A regional wind farm multi-device real-time data processing method and system

Abstract

The application provides a regional wind farm multi-device real-time data processing method and system, the method comprises the following steps: obtaining monitoring operation data of a regional wind farm multi-device, performing data preprocessing on the monitoring operation data to obtain processing data; performing first anomaly identification on the processing data to obtain a first anomaly set; performing second anomaly identification on the processing data to obtain a second anomaly set; determining a comprehensive anomaly set based on the first anomaly set and the second anomaly set, and determining elimination data based on the comprehensive anomaly set; and performing conditional diffusion filling on the elimination data to output final processing data. Through twice anomaly identification on the data, the accuracy, recall rate and precision of the identification can be effectively improved. Then, the conditional diffusion filling is performed on the data, the long-term relationship in the time series data can be effectively captured, the missing values are estimated by selecting normal values as observation values, and the effectiveness and accuracy of the data filling can be improved.

Description

Technical Field

[0001] This invention belongs to the technical field of data processing, and specifically relates to a method and system for real-time data processing of multiple devices in a regional wind farm. Background Technology

[0002] A wind farm is a place that uses wind energy to generate electricity, including onshore wind farms and offshore wind farms. Onshore wind farms refer to the complete set of facilities that use wind on land to obtain electricity. Most onshore wind farms use series capacitor compensation to transmit power. Offshore wind farms mostly refer to near-shore wind power in water depths of about 10 meters. Compared with onshore wind farms, the advantages of offshore wind farms are that they do not occupy land resources, are basically unaffected by topography, have higher wind speeds, and have larger single-unit capacity of wind turbines.

[0003] For the operation and maintenance management of wind farms, it is usually necessary to collect operating data from multiple devices in the wind farm and conduct corresponding analysis based on the operating data, such as power generation, rotor rotation speed, whether there is a fault, remaining service life, etc. However, in reality, due to sensor damage or other external factors, the raw data collected may contain abnormal or missing data, which will directly affect the data analysis results in the subsequent data analysis process. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides a method and system for real-time data processing of multiple devices in a regional wind farm, which solves the technical problems in the prior art.

[0005] In a first aspect, the present invention provides the following technical solution: a method for real-time data processing of multiple devices in a regional wind farm, comprising:

[0006] Monitoring and operation data of multiple devices in a regional wind farm are acquired through preset sensors. The monitoring and operation data is preprocessed to obtain processed data. The preset sensors include wind measurement tower sensors, noise sensors, meteorological sensors, and power generation operation sensors. The monitoring and operation data includes wind farm wind measurement tower data, wind farm noise data, wind farm meteorological data, and wind farm power generation operation data.

[0007] The processed data is subjected to a first anomaly identification to obtain a first anomaly set;

[0008] The processed data is subjected to a second anomaly identification to obtain a second set of anomalies;

[0009] A comprehensive anomaly set is determined based on the first anomaly set and the second anomaly set, and data to be removed is determined based on the comprehensive anomaly set.

[0010] The discarded data is then filled with conditional diffusion to output the final processed data.

[0011] The step of performing a first anomaly identification on the processed data to obtain a first set of anomalies includes:

[0012] The processed data is subjected to a logarithmic ratio transformation and mapped to Euclidean space to obtain a first data matrix;

[0013] The first data matrix is ​​subjected to eigenvalue decomposition to obtain the covariance matrix and principal component sequence;

[0014] Extract the eigenvalues ​​of the covariance matrix, and calculate the reference values ​​of different numbers of principal components based on the eigenvalues. ;

[0015] Select the first few principal components such that their corresponding reference values ​​are greater than the preset reference values ​​to obtain the target principal components;

[0016] Based on the target principal components, the original data is reconstructed and mapped back to the dimension of the processed data to obtain a mapped data matrix;

[0017] The first set of anomalies is determined based on the mapping data matrix.

[0018] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention first acquires monitoring and operation data of multiple devices in a regional wind farm, performs data preprocessing on the monitoring and operation data to obtain processed data; performs a first anomaly identification on the processed data to obtain a first anomaly set; performs a second anomaly identification on the processed data to obtain a second anomaly set; determines a comprehensive anomaly set based on the first and second anomaly sets, and determines data to be removed based on the comprehensive anomaly set; performs conditional diffusion imputation on the removed data to output the final processed data. This invention can effectively improve the accuracy, recall, and precision of anomaly identification by performing anomaly identification twice. Then, performing conditional diffusion imputation on the data can effectively capture long-term relationships in time-series data. By selecting normal values ​​as observations to estimate missing values, the effectiveness and accuracy of data imputation can be improved.

[0019] Preferably, the step of determining the first anomaly set based on the mapping data matrix includes:

[0020] The mapped data matrix is ​​processed using the minimum regularized covariance determinant to obtain the first robust mean estimate. First robust covariance estimation ;

[0021] Based on the first robust mean estimation Compared with the first robust covariance estimate Calculate the judgment value for each piece of processed data. :

[0022] ;

[0023] In the formula, For the first One data processing unit;

[0024] Based on the judgment value Calculate the judgment threshold :

[0025] ;

[0026] In the formula, The median of all judgment values. The absolute deviation of the median of all judgment values;

[0027] Data whose judgment value is greater than the judgment threshold is treated as first abnormal data and stored in the first abnormal set.

[0028] Preferably, the step of performing a second anomaly identification on the processed data to obtain a second anomaly set includes:

[0029] The processed data is subjected to a logarithmic transformation and mapped to Euclidean space to obtain a second data matrix. ;

[0030] The second data matrix is ​​processed using the minimum regularized covariance determinant to obtain the second robust mean estimate. Second robust covariance estimation ;

[0031] For the second robust covariance estimation Perform singular value decomposition:

[0032] ;

[0033] In the formula, It is a left singular matrix containing orthogonal vectors. It is a diagonal matrix with singular values ​​on its diagonal. The eigenvector matrix;

[0034] Calculate the target principal component matrix :

[0035] ;

[0036] In the formula, It is the identity matrix;

[0037] Calculate the reference values ​​of the first few columns of the target principal component matrix, and determine the reference principal components based on the reference values ​​of the first few columns of the target principal component matrix and the preset reference values;

[0038] Calculate the stability score of the reference principal component. :

[0039] ;

[0040] In the formula, To reference the total number of columns in the principal components, For reference principal components, the first The singular values ​​corresponding to the column, For reference principal components, the first Line number Column elements;

[0041] The processed data corresponding to the column elements of the reference principal components whose stable scores are greater than the preset scores are regarded as the second abnormal data and stored in the second abnormal set.

[0042] Preferably, the step of determining a comprehensive anomaly set based on the first anomaly set and the second anomaly set, and determining the data to be removed based on the comprehensive anomaly set, includes:

[0043] The first anomaly set and the second anomaly set are combined to obtain a comprehensive anomaly set. Data in the comprehensive anomaly set is then removed from the processed data to obtain the removed data.

[0044] Preferably, the step of performing conditional diffusion imputation on the discarded data to output the final processed data includes:

[0045] Identify missing values ​​in the removed data and divide the removed data into several subsets according to the location of the missing values;

[0046] Calculate the density ratio of each of the said subsets. :

[0047] ;

[0048] In the formula, They represent the first Estimated probability density of each subset of data;

[0049] Calculate the probability score based on the density ratio. :

[0050] ;

[0051] In the formula, The number of data points in the subset. For the first in the subset of data The density ratio of each data point;

[0052] Based on the probability score Calculate data trust value :

[0053] ;

[0054] In the formula, Represents the first in the subset of data The distance between each data point and the previous data point These are the probability scores for the corresponding subsets of data. The average of the probability scores across all subsets;

[0055] Based on the data trust value, the data in the subset are sorted in descending order, and the first few data points in the sorted subset are selected as the benchmark data.

[0056] Obtaining Gaussian noise The reference data is obtained by using bicubic interpolation. Adjust it to the same magnitude as the Gaussian noise to obtain the adjusted data. ;

[0057] Based on the Gaussian noise The reference data The adjusted data Determine the final data to be processed.

[0058] Preferably, the Gaussian noise-based The reference data The adjusted data The steps to determine the final processed data include:

[0059] The adjusted data is combined with the Gaussian noise to obtain combined data. :

[0060] ;

[0061] In the formula, It is a binary sequence. To adjust the weights, It is the XOR operator;

[0062] The combined data is subjected to initial inverse diffusion to obtain initial data for the next step. :

[0063] ;

[0064] In the formula, For the first The inverse diffusion hyperparameter of the step, For the preset neural network, It is a random number. It is a fixed constant;

[0065] For the initial next step data Conditional injection and precise reverse diffusion are performed to obtain accurate data for the next step. :

[0066] ;

[0067] In the formula, These are the initial parameters. The attenuation coefficient;

[0068] The initial and precise reverse diffusion processes are iteratively repeated until the data from the final reverse diffusion step is output to obtain the target filling data. The target filling data is then used to fill in the missing values ​​according to their positions to obtain the final processed data.

[0069] Secondly, the present invention provides the following technical solution: a real-time data processing system for multiple devices in a regional wind farm, the system comprising:

[0070] The preprocessing module is used to acquire monitoring and operation data of multiple devices in a regional wind farm through preset sensors, and to preprocess the monitoring and operation data to obtain processed data. The preset sensors include wind measurement tower sensors, noise sensors, meteorological sensors and power generation operation sensors. The monitoring and operation data includes wind farm wind measurement tower data, wind farm noise data, wind farm meteorological data and wind farm power generation operation data.

[0071] The first anomaly identification module is used to perform a first anomaly identification on the processed data to obtain a first anomaly set;

[0072] The second anomaly identification module is used to perform second anomaly identification on the processed data to obtain a second anomaly set;

[0073] The elimination module is used to determine a comprehensive anomaly set based on the first anomaly set and the second anomaly set, and to determine the data to be eliminated based on the comprehensive anomaly set.

[0074] The output module is used to perform conditional diffusion filling on the discarded data to output the final processed data.

[0075] Thirdly, the present invention provides the following technical solution: a computer, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-described method for real-time data processing of multiple devices in a regional wind farm.

[0076] Fourthly, the present invention provides the following technical solution: a storage medium storing a computer program, wherein when the computer program is executed by a processor, it implements the above-described method for real-time data processing of multiple devices in a regional wind farm. Attached Figure Description

[0077] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0078] Figure 1 This is a flowchart of a method for real-time data processing of multiple devices in a regional wind farm, provided in Embodiment 1 of the present invention.

[0079] Figure 2 This is a structural block diagram of the regional wind farm multi-device real-time data processing system provided in Embodiment 2 of the present invention;

[0080] Figure 3 This is a schematic diagram of the hardware structure of a computer provided for another embodiment of the present invention.

[0081] The embodiments of the present invention will be further described below with reference to the accompanying drawings. Detailed Implementation

[0082] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain embodiments of the present invention, and should not be construed as limiting the present invention.

[0083] Example 1

[0084] In Embodiment 1 of the present invention, as Figure 1 As shown, a method for real-time data processing of multiple devices in a regional wind farm includes:

[0085] S1. Acquire monitoring and operation data of multiple devices in the regional wind farm through preset sensors, and perform data preprocessing on the monitoring and operation data to obtain processed data. The preset sensors include wind measurement tower sensors, noise sensors, meteorological sensors and power generation operation sensors. The monitoring and operation data includes wind farm wind measurement tower data, wind farm noise data, wind farm meteorological data and wind farm power generation operation data.

[0086] The monitoring and operational data here can be acquired through several different sensors, including wind tower sensors, noise sensors, meteorological sensors, and power generation operation sensors. Wind tower sensors include ultrasonic anemometers, ultrasonic wind direction sensors, platinum resistance temperature sensors, barometric pressure sensors, and resistive humidity sensors. Noise sensors are sound level meters. Meteorological sensors include tipping bucket rain gauges, total radiation meters, humidity / temperature / barometric pressure sensors, icing sensors, forward scattering visibility meters, and lightning monitoring systems. Power generation operation sensors include voltage transformers, current transformers, yaw encoders, pitch angle encoders, speed sensors, and more. The equipment includes speedometers, temperature sensors, fiber optic grating sensors, strain gauges, etc. Among them, the monitoring and operation data includes wind farm anemometer data, wind farm noise data, wind farm meteorological data, and wind farm power generation operation data. The wind farm anemometer data mainly includes wind speed, wind direction, temperature, air pressure, and humidity data. The wind farm noise data includes noise data during power generation operation. The wind farm meteorological data mainly includes precipitation, radiation, visibility, icing, and lightning data. The wind farm power generation operation data includes power, voltage, current, yaw / pitch angle, speed, vibration, temperature, etc. The data preprocessing process here includes format conversion, data normalization, etc., as used in existing technologies.

[0087] S2. Perform a first anomaly identification on the processed data to obtain a first anomaly set;

[0088] Step S2 includes:

[0089] S21. Perform a logarithmic transformation on the processed data and map the processed data into Euclidean space to obtain a first data matrix;

[0090] Specifically, the purpose of the logarithmic ratio transformation here is to map the data to Euclidean space, specifically in matrix form.

[0091] S22. Perform eigenvalue decomposition on the first data matrix to obtain the covariance matrix and principal component sequence;

[0092] Specifically, the eigenvalue decomposition here uses the prcomp function.

[0093] S23. Extract the eigenvalues ​​of the covariance matrix, and calculate the reference values ​​of different numbers of principal components based on the eigenvalues. :

[0094] ;

[0095] In the formula, For the front Reference values ​​for each principal component, For the first 1 eigenvalue, This represents the total number of eigenvalues.

[0096] S24. Select the first few principal components such that their corresponding reference values ​​are greater than the preset reference values ​​to obtain the target principal components;

[0097] Specifically, the default reference value here is 0.8.

[0098] S25. Based on the target principal components, reconstruct the original data and map it back to the dimension of the processed data to obtain a mapped data matrix.

[0099] S26. Determine the first anomaly set based on the mapping data matrix;

[0100] Step S26 includes:

[0101] S261. Perform minimum regularization covariance determinant processing on the mapped data matrix to obtain the first robust mean estimate. First robust covariance estimation ;

[0102] Specifically, robust mean estimates and robust covariance estimates can be obtained through the generalized C-step algorithm and the MRCD algorithm.

[0103] S262, Based on the first robust mean estimation Compared with the first robust covariance estimate Calculate the judgment value for each piece of processed data. :

[0104] ;

[0105] In the formula, For the first One data processing unit;

[0106] Specifically, the purpose of calculating the judgment value is to effectively identify outliers in the data.

[0107] S263, Based on the judgment value Calculate the judgment threshold :

[0108] ;

[0109] In the formula, The median of all judgment values. The absolute deviation of the median of all judgment values;

[0110] S264. Processed data whose judgment value is greater than the judgment threshold are regarded as first abnormal data and stored in the first abnormal set.

[0111] S3. Perform a second anomaly identification on the processed data to obtain a second anomaly set;

[0112] Step S3 includes:

[0113] S31. Perform a logarithmic transformation on the processed data and map the processed data to Euclidean space to obtain a second data matrix. ;

[0114] Specifically, the steps here are the same as those in step S21.

[0115] S32. Perform minimum regularization covariance determinant processing on the second data matrix to obtain the second robust mean estimate. Second robust covariance estimation ;

[0116] Specifically, the steps here are the same as those in step S261 above.

[0117] S33, Estimation of the second robust covariance Perform singular value decomposition:

[0118] ;

[0119] In the formula, It is a left singular matrix containing orthogonal vectors. It is a diagonal matrix with singular values ​​on its diagonal. This is the eigenvector matrix.

[0120] S34. Calculate the target principal component matrix. :

[0121] ;

[0122] In the formula, It is the identity matrix;

[0123] Specifically, the identity matrix here is a matrix whose elements are all 1s.

[0124] S35. Calculate the reference values ​​of the first few columns of the target principal component matrix, and determine the reference principal components based on the reference values ​​of the first few columns of the target principal component matrix and the preset reference values.

[0125] Specifically, the reference values ​​and preset reference values ​​here are the same as in steps S23 and S24.

[0126] S36. Calculate the stability score of the reference principal component. :

[0127] ;

[0128] In the formula, To reference the total number of columns in the principal components, For reference principal components, the first The singular values ​​corresponding to the column, For reference principal components, the first Line number Column elements;

[0129] S37. The processed data corresponding to the column elements of the reference principal components whose stable scores are greater than the preset scores are taken as the second abnormal data and stored in the second abnormal set.

[0130] S4. Determine a comprehensive anomaly set based on the first anomaly set and the second anomaly set, and determine the data to be removed based on the comprehensive anomaly set;

[0131] Specifically, step S4 is as follows:

[0132] Perform a union operation on the first anomaly set and the second anomaly set to obtain a comprehensive anomaly set. Remove the data in the comprehensive anomaly set from the processed data to obtain the removed data.

[0133] Specifically, after the first and second anomaly identification steps described above, two anomaly results will be output. To improve the accuracy of anomaly identification, the two results are combined to obtain a comprehensive anomaly set. The data in the set is then removed from the original processed data to obtain the removed data.

[0134] S5. Perform conditional diffusion filling on the removed data to output the final processed data;

[0135] Specifically, after the above steps, since abnormal data has been removed, there will be gaps in the original data positions. At the same time, due to external factors, there may be missing values ​​in the original data sequence. Therefore, in this step, it is necessary to fill in the missing values.

[0136] Step S5 includes:

[0137] S51. Identify missing values ​​in the removed data and divide the removed data into several subsets according to the position of the missing values;

[0138] Specifically, missing values ​​can be identified using machine learning algorithms, such as random forest networks, neural networks, and so on.

[0139] S52. Calculate the density ratio of each subset of data. :

[0140] ;

[0141] In the formula, They represent the first Estimated probability density of each subset of data.

[0142] S53. Calculate the probability score based on the density ratio. :

[0143] ;

[0144] In the formula, The number of data points in the subset. For the first in the subset of data The density ratio of the data.

[0145] S54, Based on the probability score Calculate data trust value :

[0146] ;

[0147] In the formula, Represents the first in the subset of data The distance between each data point and the previous data point These are the probability scores for the corresponding subsets of data. The average of the probability scores across all subsets;

[0148] Specifically, for the data confidence value, the higher the value, the greater the probability that the data point will be selected as the benchmark data. Therefore, when selecting benchmark data, two constraints need to be met: the probability score of the selected point should be small, indicating that the reliability of the data should be high; and the interval between two adjacent benchmarks should not be too large, because a large interval may increase the difficulty of estimating missing values.

[0149] S55. Based on the data trust value, sort the data in the subset in descending order, and select the first few data in the sorted subset as the benchmark data.

[0150] S56. Obtain Gaussian noise The reference data is obtained by using bicubic interpolation. Adjust it to the same magnitude as the Gaussian noise to obtain the adjusted data. ;

[0151] Specifically, the Gaussian noise here follows a normal distribution.

[0152] S57, Based on the Gaussian noise The reference data The adjusted data Determine the final processed data;

[0153] Step S57 includes:

[0154] S571. Combine the adjusted data with the Gaussian noise to obtain combined data. :

[0155] ;

[0156] In the formula, It is a binary sequence. To adjust the weights, It is the XOR operator;

[0157] Specifically, the binary sequence here is used to represent the position of the reference data, which is 1 if it is the reference data and 0 otherwise. The weight adjustment is used to adjust the relative importance of the reference data and Gaussian noise.

[0158] S572. Perform initial reverse diffusion on the combined data to obtain initial data for the next step. :

[0159] ;

[0160] In the formula, For the first The inverse diffusion hyperparameter of the step, For the preset neural network, It is a random number. It is a fixed constant;

[0161] Specifically, the inverse diffusion hyperparameter is used to control the weights of the current combined data and the weights of the correction terms of the denoising network during the inverse diffusion process. The preset neural network is specifically used to denoise the data, and it is actually a convolutional neural network in the existing technology. The purpose of adding random numbers is to ensure that each generation step is random.

[0162] S573, Regarding the initial next step data Conditional injection and precise reverse diffusion are performed to obtain accurate data for the next step. :

[0163] ;

[0164] In the formula, These are the initial parameters. The attenuation coefficient;

[0165] Specifically, in practice, injecting conditions into a denoising network and expecting the network to automatically generate the desired data using the conditional information may encounter deviation problems. After multiple iterations, the generated values ​​of the observation points may deviate from their true values. One possible solution is to forcibly apply the conditional factor to the generated data (i.e., replace the generated values ​​with the true observation values ​​at each time step), but this may lead to data distortion problems. Therefore, this application introduces a decay coefficient, which, as the time step decreases and back diffusion continues, should make the generated values ​​of the observation points closer and closer to the true observation values, thus ensuring the consistency between the generated values ​​and the observed values.

[0166] S574. Iteratively repeat the above initial and precise reverse diffusion process until the data of the last reverse diffusion step is output to obtain the target filling data. Fill the missing values ​​with the target filling data according to their positions to obtain the final processed data.

[0167] The real-time data processing method for multiple devices in a regional wind farm provided in Embodiment 1 of this invention first acquires monitoring and operation data of multiple devices in the regional wind farm, performs data preprocessing on the monitoring and operation data to obtain processed data; performs a first anomaly identification on the processed data to obtain a first anomaly set; performs a second anomaly identification on the processed data to obtain a second anomaly set; determines a comprehensive anomaly set based on the first and second anomaly sets, and determines data to be removed based on the comprehensive anomaly set; and performs conditional diffusion imputation on the removed data to output the final processed data. This invention can effectively improve the accuracy, recall, and precision of anomaly identification by performing anomaly identification twice. The subsequent conditional diffusion imputation can effectively capture long-term relationships in time-series data. By selecting normal values ​​as observations to estimate missing values, the effectiveness and accuracy of data imputation can be improved.

[0168] Example 2

[0169] like Figure 2 As shown, in Embodiment 2 of the present invention, a multi-device real-time data processing system for a regional wind farm is provided. The system includes:

[0170] The preprocessing module 1 is used to acquire monitoring and operation data of multiple devices in a regional wind farm through preset sensors, and to preprocess the monitoring and operation data to obtain processed data. The preset sensors include wind measurement tower sensors, noise sensors, meteorological sensors and power generation operation sensors. The monitoring and operation data includes wind farm wind measurement tower data, wind farm noise data, wind farm meteorological data and wind farm power generation operation data.

[0171] The first anomaly identification module 2 is used to perform a first anomaly identification on the processed data to obtain a first anomaly set;

[0172] The second anomaly identification module 3 is used to perform second anomaly identification on the processed data to obtain a second anomaly set;

[0173] Elimination module 4 is used to determine a comprehensive anomaly set based on the first anomaly set and the second anomaly set, and to determine the data to be eliminated based on the comprehensive anomaly set;

[0174] Output module 5 is used to perform conditional diffusion filling on the rejected data to output the final processed data;

[0175] The first anomaly identification module 2 includes:

[0176] The first mapping submodule is used to perform a logarithmic ratio transformation on the processed data and map the processed data to Euclidean space to obtain a first data matrix;

[0177] The first decomposition submodule is used to perform feature decomposition on the first data matrix to obtain the covariance matrix and the principal component sequence;

[0178] The extraction submodule is used to extract the eigenvalues ​​of the covariance matrix and calculate the reference values ​​of different numbers of principal components based on the eigenvalues. :

[0179] ;

[0180] In the formula, For the front Reference values ​​for each principal component, For the first 1 eigenvalue, The total number of eigenvalues;

[0181] The first principal component submodule is used to select the first few principal components such that the corresponding reference values ​​are greater than the preset reference values, so as to obtain the target principal components.

[0182] The second mapping submodule is used to reconstruct the original data based on the target principal components and map it back to the dimension of the processed data to obtain a mapped data matrix;

[0183] The first anomaly output submodule is used to determine the first anomaly set based on the mapping data matrix.

[0184] The first exception output submodule includes:

[0185] The determinant processing unit is used to perform minimum regularized covariance determinant processing on the mapped data matrix to obtain a first robust mean estimate. First robust covariance estimation ;

[0186] Judgment value unit, used to determine the value based on the first robust mean estimate. Compared with the first robust covariance estimate Calculate the judgment value for each piece of processed data. :

[0187] ;

[0188] In the formula, For the first One data processing unit;

[0189] Threshold unit, used for determining the value Calculate the judgment threshold :

[0190] ;

[0191] In the formula, The median of all judgment values. The absolute deviation of the median of all judgment values;

[0192] The output unit is used to take the processed data whose judgment value is greater than the judgment threshold as the first abnormal data and store it in the first abnormal set.

[0193] The second anomaly detection module 3 includes:

[0194] The third mapping submodule is used to perform a logarithmic transformation on the processed data and map the processed data to Euclidean space to obtain a second data matrix. ;

[0195] The processing submodule is used to perform minimum regularized covariance determinant processing on the second data matrix to obtain a second robust mean estimate. Second robust covariance estimation ;

[0196] The second decomposition submodule is used for the second robust covariance estimate. Perform singular value decomposition:

[0197] ;

[0198] In the formula, It is a left singular matrix containing orthogonal vectors. It is a diagonal matrix with singular values ​​on its diagonal. The eigenvector matrix;

[0199] The principal component matrix submodule is used to calculate the target principal component matrix. :

[0200] ;

[0201] In the formula, It is the identity matrix;

[0202] The reference submodule is used to calculate the reference values ​​of the first few columns of the target principal component matrix and determine the reference principal components based on the reference values ​​of the first few columns of the target principal component matrix and the preset reference values.

[0203] The scoring module is used to calculate the stability score of the reference principal component. :

[0204] ;

[0205] In the formula, To reference the total number of columns in the principal components, For reference principal components, the first The singular values ​​corresponding to the column, For reference principal components, the first Line number Column elements;

[0206] The anomaly output submodule is used to take the processed data corresponding to the column elements of the reference principal components whose stable scores are greater than the preset scores as the second anomaly data and store them in the second anomaly set.

[0207] The rejection module 4 is specifically used for:

[0208] The first anomaly set and the second anomaly set are combined to obtain a comprehensive anomaly set. Data in the comprehensive anomaly set is then removed from the processed data to obtain the removed data.

[0209] The output module 5 includes:

[0210] The identification submodule is used to identify missing values ​​in the removed data and divide the removed data into several subsets according to the position of the missing values.

[0211] The density ratio submodule is used to calculate the density ratio of each of the said subsets. :

[0212] ;

[0213] In the formula, They represent the first Estimated probability density of each subset of data;

[0214] The fractional submodule is used to calculate the probability fraction based on the density ratio. :

[0215] ;

[0216] In the formula, The number of data points in the subset. For the first in the subset of data The density ratio of each data point;

[0217] The trust submodule is used to base its decisions on the probability score. Calculate data trust value :

[0218] ;

[0219] In the formula, Represents the first in the subset of data The distance between each data point and the previous data point These are the probability scores for the corresponding subsets of data. The average of the probability scores across all subsets;

[0220] The sorting submodule is used to sort the data in the subset in descending order based on the data trust value, and select the first few data in the sorted subset as the benchmark data.

[0221] The noise submodule is used to obtain Gaussian noise. The reference data is obtained by using bicubic interpolation. Adjust it to the same magnitude as the Gaussian noise to obtain the adjusted data. ;

[0222] Filling submodule, used for filling based on the Gaussian noise The reference data The adjusted data Determine the final data to be processed.

[0223] The filling submodule includes:

[0224] A combination unit is used to combine the adjusted data with the Gaussian noise to obtain combined data. :

[0225] ;

[0226] In the formula, It is a binary sequence. To adjust the weights, It is the XOR operator;

[0227] The first reverse diffusion unit is used to perform initial reverse diffusion on the combined data to obtain initial data for the next step. :

[0228] ;

[0229] In the formula, For the first The inverse diffusion hyperparameter of the step, For the preset neural network, It is a random number. It is a fixed constant;

[0230] The second anti-diffusion unit is used to process the initial next step data. Conditional injection and precise reverse diffusion are performed to obtain accurate data for the next step. :

[0231] ;

[0232] In the formula, These are the initial parameters. The attenuation coefficient;

[0233] The filling unit is used to iteratively repeat the above initial backdiffusion and precise backdiffusion process until the data of the last backdiffusion step is output to obtain the target filling data. The target filling data is then filled into the missing values ​​according to their positions to obtain the final processed data.

[0234] In other embodiments of the present invention, the present invention provides the following technical solution: a computer, including a memory 102, a processor 101, and a computer program stored in the memory 102 and executable on the processor 101, wherein the processor 101 executes the computer program to implement the regional wind farm multi-device real-time data processing method as described above.

[0235] Specifically, the processor 101 may include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of the present invention.

[0236] The memory 102 may include a large-capacity memory for data or instructions. For example, and not limitingly, the memory 102 may include a hard disk drive (HDD), a floppy disk drive, a solid-state drive (SSD), flash memory, an optical disk drive, a magneto-optical disk drive, magnetic tape, or a Universal Serial Bus (USB) drive, or a combination of two or more of these. Where appropriate, the memory 102 may include removable or non-removable (or fixed) media. Where appropriate, the memory 102 may be internal or external to a data processing device. In a particular embodiment, the memory 102 is non-volatile memory. In a particular embodiment, the memory 102 includes read-only memory (ROM) and random access memory (RAM). Where appropriate, the ROM may be a mask-programmed ROM, a programmable read-only memory (PROM), an erasable read-only memory (EPROM), an electrically erasable read-only memory (EEPROM), an electrically alterable read-only memory (EAROM), or flash memory, or a combination of two or more of these. Where appropriate, the RAM can be Static Random-Access Memory (SRAM) or Dynamic Random-Access Memory (DRAM). DRAM can be Fast Page Mode Dynamic Random Access Memory (FPMDRAM), Extended Data Out Dynamic Random Access Memory (EDODRAM), Synchronous Dynamic Random-Access Memory (SDRAM), etc.

[0237] The memory 102 can be used to store or cache various data files that need to be processed and / or used for communication, as well as possible computer program instructions executed by the processor 101.

[0238] The processor 101 reads and executes computer program instructions stored in the memory 102 to implement the above-mentioned real-time data processing method for multiple devices in the regional wind farm.

[0239] In some embodiments, the computer may further include a communication interface 103 and a bus 100. For example, Figure 3 As shown, the processor 101, memory 102, and communication interface 103 are connected through bus 100 and complete communication with each other.

[0240] The communication interface 103 is used to enable communication between the various modules, devices, units, and / or equipment in the embodiments of the present invention. The communication interface 103 can also enable data communication with other components such as external devices, image / data acquisition devices, databases, external storage, and image / data processing workstations.

[0241] Bus 100 includes hardware, software, or both, that couples components of a computer device together. Bus 100 includes, but is not limited to, at least one of the following: data bus, address bus, control bus, expansion bus, and local bus. For example, and not as a limitation, bus 100 may include an Accelerated Graphics Port (AGP) or other graphics bus, an Extended Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), a Hyper Transport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an InfiniBand interconnect, a Low Pin Count (LPC) bus, a memory bus, a Micro Channel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local Bus (VLB) bus, or other suitable buses, or a combination of two or more of these. Where appropriate, bus 100 may include one or more buses. Although specific buses are described and illustrated in the embodiments of the present invention, the present invention is contemplated by any suitable bus or interconnect.

[0242] The computer can execute the real-time data processing method for multiple devices in a regional wind farm based on the acquired real-time data processing system for multiple devices in a regional wind farm, thereby realizing real-time data processing for multiple devices in a regional wind farm.

[0243] In some further embodiments of the present invention, in conjunction with the above-described method for real-time data processing of multiple devices in a regional wind farm, the present invention provides the following technical solution: a storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the above-described method for real-time data processing of multiple devices in a regional wind farm.

[0244] Those skilled in the art will understand that the logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can mean any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0245] More specific examples of readable media (a non-exhaustive list) include: electrical connections (electronic devices) with one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0246] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0247] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0248] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this invention patent should be determined by the appended claims.

Claims

1. A method for real-time data processing of multiple devices in a regional wind farm, characterized in that, include: Monitoring and operation data of multiple devices in a regional wind farm are acquired through preset sensors. The monitoring and operation data is preprocessed to obtain processed data. The preset sensors include wind measurement tower sensors, noise sensors, meteorological sensors, and power generation operation sensors. The monitoring and operation data includes wind farm wind measurement tower data, wind farm noise data, wind farm meteorological data, and wind farm power generation operation data. The processed data is subjected to a first anomaly identification to obtain a first anomaly set; The processed data is subjected to a second anomaly identification to obtain a second set of anomalies; A comprehensive anomaly set is determined based on the first anomaly set and the second anomaly set, and data to be removed is determined based on the comprehensive anomaly set. The discarded data is then filled with conditional diffusion to output the final processed data. The step of performing a first anomaly identification on the processed data to obtain a first set of anomalies includes: The processed data is subjected to a logarithmic ratio transformation and mapped to Euclidean space to obtain a first data matrix; The first data matrix is ​​subjected to eigenvalue decomposition to obtain the covariance matrix and principal component sequence; Extract the eigenvalues ​​of the covariance matrix, and calculate the reference values ​​of different numbers of principal components based on the eigenvalues. ; Select the first few principal components such that their corresponding reference values ​​are greater than the preset reference values ​​to obtain the target principal components; Based on the target principal components, the original data is reconstructed and mapped back to the dimension of the processed data to obtain a mapped data matrix; The first set of anomalies is determined based on the mapping data matrix; The step of performing conditional diffusion imputation on the discarded data to output the final processed data includes: Identify missing values ​​in the removed data and divide the removed data into several subsets according to the location of the missing values; Calculate the density ratio of each of the said subsets. : ； In the formula, They represent the first Estimated probability density of each subset of data; Calculate the probability score based on the density ratio. : ； In the formula, The number of data points in the subset. For the first in the subset of data The density ratio of each data point; Based on the probability score Calculate data trust value : ； In the formula, Represents the first in the subset of data The distance between each data point and the previous data point These are the probability scores for the corresponding subsets of data. The average of the probability scores across all subsets; Based on the data trust value, the data in the subset are sorted in descending order, and the first few data points in the sorted subset are selected as the benchmark data. Obtaining Gaussian noise The reference data is obtained by using bicubic interpolation. Adjust it to the same magnitude as the Gaussian noise to obtain the adjusted data. ; Based on the Gaussian noise The reference data The adjusted data Determine the final data to be processed.

2. The method for real-time data processing of multiple devices in a regional wind farm according to claim 1, characterized in that, The step of determining the first anomaly set based on the mapping data matrix includes: The mapped data matrix is ​​processed using the minimum regularized covariance determinant to obtain the first robust mean estimate. First robust covariance estimation ; Based on the first robust mean estimation Compared with the first robust covariance estimate Calculate the judgment value for each of the processed data. : ； In the formula, For the first One data processing unit; Based on the judgment value Calculate the judgment threshold : ； In the formula, The median of all judgment values. The absolute deviation of the median of all judgment values; Data whose judgment value is greater than the judgment threshold is treated as first abnormal data and stored in the first abnormal set.

3. The method for real-time data processing of multiple devices in a regional wind farm according to claim 1, characterized in that, The step of performing a second anomaly identification on the processed data to obtain a second set of anomalies includes: The processed data is subjected to a logarithmic transformation and mapped to Euclidean space to obtain a second data matrix. ; The second data matrix is ​​processed using the minimum regularized covariance determinant to obtain the second robust mean estimate. Second robust covariance estimation ; For the second robust covariance estimation Perform singular value decomposition: ； In the formula, It is a left singular matrix containing orthogonal vectors. It is a diagonal matrix with singular values ​​on its diagonal. The eigenvector matrix; Calculate the target principal component matrix : ； In the formula, It is the identity matrix; Calculate the reference values ​​of the first few columns of the target principal component matrix, and determine the reference principal components based on the reference values ​​of the first few columns of the target principal component matrix and the preset reference values; Calculate the stability score of the reference principal component. : ； In the formula, To reference the total number of columns in the principal components, For reference principal components, the first The singular values ​​corresponding to the column, For reference principal components, the first Line number Column elements; The processed data corresponding to the column elements of the reference principal components whose stable scores are greater than the preset scores are regarded as the second abnormal data and stored in the second abnormal set.

4. The method for real-time data processing of multiple devices in a regional wind farm according to claim 1, characterized in that, The step of determining a comprehensive anomaly set based on the first anomaly set and the second anomaly set, and determining the data to be removed based on the comprehensive anomaly set, includes: The first anomaly set and the second anomaly set are combined to obtain a comprehensive anomaly set. Data in the comprehensive anomaly set is then removed from the processed data to obtain the removed data.

5. The method for real-time data processing of multiple devices in a regional wind farm according to claim 1, characterized in that, The Gaussian noise The reference data The adjusted data The steps to determine the final processed data include: The adjusted data is combined with the Gaussian noise to obtain combined data. : ； In the formula, It is a binary sequence. To adjust the weights, It is the XOR operator; The combined data is subjected to initial inverse diffusion to obtain initial data for the next step. : ； In the formula, For the first The inverse diffusion hyperparameter of the step, For the preset neural network, It is a random number. It is a fixed constant; For the initial next step data Conditional injection and precise reverse diffusion are performed to obtain accurate data for the next step. : ； In the formula, These are the initial parameters. The attenuation coefficient; The initial and precise reverse diffusion processes are iteratively repeated until the data from the final reverse diffusion step is output to obtain the target filling data. The target filling data is then used to fill in the missing values ​​according to their positions to obtain the final processed data.

6. A multi-device real-time data processing system for a regional wind farm, wherein the system employs the multi-device real-time data processing method for a regional wind farm as described in claim 1, characterized in that, The system includes: The preprocessing module is used to acquire monitoring and operation data of multiple devices in a regional wind farm through preset sensors, and to preprocess the monitoring and operation data to obtain processed data. The preset sensors include wind measurement tower sensors, noise sensors, meteorological sensors and power generation operation sensors. The monitoring and operation data includes wind farm wind measurement tower data, wind farm noise data, wind farm meteorological data and wind farm power generation operation data. The first anomaly identification module is used to perform a first anomaly identification on the processed data to obtain a first anomaly set; The second anomaly identification module is used to perform second anomaly identification on the processed data to obtain a second anomaly set; The elimination module is used to determine a comprehensive anomaly set based on the first anomaly set and the second anomaly set, and to determine the data to be eliminated based on the comprehensive anomaly set. The output module is used to perform conditional diffusion filling on the discarded data to output the final processed data.

7. A computer comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the real-time data processing method for multiple devices in a regional wind farm as described in any one of claims 1 to 5.

8. A storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, implements the real-time data processing method for multiple devices in a regional wind farm as described in any one of claims 1 to 5.

Images