Wind power digital twinning real-time detection method based on cross correlation and dynamic programming

Abstract

The invention discloses a wind power digital twinning real-time detection method based on cross correlation and dynamic programming, which comprises the following steps: obtaining twinning fan data generated by digital twinning and corresponding real fan data, performing data preprocessing on the two types of data to obtain a time sequence, and calculating cross correlation; finding out the maximum cross correlation position and the corresponding time delay; correcting the time sequence of the twin fan data by using time delay, and calculating a data error index for the time sequence of the real fan data and the corrected time sequence of the twin fan data so as to judge whether real-time abnormality exists or not; and if the real-time abnormity exists, calculating a dynamic time warping distance matrix and a matching path according to the time sequence of the real fan data and the time sequence of the corrected twin fan data. The method aims at achieving digital twinning state evaluation of the real-time scale and the accuracy scale, and the comprehensiveness and the accuracy in the operation state detection of the digital twinning system of the wind turbine group are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of digital twins of wind turbine group service quality, and in particular to a real-time detection method for wind power digital twins based on cross-correlation and dynamic programming. Background Art

[0002] In the past few years, digital twin technology has been applied to aerospace, manufacturing, healthcare, cities, energy and other fields. These applications and explorations have shown the great advantages of digital twin technology. The core idea of ​​digital twin is to make full use of physical data and mechanisms to build ultra-fidelity virtual copies, and monitor, control and optimize the physical world through interaction and iteration between the virtual world and the physical world. During the long-term operation of wind turbines, various faults may occur, affecting the efficiency of power production and causing the stability of the power grid to decrease. With the development of the wind power industry, the number and scale of wind turbines continue to expand, and how to effectively evaluate and improve the service quality of wind turbines has become a problem that needs to be solved. The detection and verification technology of the digital twin system of wind turbine service quality is crucial to the reliability and performance improvement of wind turbines. It can predict faults, optimize operation, diagnose problems, and provide intelligent management to improve the reliability and performance of wind turbines, reduce downtime and maintenance costs, and ultimately achieve more efficient wind energy utilization.

[0003] As an emerging technical means, digital twin technology can achieve the full life cycle management of wind turbine groups. However, during the long-term operation of wind turbines, the speed changes due to the changing operating conditions, making the signals used for the evaluation of digital twin systems more complex. The change in speed can also lead to problems such as frequency modulation and spectrum blurring of vibration signals, making the evaluation of digital twin systems more difficult. In order to ensure the accuracy of the detection and verification system, a high sampling rate is usually used when collecting signals. At the same time, the data processing capacity of the evaluation index algorithm of the digital twin system is limited. Due to the large amount of data, the detection and verification method requires a large amount of storage memory, which makes the evaluation process very slow. This phenomenon brings data pressure to the entire process. Therefore, it is of great significance to study a method that can improve the accuracy and real-time performance of the service quality assessment of wind turbine groups.

[0004] The research on the credibility assessment theory and methods of the digital twin system of the service quality of wind turbines can refer to the theoretical achievements of credibility assessment, which are mainly concentrated in the field of modeling and simulation. VV&A (verification, validation and confirmation) is an evaluation and analysis technology generated to ensure the credibility of simulation models, and is an important field that the modeling and simulation community has been exploring and researching. Since there is real-time data exchange between the digital twin system of the service quality of wind turbines and the physical system, the quality of the data will also affect the credibility of the twin model. The real-time data obtained by sensors is large in scale and multi-type, so the data quality of sensors needs to be evaluated from multiple dimensions. Therefore, it is necessary to propose evaluation indicators from different angles so that the digital twin system can fully and accurately reflect the operating status of wind turbines. Summary of the invention

[0005] Technical problem to be solved by the present invention: In view of the above-mentioned problems of the prior art, a real-time detection method for wind power digital twins based on cross-correlation and dynamic programming is provided. The present invention aims to realize the digital twin status evaluation on the real-time scale and the accuracy scale, and improve the comprehensiveness and accuracy of the operation status detection of the digital twin system of the wind turbine group.

[0006] In order to solve the above technical problems, the technical solution adopted by the present invention is: A wind power digital twin real-time detection method based on cross-correlation and dynamic programming, comprising: S1, obtaining twin wind turbine data generated by the digital twin and real wind turbine data collected by a wind turbine sensor of the corresponding wind turbine, wherein the twin wind turbine data and the real wind turbine data are temperature, current, speed, power or vibration data; S2, preprocessing the twin wind turbine data and the real wind turbine data to obtain time series, and then calculating the cross-correlation to find the location of the maximum cross-correlation and its corresponding time delay; S3, using the time delay to correct the time series of the twin wind turbine data, and calculating the data error index for the time series of the real wind turbine data and the time series of the corrected twin wind turbine data to determine whether there is a real-time anomaly. If there is a real-time anomaly, jump to step S4; otherwise, end and exit; S4, converting the time series of the real wind turbine data and the time series of the corrected twin wind turbine data into two-dimensional data, and calculating the dynamic time warping distance matrix and matching path between the twin wind turbine data and the real wind turbine data based on the two-dimensional data as the real-time detection result output of the wind power digital twin.

[0007] Optionally, the temperature refers to the generator temperature or the cabin temperature, the current refers to the generator current, the speed refers to the hub speed of the wind turbine, the power refers to the active power of the grid on the grid side of the inverter, and the vibration refers to the vibration of the wind turbine.

[0008] Optionally, in step S2, preprocessing the twin wind turbine data and the real wind turbine data to obtain a time series means respectively performing mean centering on the time series of the twin wind turbine data and the real wind turbine data according to the following formula to obtain the time series: , In the above formula, for The mean centering result of and are the i-th and j-th data in the time series respectively, is the number of data in the time series, and the obtained time series consists of mean-centered results.

[0009] Optionally, the function expression for calculating the cross-correlation in step S2 is: , In the above formula, Represents the time series of twin wind turbine data and the time series of real wind turbine data Based on the number of shifts The cross-correlation calculation results of The value range of is from the negative maximum delay to the positive maximum delay between the twin wind turbine data and the real wind turbine data. def means definition. is the time series of twin wind turbine data The mth data The complex conjugate of is the time series of real wind turbine data Middle data, is an integer variable; the finding of the maximum cross-correlation position and its corresponding time delay refers to finding the number of different shifts Corresponding The maximum value and the corresponding shift number , and the number of shifts as the time delay found.

[0010] Optionally, using time delay to correct the time series of the twin wind turbine data in step S3 means subtracting the time delay from the time of each data in the time series of the twin wind turbine data to eliminate the lateral error of the time series of the twin wind turbine data, so that when calculating the data error index for the time series of the twin wind turbine data, only the longitudinal error composed of the relative error of the time series is calculated.

[0011] Optionally, the data error index in step S3 includes part or all of mean square error MSE, root mean square error RMSE, mean absolute error MAR and mean absolute percentage error MAPE. If any data error index exceeds a preset threshold, the real-time performance is determined to be abnormal, otherwise the real-time performance is determined to be normal; the calculation function expressions of the mean square error MSE, root mean square error RMSE, mean absolute error MAR and mean absolute percentage error MAPE are: , , , , In the above formula, is the i-th parameter in the real wind turbine data, The i-th parameter in the corrected twin wind turbine data, is the number of parameters in the time series.

[0012] Optionally, in step S4, calculating the dynamic time warping distance matrix and the matching path between the twin wind turbine data and the real wind turbine data according to the two-dimensional data includes: S4.1, respectively calculating the Euclidean distance between corresponding points in the time series of the real wind turbine data and the time series of the corrected twin wind turbine data according to the following formula; , In the above formula, for and The Euclidean distance between is the i-th data point in the time series g of the real wind turbine data, is the j-th data point of C in the time series of the corrected twin wind turbine data; S4.2, generating a dynamic time warping distance matrix according to the Euclidean distances between the corresponding points, wherein each element in the dynamic time warping distance matrix represents the dynamic time warping distance between two time series at the corresponding time point; S4.3, using a dynamic path planning algorithm, find a shortest path from the lower left corner to the upper right corner in the dynamic time warping distance matrix as a matching path between the twin wind turbine data and the real wind turbine data.

[0013] Optionally, in step S4.3, when a shortest path from the lower left corner to the upper right corner is found in the dynamic time warping distance matrix by a dynamic path planning algorithm, the constraints adopted include boundary conditions, continuity and monotonicity, and the following function expression is used to solve the dynamic time warping distance matrix to find the shortest path from the lower left corner to the upper right corner:

[0014] In the above formula, represents the cumulative distance between the i-th data point on the time series g of the real wind turbine data and the j-th data point on the time series C of the corrected twin wind turbine data, for and The Euclidean distance between is the i-th data point in the time series g of the real wind turbine data, is the jth data point on the time series C of the corrected twin wind turbine data; min means taking the minimum value; is the cumulative distance between the i-1th data point on the time series g of the real wind turbine data and the j-1th data point C on the time series of the corrected twin wind turbine data, is the cumulative distance between the i-1th data point on the time series g of the real wind turbine data and the jth data point C on the time series of the corrected twin wind turbine data, is the cumulative distance between the i-th data point on the time series g of the real wind turbine data and the j-1-th data point on the time series C of the corrected twin wind turbine data; and when solving the dynamic time warping distance matrix to find the shortest path from the lower left corner to the upper right corner, starting from the lower left corner where i=0 and j=0, and ending at the upper right corner where i=g and j=C, the cumulative distance at the upper right corner is finally obtained. is the shortest path.

[0015] In addition, the present invention also provides a wind power digital twin real-time detection system based on cross-correlation and dynamic programming, comprising an interconnected microprocessor and a memory, wherein the microprocessor is programmed or configured to execute the wind power digital twin real-time detection method based on cross-correlation and dynamic programming.

[0016] In addition, the present invention also provides a computer-readable storage medium, which stores a computer program or instruction, and the computer program or instruction is programmed or configured to execute the wind power digital twin real-time detection method based on cross-correlation and dynamic programming through a processor.

[0017] In addition, the present invention also provides a computer program product, including a computer program or instructions, which are programmed or configured to execute the wind power digital twin real-time detection method based on cross-correlation and dynamic programming through a processor.

[0018] Compared with the prior art, the present invention mainly has the following advantages: 1. Due to the influence of factors such as calculation pressure, the digital twin value will produce a time delay compared with the actual measurement value at the same time node. The size of the time delay is a key value for evaluating the real-time performance of the digital twin system of the wind turbine group. The present invention evaluates the real-time index of the digital twin system by adopting the method of cross-correlation and dynamic programming. The present invention combines multiple dimensions such as correlation, error and time planning to improve the comprehensiveness of the digital twin system operation status detection of the wind turbine group, and can realize the digital twin status evaluation of the real-time scale and accuracy scale, and improve the comprehensiveness of the digital twin system operation status detection of the wind turbine group.

[0019] 2. Strong real-time performance. Due to the reference of the cross-correlation algorithm of the present invention, the real-time performance is strong. The present invention can timely evaluate the real-time performance of the system and timely discover the time delay between the actual data and the twin data.

[0020] 3. High accuracy. The present invention adopts multiple error indicators to comprehensively measure the accuracy of the digital twin system.

[0021] 4. Strong versatility. The dynamic time warping distance algorithm used in the present invention can process time series of different lengths and different sampling rates, and is suitable for various wind farm scenarios.

[0022] 5. The present invention has visualization support and can generate relevant icons to more intuitively display the evaluation results to facilitate decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 Schematic diagram of the basic flow of the method of the embodiment of the present invention.

[0024] Figure 2 Schematic diagram of cross-correlation of current calculated in an embodiment of the present invention.

[0025] Figure 3 Schematic diagram of current peak amplification in an embodiment of the present invention.

[0026] Figure 4 Schematic diagram of the cross-correlation of the temperature calculated in the embodiment of the present invention.

[0027] Figure 5 It is a schematic diagram of enlarging the temperature peak in an embodiment of the present invention.

[0028] Figure 6 Schematic diagram of dynamic time warping distance in an embodiment of the present invention.

[0029] Figure 7 Schematic diagram of the matching path in the embodiment of the present invention.

[0030] Figure 8 Schematic diagram of the dynamic time-warping distance of current in an embodiment of the present invention.

[0031] Fig. 9 Schematic diagram of the current matching path in an embodiment of the present invention.

[0032] Fig.10 Schematic diagram of the dynamic time-warping distance of temperature in an embodiment of the present invention.

[0033] Fig.11 Schematic diagram of the temperature matching path in an embodiment of the present invention. DETAILED DESCRIPTION

[0034] like Figure 1 As shown, the wind power digital twin real-time detection method based on cross-correlation and dynamic programming in this embodiment includes: S1, obtaining twin wind turbine data generated by the digital twin and real wind turbine data collected by a wind turbine sensor of the corresponding wind turbine, wherein the twin wind turbine data and the real wind turbine data are temperature, current, speed, power or vibration data; S2, preprocessing the twin wind turbine data and the real wind turbine data to obtain time series, and then calculating the cross-correlation to find the location of the maximum cross-correlation and its corresponding time delay; S3, using the time delay to correct the time series of the twin wind turbine data, and calculating the data error index for the time series of the real wind turbine data and the time series of the corrected twin wind turbine data to determine whether there is a real-time anomaly. If there is a real-time anomaly, jump to step S4; otherwise, end and exit; S4, converting the time series of the real wind turbine data and the time series of the corrected twin wind turbine data into two-dimensional data, and calculating the dynamic time warping distance matrix and matching path between the twin wind turbine data and the real wind turbine data based on the two-dimensional data as the real-time detection result output of the wind power digital twin.

[0035] As an optional implementation, in this embodiment, the temperature refers to the generator temperature or the cabin temperature, the current refers to the generator current, the speed refers to the hub speed of the wind turbine, the power refers to the active power of the grid on the grid side of the inverter, and the vibration refers to the vibration of the wind turbine.

[0036] See also Figure 1 It can be seen that in this embodiment, the acquired twin wind turbine data and real wind turbine data are mainly processed in three ways, and the real-time performance, accuracy, and comprehensive results of real-time and accuracy are calculated respectively.

[0037] In the calculation of real-time processing, the method of calculating the cross-correlation between the digital twin generation sequence and the sensor acquisition sequence is adopted in this embodiment, that is, the original signal is preprocessed, the two sets of data are respectively mean-centered, and then the cross-correlation is calculated. In step S2 of this embodiment, the twin wind turbine data and the real wind turbine data are preprocessed to obtain a time series, which means that the time series of the twin wind turbine data and the real wind turbine data are respectively mean-centered according to the following formula to obtain a time series: , In the above formula, for The mean centering result of and are the i-th and j-th data in the time series respectively, is the number of data in the time series, and the obtained time series consists of mean-centered results. Through the above steps, the mean of the sensor data and digital twin data is adjusted to zero, thereby eliminating the influence of the signal mean on subsequent calculations and ensuring the accuracy of data processing.

[0038] In this embodiment, the time delay corresponding to the maximum correlation is found by calculating the cross-correlation. Specifically, the function expression for calculating the cross-correlation in step S2 in this embodiment is: , In the above formula, Represents the time series of twin wind turbine data and the time series of real wind turbine data Based on the number of shifts The cross-correlation calculation results of The value range of is from the negative maximum delay to the positive maximum delay between the twin wind turbine data and the real wind turbine data. def means definition. is the time series of twin wind turbine data The mth data The complex conjugate of is the time series of real wind turbine data Middle data, is an integer variable; the finding of the maximum cross-correlation position and its corresponding time delay refers to finding the number of different shifts Corresponding The maximum value and the corresponding shift number , and the number of shifts As the time delay found. The result of the above calculation can be used to obtain a new time series, and the new time series can be plotted as follows Figure 2 to Figure 5 As shown, Figure 2is a schematic diagram of the cross-correlation of the calculated current, Figure 3 This is a schematic diagram of current peak amplification. Figure 4 is a schematic diagram of the cross-correlation of the calculated temperature, Figure 5 This is a schematic diagram of the enlarged temperature peak. Figure 3 and Figure 5 The highest point corresponding to the middle dotted line is the maximum correlation position, and its horizontal axis is the delay time. Taking the current and temperature indicators as an example, the calculation results are shown in Table 1.

[0039] Table 1 Delay time results obtained by cross-correlation of current and temperature time delay

[0040] In the calculation of real-time processing, the digital twin value will first be adjusted according to the lateral error obtained above. After adjustment, the longitudinal error between the digital twin system adjustment value and the actual measurement value is extracted, and the relative error of each time node is calculated. According to the relative error of each node, the accuracy of the digital twin system of the wind turbine group is evaluated. In step S3 of this embodiment, the time series of the twin wind turbine data is corrected using time delay, which means that the time of each data in the time series of the twin wind turbine data is subtracted from the time delay to eliminate the lateral error of the time series of the twin wind turbine data, so that when calculating the data error index for the time series of the twin wind turbine data, only the longitudinal error composed of the relative error of the time series is calculated.

[0041] As an optional implementation, the data error index in step S3 of this embodiment includes part or all of mean square error MSE (Mean Squared Error), root mean square error RMSE (Root Mean Squared Error), mean absolute error MAR (Mean Absolute Error) and mean absolute percentage error MAPE (Mean Absolute Percentage Error). If any data error index exceeds a preset threshold, the real-time performance is determined to be abnormal, otherwise the real-time performance is determined to be normal; the calculation function expressions of the mean square error MSE, root mean square error RMSE, mean absolute error MAR and mean absolute percentage error MAPE are: , , , , In the above formula, is the i-th parameter in the real wind turbine data, The i-th parameter in the corrected twin wind turbine data, is the number of parameters in the time series. This example compares the original data and the data error indicators after lateral correction. The calculation results of the current and temperature indicators are shown in Table 2.

[0042] Table 2 Comparison of error indicators

[0043] The method of using cross-correlation and data error indicators can have a real-time effect.

[0044] When the real-time anomaly is clearly known, a more accurate DTW (Dynamic Time Warping) distance and matching path method can be used to see through the curve whether the time between each two data points generated by the digital twin is uniform, and the difference between the data point precision can be seen through the distance parameter. DTW is an algorithm for measuring the similarity between two time series. It calculates the distance between two time series (time series 1 and time series 2, representing the time series of real wind turbine data and the time series of corrected twin wind turbine data) by finding the best alignment between them, that is, the dynamic time warping distance (DTW distance). The DTW curve shows the corresponding relationship between the two time series at different time points. In this embodiment, two random sequences are generated such as Figure 6 and Figure 7 As shown in , through the DTW algorithm, the corresponding points can be found by comparing the minimum distance. Figure 6 In the above figure, A A', B B', C C'. Figure 6 Three points A, B, and C on time series 1 were selected, and their corresponding points A', B', and C' were found on time series 2. These corresponding points are connected by red dotted lines to form a DTW curve. The shape of the DTW curve reflects the degree of similarity between the two time series at different time scales. The flatter the curve, the more consistent the changing trends of the two time series; the more curved the curve, the different changing trends of the two time series in certain time periods. The dynamic time warping distance matrix (dynamic time warping distance matrix) shows the process of calculating the dynamic time warping distance. Each element in the dynamic time warping distance matrix represents the cumulative distance between the two time series at the corresponding time point. Through the dynamic programming algorithm, a shortest path from the lower left corner to the upper right corner can be found in the matrix, such as Figure 7As shown in the figure, the path represents the best alignment between the two time series, and its length is the dynamic time warping distance. The smaller the dynamic time warping distance, the more similar the two time series are; conversely, the larger the dynamic time warping distance, the greater the difference between the two time series. In practical applications, the DTW algorithm can be used to compare digital twin data and to compare time series data of different lengths, speeds, or forms. By calculating the dynamic time warping distance, it is possible to determine whether two time series belong to the same category or to evaluate the degree of similarity between them.

[0045] In this embodiment, calculating the dynamic time warping distance matrix and matching path between the twin wind turbine data and the real wind turbine data according to the two-dimensional data in step S4 includes: S4.1, respectively calculating the Euclidean distance between corresponding points in the time series of the real wind turbine data and the time series of the corrected twin wind turbine data according to the following formula; , In the above formula, for and The Euclidean distance between is the i-th data point in the time series g of the real wind turbine data, is the j-th data point of C in the time series of the corrected twin wind turbine data; S4.2, generating a dynamic time warping distance matrix according to the Euclidean distances between the corresponding points, wherein each element in the dynamic time warping distance matrix represents the dynamic time warping distance between two time series at the corresponding time point; S4.3, using a dynamic path planning algorithm, find a shortest path from the lower left corner to the upper right corner in the dynamic time warping distance matrix as a matching path between the twin wind turbine data and the real wind turbine data.

[0046] In this embodiment, when the dynamic path planning algorithm is used in step S4.3 to find the shortest path from the lower left corner to the upper right corner in the dynamic time warping distance matrix, the constraints adopted include boundary conditions, continuity and monotonicity, and the following function expression is used to solve the dynamic time warping distance matrix to find the shortest path from the lower left corner to the upper right corner:

[0047] In the above formula, represents the cumulative distance between the i-th data point on the time series g of the real wind turbine data and the j-th data point on the time series C of the corrected twin wind turbine data, for and The Euclidean distance between is the i-th data point in the time series g of the real wind turbine data, is the jth data point on the time series C of the corrected twin wind turbine data; min means taking the minimum value; is the cumulative distance between the i-1th data point on the time series g of the real wind turbine data and the j-1th data point C on the time series of the corrected twin wind turbine data, is the cumulative distance between the i-1th data point on the time series g of the real wind turbine data and the jth data point C on the time series of the corrected twin wind turbine data, is the cumulative distance between the i-th data point on the time series g of the real wind turbine data and the j-1-th data point on the time series C of the corrected twin wind turbine data; and when solving the dynamic time warping distance matrix to find the shortest path from the lower left corner to the upper right corner, starting from the lower left corner where i=0 and j=0, and ending at the upper right corner where i=g and j=C, the cumulative distance at the upper right corner is finally obtained. According to the above formula, the cumulative distance between the i-th data point on the time series g of the real wind turbine data and the j-th data point on the time series C of the corrected twin wind turbine data is is the current grid distance , that is and The sum of the Euclidean distance (similarity) between the points and the cumulative distance of the smallest neighboring element that can reach the point.

[0048] In this embodiment, the digital twin of temperature and current and the actual signal are used to perform DTW analysis, and the results obtained are as follows: Figure 8 to Figure 11 As shown, Figure 8 is a schematic diagram of the dynamic time-warping distance of the current in this embodiment, Fig. 9 is a schematic diagram of the current matching path in this embodiment, Fig.10 Schematic diagram of the dynamic time-warping distance of temperature in this embodiment, Fig.11 The figure is a schematic diagram of the temperature matching path in this embodiment. However, due to the different data scale ranges, the dynamic time warping distance obtained does not have a reference value for comparison. The normalization is shown in Table 3.

[0049] Table 3 Comparison of dynamic time warping distance normalization before and after

[0050] The DTW curve reflects the trend consistency. The DTW curve shows the optimal alignment path of two time series after dynamic warping. If the DTW curve is close to a straight line, that is, the alignment path is almost along the main diagonal of the matrix, then it indicates that the two time series are highly consistent in trend. The two time series show similar change patterns on their respective time axes, and their shapes and trends are basically consistent. The dynamic time warping distance reflects numerical consistency: the dynamic time warping distance is the minimum cumulative distance between two time series obtained by dynamic warping. The smaller the dynamic time warping distance, the closer the two time series are in value, and the smaller the difference in their values ​​at each corresponding time point. These two indicators can be used to evaluate the working quality of the digital twin system as a whole.

[0051] To summarize, the method of this embodiment includes obtaining the twin wind turbine data generated by the digital twin and the real wind turbine data collected by the wind turbine sensor of the corresponding wind turbine, performing data preprocessing on the twin wind turbine data and the real wind turbine data to obtain a time series, and then calculating the cross-correlation to find the position of the largest cross-correlation and its corresponding time delay; using the time delay to correct the time series of the twin wind turbine data, calculating the data error index for the time series of the real wind data and the time series of the corrected twin wind turbine data to determine whether there is a real-time anomaly; if there is a real-time anomaly, the time series of the real wind data and the time series of the corrected twin wind turbine data are corrected. The sequence is converted into two-dimensional data, and the dynamic time warping distance matrix and matching path between the twin wind turbine data and the real wind turbine data are calculated based on the two-dimensional data to output the real-time detection result of the wind power digital twin. The method of this embodiment introduces the dynamic time warping distance matrix of cross-correlation, data error index and dynamic time warping distance on the basis of the original measurement signal and the signal of the digital twin, which improves the comprehensiveness and accuracy of the digital twin system operation status detection of the wind turbine group, and can realize the digital twin state evaluation of the real-time scale and the accuracy scale, and improve the comprehensiveness and accuracy of the digital twin system operation status detection of the wind turbine group. The method of this embodiment comprehensively proposes the evaluation indicators of the real-time and accuracy of the digital twin embodiment method of the wind turbine group in the operation status detection by introducing various technical means such as correlation coefficient, error calculation index and DTW dynamic time warping. The method of this embodiment method can not only evaluate the operation status of the wind turbine group in real time, but also comprehensively measure the accuracy of the digital twin embodiment method through various error indicators, thereby realizing the intelligent management of the wind turbine group. By generating relevant icons, the method of this embodiment also provides intuitive visualization support, which is helpful for decision analysis. In addition, the method of this embodiment is highly versatile, and the DTW algorithm can process time series of different lengths and different sampling rates, and is applicable to various wind farm scenarios. In summary, the method of this embodiment has significant advantages and broad application prospects in improving the accuracy and real-time performance of wind turbine service quality assessment.

[0052] In addition, this embodiment also provides a wind power digital twin real-time detection system based on cross-correlation and dynamic programming, including an interconnected microprocessor and a memory, and the microprocessor is programmed or configured to execute the wind power digital twin real-time detection method based on cross-correlation and dynamic programming.

[0053] In addition, this embodiment also provides a computer-readable storage medium, which stores a computer program or instructions, and the computer program or instructions are programmed or configured to execute the wind power digital twin real-time detection method based on cross-correlation and dynamic programming through a processor.

[0054] In addition, this embodiment also provides a computer program product, including a computer program or instructions, which are programmed or configured to execute the wind power digital twin real-time detection method based on cross-correlation and dynamic programming through a processor.

[0055] Those skilled in the art should understand that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application can take the form of a computer program product implemented on one or more computer-readable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the process Figure 1 A process or multiple processes and / or boxes Figure 1 These computer program instructions can also be stored in a computer-readable memory that can guide a computer or other programmable data processing device to work in a specific way, so that the instructions stored in the computer-readable memory produce a product including an instruction device, which implements the functions specified in the process. Figure 1 A process or multiple processes and / or boxes Figure 1These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide for implementing the process in the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0056] The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. A wind power digital twin real-time detection method based on cross-correlation and dynamic programming, characterized in that: include: S1, obtaining twin wind turbine data generated by the digital twin and real wind turbine data collected by a wind turbine sensor of the corresponding wind turbine, wherein the twin wind turbine data and the real wind turbine data are temperature, current, speed, power or vibration data; S2, preprocessing the twin wind turbine data and the real wind turbine data to obtain time series, and then calculating the cross-correlation to find the location of the maximum cross-correlation and its corresponding time delay; S3, using the time delay to correct the time series of the twin wind turbine data, and calculating the data error index for the time series of the real wind turbine data and the time series of the corrected twin wind turbine data to determine whether there is a real-time anomaly. If there is a real-time anomaly, jump to step S4; otherwise, end and exit; S4, converting the time series of the real wind turbine data and the time series of the corrected twin wind turbine data into two-dimensional data, and calculating the dynamic time warping distance matrix and matching path between the twin wind turbine data and the real wind turbine data based on the two-dimensional data as the real-time detection result output of the wind power digital twin.

2. The wind power digital twin real-time detection method based on cross-correlation and dynamic programming according to claim 1 is characterized in that: The temperature refers to the generator temperature or the cabin temperature, the current refers to the generator current, the speed refers to the hub speed of the fan, the power refers to the active power of the inverter grid side grid, and the vibration refers to the vibration of the fan.

3. The wind power digital twin real-time detection method based on cross-correlation and dynamic programming according to claim 1 is characterized in that: In step S2, preprocessing the twin wind turbine data and the real wind turbine data to obtain a time series means respectively performing mean centering on the time series of the twin wind turbine data and the real wind turbine data according to the following formula to obtain a time series: ， In the above formula, for The mean centering result of and are the i-th and j-th data in the time series respectively, is the number of data in the time series, and the obtained time series is composed of mean-centered results; the function expression for calculating the cross-correlation in step S2 is: ， In the above formula, Represents the time series of twin wind turbine data and the time series of real wind turbine data Based on the number of shifts The cross-correlation calculation results of The value range of is from the negative maximum delay to the positive maximum delay between the twin wind turbine data and the real wind turbine data. def means definition. is the time series of twin wind turbine data The mth data The complex conjugate of is the time series of real wind turbine data Middle data, is an integer variable; the finding of the maximum cross-correlation position and its corresponding time delay refers to finding the number of different shifts Corresponding The maximum value and the corresponding shift number , and the number of shifts as the time delay found.

4. The wind power digital twin real-time detection method based on cross-correlation and dynamic programming according to claim 1 is characterized in that: Using time delay to correct the time series of the twin wind turbine data in step S3 means subtracting the time delay from the time of each data in the time series of the twin wind turbine data to eliminate the lateral error of the time series of the twin wind turbine data, so that when calculating the data error index for the time series of the twin wind turbine data, only the longitudinal error composed of the relative error of the time series is calculated.

5. The wind power digital twin real-time detection method based on cross-correlation and dynamic programming according to claim 1 is characterized in that: The data error indicators in step S3 include part or all of the mean square error MSE, the root mean square error RMSE, the mean absolute error MAR and the mean absolute percentage error MAPE. If any one of the data error indicators exceeds the preset threshold, the real-time performance is judged to be abnormal, otherwise the real-time performance is judged to be normal; the calculation function expressions of the mean square error MSE, the root mean square error RMSE, the mean absolute error MAR and the mean absolute percentage error MAPE are: ， ， ， ， In the above formula, is the i-th parameter in the real wind turbine data, The i-th parameter in the corrected twin wind turbine data, is the number of parameters in the time series.

6. The wind power digital twin real-time detection method based on cross-correlation and dynamic programming according to claim 1 is characterized in that: In step S4, the dynamic time warping distance matrix and matching path between the twin wind turbine data and the real wind turbine data are calculated based on the two-dimensional data, including: S4.1, respectively calculating the Euclidean distance between corresponding points in the time series of the real wind turbine data and the time series of the corrected twin wind turbine data according to the following formula; ， In the above formula, for and The Euclidean distance between is the i-th data point in the time series g of the real wind turbine data, is the jth data point in the time series C of the corrected twin wind turbine data; S4.2, generating a dynamic time warping distance matrix according to the Euclidean distances between the corresponding points, wherein each element in the dynamic time warping distance matrix represents the dynamic time warping distance between two time series at the corresponding time point; S4.3, using a dynamic path planning algorithm, find a shortest path from the lower left corner to the upper right corner in the dynamic time warping distance matrix as a matching path between the twin wind turbine data and the real wind turbine data.

7. The wind power digital twin real-time detection method based on cross-correlation and dynamic programming according to claim 6 is characterized in that: In step S4.3, when a shortest path from the lower left corner to the upper right corner is found in the dynamic time warping distance matrix by the dynamic path planning algorithm, the constraints adopted include boundary conditions, continuity and monotonicity, and the following function expression is used to solve the dynamic time warping distance matrix to find the shortest path from the lower left corner to the upper right corner: In the above formula, represents the cumulative distance between the i-th data point on the time series g of the real wind turbine data and the j-th data point on the time series C of the corrected twin wind turbine data, for and The Euclidean distance between is the i-th data point in the time series g of the real wind turbine data, is the jth data point in the time series C of the corrected twin wind turbine data; min means taking the minimum value; is the cumulative distance between the i-1th data point on the time series g of the real wind turbine data and the j-1th data point C on the time series of the corrected twin wind turbine data, is the cumulative distance between the i-1th data point on the time series g of the real wind turbine data and the jth data point C on the time series of the corrected twin wind turbine data, is the cumulative distance between the i-th data point on the time series g of the real wind turbine data and the j-1-th data point on the time series C of the corrected twin wind turbine data; and when solving the dynamic time warping distance matrix to find the shortest path from the lower left corner to the upper right corner, starting from the lower left corner where i=0 and j=0, and ending at the upper right corner where i=g and j=C, the cumulative distance at the upper right corner is finally obtained. is the shortest path.

8. A wind power digital twin real-time detection system based on cross-correlation and dynamic programming, comprising an interconnected microprocessor and a memory, characterized in that: The microprocessor is programmed or configured to execute the wind power digital twin real-time detection method based on cross-correlation and dynamic programming as described in any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program or instruction stored therein, characterized in that: The computer program or instruction is programmed or configured to execute the wind power digital twin real-time detection method based on cross-correlation and dynamic programming as described in any one of claims 1 to 7 through a processor.

10. A computer program product comprising a computer program or instructions, characterized in that The computer program or instruction is programmed or configured to execute the wind power digital twin real-time detection method based on cross-correlation and dynamic programming as described in any one of claims 1 to 7 through a processor.

Images