A wind turbine operation linkage analysis method based on Vine Copula model

Abstract

The present invention relates to a method for analyzing the operational linkage of wind turbines based on a Vine Copula model. The SCADA data of several wind turbines are obtained respectively, the SCADA data are cleaned and integrated, and an operational linkage analysis data set is generated. The Vine Copula model is used to generate three dependency structures of R-Vine, C-Vine, and D-Vine. The AIC value, BIC value, Loglik value, and Vuong test value of the dependency structure are calculated, and the optimal dependency structure is obtained by comparison. The rank correlation coefficient, upper tail correlation coefficient, and lower tail correlation coefficient between each two wind turbines are calculated. The operational linkage of several wind turbines is determined based on the optimal dependency structure and the rank correlation coefficient, upper tail correlation coefficient, and lower tail correlation coefficient. The method of the present invention introduces the Vine Copula model from the wind farm level to perform operational linkage analysis on multiple wind turbines, and macroscopically analyzes the output of each unit, so as to facilitate the monitoring and diagnosis of a single faulty wind turbine during the operation and maintenance of the wind turbine.

Description

Technical Field

[0001] The present invention belongs to the technical field of wind power operation and maintenance, and particularly relates to a wind turbine operation linkage analysis method based on a Vine Copula model. Background Art

[0002] With the introduction of the "dual carbon" development goals, clean and renewable energy sources, represented by wind and solar energy, are experiencing new development opportunities. Wind power generation technology continues to develop and improve, the operational reliability of wind turbines continues to improve, and annual installed capacity has steadily increased. By the end of 2021, China's cumulative installed wind power capacity had reached 329 GW, ranking first globally. With the large-scale grid connection of wind farms and the uncontrollable nature of wind power generation, the operation and maintenance of large wind farms has become increasingly difficult, reducing the stability of the power grid system. Furthermore, as wind farms age, the components of individual turbines inevitably wear out and their resistance decreases, leading to reduced power generation efficiency and potentially even failures and damage to the turbines.

[0003] Because multiple wind turbines in the same wind farm are geographically adjacent and share highly similar external environments, their operating states inevitably share similarities: the output of each turbine often increases or decreases simultaneously. Analyzing the operational linkage of wind turbines allows for macroscopic determination of the operating status of each turbine by comparing the power generation efficiency of each turbine, enabling tasks such as fleet operational status assessment and abnormal turbine screening. From the perspective of improving the operational efficiency, safety, and stability of grid-connected systems, correlation analysis of wind turbines within the same wind farm is of great theoretical and practical value when optimizing power system operations. Therefore, a method for analyzing the operational linkage of wind turbines is needed. Summary of the Invention

[0004] Based on the above-mentioned shortcomings and deficiencies in the prior art, one of the objects of the present invention is to at least solve one or more of the above-mentioned problems in the prior art. In other words, one of the objects of the present invention is to provide a wind turbine operation linkage analysis method based on the Vine Copula model that meets one or more of the above-mentioned needs.

[0005] In order to achieve the above-mentioned object of the invention, the present invention adopts the following technical solutions:

[0006] A method for analyzing the operational linkage of wind turbines based on a Vine Copula model comprises the following steps:

[0007] S1. Obtain SCADA data of several wind turbines respectively;

[0008] S2. Clean and integrate SCADA data to generate operational linkage analysis data sets;

[0009] S3. Use the Vine Copula model to generate the dependency structure of several wind turbines based on the operational linkage analysis data set. The Vine Copula model includes three types: R-Vine, C-Vine, and D-Vine.

[0010] S4. Calculate the AIC value, BIC value, Loglik value, and Vuong test value of the dependency structure and compare them to obtain the optimal dependency structure;

[0011] S5. Calculate the rank correlation coefficient, upper tail correlation coefficient, and lower tail correlation coefficient between every two of the wind turbines according to the operation linkage analysis data set;

[0012] S6. Determine the operational linkage of several wind turbines based on the optimal dependency structure and the rank correlation coefficient, the upper tail correlation coefficient, and the lower tail correlation coefficient.

[0013] As a preferred solution, step S3 specifically includes:

[0014] S31, calculating the correlation between every two wind turbines based on the operation linkage analysis data set and the Copula function set, and determining the optimal Copula function for every two wind turbines;

[0015] S32, calculating the rank correlation coefficient of every two wind turbines based on the operation linkage analysis data set and the optimal Copula function, generating a correlation coefficient matrix, and generating a C-Vine dependency structure based on the correlation coefficient matrix;

[0016] S33, generating a D-Vine dependency structure based on the operation linkage analysis data set, the optimal Copula function, and the geographical locations of several wind turbines;

[0017] S34. Generate an R-Vine dependency structure based on the running linkage analysis data set, the optimal Copula function and the maximum spanning tree principle.

[0018] As a preferred solution, step S4 specifically includes:

[0019] S41. Calculate the AIC value, BIC value, Loglik value, and Vuong test value for each dependency structure;

[0020] S42, calculating the goodness-of-fit value of each dependency structure according to the AIC value, BIC value, and Loglik value of the dependency structure;

[0021] S43. Compare the goodness-of-fit value and Vuong test value of each dependency structure to obtain the optimal dependency structure.

[0022] As a preferred solution, step S5 specifically includes:

[0023] S51. Use t-Copula to calculate the rank correlation coefficient between the wind turbine with the best power generation capacity and other wind turbines;

[0024] S52. Use BB1 ​​Copula to calculate the upper tail correlation coefficient between the wind turbine with the best power generation capacity and other wind turbines;

[0025] S53. Use BB7 Copula to calculate the lower tail correlation coefficient between the wind turbine with the best power generation capacity and other wind turbines.

[0026] As a preferred solution, step S6 specifically includes:

[0027] S61. Obtaining overall operational linkage among a plurality of wind turbines according to an optimal interdependence structure;

[0028] S62. Obtaining the local operation linkage between the wind turbine with the best power generation capacity and other wind turbines based on the rank correlation coefficient, the upper tail correlation coefficient, and the lower tail correlation coefficient;

[0029] S63. Determine the operational linkage of several wind turbines by combining the overall operational linkage and the local operational linkage.

[0030] As a preferred solution, step S2 specifically includes:

[0031] S21. Delete data in the SCADA data where the wind speed is less than the cut-in wind speed, the wind speed is greater than the cut-out wind speed, the wind speed is greater than the cut-in wind speed and the active power is equal to 0, or the wind speed power limit setting value is less than the rated power;

[0032] S22, using the bin method to delete the outlier scattered data in the SCADA data after step S21;

[0033] S23, deleting the SCADA data that is out of sync with the time in the SCADA data after step S22;

[0034] S24: Time axis alignment and normalization are performed on the SCADA data after step S23 to generate an operation linkage analysis data set.

[0035] As a preferred solution, step S6 further includes step S7, determining abnormal wind turbines among the plurality of wind turbines according to their operation linkage.

[0036] As a further preferred solution, step S7 specifically includes:

[0037] S71. Calculate the deviation between each wind turbine generator set and the predicted value of the operational linkage;

[0038] S72. Determine that the wind turbine generator set whose deviation is greater than a preset threshold is an abnormal turbine generator set.

[0039] Compared with the prior art, the present invention has the following beneficial effects:

[0040] The method of the present invention introduces the Vine Copula model from the wind farm level to analyze the operational linkage of multiple wind turbines, and macroscopically analyzes the output of each unit, which can achieve preliminary screening of inefficient units and capture wind turbines with weaker power generation capacity, thereby facilitating the monitoring and diagnosis of single faulty wind turbines during wind turbine operation and maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 This is a flow chart of a method for analyzing wind turbine operation linkage based on a Vine Copula model according to an embodiment of the present invention;

[0042] Figure 2 This is a wind speed-active power scatter plot of a wind turbine operation linkage analysis method based on a Vine Copula model according to an embodiment of the present invention;

[0043] Figure 3 1 is a schematic diagram of a dependent structure of a method for analyzing wind turbine operation linkage based on a Vine Copula model according to an embodiment of the present invention;

[0044] Figure 4 The present invention provides an active power correlation coefficient heat map matrix of a wind turbine operation linkage analysis method based on a Vine Copula model. DETAILED DESCRIPTION

[0045] To more clearly illustrate the embodiments of the present invention, specific embodiments of the present invention will be described below with reference to the accompanying drawings. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings and other embodiments can be obtained based on these drawings without inventive efforts.

[0046] Example: This example provides a method for analyzing the operation linkage of wind turbines based on the Vine Copula model. The operation linkage analysis is performed on six wind turbines No. 1 to No. 6. The flow chart is as follows: Figure 1 Here is an example of the following steps:

[0047] When performing operation linkage analysis on the six wind turbines No. 1 to No. 6, step S1 is first executed to read the SCADA data of the wind turbines No. 1 to No. 6 respectively according to the geographical locations.

[0048] S2. Generate a wind speed-active power scatter plot using the SCADA data of wind turbines 1-6. Clean and integrate the SCADA data using the wind speed-active power scatter plot to generate an operation linkage analysis data set.

[0049] Specifically, step S2 can be implemented as follows:

[0050] S21. Generate a wind speed-active power scatter plot based on the SCADA data read in step S1, perform a cleanup, and select and delete the points in the scatter plot that have the following characteristics:

[0051] The wind speed is less than the cut-in wind speed; the wind speed is greater than the cut-out wind speed;

[0052] The wind speed is greater than the cut-in wind speed and the active power is equal to 0;

[0053] The wind speed power limit setting value is less than the rated power.

[0054] The wind speed-active power scatter plot generated by the original SCADA data read in step S1 is as follows: Figure 2 As shown in Figure (a), the scatter plot after cleaning in step S21 is as follows Figure 2 As shown in Figure (b).

[0055] S22, perform a second cleaning, use the bin method to delete the outlier scattered data in the scatter plot generated by the remaining SCADA data after step S21, specifically, the normal operation wind speed range of the unit [3m / s, 20m / s] is divided into 68 sub-ranges with a step size of 0.25 m / s, retain the data from the 10% quantile to the 90% quantile in each sub-range, and delete the unretained data. The cleaning process of step S22 is as follows Figure 2 As shown in the middle (C) figure, the scattered points are divided into retained scattered points and edge scattered points, and then the edge scattered points are deleted.

[0056] S23. In order to ensure that all wind turbines have normal data for analysis at each detection moment of the operation linkage analysis, a third cleaning is performed. The SCADA data that is not synchronized with the time is found in the remaining SCADA data after the cleaning in step S22. If not all wind turbines have SCADA data at a certain moment, then all SCADA data at this moment are SCADA data that is not synchronized with the time. The scattered points remaining after deleting the SCADA data that are not synchronized with the time are as follows: Figure 2 As shown in Figure (d), the remaining SCADA data have data of all wind turbines at the corresponding time for analysis.

[0057] S24: Time axis alignment and normalization are performed on the SCADA data remaining after the third cleaning in step S23 to generate an operation linkage analysis data set.

[0058] After data cleaning to generate the operational linkage analysis dataset, step S3 is performed to use the Vine Copula model to model the monthly data for wind turbines 1-6 based on the operational linkage analysis dataset, generating a dependency structure between each wind turbine in each month. The Vine Copula model includes three types: R-Vine, C-Vine, and D-Vine. Therefore, for each month's data, three dependency structures are generated: R-Vine, C-Vine, and D-Vine.

[0059] Specifically, step S3 includes the following steps:

[0060] S31. Select Gaussian, t, Clayton, Gumbel, Frank, Joe, BB1, BB6, BB7, BB8 and TawnCopula functions as alternative Copula functions.

[0061] Based on the active power, wind speed, and ambient temperature in the operational linkage analysis data set as input, the correlation between each two wind turbines in wind turbines 1-6 is calculated using a set of alternative Copula functions. The optimal Copula function between each two wind turbines is then determined based on the fitting accuracy.

[0062] S32, taking the active power, wind speed and ambient temperature in the operation linkage analysis data set as input, using the optimal Copula function obtained in S31 to calculate the rank correlation coefficient of every two wind turbines, generating a correlation coefficient matrix, and generating a C-Vine dependency structure based on the correlation coefficient matrix;

[0063] S33, based on the active power, wind speed and ambient temperature in the operation linkage analysis data set as input, using the optimal Copula function obtained in S31 and the geographical locations of wind turbines 1-6 to generate a D-Vine dependency structure;

[0064] S34. Based on the active power, wind speed and ambient temperature in the operation linkage analysis data set as input, the optimal Copula function is used to generate the R-Vine dependency structure according to the maximum spanning tree principle (MST-prim).

[0065] After obtaining the three dependency structures of R-Vine, C-Vine, and D-Vine among wind turbines for each month, step S4 is performed to calculate the AIC value, BIC value, Loglik value, and Vuong test value of the dependency structure. The one with the best fitting performance is selected from the three dependency structures of R-Vine, C-Vine, and D-Vine for each month as the optimal dependency structure for that month.

[0066] Specifically, step S4 is implemented as follows:

[0067] S41. Calculate the AIC value, BIC value, Loglik value, and Vuong test value for each dependency structure;

[0068] S42. Calculate the goodness-of-fit value of the dependency structure based on the AIC value, BIC value, and Loglik value of each dependency structure. The AIC value and BIC value are negatively correlated with the goodness-of-fit value, and the Loglik value is positively correlated with the goodness-of-fit value. The larger the goodness-of-fit value, the better the fitting performance of the dependency structure.

[0069] S43. Comprehensively compare the goodness-of-fit value and Vuong test value of each dependency structure to obtain the optimal dependency structure.

[0070] After obtaining the dependency structure, the correlations of the features between the wind turbines in the dependency structure are compared. S5. Calculate the rank correlation coefficient, upper tail correlation coefficient, and lower tail correlation coefficient between each two wind turbines No. 1-6 based on the operation linkage analysis data set.

[0071] Specifically, step S5 is divided into the following steps to respectively calculate the rank correlation coefficient, the upper tail correlation coefficient and the lower tail correlation coefficient.

[0072] S51. Use t-Copula to calculate the rank correlation coefficient between the wind turbine with the best power generation capacity and other wind turbines;

[0073] S52. Use BB1 ​​Copula to calculate the upper tail correlation coefficient between the wind turbine with the best power generation capacity and other wind turbines;

[0074] S53. Use BB7 Copula to calculate the lower tail correlation coefficient between the wind turbine with the best power generation capacity and other wind turbines.

[0075] Then, the correlation between the dependency structure and the characteristics of the wind turbines is integrated to determine the operational linkage of the wind turbines, and step S6 is executed to determine the operational linkage of several wind turbines based on the optimal dependency structure and the rank correlation coefficient, the upper tail correlation coefficient, and the lower tail correlation coefficient.

[0076] Specifically, step S6 includes:

[0077] S61. Through the analysis of the optimal interdependence structure among the three Vine structures of R-Vine, C-Vine and D-Vine, the overall operation linkage between wind turbines 1-6 is obtained. Figure 3As shown in the figure, the R-Vine structure shows that unit 3 has the highest correlation with the other five units, while units 4 and 5 are located at the edges of the R-Vine structure. The C-Vine structure shows that unit 3 is the root node and has weaker correlation with units 5 and 6 at the tail. The D-Vine structure shows that the six wind turbines in its dependency structure are arranged according to geographical location. The three Vine structures show different relationships between the six wind turbines.

[0078] S62. Based on the analysis of the rank correlation coefficient, the upper tail correlation coefficient, and the lower tail correlation coefficient, we can further obtain the local operation linkage between the wind turbine with the best power generation capacity and the other five wind turbines. Figure 4 The active power correlation coefficient heat map matrix is ​​shown in Figure 1, where (a) is the rank correlation coefficient, (b) is the upper tail correlation coefficient, and (c) is the lower tail correlation coefficient. The rank, upper tail, and lower tail correlation coefficients of units 1, 2, and 3 are all high, while the correlation coefficients between the other units are low, reflecting the good operational linkage between units 1, 2, and 3.

[0079] S63. Determine the operational linkage of several wind turbines by combining the overall operational linkage and the local operational linkage.

[0080] This paper introduces the Copula correlation analysis method into the field of wind turbine operational linkage analysis. Based on the accumulated SCADA data from wind farm operations, the original SCADA data of multiple wind turbines within the same wind farm is cleaned using a synchronous cleaning strategy. An analysis dataset is constructed, and a dependency structure model with different characteristics between multiple wind turbines is established. The optimal Vine Copula model is determined through goodness-of-fit comparison. Furthermore, the operational linkage of the units is determined by analyzing the three Vine structures (R-Vine, C-Vine, and D-Vine) and combining the rank correlation coefficient, upper tail correlation coefficient, and lower tail correlation coefficient to analyze the correlation of active power between the six units. This results in a differential output of units operating in similar external environments.

[0081] Furthermore, in order to determine the abnormal unit based on the operation linkage, step S6 is followed by step S7, determining whether there is an abnormal unit among wind turbines No. 1-6 based on the operation linkage, and determining the abnormal unit.

[0082] Specifically, step S7 can be implemented by the following means to determine the abnormal unit.

[0083] S71. Calculate the deviation between each wind turbine generator set and the predicted value of the operational linkage;

[0084] S72. Determine that the wind turbine generator set whose deviation is greater than a preset threshold is an abnormal turbine generator set.

[0085] It should be noted that the above embodiments are only detailed descriptions of the preferred embodiments and principles of the present invention. For ordinary technicians in this field, there will be changes in the specific implementation methods based on the ideas provided by the present invention, and these changes should also be regarded as the scope of protection of the present invention.

Claims

1. A wind turbine operation linkage analysis method based on the Vine Copula model, characterized in that: The method comprises the following steps: S1. Obtain SCADA data of several wind turbines respectively; S2. Clean and integrate the SCADA data to generate an operation linkage analysis data set; the step S2 specifically includes: S21. Delete the data in the SCADA data where the wind speed is less than the cut-in wind speed, the wind speed is greater than the cut-out wind speed, the wind speed is greater than the cut-in wind speed and the active power is equal to 0, and the wind speed power limit setting value is less than the rated power; S22. Use the bin method to delete the outlier scattered data in the SCADA data after the step S21; S23. Delete the time-unsynchronized SCADA data in the SCADA data after the step S22; S24. Time axis alignment and normalization are performed on the SCADA data after the step S23 to generate an operation linkage analysis data set; S3. Using a Vine Copula model to model the operation linkage analysis data set to generate a dependency structure of the plurality of wind turbines, wherein the Vine Copula model includes three types: R-Vine, C-Vine, and D-Vine; S4, calculating the AIC value, BIC value, Loglik value and Vuong test value of the dependency structure, and comparing to obtain the optimal dependency structure; the step S4 specifically includes: S41, calculating the AIC value, BIC value, Loglik value and Vuong test value of each dependency structure; S42, calculating the goodness of fit value of the dependency structure according to the AIC value, BIC value and Loglik value of each dependency structure; S43, comparing the goodness of fit value and the Vuong test value of each dependency structure to obtain the optimal dependency structure; S5. Calculating the rank correlation coefficient, the upper tail correlation coefficient, and the lower tail correlation coefficient between every two of the plurality of wind turbines according to the operation linkage analysis data set; S6. Determine the operational linkage of the plurality of wind turbines according to the optimal dependency structure and the rank correlation coefficient, the upper tail correlation coefficient, and the lower tail correlation coefficient.

2. The method for analyzing the wind turbine operation linkage based on the Vine Copula model according to claim 1, characterized in that: The step S3 specifically includes: S31, calculating the correlation between each two wind turbines according to the operation linkage analysis data set and the Copula function set, and determining the optimal Copula function for each two wind turbines; S32, calculating the rank correlation coefficient of every two wind turbines according to the operation linkage analysis data set and the optimal Copula function, generating a correlation coefficient matrix, and generating a C-Vine dependency structure according to the correlation coefficient matrix; S33, generating a D-Vine dependency structure according to the operation linkage analysis data set, the optimal Copula function, and the geographical locations of the plurality of wind turbines; S34. Generate an R-Vine dependency structure according to the operation linkage analysis data set, the optimal Copula function, and the maximum spanning tree principle.

3. The method for analyzing the operation linkage of wind turbines based on the Vine Copula model according to claim 1, characterized in that: The step S5 specifically includes: S51. Calculate the rank correlation coefficient between the wind turbine with the best power generation capacity and other wind turbines using t-Copula; S52. Calculate the upper tail correlation coefficient of the wind turbine with the best power generation capacity and other wind turbines using BB1 Copula; S53. Calculate the lower tail correlation coefficient between the wind turbine generator set with the best power generation capacity and other wind turbine generator sets using BB7 Copula.

4. The method for analyzing wind turbine operation linkage based on the Vine Copula model according to claim 1, characterized in that: The step S6 specifically includes: S61. Obtaining overall operational linkage among the plurality of wind turbine generator sets according to the optimal interdependence structure; S62. Obtaining local operation linkage between the wind turbine with the best power generation capacity and other wind turbines based on the rank correlation coefficient, the upper tail correlation coefficient, and the lower tail correlation coefficient; S63: Determine the operation linkage of the plurality of wind turbines by combining the overall operation linkage and the local operation linkage.

5. The method for analyzing wind turbine operation linkage based on the Vine Copula model according to claim 1, characterized in that: After step S6, the method further includes step S7 of determining abnormal wind turbines among the plurality of wind turbines according to the operation linkage.

6. The method for analyzing wind turbine operation linkage based on the Vine Copula model according to claim 5, characterized in that: The step S7 specifically includes: S71, calculating the deviation between each wind turbine generator set and the operation linkage prediction value; S72: Determine that the wind turbine generator set whose deviation is greater than a preset threshold is an abnormal turbine generator set.

Images