actual temperature value, and the warning threshold information and the corresponding offset statistics are determined by the exponentially weighted moving average control chart generated by the residual sequence. By judging the warning threshold information and the corresponding offset statistics, it is determined whether the preset warning trigger condition is met at the current moment, and the fault warning information of the wind turbine is obtained in advance. In this way, in the process of monitoring the filter board temperature of the wind turbine based on the target multivariate filter board temperature model, it is determined whether the preset warning trigger condition is met at present through the current warning threshold information and the local residual sequence to complete the corresponding fault warning operation. Fault warning can be effectively realized based on EWMA and residual sequence, and the timeliness of the warning is improved while ensuring the accuracy of the warning.

[0097] Accordingly, see Figure 5 As shown, the present application also provides a filter board temperature monitoring device, comprising:

[0098] The variable screening module 11 is used to obtain the correlation coefficient between each variable in the data to be processed and the filter plate temperature based on the Pearson correlation coefficient method, and perform corresponding variable screening according to each of the correlation coefficients and the preset variable screening conditions to obtain corresponding screened data; wherein the data to be processed is the state monitoring data corresponding to the wind turbine;

[0099] The target model acquisition module 12 is used to determine the corresponding training set and test set through the screened data and the preset data processing rules, and use the training set and the test set to perform corresponding model training and model testing on the constructed initial multivariable filter board temperature model, so as to obtain the corresponding target multivariable filter board temperature model; the initial multivariable filter board temperature model is a model constructed based on a deep neural network;

[0100] The fault warning module 13 is used to determine whether the preset warning trigger conditions are currently met through the current warning threshold information and the local residual sequence during the process of monitoring the filter board temperature of the wind turbine based on the target multivariable filter board temperature model, so as to complete the corresponding fault warning operation; wherein the residual sequence includes a number of values ​​used to characterize the difference between the model prediction value and the corresponding actual temperature value.

[0101] As can be seen from the above, before establishing the target multivariable filter board temperature model, the present application first screens the various variables to be processed based on the Pearson correlation coefficient method to obtain the training set and test set corresponding to the initial multivariable filter board temperature model for training and testing, and performs corresponding model training and model testing on the initial multivariable filter board temperature model through the training set and the test set to obtain the corresponding target multivariable filter board temperature model, and then when using the multivariable filter board temperature model to monitor the filter board temperature, it is determined whether to perform a fault warning for the wind turbine based on the current local residual sequence. In this way, it is possible to effectively monitor the filter board temperature based on multiple variables, and then effectively realize the fault warning of the wind turbine, thereby improving the accuracy and effectiveness of monitoring.

[0102] In some specific implementations, the variable screening module 11 may specifically include:

[0103] A data acquisition unit, used for acquiring the status monitoring data of the wind turbine to obtain the data to be processed;

[0104] A correlation calculation unit, used for calculating the correlation coefficient between each variable in the data to be processed and the filter plate temperature based on the Pearson correlation coefficient method;

[0105] A coefficient judgment unit, used to judge whether each of the correlation coefficients is greater than a corresponding preset coefficient threshold value, so as to obtain a corresponding judgment result;

[0106] The variable screening completion unit is used to determine the corresponding variable as the target input variable for the judgment result indicating that it is greater than the preset coefficient threshold, so as to complete the corresponding variable screening operation and obtain the corresponding screened data.

[0107] In some specific implementations, the target model acquisition module 12 may specifically include:

[0108] A model training unit, used to perform corresponding model training on the constructed initial multivariable filter plate temperature model using the training set until a preset model training termination condition is met, thereby obtaining a corresponding trained model;

[0109] A model evaluation unit is used to perform corresponding model performance evaluation operations in the process of model testing the trained model based on the test set through a root mean square error calculation formula, a predicted value to be evaluated output by the trained model, and a corresponding actual temperature value.

[0110] In some specific implementations, the target model acquisition module 12 may specifically include:

[0111] A data scrambling unit, configured to randomly scramble the cleaned data after cleaning the screened data, so as to obtain corresponding scrambled data;

[0112] A data normalization unit, used for normalizing the shuffled data accordingly based on a Min-Max Scaling function to obtain normalized data;

[0113] The data partitioning unit is used to partition the normalized data based on a preset data set partitioning ratio to obtain corresponding training sets and test sets.

[0114] In some specific implementations, the target model acquisition module 12 may specifically include:

[0115] The initial model building unit is used to establish a deep neural network with a preset number of layers, and perform corresponding node number configuration, activation function configuration, loss function configuration, optimizer configuration, and callback function configuration on the deep neural network to complete the corresponding model building operation and obtain the corresponding initial multivariable filter board temperature model.

[0116] In some specific implementations, the fault warning module 13 may specifically include:

[0117] a difference determination unit, used for determining the difference between the current model prediction value and the corresponding actual temperature value during the process of monitoring the filter board temperature of the wind turbine based on the target multivariable filter board temperature model, and saving the difference to a local residual sequence;

[0118] An offset statistic determination unit, used to determine current warning threshold information and offset statistics using an exponentially weighted moving average control chart generated based on the residual sequence;

[0119] The fault warning module is used to determine whether the preset warning triggering condition is currently met by comparing the offset statistics and the warning threshold information, so as to complete the corresponding fault warning operation.

[0120] Furthermore, the present application also discloses an electronic device. Figure 6It is a structural diagram of an electronic device 20 shown according to an exemplary embodiment, and the content in the figure cannot be regarded as any limitation on the scope of use of this application. The electronic device 20 may specifically include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input and output interface 25 and a communication bus 26. Among them, the memory 22 is used to store a computer program, and the computer program is loaded and executed by the processor 21 to implement the relevant steps in the filter board temperature monitoring method disclosed in any of the aforementioned embodiments. In addition, the electronic device 20 in this embodiment may specifically be an electronic computer.

[0121] In this embodiment, the power supply 23 is used to provide working voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and the external device, and the communication protocol it follows is any communication protocol that can be applied to the technical solution of the present application, and is not specifically limited here; the input and output interface 25 is used to obtain external input data or output data to the outside world, and its specific interface type can be selected according to specific application needs and is not specifically limited here.

[0122] In addition, the memory 22 as a carrier for resource storage may be a read-only memory, a random access memory, a disk or an optical disk, etc. The resources stored thereon may include an operating system 221, a computer program 222, etc., and the storage method may be temporary storage or permanent storage.

[0123] The operating system 221 is used to manage and control the hardware devices and computer program 222 on the electronic device 20, which can be Windows Server, Netware, Unix, Linux, etc. In addition to including a computer program that can be used to complete the filter board temperature monitoring method performed by the electronic device 20 disclosed in any of the aforementioned embodiments, the computer program 222 can further include a computer program that can be used to complete other specific tasks.

[0124] Furthermore, the present application also discloses a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, the filter board temperature monitoring method disclosed above is implemented. For the specific steps of the method, reference may be made to the corresponding contents disclosed in the above embodiments, and no further description will be given here.

[0125] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0126] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0127] The steps of the method or algorithm described in conjunction with the embodiments disclosed herein may be implemented directly using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0128] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0129] The technical solution provided by the present application is introduced in detail above. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea. At the same time, for general technicians in this field, according to the idea of ​​the present application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A filter board temperature monitoring method, characterized in that: include: Obtaining the correlation coefficient between each variable in the data to be processed and the temperature of the filter plate based on the Pearson correlation coefficient method, and performing corresponding variable screening according to each of the correlation coefficients and the preset variable screening conditions to obtain corresponding screened data; wherein the data to be processed is the status monitoring data corresponding to the wind turbine; Determine the corresponding training set and test set by the screened data and the preset data processing rules, and use the training set and the test set to perform corresponding model training and model testing on the constructed initial multivariable filter board temperature model, so as to obtain the corresponding target multivariable filter board temperature model; the initial multivariable filter board temperature model is a model constructed based on a deep neural network; In the process of monitoring the filter board temperature of the wind turbine based on the target multivariable filter board temperature model, whether the preset warning trigger condition is currently met is judged by the current warning threshold information and the local residual sequence to complete the corresponding fault warning operation; wherein the residual sequence includes a plurality of values ​​for representing the difference between the model prediction value and the corresponding actual temperature value; The method of obtaining the correlation coefficient between each variable in the data to be processed and the filter board temperature based on the Pearson correlation coefficient method, and performing corresponding variable screening according to each of the correlation coefficients and the preset variable screening conditions, includes: Acquire status monitoring data of the wind turbine to obtain the data to be processed; Calculate the correlation coefficient between each variable in the data to be processed and the filter plate temperature based on the Pearson correlation coefficient method; Determine whether each of the correlation coefficients is greater than a corresponding preset coefficient threshold to obtain a corresponding determination result; For the judgment result indicating that the value is greater than the preset coefficient threshold, determining the corresponding variable as the target input variable to complete the corresponding variable screening operation and obtain the corresponding screened data; The training set and the test set respectively perform corresponding model training and model testing on the constructed initial multivariable filter board temperature model, including: Using the training set to perform corresponding model training on the constructed initial multivariable filter plate temperature model until a preset model training termination condition is met, thereby obtaining a corresponding trained model; In the process of model testing the trained model based on the test set, a corresponding model performance evaluation operation is performed through a root mean square error calculation formula, a predicted value to be evaluated output by the trained model, and a corresponding actual temperature value; wherein the predicted value to be evaluated is a temperature prediction value obtained after the trained model processes the target input variable in any sample in the training set; The determining of the corresponding training set and test set by using the screened data and preset data processing rules includes: After cleaning the screened data, randomly shuffling the cleaned data to obtain corresponding shuffled data; Based on the Min-Max Scaling function, the shuffled data is normalized accordingly to obtain normalized data; Dividing the normalized data based on a preset data set division ratio to obtain corresponding training sets and test sets; In the process of monitoring the filter board temperature of the wind turbine based on the target multivariable filter board temperature model, judging whether the preset warning triggering condition is currently met through the current warning threshold information and the local residual sequence to complete the corresponding fault warning operation, including: In the process of monitoring the filter plate temperature of the wind turbine based on the target multivariable filter plate temperature model, determining the difference between the current model predicted value and the corresponding actual temperature value, and saving the difference in a local residual sequence; Determine current warning threshold information and offset statistics using an exponentially weighted moving average control chart generated based on the residual sequence; By comparing the offset statistics and the warning threshold information, it is determined whether the preset warning triggering condition is currently met to complete the corresponding fault warning operation.

2. The filter board temperature monitoring method according to claim 1, characterized in that: Before using the training set and the test set to perform corresponding model training and model testing on the constructed initial multivariable filter board temperature model to obtain the corresponding target multivariable filter board temperature model, the method further includes: A deep neural network with a preset number of layers is established, and the deep neural network is configured with the corresponding number of nodes, activation function, loss function, optimizer, and callback function to complete the corresponding model building operation and obtain the corresponding initial multivariable filter board temperature model.

3. A filter board temperature monitoring device, characterized in that: include: A variable screening module, used for obtaining the correlation coefficient between each variable in the data to be processed and the temperature of the filter plate based on the Pearson correlation coefficient method, and performing corresponding variable screening according to each of the correlation coefficients and preset variable screening conditions to obtain corresponding screened data; wherein the data to be processed is the status monitoring data corresponding to the wind turbine; A target model acquisition module is used to determine a corresponding training set and a test set through the screened data and preset data processing rules, and use the training set and the test set to perform corresponding model training and model testing on the constructed initial multivariable filter board temperature model, so as to obtain a corresponding target multivariable filter board temperature model; the initial multivariable filter board temperature model is a model constructed based on a deep neural network; A fault warning module is used to determine whether a preset warning trigger condition is currently met through current warning threshold information and a local residual sequence during the process of monitoring the filter board temperature of the wind turbine based on the target multivariable filter board temperature model, so as to complete the corresponding fault warning operation; wherein the residual sequence includes a plurality of values ​​for representing the difference between the model prediction value and the corresponding actual temperature value; Wherein, the variable screening module includes: A data acquisition unit, used for acquiring the status monitoring data of the wind turbine to obtain the data to be processed; A correlation calculation unit, used for calculating the correlation coefficient between each variable in the data to be processed and the filter plate temperature based on the Pearson correlation coefficient method; A coefficient judgment unit, used to judge whether each of the correlation coefficients is greater than a corresponding preset coefficient threshold value, so as to obtain a corresponding judgment result; A variable screening completion unit, for determining the corresponding variable as a target input variable for the judgment result indicating that the value is greater than the preset coefficient threshold, so as to complete the corresponding variable screening operation and obtain the corresponding screened data; The target model acquisition module includes: A model training unit, used to perform corresponding model training on the constructed initial multivariable filter plate temperature model using the training set until a preset model training termination condition is met, thereby obtaining a corresponding trained model; A model evaluation unit, used for performing corresponding model performance evaluation operations according to a root mean square error calculation formula, a predicted value to be evaluated output by the trained model, and a corresponding actual temperature value during a process of performing model testing on the trained model based on the test set; wherein the predicted value to be evaluated is a temperature prediction value obtained after the trained model processes the target input variable in any sample in the training set; The target model acquisition module includes: A data scrambling unit, configured to randomly scramble the cleaned data after cleaning the screened data, so as to obtain corresponding scrambled data; A data normalization unit, used for normalizing the shuffled data accordingly based on a Min-Max Scaling function to obtain normalized data; A data partitioning unit, used to partition the normalized data based on a preset data set partitioning ratio to obtain corresponding training sets and test sets; The fault warning module comprises: a difference determination unit, used for determining the difference between the current model prediction value and the corresponding actual temperature value during the process of monitoring the filter board temperature of the wind turbine based on the target multivariable filter board temperature model, and saving the difference to a local residual sequence; An offset statistic determination unit, used to determine current warning threshold information and offset statistics using an exponentially weighted moving average control chart generated based on the residual sequence; The fault warning module is used to determine whether the preset warning triggering condition is currently met by comparing the offset statistics and the warning threshold information, so as to complete the corresponding fault warning operation.

4. An electronic device, characterized in that: include: Memory, used to store computer programs; A processor, used to execute the computer program to implement the filter board temperature monitoring method as described in claim 1 or 2.

5. A computer-readable storage medium, characterized in that: Used to store a computer program, wherein when the computer program is executed by a processor, the filter board temperature monitoring method as claimed in claim 1 or 2 is implemented.

Citation Information

Patent Citations

  • Fan fault detecting method based on depth belief network model

    CN107725283A

  • Fault early warning method for double-fed fan main shaft

    CN112734130A