A method and device for processing fan abnormal data cleaning in a multi-wind rejection condition

Abstract

The application relates to the field of wind power technology and discloses a wind turbine abnormal data cleaning method and device for processing multiple wind rejection conditions, which comprises the following steps: collecting wind turbine operation data in a continuous time period, preprocessing the wind turbine operation data in the continuous time period to generate a first wind turbine data set; constructing a multiple wind rejection data identification boundary based on the first wind turbine data set, deleting multiple wind rejection data based on the multiple wind rejection data identification boundary, and generating a second wind turbine data set; determining high-density area accumulated data by using an abnormality detection algorithm based on the second wind turbine data set, deleting the high-density area accumulated data, and generating a third wind turbine data set; determining wind rejection accumulated data based on the third wind turbine data set, deleting the wind rejection accumulated data, and generating cleaned wind turbine operation data. The application realizes effective cleaning of wind turbine abnormal data under multiple wind rejection conditions.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of wind power technology, and in particular to a wind turbine abnormal data cleaning method and device for handling multiple abandoned wind conditions. BACKGROUND

[0002] Wind power is one of the currently large-scale renewable energy sources, and wind power has the characteristics of safety, environmental protection, cleanliness, and abundance. Wind turbine condition monitoring can effectively and timely find the operation problems of wind turbine generators, and establishing a real wind speed-power output curve can effectively monitor the state of the wind turbine generator and predict the power. There are several situations for the common wind turbine operation power-wind speed curve: 1, data accumulation caused by abandoned wind and power limiting; 2, data anomalies caused by sensor collection errors; 3, data problems caused by actual wind turbine operation.

[0003] Some methods related to processing wind turbine abnormal data use a single clustering anomaly processing method, but the single clustering anomaly processing method is difficult to clean abnormal data under long-term, multiple abandoned wind, and different power limiting conditions. Under long-term, multiple abandoned wind, and different power limiting conditions, a large amount of wind turbine operation data is in a state of accumulation in the middle of the P-V curve, and it is difficult to distinguish from normal operation data. SUMMARY

[0004] Therefore, the present application provides a wind turbine abnormal data cleaning method and device for handling multiple abandoned wind conditions to solve the problem that a single clustering anomaly processing method is difficult to clean wind turbine abnormal data under multiple abandoned wind conditions.

[0005] In a first aspect, the present application provides a wind turbine abnormal data cleaning method for handling multiple abandoned wind conditions, comprising:

[0006] Collecting wind turbine operation data in a continuous time period, preprocessing the wind turbine operation data in the continuous time period, and generating a first wind turbine data set;

[0007] Based on the first wind turbine data set, a multiple abandoned wind data recognition boundary is constructed, and multiple abandoned wind data is deleted based on the multiple abandoned wind data recognition boundary to generate a second wind turbine data set;

[0008] Based on the second wind turbine data set, an abnormal detection algorithm is used to determine high-density area accumulation data, and the high-density area accumulation data is deleted to generate a third wind turbine data set;

[0009] Based on the third wind turbine data set, abandoned wind accumulation data is determined, and the abandoned wind accumulation data is deleted to generate cleaned wind turbine operation data.

[0010] The embodiment provides a wind turbine abnormal data cleaning method for processing the multiple abandoned wind condition, the wind turbine operation data in the continuous time period is preprocessed, and then a multiple abandoned wind data identification boundary is constructed, separation of normal data and multiple abandoned wind data in a high-dimensional space is realized, the abnormal data detection algorithm is used to delete the high-density area accumulated data, and the abandoned wind accumulated data in the third wind turbine data set is deleted in the high-dimensional space, and effective cleaning of the wind turbine abnormal data under the multiple abandoned wind condition is realized.

[0011] In an optional implementation, the wind turbine operation data in the continuous time period is preprocessed to generate the first wind turbine data set, including:

[0012] The wind turbine output power and the wind turbine wind speed corresponding to different moments are determined based on the wind turbine operation data in the continuous time period;

[0013] The abnormal data is determined based on the wind turbine output power;

[0014] The abnormal data in the wind turbine operation data in the continuous time period is deleted to generate the first wind turbine data set.

[0015] The embodiment provides a wind turbine abnormal data cleaning method for processing the multiple abandoned wind condition, the abnormal data in the wind turbine operation data in the continuous time period is deleted, and preliminary cleaning of the wind turbine operation data in the continuous time period is realized.

[0016] In an optional implementation, the abnormal data is determined based on the wind turbine output power, including:

[0017] The output power mean and the output power standard deviation are calculated based on the wind turbine output power;

[0018] The standard score is calculated based on the wind turbine output power, the output power mean and the output power standard deviation;

[0019] The standard score is compared with the first preset threshold value, when the standard score is greater than the first preset threshold value, the wind turbine output power and the wind turbine wind speed are taken as the abnormal data.

[0020] The embodiment provides a wind turbine abnormal data cleaning method for processing the multiple abandoned wind condition, the standard score is calculated, and then the abnormal data is determined based on the comparison result of the standard score and the first preset threshold value, accurate identification of the abnormal data is realized, and a foundation for effectively cleaning the abnormal data is laid.

[0021] In an optional implementation, the multiple abandoned wind data identification boundary is constructed based on the first wind turbine data set, and the multiple abandoned wind data is deleted based on the multiple abandoned wind data identification boundary to generate the second wind turbine data set, including:

[0022] determine the first identification function coefficient and the first offset based on the first wind turbine data set;

[0023] obtain a kernel function, and calculate the multiple-rejected-wind data identification function value based on the wind turbine output power in the first wind turbine data set, the first identification function coefficient, the first offset, and the kernel function;

[0024] compare the multiple-rejected-wind data identification function value with a second preset threshold, and construct a multiple-rejected-wind data identification boundary based on a comparison result;

[0025] delete the multiple-rejected-wind data based on the multiple-rejected-wind data identification boundary, and generate a second wind turbine data set.

[0026] The wind turbine abnormal data cleaning method provided in the embodiment can realize accurate separation of normal data and multiple-rejected-wind data in a high-dimensional space, and thus effectively clean the multiple-rejected-wind data.

[0027] In an optional implementation, the determination of the first identification function coefficient and the first offset based on the first wind turbine data set comprises:

[0028] input the wind turbine output power in the first wind turbine data set into the first target optimization function, and solve the first target optimization function to generate the first identification function coefficient and the first offset.

[0029] The wind turbine abnormal data cleaning method provided in the embodiment can generate the first identification function coefficient and the first offset by solving the first target optimization function, and lay a foundation for subsequent construction of a multiple-rejected-wind data identification boundary.

[0030] In an optional implementation, the input of the wind turbine output power in the first wind turbine data set into the first target optimization function, and the solving of the first target optimization function to generate the first identification function coefficient and the first offset comprise:

[0031] The first identification function coefficient and the first offset are generated by solving the first target optimization function using a sequential minimal optimization algorithm.

[0032] In an optional implementation, the multiple-rejected-wind data identification function value is calculated based on the wind turbine output power in the first wind turbine data set, the first identification function coefficient, the first offset, and a kernel function, and a calculation formula of the multiple-rejected-wind data identification function value is as follows:

[0033]

[0034] wherein f(p) represents the multiple-rejected-wind data identification function value, a represents the first identification function coefficient, b represents the first offset, and K represents the kernel function. irepresents a first identification function coefficient, K1 represents a kernel function, p represents a kernel function variable, p i represents the wind turbine output power of the first wind turbine data set at the i-th moment, and p1 represents a first offset.

[0035] In an optional implementation, based on the second wind turbine data set, the high-density area accumulated data is determined by using the anomaly detection algorithm, and the high-density area accumulated data is deleted to generate a third wind turbine data set, including:

[0036] Based on the second wind turbine data set, the anomaly detection algorithm is used to determine the negative outlier factor value;

[0037] The negative outlier factor value is compared with a third preset threshold value, and when the negative outlier factor value is greater than the third preset threshold value, the wind turbine operation data corresponding to the negative outlier factor value is the high-density area accumulated data;

[0038] The high-density area accumulated data in the second wind turbine data set is deleted to generate a third wind turbine data set.

[0039] The wind turbine abnormal data cleaning method provided in the embodiment for processing the multi-rejected wind condition realizes accurate identification of the high-density area accumulated data by using the anomaly detection algorithm, and realizes effective cleaning of the high-density area accumulated data in the second wind turbine data set.

[0040] In an optional implementation, based on the second wind turbine data set, the negative outlier factor value is determined by using the anomaly detection algorithm, including:

[0041] Based on the second wind turbine data set, wind turbine operation data at multiple moments is determined, and the wind turbine operation data at multiple moments is taken as multiple data points;

[0042] Based on the multiple data points, the neighborhood data of the data points and the Euclidean distance between the data points are determined;

[0043] Based on the neighborhood data of the data points and the Euclidean distance between the data points, the reachable distance between the data points is determined;

[0044] Based on the neighborhood data of the data points and the reachable distance between the data points, the local reachable density of the data points is determined;

[0045] Based on the neighborhood data of the data points, the Euclidean distance between the data points, and the local reachable density of the data points, the negative outlier factor value is calculated.

[0046] The wind turbine abnormal data cleaning method provided in the embodiment for processing the multi-rejected wind condition increases the elimination effect of the outlier points by improving the anomaly detection algorithm, that is, determining the negative outlier factor value based on the local reachable density of the data points, and realizes effective cleaning of the high-density area accumulated data.

[0047] In an optional embodiment, the local outlier factor value is calculated based on the neighborhood data of the data points, the Euclidean distance between the data points and the local reachable density of the data points, and the formula for calculating the local outlier factor value is as follows:

[0048]

[0049] wherein LOF k (w) represents the local outlier factor value, N k (w) represents the k-distance neighborhood data of the data point w, lrd k (o) represents the k-local reachable density of the data point o, lrd k (w) represents the k-local reachable density of the data point w, d k (o) represents the k-distance neighborhood data of the data point o, d k (o, w) represents the Euclidean distance between the data point o and the data point w.

[0050] In an optional embodiment, the wind curtailment accumulation data is determined based on the third wind turbine data set, the wind curtailment accumulation data is deleted, and the cleaned wind turbine operation data is generated, including:

[0051] The second identification function coefficient and the second offset are determined based on the third wind turbine data set;

[0052] The radial basis function is obtained, and the redundant data anomaly identification function value is determined based on the wind turbine output power in the third wind turbine data set, the second identification function coefficient, the second offset and the radial basis function;

[0053] The redundant data anomaly identification function value is compared with the fourth preset threshold, and the redundant data anomaly identification boundary is constructed based on the comparison result;

[0054] The wind curtailment accumulation data is deleted based on the constructed redundant data anomaly identification boundary, and the cleaned wind turbine operation data is generated.

[0055] The wind turbine abnormal data cleaning method provided in the embodiment can accurately identify and effectively clean the wind curtailment accumulation data in a high-dimensional space by constructing the redundant data anomaly identification boundary.

[0056] In an optional embodiment, the second identification function coefficient and the second offset are determined based on the third wind turbine data set, including:

[0057] The wind turbine output power in the third wind turbine data set is input into the second target optimization function, and the second target optimization function is solved to generate the second identification function coefficient and the second offset.

[0058] In a second aspect, the present application provides a wind turbine abnormal data cleaning device for processing multiple abandoned wind conditions, comprising:

[0059] a preprocessing module, configured to collect wind turbine operation data in a continuous time period, preprocess the wind turbine operation data in the continuous time period, and generate a first wind turbine data set;

[0060] a multiple abandoned wind data deletion module, configured to construct a multiple abandoned wind data identification boundary based on the first wind turbine data set, delete multiple abandoned wind data based on the multiple abandoned wind data identification boundary, and generate a second wind turbine data set;

[0061] a high-density area accumulated data deletion module, configured to determine high-density area accumulated data based on the second wind turbine data set by using an abnormality detection algorithm, delete the high-density area accumulated data, and generate a third wind turbine data set;

[0062] abandoned wind accumulated data deletion module, configured to determine abandoned wind accumulated data based on the third wind turbine data set, delete the abandoned wind accumulated data, and generate cleaned wind turbine operation data.

[0063] In a third aspect, the present application provides a computer device, comprising a memory and a processor, the memory and the processor are in communication connection with each other, the memory stores computer instructions, and the processor executes the computer instructions to perform the wind turbine abnormal data cleaning method of the first aspect or any of the corresponding embodiments thereof.

[0064] In a fourth aspect, the present application provides a computer readable storage medium, which stores computer instructions, and the computer instructions are used to make the computer execute the wind turbine abnormal data cleaning method of the first aspect or any of the corresponding embodiments thereof. BRIEF DESCRIPTION OF DRAWINGS

[0065] In order to more clearly illustrate the specific embodiments of the present application or the technical solutions in the prior art, the drawings needed in the specific embodiments or prior art description will be briefly introduced below. Obviously, the drawings in the following description are some embodiments of the present application, and those skilled in the art can obtain other drawings according to these drawings without creative labor.

[0066] Figure 1 is a flowchart of a wind turbine abnormal data cleaning method for processing multiple abandoned wind conditions according to an embodiment of the present application;

[0067] Figure 2 is a flowchart of another wind turbine abnormal data cleaning method for processing multiple abandoned wind conditions according to an embodiment of the present application;

[0068] Figure 3 is a flowchart of another method for processing fan abnormal data cleaning in the case of multiple abandoned wind according to an embodiment of the application;

[0069] Figure 4 is a flowchart of another method for processing fan abnormal data cleaning in the case of multiple abandoned wind according to an embodiment of the application;

[0070] Figure 5 is a comparative effect diagram of the existing LOF algorithm and the improved LOF algorithm according to an embodiment of the application;

[0071] Figure 6 is a flowchart of another method for processing fan abnormal data cleaning in the case of multiple abandoned wind according to an embodiment of the application;

[0072] Figure 7 is a flowchart of a method for processing fan abnormal data cleaning in the case of multiple abandoned wind according to an embodiment 1 of the application;

[0073] Figure 8 is a fan operation data diagram of a power plant in Wulanchabu, Inner Mongolia Autonomous Region according to an embodiment of the application;

[0074] Figure 9 is a diagram of a first data set after data preprocessing according to an embodiment of the application;

[0075] Figure 10 is a diagram of a second data set according to an embodiment of the application;

[0076] Figure 11 is a diagram of a third data set after cleaning the accumulated data using the improved LOF algorithm according to an embodiment of the application;

[0077] Figure 12 is a diagram of cleaned wind turbine operation data according to an embodiment of the application;

[0078] Figure 13 is a structural block diagram of a device for processing fan abnormal data cleaning in the case of multiple abandoned wind according to an embodiment of the application;

[0079] Figure 14 is a hardware structure diagram of a computer device according to an embodiment of the application. DETAILED DESCRIPTION

[0080] In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some but not all of the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0081] The embodiment of the present application provides a fan abnormal data cleaning method for processing multiple abandoned wind conditions, which separates the boundary of accumulated data and normal data by using One-Class (single classification algorithm), and deletes a large amount of abandoned wind data by using an improved LOF algorithm and OneClassSVM (Support Vector Machine, an unsupervised learning algorithm), so as to effectively clean the multiple abandoned wind data.

[0082] The One-Class adopts a sigmoid kernel (a kind of nonlinear kernel function), and the OneClassSVM adopts an RBF kernel (Radial Basis Function kernel).

[0083] According to the embodiment of the present application, a fan abnormal data cleaning method for processing multiple abandoned wind conditions is provided. It should be noted that the steps shown in the flowchart of the drawings can be executed in a computer system such as a group of computer executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0084] In the present embodiment, a fan abnormal data cleaning method for processing multiple abandoned wind conditions is provided, which can be used in the mobile terminal described above, such as a mobile phone, a tablet computer and the like (the execution subject is described in combination with the actual situation), Figure 1 The flowchart of the fan abnormal data cleaning method for processing multiple abandoned wind conditions according to the embodiment of the present application is shown in FIG. 1, which includes the following steps: Figure 1

[0085] In step S101, the running data of the wind turbine in a continuous time period is collected, the running data of the wind turbine in the continuous time period is preprocessed, and a first fan data set is generated.

[0086] Specifically, the running data corresponding to N time points of a single wind turbine in a continuous time period is collected by using a wind speed sensor and a power sensor Wherein, i = 0, 1, 2,..., N, P i is the output power of the wind turbine, V i is the wind speed of the fan generator.​

[0087] Step S102, constructing a multi-wind rejection data recognition boundary based on the first wind turbine data set, and deleting the multi-wind rejection data based on the multi-wind rejection data recognition boundary to generate a second wind turbine data set.

[0088] Step S103, determining high-density area accumulated data by using an anomaly detection algorithm based on the second wind turbine data set, and deleting the high-density area accumulated data to generate a third wind turbine data set.

[0089] Step S104, determining wind rejection accumulated data based on the third wind turbine data set, and deleting the wind rejection accumulated data to generate cleaned wind turbine operation data.

[0090] The wind turbine abnormal data cleaning method provided in the embodiment can effectively clean the abnormal data of the wind turbine by preprocessing the wind turbine operation data in a continuous time period, constructing a multi-wind rejection data recognition boundary, separating the normal data and the multi-wind rejection data in a high-dimensional space, deleting the high-density area accumulated data by using an anomaly detection algorithm, and deleting the wind rejection accumulated data in the third wind turbine data set in the high-dimensional space.

[0091] In the embodiment, a wind turbine abnormal data cleaning method for processing the multi-wind rejection condition is provided, which can be applied to the mobile terminal such as a mobile phone, a tablet computer and the like. Figure 2 The flowchart of the wind turbine abnormal data cleaning method for processing the multi-wind rejection condition according to the embodiment of the present application is shown in FIG. 1, which includes the following steps. Figure 2

[0092] Step S201, collecting wind turbine operation data in a continuous time period, preprocessing the wind turbine operation data in the continuous time period, and generating a first wind turbine data set.

[0093] Specifically, the step S201 includes the following steps.

[0094] Step S2011, determining the wind turbine output power and the wind turbine wind speed corresponding to different time based on the wind turbine operation data in the continuous time period.

[0095] Step S2012, determining abnormal data based on the wind turbine output power.

[0096] In some optional embodiments, the step S2012 includes the following steps.

[0097] Step a1, calculating the output power mean value and the output power standard deviation based on the wind turbine output power.

[0098] Step a2, calculating the standard score based on the wind turbine output power, the output power mean value and the output power standard deviation.​

[0099] Specifically, a standard score is calculated for the wind turbine output power at each moment. The formula for calculating the standard score is as follows:

[0100]

[0101] Where z represents the standard score, P i Let represent the output power of the wind turbine at time i, μ represent the mean output power of the wind turbine over N times, and σ represent the standard deviation of the output power of the wind turbine over N times.

[0102] Step a3: Compare the standard score with the first preset threshold. If the standard score is greater than the first preset threshold, the wind turbine output power and wind turbine wind speed are considered abnormal data.

[0103] For example, if the first preset threshold is set to 2.2, and the standard score is greater than 2.2, then the wind turbine operation data at time i is abnormal data.

[0104] Step S2013: Delete abnormal data in the wind turbine operation data within a continuous time period to generate the first wind turbine dataset.

[0105] Specifically, after deleting outlier data from wind turbine operation data within consecutive time periods, the remaining M wind turbine operation data constitute the first wind turbine dataset, i.e. i = 0, 1, 2, ..., M, and M <N。

[0106] Step S202: Construct a multi-curtailment data identification boundary based on the first wind turbine dataset, and delete multi-curtailment data based on the multi-curtailment data identification boundary to generate a second wind turbine dataset. For details, please refer to [link to relevant documentation]. Figure 1 Step S102 of the illustrated embodiment will not be described again here.

[0107] Step S203: Based on the second wind turbine dataset, an anomaly detection algorithm is used to identify high-density accumulation data, and this high-density accumulation data is deleted to generate the third wind turbine dataset. For details, please refer to [link to relevant documentation]. Figure 1 Step S103 of the illustrated embodiment will not be described again here.

[0108] Step S204: Based on the third wind turbine dataset, determine the accumulated wind curtailment data, delete the accumulated wind curtailment data, and generate cleaned wind turbine operating data. For details, please refer to [link to relevant documentation]. Figure 1 Step S104 of the illustrated embodiment will not be described again here.

[0109] The embodiment provides a wind fan abnormal data cleaning method for processing a multiple-wind-rejection condition.

[0110] In the embodiment, a wind fan abnormal data cleaning method for processing a multiple-wind-rejection condition is provided, and can be applied to the mobile terminal such as a mobile phone, a tablet computer and the like. Figure 3 The embodiment provides a flowchart of the wind fan abnormal data cleaning method for processing a multiple-wind-rejection condition according to the application, as shown in the figure, and the flowchart comprises the following steps. Figure 3

[0111] In step S301, wind turbine operation data in a continuous time period is collected, the wind turbine operation data in the continuous time period is preprocessed, and a first wind fan data set is generated. For details, refer to step S201 in the embodiment shown in the figure, which will not be repeated here. Figure 2

[0112] In step S302, a multiple-wind-rejection data identification boundary is constructed based on the first wind fan data set, and multiple-wind-rejection data is deleted based on the multiple-wind-rejection data identification boundary, to generate a second wind fan data set.

[0113] Specifically, step S302 comprises the following steps.

[0114] In step S3021, first identification function coefficients and a first offset are determined based on the first wind fan data set.

[0115] Specifically, the wind turbine output power in the first wind fan data set is input into a first target optimization function, and the first target optimization function is solved to generate the first identification function coefficients and the first offset.

[0116] Further, the expression of the first target optimization function is as follows.

[0117]

[0118]

[0119] Wherein, v1 represents an abnormal data estimation ratio, α i represents the first identification function coefficients corresponding to the i th moment, α j represents the first identification function coefficients corresponding to the j th moment, K1 represents a kernel function, p i represents the wind turbine output power in the first wind fan data set at the i th moment, p j represents the wind turbine output power in the first wind fan data set at the j th moment, ρ1 represents the first offset, and M represents the moment.

[0120] ​​Further, the first target optimization function is solved by using a sequential minimal optimization (SMO) algorithm to generate the first identification function coefficient and the first offset.

[0121] Further, the target optimization function is solved by using an SMO algorithm to obtain α i and ρ1.

[0122] In step S3022, the kernel function is obtained, and the multiple curtailment data identification function value is calculated based on the wind turbine output power in the first wind turbine data set, the first identification function coefficient, the first offset, and the kernel function.

[0123] Specifically, the calculation formula of the multiple curtailment data identification function value is as follows:

[0124]

[0125] wherein f(p) represents the multiple curtailment data identification function value, α i represents the first identification function coefficient, K1 represents the kernel function, p represents the kernel function variable, p i represents the wind turbine output power at the i-th moment in the first wind turbine data set, and ρ1 represents the first offset.

[0126] Further, the expression of the kernel function is as follows:

[0127] K1(p,p i )=tanh(γp T p i +r)i=1,2,...M(4)

[0128] wherein p is the independent variable of the K1 kernel function, γ is the K1 kernel function coefficient, and r is the offset.

[0129] In step S3023, the multiple curtailment data identification function value is compared with the second preset threshold, and the multiple curtailment data identification boundary is constructed based on the comparison result.

[0130] Specifically, the judgment standard of the multiple curtailment data, i.e., the multiple curtailment data identification boundary, is represented as follows:

[0131]

[0132] In step S3024, the multiple curtailment data is deleted based on the multiple curtailment data identification boundary to generate the second wind turbine data set.

[0133] Specifically, when f(p) < 0, the wind turbine operation data corresponding to the moment of f(p) is the multiple curtailment data, the multiple curtailment data is removed, and the remaining T data constitute a new data set, i.e., the second wind turbine data set (Pi (2) V i (2) ), i = 0, 1, 2,..., T, and T < M.

[0134] Step S303, based on the second wind turbine data set, using an anomaly detection algorithm to determine high-density area accumulation data, and deleting the high-density area accumulation data, to generate a third wind turbine data set. For details, please refer to step S203 of the embodiment shown in Figure 2 Step S203 of the embodiment shown in

[0135] Step S304, based on the third wind turbine data set, determining the abandoned wind accumulation data, and deleting the abandoned wind accumulation data, to generate the cleaned wind turbine operation data. For details, please refer to step S204 of the embodiment shown in Figure 2 Step S204 of the embodiment shown in

[0136] The wind turbine abnormal data cleaning method provided in the embodiment can realize accurate separation of normal data and multi-abandoned wind data in high-dimensional space by constructing a multi-abandoned wind data recognition boundary, and further realize effective cleaning of multi-abandoned wind data.

[0137] In the embodiment, a wind turbine abnormal data cleaning method for handling the multi-abandoned wind condition is provided, which can be used in the mobile terminal as described above, such as a mobile phone, a tablet computer, etc. Figure 4 is a flowchart of a wind turbine abnormal data cleaning method for handling the multi-abandoned wind condition according to an embodiment of the present application, as shown in Figure 4 The flowchart includes the following steps:

[0138] Step S401, collecting wind turbine operation data in a continuous time period, and pre-processing the wind turbine operation data in the continuous time period to generate a first wind turbine data set. For details, please refer to step S301 of the embodiment shown in Figure 3 Step S301 of the embodiment shown in

[0139] Step S402, constructing a multi-abandoned wind data recognition boundary based on the first wind turbine data set, and deleting the multi-abandoned wind data based on the multi-abandoned wind data recognition boundary to generate a second wind turbine data set. For details, please refer to step S302 of the embodiment shown in Figure 3 Step S302 of the embodiment shown in

[0140] Step S403, based on the second wind turbine data set, using an anomaly detection algorithm to determine high-density area accumulation data, and deleting the high-density area accumulation data, to generate a third wind turbine data set.

[0141] Specifically, the above step S403 includes:

[0142] Step S4031, determining a negative outlier factor value based on the second wind turbine dataset by using the anomaly detection algorithm.

[0143] In some optional embodiments, the step S4031 comprises:

[0144] Step b1, determining wind turbine operation data at multiple time points based on the second wind turbine dataset, and taking the wind turbine operation data at the multiple time points as multiple data points.

[0145] Step b2, determining neighborhood data of the data points and Euclidean distances between the data points based on the multiple data points.

[0146] Step b3, determining reachable distances between the data points based on the neighborhood data of the data points and the Euclidean distances between the data points.

[0147] Specifically, the reachable distance between the data points reach_dist k The calculation formula of reach_dist

[0148] reach_dist k (o,w)=max(d k (o),d(o,w))(6)

[0149] wherein d k (o) represents the k-distance neighborhood data of the data point o, and d k (o,w) represents the Euclidean distance between the data point o and the data point w.

[0150] Step b4, determining local reachable density of the data points based on the neighborhood data of the data points and the reachable distances between the data points.

[0151] Specifically, the calculation formula of the local reachable density lrd k (w) of the data points is as follows:

[0152]

[0153] wherein N k (w) represents the k-distance neighborhood data of the data point w, i.e. the set of all data points within the k-distance of the data point w.

[0154] Step b5, calculating the negative outlier factor value based on the neighborhood data of the data points, the Euclidean distances between the data points and the local reachable density of the data points.

[0155] Specifically, the calculation formula of the negative outlier factor value is as follows:

[0156]

[0157] wherein, LOF k (w) represents the kth distance neighborhood data of the data point w, lrd k (w) represents the kth local reachable density of the data point w, lrd k (o) represents the kth local reachable density of the data point o, lrd k (w) represents the kth local reachable density of the data point w, d k (o) represents the kth distance neighborhood data of the data point o, d k (o, w) represents the Euclidean distance between the data point o and the data point w.

[0158] Step S4032, the LOF value is compared with the third preset threshold value, when the LOF value is greater than the third preset threshold value, the wind turbine operation data corresponding to the LOF value is high-density area accumulated data.

[0159] Specifically, when the third preset threshold value is 1, when the LOF k (w) > 1, the LOF k (w) corresponding wind turbine operation data at the moment is high-density area accumulated data.

[0160] Step S4033, the high-density area accumulated data in the second wind turbine data set is deleted to generate a third wind turbine data set.

[0161] Specifically, after the high-density area accumulated data in the second wind turbine data set is deleted, the third wind turbine data set is constructed based on the remaining Q wind turbine operation data at the moment, that is, i = 0, 1, 2,..., Q, and Q < T.

[0162] Step S404, the wind turbine operation data after cleaning is generated by determining the abandoned wind accumulated data based on the third wind turbine data set and deleting the abandoned wind accumulated data. For details, please refer to Figure 3 Step S304 of the embodiment shown in the figure will not be repeated here.

[0163] The wind turbine abnormal data cleaning method provided in the embodiment for processing the multiple abandoned wind condition increases part as a density adjustment factor, the average ratio of the reachable distance in the K neighborhood of the data point is calculated, such as Figure 5 The improved LOF algorithm increases the outlier elimination effect.

[0164] A wind turbine abnormal data cleaning method for processing the multiple abandoned wind condition is provided in the embodiment, which can be used in the mobile terminal such as the mobile phone, the tablet computer and the like, Figure 6 is a flowchart of a wind turbine abnormal data cleaning method for processing the multiple abandoned wind condition according to the embodiment of the application, as shown in Figure 6As shown, the flow includes the following steps:

[0165] At step S601, wind turbine operation data in a continuous time period is collected, the wind turbine operation data in the continuous time period is preprocessed, and a first wind turbine data set is generated. For details, please refer to Figure 4 At step S401 of the embodiment shown, further description is not repeated here.

[0166] At step S602, a multi-wind rejection data recognition boundary is constructed based on the first wind turbine data set, and multi-wind rejection data is deleted based on the multi-wind rejection data recognition boundary, to generate a second wind turbine data set. For details, please refer to Figure 4 At step S402 of the embodiment shown, further description is not repeated here.

[0167] At step S603, based on the second wind turbine data set, a high-density area accumulation data is determined by using an anomaly detection algorithm, and the high-density area accumulation data is deleted, to generate a third wind turbine data set. For details, please refer to Figure 4 At step S403 of the embodiment shown, further description is not repeated here.

[0168] At step S604, wind rejection accumulation data is determined based on the third wind turbine data set, and the wind rejection accumulation data is deleted, to generate cleaned wind turbine operation data.

[0169] Specifically, the above step S604 includes:

[0170] At step S6041, a second recognition function coefficient and a second offset are determined based on the third wind turbine data set.

[0171] Specifically, the wind turbine output power in the third wind turbine data set is input into the second target optimization function, and the second target optimization function is solved, to generate the second recognition function coefficient and the second offset.

[0172] Further, the expression of the second target optimization function is as shown below:

[0173]

[0174]

[0175] Wherein, v2 represents wind rejection accumulation data estimation ratio, β i represents the second recognition function coefficient corresponding to the i-th moment, β j represents the second recognition function coefficient corresponding to the j-th moment, K2 represents the radial basis function, p i represents the wind turbine output power at the i-th moment in the third wind turbine data set, p j represents the wind turbine output power at the j-th moment in the third wind turbine data set, ρ2 represents the second offset, and Q represents the moment.

[0176] Further, the second objective optimization function is solved by using a sequential minimal optimization (SMO) algorithm.

[0177] Further, the objective optimization function is solved by using the SMO algorithm to obtain β i and ρ2.

[0178] In step S6042, a radial basis function is obtained, and the redundant data anomaly identification function value is determined based on the wind turbine output power in the third wind turbine data set, the second identification function coefficient, the second offset, and the radial basis function.

[0179] Specifically, the calculation formula of the redundant data anomaly identification function value is as follows:

[0180]

[0181] wherein g(x) represents the multi-wind rejection data identification function value, and x represents the radial basis function variable.

[0182] Further, the expression of the radial basis function is as follows:

[0183]

[0184] wherein x, x i is the independent variable of the K2 radial basis function, and σ is the parameter of the K2 radial basis function.

[0185] In step S6043, the redundant data anomaly identification function value is compared with a fourth preset threshold, and a redundant data anomaly identification boundary is constructed based on the comparison result.

[0186] Specifically, the judgment standard of the wind rejection accumulation data, i.e., the redundant data anomaly identification boundary, is represented as follows:

[0187]

[0188] In step S6044, the wind rejection accumulation data is deleted based on the constructed redundant data anomaly identification boundary, and cleaned wind turbine operation data is generated.

[0189] Specifically, when g(x) < 0, the wind turbine operation data corresponding to the time of g(x) is wind rejection accumulation data, and the wind rejection accumulation data is deleted to generate cleaned wind turbine operation data, i.e., and U < Q.

[0190] The wind turbine abnormal data cleaning method provided in the embodiment can accurately identify and effectively clean the wind rejection accumulation data in a high-dimensional space by constructing a redundant data anomaly identification boundary.

[0191] As Figure 7 shown below by a specific example to illustrate a method of processing fan abnormal data cleaning in the case of multiple abandoned wind.

[0192] Example 1:

[0193] As Figure 8 shown, this embodiment selects SCADA (Supervisory Control And Data Acquisition system, i.e. data acquisition and monitoring control system) data of a certain wind farm in Ulanqab, Inner Mongolia Autonomous Region from January to February 2020, the data amount is about 50000, and the collection accuracy is 15 min (minute) level.

[0194] Step one, using wind speed sensor, power sensor to collect N = 50000 data of a single wind turbine in a continuous time period as a data set, i = 0, 1, 2,..., N, P i is the output power of the wind turbine, V i is the wind speed of the fan generator.

[0195] First, preprocess the obtained multiple abandoned wind condition data, the preprocessing method is: for each data, calculate Where μ is the mean of the output power of N data, and is the standard deviation. If the calculation result is greater than 2.2, delete the data, as Figure 9 shown, the remaining M data after deletion constitute a new data set

[0196] Step two, M data constitute a new data set Data build abnormal data identification boundary, the identification function is:

[0197]

[0198] The expression of the first target optimization function is as follows:

[0199]

[0200]

[0201] Where v1 represents the abnormal data estimation ratio, taking 0.5, α i represents the first identification function coefficient corresponding to the i-th moment, α j represents the first identification function coefficient corresponding to the j-th moment, K1 represents the kernel function, p i represents the output power of the wind turbine in the first fan data set at the i-th moment, p jPj represents the wind turbine output power of the first wind fan data set at the jth moment, ρ1 represents the first offset, and M represents the moment.

[0202] Solving the optimization objective to obtain α i (i = 1, 2,..., M) and ρ1, to obtain:

[0203]

[0204] For data points in the data set f(p) < 0, remove them, as shown in Figure 10 The remaining T data constitute a new data set

[0205] Step three, for the remaining T data Calculate the LOF value of each data:

[0206]

[0207] Where, LOF k (w) represents the negative outlier factor value, N k (w) represents the k-distance neighborhood data of data point w, lrd k (o) represents the k-local reachable density of data point o, lrd k (w) represents the k-local reachable density of data point w, d k (o) represents the k-distance neighborhood data of data point o, d k (o, w) represents the Euclidean distance between data point o and data point w.

[0208] Where, lrd k (w) is calculated as follows:

[0209]

[0210] Where, reach_dist k (o, w) is the reachable distance of data point o to w.

[0211] Further, the data points with LOF value > 1 are removed, as shown in Figure 11 For the remaining Q data, construct a data set

[0212] Step four, for the remaining Q data, construct a data set Construct a redundant data anomaly recognition boundary, and the recognition function is:

[0213]

[0214] Specifically, the expression of the second target optimization function is as follows:

[0215]

[0216]

[0217] Where v2 represents the estimated ratio of wind curtailment accumulation data, with a value of 0.48, and β i β represents the coefficient of the second identification function at time i. j Let K2 represent the coefficients of the second identification function at time j, and let p represent the radial basis functions. i p represents the output power of the wind turbine at time i in the third wind turbine dataset. j Let ρ2 represent the output power of the wind turbine at time j in the third wind turbine dataset, ρ2 represent the second offset, and Q represent the time.

[0218] The second objective function yields β. i (i = 1, 2, ..., Q) and ρ2, thus obtaining:

[0219]

[0220] like Figure 12 As shown, for the dataset Points where g(p) < 0 are removed.

[0221] This embodiment also provides a device for cleaning abnormal wind turbine data under multiple wind curtailment scenarios. This device is used to implement the above embodiments and preferred embodiments, and details already described will not be repeated. As used below, the term "module" can refer to a combination of software and / or hardware that performs a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0222] This embodiment provides a device for cleaning abnormal wind turbine data under multiple wind curtailment scenarios, such as... Figure 13 As shown, it includes:

[0223] Preprocessing module 131 is used to collect wind turbine operation data within a continuous time period, preprocess the wind turbine operation data within the continuous time period, and generate the first wind turbine dataset.

[0224] The multi-curtailment data deletion module 132 is used to construct a multi-curtailment data identification boundary based on the first wind turbine dataset, and delete multi-curtailment data based on the multi-curtailment data identification boundary to generate a second wind turbine dataset.

[0225] The high-density area accumulation data deletion module 133 is used to determine the high-density area accumulation data based on the second wind turbine dataset using an anomaly detection algorithm, delete the high-density area accumulation data, and generate the third wind turbine dataset.

[0226] The wind accumulation data deletion module 134 is configured to determine wind accumulation data based on the third wind turbine data set, delete the wind accumulation data, and generate cleaned wind turbine operation data.

[0227] In some optional embodiments, the preprocessing module 131 comprises:

[0228] The first determination sub-module is configured to determine the wind turbine output power and the wind speed of the wind turbine corresponding to different time points based on the wind turbine operation data in the continuous time period.

[0229] The second determination sub-module is configured to determine abnormal data based on the wind turbine output power.

[0230] The first deletion sub-module is configured to delete the abnormal data in the wind turbine operation data in the continuous time period, and generate the first wind turbine data set.

[0231] In some optional embodiments, the second determination sub-module comprises:

[0232] The first calculation unit is configured to calculate the mean value and the standard deviation of the output power based on the wind turbine output power.

[0233] The second calculation unit is configured to calculate the standard score based on the wind turbine output power, the mean value of the output power, and the standard deviation of the output power.

[0234] The first comparison unit is configured to compare the standard score with a first preset threshold value, and when the standard score is greater than the first preset threshold value, the wind turbine output power and the wind speed of the wind turbine are determined as the abnormal data.

[0235] In some optional embodiments, the multiple wind curtailment data deletion module 132 comprises:

[0236] The third determination sub-module is configured to determine the first identification function coefficient and the first offset based on the first wind turbine data set.

[0237] The first calculation sub-module is configured to obtain a kernel function, and calculate the multiple wind curtailment data identification function value based on the wind turbine output power in the first wind turbine data set, the first identification function coefficient, the first offset, and the kernel function.

[0238] The first comparison sub-module is configured to compare the multiple wind curtailment data identification function value with a second preset threshold value, and construct a multiple wind curtailment data identification boundary based on the comparison result.

[0239] The second deletion sub-module is configured to delete the multiple wind curtailment data based on the multiple wind curtailment data identification boundary, and generate the second wind turbine data set.

[0240] In some optional embodiments, the third determining sub-module is specifically configured to input the wind turbine output power in the first fan data set into the first target optimization function, and solve the first target optimization function to generate the first identification function coefficient and the first offset.

[0241] In some optional embodiments, the third determining sub-module is specifically configured to solve the first target optimization function by using a sequential minimal optimization algorithm to generate the first identification function coefficient and the first offset.

[0242] In some optional embodiments, the calculation formula of the multiple-wind-rejection data identification function value in the first calculation sub-module is as follows:

[0243]

[0244] wherein f(p) represents the multiple-wind-rejection data identification function value, a i represents the first identification function coefficient, K1 represents a kernel function, p represents a kernel function variable, and p i represents the wind turbine output power at the i-th moment in the first fan data set, and p1 represents the first offset.

[0245] In some optional embodiments, the high-density area accumulation data deletion module 133 comprises:

[0246] The fourth determining sub-module is configured to determine a negative outlier factor value based on the second fan data set by using an anomaly detection algorithm;

[0247] The second comparing sub-module is configured to compare the negative outlier factor value with a third preset threshold value, and when the negative outlier factor value is greater than the third preset threshold value, the wind turbine operation data corresponding to the negative outlier factor value is high-density area accumulation data;

[0248] The third deleting sub-module is configured to delete the high-density area accumulation data in the second fan data set to generate a third fan data set.

[0249] In some optional embodiments, the fourth determining sub-module comprises:

[0250] The first determining unit is configured to determine wind turbine operation data at multiple moments based on the second fan data set, and take the wind turbine operation data at the multiple moments as multiple data points;

[0251] The second determining unit is configured to determine neighborhood data of the data points and Euclidean distances between the data points based on the multiple data points;

[0252] The third determining unit is configured to determine reachable distances between the data points based on the neighborhood data of the data points and the Euclidean distances between the data points;

[0253] The fourth determining unit is configured to determine the local reachable density of the data point based on the neighborhood data of the data point and the reachable distance between the data points.

[0254] The fifth determining unit is configured to calculate the local outlier factor value based on the neighborhood data of the data point, the Euclidean distance between the data points and the local reachable density of the data point.

[0255] In some optional embodiments, the formula for calculating the local outlier factor value in the fifth determining unit is as follows:

[0256]

[0257] wherein, LOF k (w) represents the local outlier factor value, N k (w) represents the k-distance neighborhood data of the data point w, lrd k (o) represents the k-local reachable density of the data point o, lrd k (w) represents the k-local reachable density of the data point w, d k (o) represents the k-distance neighborhood data of the data point o, d k (o, w) represents the Euclidean distance between the data point o and the data point w.

[0258] In some optional embodiments, the abandoned wind accumulation data deletion module 134 comprises:

[0259] The fifth determining sub-module is configured to determine the second identification function coefficient and the second offset based on the third wind turbine data set;

[0260] The sixth determining sub-module is configured to obtain a radial basis function, and determine the redundant data anomaly identification function value based on the wind turbine output power in the third wind turbine data set, the second identification function coefficient, the second offset and the radial basis function;

[0261] The third comparing sub-module is configured to compare the redundant data anomaly identification function value with a fourth preset threshold, and construct a redundant data anomaly identification boundary based on a comparison result.

[0262] The fourth deleting sub-module is configured to delete the abandoned wind accumulation data based on the constructed redundant data anomaly identification boundary, and generate cleaned wind turbine operation data.

[0263] In some optional embodiments, the fifth determining sub-module is specifically configured to input the wind turbine output power in the third wind turbine data set into the second target optimization function, and solve the second target optimization function to generate the second identification function coefficient and the second offset.

[0264] Further function descriptions of the above various modules and units are the same as those of the above corresponding embodiments, and will not be repeated here.

[0265] In this embodiment, a wind turbine abnormal data cleaning device for handling multiple wind curtailment situations is presented in the form of a functional unit. Here, a unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that execute one or more software or fixed programs, and / or other devices that can provide the above functions.

[0266] This invention also provides a computer device having the above-described features. Figure 13 The diagram shows a device for cleaning abnormal wind turbine data under multiple wind curtailment scenarios.

[0267] Please see Figure 14 , Figure 14 This is a schematic diagram of the structure of a computer device provided in an optional embodiment of the present invention, such as... Figure 14 As shown, the computer device includes one or more processors 10, memory 20, and interfaces for connecting the components, including high-speed interfaces and low-speed interfaces. The components communicate with each other via different buses and can be mounted on a common motherboard or otherwise installed as needed. The processors can process instructions executed within the computer device, including instructions stored in or on memory to display graphical information of a GUI on external input / output devices (such as display devices coupled to the interfaces). In some alternative implementations, multiple processors and / or multiple buses can be used with multiple memories and multiple memory modules, if desired. Similarly, multiple computer devices can be connected, each providing some of the necessary operations (e.g., as a server array, a group of blade servers, or a multiprocessor system). Figure 14 Take a processor 10 as an example.

[0268] Processor 10 may be a central processing unit, a network processor, or a combination thereof. Processor 10 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof. The programmable logic device may be a complex programmable logic device (CAMP), a field-programmable gate array (FPGA), a general-purpose array logic (GDA), or any combination thereof.

[0269] The memory 20 stores instructions executable by at least one processor 10 to cause the at least one processor 10 to perform the method shown in the above embodiments.

[0270] The memory 20 can include a program storage area and a data storage area, where the program storage area can store an operating system, application programs required for at least one function, and the data storage area can store data created according to the use of the computer device, etc. In addition, the memory 20 can include a high-speed random access memory, and can also include a non-transitory memory, such as at least one magnetic disk storage device, a flash memory device, or other non-transitory solid-state memory device. In some alternative embodiments, the memory 20 can optionally include a memory disposed remotely from the processor 10, which can be connected to the computer device through a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0271] The memory 20 can include a volatile memory, such as a random access memory, and can also include a non-volatile memory, such as a flash memory, a hard disk, or a solid state disk, and can also include a combination of the above-mentioned kinds of memories.

[0272] The computer device also includes an input device 30 and an output device 40. The processor 10, the memory 20, the input device 30, and the output device 40 can be connected through a bus or other means, Figure 14 The above-mentioned connection through a bus is taken as an example.

[0273] The input device 30 can receive inputted digital or character information, and generate key signal inputs related to the user settings and function controls of the computer device, such as a touch screen, a keypad, a mouse, a trackpad, a touchpad, a pointing stick, one or more mouse buttons, a trackball, a joystick, etc. The output device 40 can include a display device, an auxiliary lighting device (such as an LED), a tactile feedback device (such as a vibration motor), etc. The above-mentioned display device includes, but is not limited to, a liquid crystal display, a light-emitting diode, a display, and a plasma display. In some alternative embodiments, the display device can be a touch screen.

[0274] The embodiments of the present application further provide a computer readable storage medium, and the method according to the embodiments of the present application can be implemented in hardware, firmware, or recorded in a storage medium, or be implemented as computer codes stored in a remote storage medium or a non-transitory machine readable storage medium and downloaded through a network and stored in a local storage medium, so that the method described herein can be processed by such software on a storage medium using a general purpose computer, a special purpose processor, or programmable or special hardware. The storage medium can be a magnetic disk, an optical disk, a read-only memory, a random access memory, a flash memory, a hard disk, or a solid state disk, etc. Further, the storage medium can also include a combination of the above-mentioned types of memories. It can be understood that the computer, the processor, the microprocessor controller, or the programmable hardware includes a storage component that can store or receive software or computer codes, when the software or computer codes are accessed and executed by the computer, the processor, or the hardware, the method shown in the above embodiments is implemented.

[0275] Although the embodiments of the present application are described in conjunction with the drawings, various modifications and changes can be made by those skilled in the art without departing from the spirit and scope of the present application, and such modifications and changes fall within the scope defined by the appended claims.

Claims

1. A method for cleaning abnormal wind turbine data under multiple wind curtailment scenarios, characterized in that, The method includes: Collect wind turbine operation data over a continuous time period, preprocess the wind turbine operation data over the continuous time period, and generate a first wind turbine dataset; Based on the first wind turbine dataset, a multi-curtailment data identification boundary is constructed, and multi-curtailment data is deleted based on the multi-curtailment data identification boundary to generate a second wind turbine dataset. Based on the second wind turbine dataset, an anomaly detection algorithm is used to identify high-density area accumulation data, and the high-density area accumulation data is deleted to generate a third wind turbine dataset. Based on the third wind turbine dataset, the accumulated wind curtailment data is determined, the accumulated wind curtailment data is deleted, and cleaned wind turbine operation data is generated. The step of constructing a multi-curtailment data identification boundary based on the first wind turbine dataset, and deleting multi-curtailment data based on the multi-curtailment data identification boundary to generate a second wind turbine dataset includes: The first identification function coefficients and the first offset are determined based on the first wind turbine dataset; Obtain the kernel function, and calculate the multi-curtailment data identification function value based on the wind turbine output power, the first identification function coefficients, the first offset, and the kernel function in the first wind turbine dataset; the calculation formula for the multi-curtailment data identification function value is as follows: in, This represents the identification function value for multiple wind curtailment data. Represents the coefficients of the first identification function. Represents the kernel function. Represents the kernel function variable. Indicates the first wind turbine in the dataset. The output power of the wind turbine at any given time. Indicates the first offset; The value of the multi-curtailed wind data identification function is compared with a second preset threshold, and the multi-curtailed wind data identification boundary is constructed based on the comparison result. Based on the identified boundaries of the multiple wind curtailment data, the multiple wind curtailment data is deleted to generate the second wind turbine dataset; The process of generating a third wind turbine dataset based on the second wind turbine dataset, using an anomaly detection algorithm to identify high-density accumulation data, deleting the high-density accumulation data, and then performing the following steps: Based on the second wind turbine dataset, the negative outlier factor value is determined using an anomaly detection algorithm. The negative outlier value is compared with a third preset threshold. When the negative outlier value is greater than the third preset threshold, the wind turbine operating data corresponding to the negative outlier value is the high-density area accumulation data. Delete the high-density area data accumulation in the second wind turbine dataset to generate the third wind turbine dataset; The step of determining the negative outlier value based on the second wind turbine dataset using an anomaly detection algorithm includes: Based on the second wind turbine dataset, wind turbine operation data at multiple times are determined, and the wind turbine operation data at multiple times are used as multiple data points; Based on the multiple data points, determine the neighborhood data of the data points and the Euclidean distance between the data points; The reachability distance between data points is determined based on the neighborhood data of the data points and the Euclidean distance between the data points. The local reachability density of the data points is determined based on the neighborhood data of the data points and the reachability distance between the data points. The negative outlier factor value is calculated based on the neighborhood data of the data points, the Euclidean distance between the data points, and the local reachability density of the data points; the formula for calculating the negative outlier factor value is as follows: in, Represents the negative outlier value. Representing data points The Distance neighborhood data, Representing data points The Locally achievable density, Representing data points The Locally achievable density, Representing data points The Distance neighborhood data, Representing data points and data points The Euclidean distance between them.

2. The method according to claim 1, characterized in that, The step of preprocessing the wind turbine operating data within the continuous time period to generate the first wind turbine dataset includes: Based on the wind turbine operating data within the continuous time period, determine the wind turbine output power and wind speed at different times. Abnormal data is determined based on the output power of the wind turbine; Deleting abnormal data from the wind turbine operation data within the continuous time period generates the first wind turbine dataset.

3. The method according to claim 2, characterized in that, The determination of abnormal data based on the output power of the wind turbine includes: Calculate the mean output power and standard deviation of the output power based on the output power of the wind turbine generator; A standard score is calculated based on the wind turbine's output power, the mean of the output power, and the standard deviation of the output power. The standard score is compared with a first preset threshold. When the standard score is greater than the first preset threshold, the wind turbine output power and the wind turbine wind speed are regarded as abnormal data.

4. The method according to claim 1, characterized in that, The step of determining the first identification function coefficients and the first offset based on the first wind turbine dataset includes: The wind turbine output power in the first wind turbine dataset is input into the first objective optimization function, and the first objective optimization function is solved to generate the first identification function coefficients and the first offset.

5. The method according to claim 4, characterized in that, The step of inputting the wind turbine output power from the first wind turbine dataset into the first objective optimization function, solving the first objective optimization function, and generating the first identification function coefficients and the first offset includes: The first objective optimization function is solved using a sequence minimum optimization algorithm to generate the coefficients of the first identification function and the first offset.

6. The method according to claim 1, characterized in that, The process of determining the accumulated wind curtailment data based on the third wind turbine dataset, deleting the accumulated wind curtailment data, and generating cleaned wind turbine operating data includes: The second identification function coefficients and the second offset are determined based on the third wind turbine dataset. Obtain the radial basis function, and determine the value of the redundant data anomaly identification function based on the wind turbine output power, the coefficients of the second identification function, the second offset, and the radial basis function in the third wind turbine dataset; The value of the redundant data anomaly identification function is compared with a fourth preset threshold, and a redundant data anomaly identification boundary is constructed based on the comparison result. Based on the constructed redundant data anomaly identification boundary, the abandoned wind power data is deleted, and the cleaned wind turbine operation data is generated.

7. The method according to claim 6, characterized in that, The determination of the second identification function coefficients and the second offset based on the third wind turbine dataset includes: The wind turbine output power in the third wind turbine dataset is input into the second objective optimization function, and the second objective optimization function is solved to generate the second identification function coefficients and the second offset.

8. A device for cleaning abnormal wind turbine data under multiple wind curtailment scenarios, characterized in that, The device includes: The preprocessing module is used to collect wind turbine operation data over a continuous time period, preprocess the wind turbine operation data over the continuous time period, and generate a first wind turbine dataset. The multiple wind curtailment data deletion module is used to construct a multiple wind curtailment data identification boundary based on the first wind turbine dataset, and delete multiple wind curtailment data based on the multiple wind curtailment data identification boundary to generate a second wind turbine dataset; The high-density area accumulation data deletion module is used to determine the high-density area accumulation data based on the second wind turbine dataset using an anomaly detection algorithm, delete the high-density area accumulation data, and generate a third wind turbine dataset. The wind curtailment data deletion module is used to determine the wind curtailment data based on the third wind turbine dataset, delete the wind curtailment data, and generate cleaned wind turbine operation data. The multi-curtailment wind data deletion module includes: The third determination submodule is used to determine the first identification function coefficients and the first offset based on the first wind turbine dataset; The first calculation submodule is used to obtain the kernel function and calculate the multi-curtailment data identification function value based on the wind turbine output power, first identification function coefficients, first offset, and kernel function in the first wind turbine dataset. The calculation formula for the multi-curtailment data identification function value is shown below: in, This represents the identification function value for multiple wind curtailment data. Represents the coefficients of the first identification function. Represents the kernel function. Represents the kernel function variable. Indicates the first wind turbine in the dataset. The output power of the wind turbine at any given time. Indicates the first offset; The first comparison submodule is used to compare the multi-curtailed wind data identification function value with the second preset threshold, and construct the multi-curtailed wind data identification boundary based on the comparison result. The second deletion submodule is used to identify boundaries based on multiple wind curtailment data, delete multiple wind curtailment data, and generate a second wind turbine dataset. The high-density area data deletion module includes: The fourth determination submodule is used to determine the negative outlier value based on the second wind turbine dataset using an anomaly detection algorithm; The second comparison submodule is used to compare the negative outlier value with the third preset threshold. When the negative outlier value is greater than the third preset threshold, the wind turbine operating data corresponding to the negative outlier value is high-density area accumulation data. The third deletion submodule is used to delete the high-density accumulated data in the second wind turbine dataset and generate the third wind turbine dataset. The fourth determination submodule includes: The first determining unit is used to determine the wind turbine operation data at multiple times based on the second wind turbine dataset, and to use the wind turbine operation data at multiple times as multiple data points. The second determining unit is used to determine the neighborhood data of a data point and the Euclidean distance between data points based on multiple data points; The third determining unit is used to determine the reachability distance between data points based on the neighborhood data of the data points and the Euclidean distance between the data points; The fourth determining unit is used to determine the local reachability density of a data point based on its neighborhood data and the reachability distance between data points. The fifth determining unit is used to calculate the negative outlier value based on the neighborhood data of the data points, the Euclidean distance between the data points, and the local reachability density of the data points; the formula for calculating the negative outlier value is as follows: in, Represents the negative outlier value. Representing data points The Distance neighborhood data, Representing data points The Locally achievable density, Representing data points The Locally achievable density, Representing data points The Distance neighborhood data, Representing data points and data points The Euclidean distance between them.

9. A computer device, characterized in that, include: The system includes a memory and a processor, which are interconnected. The memory stores computer instructions, and the processor executes the computer instructions to perform the wind turbine abnormal data cleaning method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing the computer to execute the wind turbine abnormal data cleaning method according to any one of claims 1 to 7.

Images