Pitch Bearing Detection Method, Device and Wind Turbine

Abstract

The present invention provides a pitch bearing detection method, device and wind turbine unit for a wind turbine, including: obtaining a time series data set of the wind turbine unit under the condition of shutdown and pitch retraction, constructing a high-dimensional correlation variable model for each target variable, and determining the target variable detection values of each grid unit in the high-dimensional correlation variable model according to the time series data set; determining the target variable reference values of each target variable detection value in the high-dimensional decision model; and determining the detection result of the pitch bearing of the wind turbine unit to be measured according to each target variable detection value and the corresponding target variable reference value. By performing multivariate non-linear fitting on the correlation variables of each target variable, the present invention forms a high-dimensional decision model, and then comprehensively uses the target variable reference values in the high-dimensional decision model and the actually collected target variable detection values for fault detection, taking into account the comprehensive influence of multiple target variables and all correlation variables of each target variable, with higher accuracy and stronger interpretability based on the data characteristics of mechanism analysis.

Description

Technical Field

[0001] The present invention relates to the technical field of new power energy, and particularly to a method and device for detecting a pitch bearing of a wind turbine and a wind turbine unit. Background Art

[0002] The pitch bearing is a slewing mechanism in the pitch drive device of a wind turbine. After a period of operation, due to reasons such as wear, corrosion, grease aging, fatigue spalling of the raceway surface, and cage fracture, abnormal noises and jams will occur during its rotation. In severe cases, it will lead to faults such as excessive torque of the pitch motor, overheating, and large deviation of the blade angle between shafts, which will cause the wind turbine to shut down. Therefore, in order to reduce the power generation loss and operation and maintenance costs caused by fault shutdowns, it is necessary to identify and monitor abnormal rotations and perform predictive interventions at an early stage.

[0003] At present, there is no direct monitoring method for the operating state of the pitch bearing in the industry. Generally, it is to analyze the relevant operating variables recorded by the Supervisory Control And Data Acquisition (SCADA) system to indirectly infer whether the pitch bearing may have abnormal rotation.

[0004] At present, the algorithms for identifying abnormal pitch bearings of wind turbine generators are generally divided into two categories: one is to establish a detection model based on time-frequency domain analysis by extracting the vibration characteristics of the bearing rotation. This method cannot identify the abnormally high current caused by jams and increased resistance in the pitch rotation mechanism, so the detection range is limited and the credibility of the detection results is poor. The other is to establish a data-driven detection model by combining mechanism analysis and data mining methods, mainly to establish a supervised detection classification model based on normal samples and fault samples. However, the disadvantage of this method is that most of the units are in a normal state, and it is difficult to implement due to the small number of fault samples. Summary of the Invention

[0005] The present invention provides a method and device for detecting a pitch bearing of a wind turbine and a wind turbine unit, which are used to solve the defect that it is difficult to implement in the prior art when constructing and pre-training a detection classification model because a large number of fault samples need to be collected, and can achieve rapid and accurate detection of the pitch bearing of the wind turbine unit.

[0006] In a first aspect, the present invention provides a method for detecting a pitch bearing of a wind turbine, including:

[0007] Obtain the time-series data of the wind turbine to be measured under the condition of shutting down and retracting the blades, and construct a time-series data set of each target variable for detecting the pitch bearing;

[0008] In the case where any target variable involves at least one associated variable, a high-dimensional associated variable model related to the any target variable is constructed, and according to the time-series data set of the any target variable, the target variable detection values of each grid cell in the high-dimensional associated variable model are determined;

[0009] A high-dimensional decision model related to the any target variable is called, and the target variable reference value corresponding to each target variable detection value in the high-dimensional decision model is determined;

[0010] According to all target variable detection values and their corresponding target variable reference values, the detection result of the pitch bearing of the to-be-tested wind turbine generator is determined.

[0011] According to a pitch bearing detection method for a wind turbine generator provided by the present invention, the determining the detection result of the pitch bearing of the to-be-tested wind turbine generator according to all target variable detection values and their corresponding target variable reference values includes:

[0012] Determine a first ratio array and a first outlier proportion related to the any target variable; the first ratio array is composed of the ratios between each target variable detection value and its corresponding target variable reference value, and the first outlier proportion is the proportion of target variable detection values greater than the corresponding target variable reference values;

[0013] According to the first ratio arrays and the first outlier proportions related to all target variables, the detection result of the pitch bearing of the to-be-tested wind turbine generator is determined.

[0014] According to a pitch bearing detection method for a wind turbine generator provided by the present invention, it further includes:

[0015] In the case where any target variable does not involve an associated variable, a variable reference threshold related to the any target variable is called;

[0016] According to the time-series data set of the any target variable and the variable reference threshold, a second ratio array and a second outlier proportion related to the any target variable are determined;

[0017] The second ratio array is composed of the ratios between each time-series data and the variable reference threshold, and the second outlier proportion is the proportion of time-series data greater than the variable reference threshold;

[0018] According to the first ratio arrays and the first outlier proportions related to all target variables involving associated variables, and the second ratio arrays and the second outlier proportions related to all target variables not involving associated variables, the detection result of the pitch bearing of the to-be-tested wind turbine generator is re-determined.

[0019] According to a pitch bearing detection method provided by the present invention, when any one of the target variables is the pitch motor current, and the associated variables related to the pitch motor current are the pitch angle and the blade azimuth angle, before calling the high-dimensional decision model related to any one of the target variables, it further includes:

[0020] Obtain the historical time-series data of other wind turbines of the same model as the wind turbine to be tested under the condition of shutdown and pitch retraction, and screen out the first time-series data related to the pitch motor current from the historical time-series data;

[0021] Divide the blade azimuth angle interval into m first equal sub-intervals according to the first step size;

[0022] Divide the pitch angle interval into n second equal sub-intervals; the n second equal sub-intervals are generated by dividing the pitch angle interval into multiple pitch angle sub-intervals and then equally dividing each pitch angle sub-interval according to the corresponding pitch speed of each pitch angle sub-interval;

[0023] The m first equal sub-intervals and the n second equal sub-intervals constitute a high-dimensional decision model related to the pitch motor current, and the high-dimensional decision model is a two-dimensional decision model;

[0024] According to the first time-series data, determine the historical data of the target variable in each grid unit within the two-dimensional decision model.

[0025] According to a pitch bearing detection method provided by the present invention, after determining the historical data of the target variable in each grid unit within the two-dimensional decision model according to the first time-series data, it further includes:

[0026] When it is determined that the number of historical data of the target variable in each grid unit in the two-dimensional decision model is greater than the first threshold, determine the reference value of the target variable of each grid unit according to the historical data of the target variable in each grid unit in the two-dimensional decision model.

[0027] According to a pitch bearing detection method provided by the present invention, the determining the reference value of the target variable of each grid unit according to the historical data of the target variable in each grid unit in the two-dimensional decision model includes:

[0028] After arranging all the historical data of the target variable in any grid unit in ascending order according to the pitch motor current, construct a first box plot from all the historical data of the target variable, and determine the first quartile and the third quartile of the first box plot;

[0029] According to the first quartile and the third quartile, determine the interquartile range of the first box plot;

[0030] Combine the interquartile range and the third quartile to determine the upper limit of the first box plot, and use the upper limit as the reference value of the target variable for any grid cell.

[0031] According to a pitch bearing detection method for a wind turbine provided by the present invention, when any target variable is the blade angle difference between axes and the blade angle difference between axes does not involve associated variables, before calling the variable reference threshold related to any target variable, it further includes:

[0032] Obtain the historical time series data of other wind turbines of the same type as the wind turbine to be tested under the condition of shutdown and blade retraction, and screen out the second time series data related to the blade angle difference between axes from the historical time series data;

[0033] After arranging the second time series data in ascending order of the blade angle difference between axes, construct a second box plot, and determine the first quartile and the third quartile of the second box plot;

[0034] Determine the interquartile range of the second box plot according to the first quartile and the third quartile;

[0035] Combine the interquartile range and the third quartile to determine the upper limit of the second box plot, and use the upper limit as the variable reference threshold related to the blade angle difference between axes.

[0036] According to a pitch bearing detection method for a wind turbine provided by the present invention, when any target variable includes the temperature rise of the pitch motor and the associated variables involved in the temperature rise of the pitch motor are the pitch angle, the hub temperature, and the outdoor temperature, before calling the high-dimensional decision model related to any target variable, it further includes:

[0037] Obtain the historical time series data of other wind turbines of the same type as the wind turbine to be tested under the condition of shutdown and blade retraction, and screen out the third time series data related to the temperature rise of the pitch motor from the historical time series data;

[0038] Construct a high-dimensional decision model related to the temperature rise of the pitch motor, and the high-dimensional decision model is a three-dimensional decision model;

[0039] Determine the reference value of the target variable for each grid cell located in the three-dimensional decision model according to the third time series data.

[0040] According to a pitch bearing detection method for a wind turbine provided by the present invention, determine the detection result of the pitch bearing of the wind turbine to be tested according to the first ratio array and the first outlier ratio related to all target variables, including:

[0041] If each ratio in the first ratio array of each of the target variables is greater than a second threshold value associated with each of the target variables, the comprehensive outlier of all the target variables is greater than a third threshold value associated with each of the target variables and greater than the historical comprehensive outlier corresponding to all the target variables in the previous detection period, then it is determined that the pitch bearing of the wind turbine to be tested has a serious fault; the comprehensive outlier is determined according to the detection weight of each of the target variables and the proportion of the first outlier of each target variable.

[0042] If each ratio in the first ratio array of each of the target variables is greater than a fourth threshold value associated with each of the target variables and not greater than the second threshold value, and the comprehensive outlier of the target variable is greater than a fifth threshold value associated with each of the target variables and not greater than the third threshold value, then it is determined that the pitch bearing of the wind turbine to be tested has a minor fault.

[0043] Otherwise, it is determined that the pitch bearing of the wind turbine to be tested is normal.

[0044] In a second aspect, the present invention further provides a pitch bearing detection device for a wind turbine, including:

[0045] A data calling unit, configured to obtain the time-series data of the wind turbine to be tested in the shutdown and pitch-retracting working condition, and construct a time-series data set of each target variable for pitch bearing detection.

[0046] A data arrangement unit, configured to construct a high-dimensional associated variable model related to any one of the target variables when any one of the target variables involves at least one associated variable, and determine the target variable detection values of each grid unit in the high-dimensional associated variable model according to the time-series data set of any one of the target variables.

[0047] A decision model calling unit, which calls a high-dimensional decision model related to any one of the target variables and determines the target variable reference values corresponding to each of the target variable detection values in the high-dimensional decision model.

[0048] A data comparison unit, configured to determine the detection result of the pitch bearing of the wind turbine to be tested according to all the target variable detection values and their corresponding target variable reference values.

[0049] In a third aspect, the present invention provides a wind turbine, including a wind turbine body, and a detection processor is arranged in the wind turbine body; it further includes a memory and a program or instruction stored on the memory and executable on the detection processor, and when the program or instruction is executed by the detection processor, it executes the steps of any one of the above-mentioned pitch bearing detection methods for the wind turbine.

[0050] Fourthly, the present invention provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, the steps of any one of the above-mentioned pitch bearing detection methods for wind turbine generators are implemented.

[0051] Fifthly, the present invention further provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of any one of the above-mentioned pitch bearing detection methods for wind turbine generators are implemented.

[0052] The pitch bearing detection method, device, and wind turbine unit provided by the present invention form a high-dimensional decision model through multivariate non-linear fitting of the associated variables of each target variable, and then comprehensively utilize the reference value of the target variable in the high-dimensional decision model and the detected value of the target variable actually collected for fault detection. Considering the comprehensive influence of multiple target variables and all associated variables of each target variable, the accuracy is higher, and based on the data characteristics of mechanism analysis, the interpretability is stronger. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0054] Figure 1 is one of the schematic flowcharts of the pitch bearing detection method for wind turbine generators provided by the present invention;

[0055] Figure 2 is another schematic flowchart of the pitch bearing detection method for wind turbine generators provided by the present invention;

[0056] Figure 3 is a schematic diagram of a high-dimensional decision model provided by the present invention;

[0057] Figure 4 is a schematic diagram of the distribution of the detected value of the target variable and the two-dimensional decision model under normal detection results provided by the present invention;

[0058] Figure 5 is a schematic diagram of the distribution of the detected value of the target variable and the two-dimensional decision model under abnormal detection results provided by the present invention;

[0059] Figure 6 is a schematic structural diagram of the pitch bearing detection device for wind turbine generators provided by the present invention;

[0060] Figure 7It is a schematic structural diagram of the electronic device provided by the present invention. Detailed implementation manners

[0061] To make the objectives, technical solutions and advantages of the present invention clearer, the technical solutions in the present invention will be clearly and completely described below with reference to the accompanying drawings in the present invention. Apparently, the described embodiments are some but not all of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present invention without making creative efforts shall fall within the protection scope of the present invention.

[0062] It should be noted that in the description of the embodiments of the present invention, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements but also includes other elements not explicitly listed, or further includes elements inherent to such process, method, article or device. Without further limitations, the element defined by the statement "including one..." does not exclude the existence of additional identical elements in the process, method, article or device including the element. The orientation or positional relationship indicated by terms such as "upper", "lower", etc. is based on the orientation or positional relationship shown in the accompanying drawings, and is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus cannot be construed as a limitation of the present invention. Unless otherwise clearly defined and limited, the terms "mount", "connect" and "couple" shall be understood in a broad sense. For example, it may be a fixed connection, a detachable connection or an integral connection; it may be a mechanical connection or an electrical connection; it may be directly connected or indirectly connected through an intermediate medium, and it may be the internal communication of two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0063] The terms "first", "second", etc. in the description and claims of the present application are used to distinguish similar objects and are not used to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments of the present application can be implemented in an order other than those illustrated or described herein, and the objects distinguished by "first", "second", etc. are generally of the same category and do not limit the number of objects. For example, the first object may be one or multiple. In addition, "and / or" in the description and claims means at least one of the connected objects, and the character " / " generally means an "or" relationship between the associated objects before and after.

[0064] Figure 1It is one of the schematic flowcharts of the pitch bearing detection method provided by the present invention. As Figure 1 shown, it includes but is not limited to the following steps:

[0065] Step 101: Obtain the time-series data of the wind turbine to be tested under the condition of shutdown and pitch retraction, and construct a time-series data set of each target variable for pitch bearing detection.

[0066] Among them, the shutdown and pitch retraction condition refers to one of the control logics such as normal pitch retraction or emergency pitch retraction, and it is required to satisfy that the pitch speed (abbreviation: rs) is constant (denoted as RS), or the pitch speed is constant within each pitch angle sub-interval after the pitch angle is segmented.

[0067] For example: According to the pitch speed, the entire pitch angle interval PitchAngelInterval n ∈{[θ1,θ2),[θ2,θ3),...,[θ n ,θ n+1 )}. Among them, [θ1,θ2), [θ2,θ3)…[θ n ,θ n+1 ) are respectively different pitch angle sub-intervals, corresponding to different pitch speeds, that is, rs∈{RS1,RS2,...,RS n}. When the pitch angle θ∈PitchAngelInterval k , the corresponding pitch speed rs = RS k , where n≥1.

[0068] For example, when the pitch angle interval of a certain wind turbine is 0°-87.5°, the shutdown and pitch retraction condition refers to that during the execution of the pitch retraction action, the pitch speed is constant throughout the process of changing the pitch angle from 0° to 87.5°.

[0069] For another example, when the pitch angle interval of a certain wind turbine is 0°-87.5°, the pitch speed remains unchanged within the three pitch angle sub-intervals of [0°, 30°), [30°, 50°), and [50°, 87.5°), which is also considered to be in the shutdown and pitch retraction condition.

[0070] In the present invention, by analyzing the time-series data related to the pitch movement, especially the time-series data under the shutdown and pitch retraction condition, selected from all the operation data retrieved from the SCADA system related to the wind turbine, the characteristics of large-angle rotation of the pitch bearing and more obvious data anomalies under the shutdown and pitch retraction condition are fully utilized. Screening specific condition data can reduce interference, so the detection result is more accurate.

[0071] Among them, the time-series data set refers to a set of operation data of the to-be-detected wind turbine collected within a certain duration according to a preset sampling frequency. For example, the time-series data collected at each sampling moment is one or more of the following: wind speed, pitch motor temperature, pitch motor temperature rise, pitch motor current, pitch motor power, pitch motor speed, pitch angle, blade azimuth angle, hub temperature, outdoor temperature, etc.

[0072] The above operation data collected at each sampling moment can be used as a set of time-series data, which constitutes a data set in terms of time sequence, and is called a time-series data set. For example:

[0073] Retrieve the second-level operation data (such as the sampling frequency is 1 second / time) of the to-be-detected wind turbine operating in the most recent week from the SCADA system, and construct a time-series data set as the to-be-detected data set for detecting the current operation state of the to-be-detected wind turbine.

[0074] Step 102: In the case where any target variable involves at least one associated variable, construct a high-dimensional associated variable model related to the any target variable, and determine the target variable detection values of each grid cell in the high-dimensional associated variable model according to the time-series data set of the any target variable.

[0075] Figure 2 This is the second flow schematic diagram of the pitch bearing detection method for wind turbines provided by the present invention. As Figure 2 shown,

[0076] Assume that in the retrieved time-series data set, the target variables used for pitch bearing detection include pitch motor current, pitch motor temperature rise, etc., and for each target variable, determine the associated variables that have a direct impact or are associated with it. For example: for the target variable pitch motor current, its associated variables can be determined as at least one or more of pitch angle, outdoor temperature, blade azimuth angle, etc.; for the target variable pitch motor temperature rise, its associated variables can be determined as at least one or more of pitch angle, hub temperature, outdoor temperature, etc.

[0077] Taking the pitch motor current as the target variable and its associated features as outdoor temperature and blade azimuth angle (hereinafter referred to as: blade angle) as an example, the method for constructing a high-dimensional associated variable model is introduced as follows:

[0078] First, the target variable collected at each sampling moment and all its corresponding associated variables can be screened out from the time-series data set, and other invalid data can be cleared for constructing a high-dimensional associated variable model related to the target variable.

[0079] Taking the blade angle and the outdoor temperature as the X-axis and Y-axis respectively, the filtered time-series data set is binned and discretized to construct a high-dimensional correlation variable model (at this time, the high-dimensional correlation variable model is a two-dimensional correlation variable model), that is, each time-series data in the time-series data set can be represented by a data point in the high-dimensional correlation variable model.

[0080] Optionally, a grid cell in the high-dimensional correlation variable model refers to a space cell formed by the blade angle, the outdoor temperature in their respective preset intervals on the coordinate axes, and the variable pitch motor current from negative infinity to positive infinity. For example, assuming that the blade angle interval is in units of 10° and the outdoor temperature interval is in units of 10°, and the high-dimensional correlation variable model is divided into grid cells, then each grid cell is a space cell composed of (10°, 10°).

[0081] After constructing the high-dimensional correlation variable model using the time-series data set, the target variable detection values of each grid cell therein can be determined.

[0082] Among them, the target variable detection value refers to the value of the variable pitch motor current corresponding to each grid cell. In the case where there are multiple data points in any grid cell, the average value or the maximum or minimum value of the variable pitch motor currents corresponding to all data points can be used as the target variable detection value; in the case where there is only 1 data point in any grid cell, the variable pitch motor current corresponding to this data point can be directly used as the target variable detection value; in the case where there are no data points in any grid cell, the target variable detection value of this grid cell is regarded as 0.

[0083] The present invention provides a variable pitch bearing detection method that filters multiple groups of correlation variables of each target variable for grouped statistics based on statistical principles, and uses statistical values for multivariate non-linear fitting to form a high-dimensional correlation variable model, taking into account the data characteristics of mechanism analysis and having stronger interpretability.

[0084] Step 103: Invoke the high-dimensional decision model related to any one of the target variables, and determine the target variable reference value corresponding to each target variable detection value in the high-dimensional decision model.

[0085] It should be noted that the present invention constructs a historical time-series data set by collecting the historical time-series data of other normal state wind turbines of the same model as the wind turbine to be measured under the condition of shutdown and blade retraction, and constructs a high-dimensional decision model by using a method similar to the method of constructing a high-dimensional correlation variable model provided in the above embodiment.

[0086] In the high-dimensional decision model of the present invention, each grid cell in the high-dimensional decision model corresponds one-to-one in space. After obtaining the target variable detection values of each grid cell in the high-dimensional decision model related to each target variable, the target variable reference values corresponding to the grid cells corresponding to each target variable detection value in the high-dimensional decision model are determined.

[0087] Step 104: Determine the detection result of the pitch bearing of the wind turbine to be tested according to all target variable detection values and their corresponding target variable reference values.

[0088] Generally speaking, each target variable detection value and its corresponding target variable reference value can be used as a comparison data group. In this way, by comparing the magnitudes between the target variable detection value and the target variable reference value in each comparison data group, and counting the number of comparison data groups where the target variable detection value is greater than the target variable reference value as the number of outliers.

[0089] Optionally, determine the detection result of the pitch bearing of the wind turbine to be tested according to the ratio between the number of outliers and the total number of comparison data groups.

[0090] For example, when the ratio is greater than 0.1, determine that the detection result is a serious fault; if the ratio is not greater than 0.1, determine that the detection result is normal.

[0091] The pitch bearing detection method for wind turbines provided by the present invention forms a high-dimensional decision model through multivariate non-linear fitting of the associated variables of each target variable, and then comprehensively uses the target variable reference values in the high-dimensional decision model and the actually collected target variable detection values for fault detection. It takes into account the comprehensive influence of multiple target variables and all associated variables of each target variable, has higher accuracy, and stronger interpretability based on the data characteristics of mechanism analysis.

[0092] Based on the content of the above embodiments, as an alternative embodiment, the above determining the detection result of the pitch bearing of the wind turbine to be tested according to all target variable detection values and their corresponding target variable reference values can be implemented in the following manner:

[0093] Determine the first ratio array and the first outlier proportion related to any one of the target variables; the first ratio array is composed of the ratios between each target variable detection value and its corresponding target variable reference value, and the first outlier proportion is the proportion of the target variable detection value being greater than the corresponding target variable reference value;

[0094] Determine the detection result of the pitch bearing of the wind turbine to be tested according to the first ratio arrays and the first outlier proportions related to all target variables.

[0095] Among them, determine the first ratio array related to any one of the target variables and the proportion of the first outliers; the first ratio array is composed of the ratio between each target variable detection value and the corresponding target variable reference value, and the proportion of the first outliers is the proportion of the target variable detection value greater than the corresponding target variable reference value.

[0096] Obtain the ratio between the target variable detection value corresponding to each grid cell and the corresponding target variable reference value, so that the first ratio array can be obtained.

[0097] Suppose the number of preset intervals of the blade angle and the outdoor temperature on their respective axes are m and n respectively, then the first ratio array is a two-dimensional array of m*n.

[0098] Furthermore, it is also possible to compare the size between the target variable detection value corresponding to each grid cell and the corresponding target variable reference value, and count the number k (k is called the first outlier) of the target variable detection value greater than the corresponding target variable reference value, and take k / N as the proportion of the first outliers of the target variable. Among them, N is the total amount of all target variable detection values.

[0099] Furthermore, according to the first ratio array and the proportion of the first outliers related to all target variables, determine the detection result of the pitch bearing of the wind turbine to be tested.

[0100] Suppose there are multiple target variables involving at least one associated variable, such as including pitch motor current, pitch motor temperature rise, etc., then the first ratio array and the proportion of the first outliers related to each target variable can be obtained to digitally and intuitively express the operating state of the pitch bearing.

[0101] For example, if each ratio in the first ratio array related to all target variables is greater than a threshold value, and the proportion of the first outliers related to all target variables is greater than another threshold value, it can be considered that there is a serious fault in the pitch bearing of the wind turbine to be tested.

[0102] The pitch bearing detection method for a wind turbine provided by the present invention forms a high-dimensional decision model by performing multivariate non-linear fitting on the associated variables of each target variable, and then comprehensively uses the target variable reference value in the high-dimensional decision model and the actually collected target variable detection value for fault detection, considering the comprehensive influence of multiple target variables and all associated variables of each target variable, with higher accuracy and stronger interpretability based on the data characteristics of mechanism analysis.

[0103] Based on the content of the above embodiments, as an alternative embodiment, the pitch bearing detection method for a wind turbine provided by the present invention may further include:

[0104] In the case where any target variable does not involve associated variables, a variable reference threshold associated with the any target variable is called.

[0105] According to the time series data set of the any target variable and the variable reference threshold, a second ratio array and a second outlier proportion associated with the any target variable are determined.

[0106] The second ratio array is composed of the ratios between each time series data and the variable reference threshold, and the second outlier proportion is the proportion of time series data greater than the variable reference threshold.

[0107] According to the first ratio array and the first outlier proportion associated with the target variables involving associated variables, and the second ratio array and the second outlier proportion associated with the target variables not involving associated variables, the detection result of the pitch bearing of the to-be-tested wind turbine generator is re-determined.

[0108] For example, it can be determined that the target variables include not only the pitch motor current and the pitch motor temperature rise involving associated variables, but also the inter-axis blade angle difference not involving associated variables. Then, for the target variables not involving associated variables, the variable reference threshold can be directly determined according to all the selected historical data (in this case, all the selected historical data is only the inter-axis blade angle difference collected at different sampling moments), for example, calculating the arithmetic mean of all the inter-axis blade angle differences as the variable reference threshold.

[0109] Further, from the time series data set of the to-be-tested wind turbine generator under the shutdown and pitch retraction condition, the inter-axis blade angle differences collected at each sampling moment are selected, and the ratios between them and the above variable reference threshold are calculated respectively to construct a second ratio array.

[0110] Then, the number of all inter-axis blade angle differences greater than the variable reference threshold (i.e., the second outliers are determined), and the ratio between this number and the total amount of the inter-axis blade angle differences is calculated as the second outlier proportion.

[0111] Finally, in the case where all the ratios in the first ratio array and the second ratio array are greater than a certain threshold, and both the first outlier proportion and the second outlier proportion are greater than another threshold, it can be determined that the pitch bearing of the to-be-tested wind turbine generator has a serious fault.

[0112] The pitch bearing detection method for a wind turbine generator provided by the present invention, when performing pitch bearing fault detection, considers both the target variables involving associated variables and the target variables not involving associated variables, and adopts different methods to extract relevant parameters for different types of associated variables, which can effectively improve the accuracy and reliability of fault detection.

[0113] Based on the content of the above embodiments, as an alternative embodiment, when any one of the target variables is the pitch motor current, and the associated variables involved in the pitch motor current are the pitch angle and the blade azimuth angle, before invoking the high-dimensional decision model related to any one of the target variables, it further includes:

[0114] Obtain the historical time-series data of other wind turbines of the same model as the wind turbine to be measured under the condition of shutdown and pitch retraction, and screen out the first time-series data related to the pitch motor current from the historical time-series data;

[0115] Divide the blade azimuth angle interval into m first equal sub-intervals according to the first step size;

[0116] Divide the pitch angle interval into n second equal sub-intervals; the n second equal sub-intervals are generated by dividing the pitch angle interval into multiple pitch angle sub-intervals and then equally dividing each pitch angle sub-interval according to the corresponding pitch speed of each pitch angle sub-interval;

[0117] The m first equal sub-intervals and the n second equal sub-intervals form a high-dimensional decision model related to the pitch motor current, and the high-dimensional decision model is a two-dimensional decision model;

[0118] According to the first time-series data, determine the historical data of the target variable of each grid cell in the two-dimensional decision model.

[0119] The present invention provides a method for constructing a high-dimensional decision model related to a certain target variable (pitch motor current) according to the historical time-series data of other wind turbines of the same model as the wind turbine to be measured under the condition of shutdown and pitch retraction.

[0120] It should be noted that the above method is only used to illustrate the general idea of constructing a high-dimensional decision model. When the target variable is different or the associated variables of the target variable are different, the constructed high-dimensional decision model is also different.

[0121] Figure 3 This is a schematic diagram of a high-dimensional decision model provided by the present invention. As Figure 3 shown, the target variable corresponding to the constructed high-dimensional decision model is the pitch motor current, and the selected associated variables are the blade angle and the outdoor temperature.

[0122] Next, still taking the target variable corresponding to the high-dimensional decision model as the pitch motor current, and the selected associated variables as the blade angle (i.e., the blade azimuth angle) and the pitch angle as an example, the construction steps of the high-dimensional decision model can include but are not limited to:

[0123] First step, screen the time-series data under the specific pitch-retracting program condition in the data of the same model as the wind turbine to be measured, filter other data, and after cleaning the invalid data, obtain the first time-series data related to the pitch motor current. Among them, the time-series data related to the pitch motor current mainly includes the continuous data of the blade angle in time series and the continuous data of the pitch angle in time series.

[0124] Second step, divide the blade angle interval [0, 360) into m equal parts with the first step length α step on average to obtain m first equal sub-intervals, then

[0125] Divide the pitch angle θ into bin k , θ k+1 second equal sub-intervals within different pitch angle intervals [θ θ_k , where then the pitch angle θ can be divided into second equal sub-intervals in total.

[0126] Construct a high-dimensional decision model related to the pitch motor current from the X-axis where the above blade angle interval is located, the Y-axis where the pitch angle is located, and the Z-axis where the pitch motor current is located.

[0127] Third step, the method provided in the above embodiment can be used to calculate the reference value of the target variable related to each grid cell in the high-dimensional decision model respectively, which will not be elaborated here.

[0128] Finally, the historical data of the target variable related to the grid cell can be recorded into the two-dimensional array limits_array[m][n] of this model, and the number of records is m * n.

[0129] As an alternative embodiment, the present invention provides a method for accurately calculating the reference value of the target variable related to each grid cell.

[0130] Specifically, determining the reference value of the target variable of each grid cell according to the historical data of the target variable in each grid cell in the two-dimensional decision model includes:

[0131] After arranging all the historical data of the target variable in any grid cell in ascending order of the pitch motor current, construct a first box plot from all the historical data of the target variable and determine the first quartile and the third quartile of the first box plot;

[0132] Determine the interquartile range of the first box plot according to the first quartile and the third quartile;

[0133] Combine the interquartile range and the third quartile to determine the upper limit of the first box plot, and use the upper limit as the target variable reference value for any grid cell.

[0134] In the present invention, the upper limit value of all historical data of the target variable in each grid cell is statistically obtained by constructing a box plot as the target variable reference value for each grid cell. For any grid cell:

[0135] First, the pitch motor current in each grid cell is taken as a sample respectively, and after arranging them in ascending order, they are divided into four equal parts.

[0136] Among them: The first quartile Q1 is the current value of the 25th percentile of the sample from small to large; the second quartile Q2 is the current value of the 50th percentile of the sample from small to large; the third quartile Q3 is the current value of the 75th percentile of the sample from small to large;

[0137] Furthermore, the interquartile range IQR of the box plot can be calculated as IQR = Q3 - Q1; finally, the upper limit of the box plot can be calculated as Q3 + 1.5IQR, which is used as the target variable reference value for any grid cell.

[0138] It should be noted that when determining the target variable detection values of each grid cell in the high-dimensional correlation variable model, the above box plot method is also used for calculation, which will not be elaborated here.

[0139] The present invention provides a method for determining the target variable reference value of each grid cell by using the box plot method, which can accurately characterize the characteristics of all historical data of the target variable in the grid cell.

[0140] In the above embodiment, a method for constructing a corresponding high-dimensional decision model for the target variable involving correlation variables is introduced. As an alternative embodiment, the present invention also provides a method for constructing a high-dimensional correlation variable model for the target variable not involving correlation variables, and determining the target variable detection values of each grid cell in the high-dimensional decision model.

[0141] Assume that any target variable is the pitch angle difference between shafts. In the case that the pitch angle difference between shafts does not involve correlation variables, before calling the variable reference threshold related to any target variable, the following steps may further be included:

[0142] Obtain the historical time series data of other wind turbines of the same type as the wind turbine to be measured under the condition of shutdown and blade retraction, and screen out the second time series data related to the pitch angle difference between shafts from the historical time series data;

[0143] After arranging the second timing data in ascending order of the blade angle difference between axes, a second box plot is constructed, and the first quartile and the third quartile of the second box plot are determined;

[0144] According to the first quartile and the third quartile, the interquartile range of the second box plot is determined;

[0145] Combining the interquartile range and the third quartile, the upper limit of the second box plot is determined, and the upper limit is used as the variable reference threshold related to the blade angle difference between axes.

[0146] Similar to the box plot method provided in the above embodiment, after screening out the second timing data related to the blade angle difference between axes, for any grid cell, a corresponding second box plot is constructed.

[0147] After sorting the second timing data in ascending order of the blade angle difference between axes and dividing it into four equal parts, the first quartile Q1, the second quartile Q2, and the third quartile Q3 of the corresponding second box plot are determined in sequence to calculate the interquartile range of the second box plot, and then calculate the value of the upper limit of the second box plot as the variable reference threshold of any grid cell.

[0148] Then, using the above method, the variable reference thresholds of each of the other grid cells are calculated respectively.

[0149] As an alternative embodiment, as Figure 2 shown, after determining the historical data of the target variable of each grid cell located in the two-dimensional decision model according to the first timing data, it further includes:

[0150] When it is determined that the number of historical data of the target variable in each grid cell in the two-dimensional decision model is greater than the first threshold size1, according to the historical data of the target variable in each grid cell in the two-dimensional decision model, the reference value of the target variable of each grid cell is determined. If the number of historical data of the target variable in each grid cell is not all greater than the first threshold, it means that the number of timing data of the wind turbine under the shutdown and blade retraction condition that can be collected by the present invention is insufficient, then the variable pitch bearing detection process of this round is ended to re-collect the timing data of the wind turbine to be measured under the shutdown and blade retraction condition in the next cycle until the number of historical data of the target variable in each grid cell in the obtained timing data set is greater than the first threshold, and then the subsequent state detection and analysis process can be continued.

[0151] Based on mechanism analysis, the present invention obtains two sets of target variables and associated variables, namely the pitch motor current and the pitch angle, outdoor temperature, impeller azimuth angle, etc., and the pitch motor temperature rise and the pitch angle, hub temperature, temperature difference between outdoor temperature, etc. And the reference values of the target variables corresponding to the target variables and the high-dimensional decision model / variable reference threshold and the upper limit of the over-limit data volume are set. And based on whether it exceeds the limit, it is judged whether the pitch bearing is abnormal, and multiple sets of associated variables are selected for multivariate non-linear fitting, considering the comprehensive influence of multiple associated variables, with higher accuracy and stronger interpretability based on the data characteristics of mechanism analysis.

[0152] In addition, the present invention uses big data technology to perform statistical modeling on wind turbine units of the same model, reducing the deviation caused by other non-critical factors such as operating environment, manufacturing and assembly processes of different units, and having strong robustness.

[0153] Based on the content of the above embodiments, as an alternative embodiment, when any one of the target variables includes the pitch motor temperature rise, and the associated variables involved in the pitch motor temperature rise are the pitch angle, hub temperature, and outdoor temperature, before calling the high-dimensional decision model related to any one of the target variables, it further includes:

[0154] Obtain the historical time-series data of other wind turbine units of the same model as the wind turbine unit to be tested under the condition of shutdown and pitch retraction, and screen out the third time-series data related to the pitch motor temperature rise from the historical time-series data;

[0155] Construct a high-dimensional decision model related to the pitch motor temperature rise, and the high-dimensional decision model is a three-dimensional decision model;

[0156] According to the third time-series data, determine the reference values of the target variables of each grid unit in the three-dimensional decision model.

[0157] In the above embodiments, the method steps for constructing a high-dimensional decision model have been introduced when any one of the target variables is an associated variable involved, and its corresponding associated variables are two. In this embodiment, the method steps for constructing a high-dimensional associated variable and its corresponding high-dimensional decision model will be further introduced when any one of the target variables is an associated variable involved, and its corresponding associated variables are three.

[0158] Taking the construction of a high-dimensional decision model with the target variable being the pitch motor temperature rise, and its associated variables including the pitch angle, hub temperature, and outdoor temperature as an example, the description is as follows:

[0159] First, retrieve the historical time-series data (second-level data) of the generator sets of the same model as the target unit, and screen out the third time-series data related to the pitch motor temperature rise from it;

[0160] Then, according to the pitch angle, hub temperature, and outdoor temperature of the pitch motor temperature rise respectively, the pitch motor temperature rise data is spatially binned and separated, and the reference values of the target variables for each grid cell in the high-dimensional decision model related to the pitch motor temperature rise are calculated.

[0161] Finally, the reference values of the target variables for each of the above grid cells are recorded in a three-dimensional array of this model type, and the number of records is a*b*c; where a, b, and c are the number of partitions of the pitch angle, hub temperature, and outdoor temperature in their respective dimensions.

[0162] It should be noted that in the case where the number of associated variables involved in the target variable is larger, the dimension of the constructed high-dimensional decision model is higher. Since the correlation relationship between each associated variable and the target variable is fully considered, its detection accuracy will be effectively improved, but at the same time, the solution complexity will be correspondingly increased.

[0163] For the pitch bearing detection method of the wind turbine provided by the present invention, on the basis of weighing the detection accuracy and the calculation complexity, generally the number of associated variables involved in each target variable is set to two or less, but this is not regarded as a specific limitation on the protection scope of the present invention.

[0164] Based on the content of the above embodiments, as an alternative embodiment, in combination with Figure 2 As shown, determining the detection result of the pitch bearing of the to-be-tested wind turbine according to the first ratio array and the first outlier ratio related to all target variables specifically includes:

[0165] If each ratio in the first ratio array of each target variable is greater than the second threshold, the comprehensive outlier value of all target variables is greater than the third threshold and greater than the historical comprehensive outlier value corresponding to all target variables in the previous detection period, then it is determined that the pitch bearing of the to-be-tested wind turbine has a serious fault; the comprehensive outlier value is determined according to the detection weight of each target variable and the first outlier ratio of each target variable.

[0166] If each ratio in the first ratio array of each target variable is greater than the fourth threshold and not greater than the second threshold, and the comprehensive outlier value of the target variable is greater than the fifth threshold and not greater than the third threshold, then it is determined that the pitch bearing of the to-be-tested wind turbine has a minor fault.

[0167] Otherwise, it is determined that the pitch bearing of the to-be-tested wind turbine is normal.

[0168] Figure 4 It is a schematic diagram of the distribution of the target variable detection value and the two-dimensional decision model under normal detection results provided by the present invention. Figure 5 It is a schematic diagram of the distribution of the target variable detection value and the two-dimensional decision model under abnormal detection results provided by the present invention, asFigure 4 as shown Figure 5 As shown, the two-dimensional decision model uses the pitch motor current as the target variable, the blade azimuth angle (i.e., the blade angle) and the outdoor temperature as its associated variables, and is constructed based on the time series data set collected within a preset time period. Figure 4 and Figure 5 respectively show the distribution diagrams of the target variable detection values and the two-dimensional decision model under different detection results of the two.

[0169] Specifically, after obtaining the high-dimensional associated variable model of the wind turbine to be measured, the distribution of all target variable detection values in each grid cell can be obtained.

[0170] Assuming that the target variables include the pitch motor current, the pitch motor temperature rise, and the blade angle difference between shafts, for each grid cell in the high-dimensional associated variable model related to any target variable, the target variable detection value is compared with the corresponding target variable reference value, and the ratio between the target variable detection value and the target variable reference value is calculated. Therefore, finally, the first ratio array d1 related to the pitch motor current, the second ratio d2 related to the blade angle difference between shafts, and the second ratio array d3 related to the pitch motor temperature rise can be obtained.

[0171] For any high-dimensional associated variable model, when the target variable detection value is greater than the target variable reference value, the corresponding data point is located above the corresponding decision surface of the two-dimensional decision model (i.e., the corresponding data point is abnormal). Thus, by counting the number of data points located above the corresponding decision surface, the proportion of abnormal values can be calculated in combination with the total number of data points. Therefore, finally, the first abnormal value proportion r1 related to the pitch motor current, the second abnormal value proportion r2 related to the blade angle difference between shafts, and the third abnormal value proportion r3 related to the pitch motor temperature rise can be obtained.

[0172] As an alternative embodiment, in the fault analysis stage, calculate whether all the ratios in the ratio arrays d1, d2, or d3 exceed the corresponding second thresholds D1, D2, or D3; and whether the proportions r1, r2, r3 of the abnormal values exceeding the decision surface (or the variable reference threshold) all exceed the corresponding third thresholds R1, R2, or R3; and whether the proportions r1, r2, r3 of the abnormal values are all increased compared with the proportions r1', r2', and r3' of the abnormal values recorded in the previous process detection operation cycle. If so, it indicates that a serious fault has occurred in the pitch bearing of the wind turbine to be measured.

[0173] Further, if not all of the ratios within the ratio arrays d1, d2, or d3 exceed the corresponding second thresholds D1, D2, or D3, but all exceed the fourth threshold (Fourth threshold = Second threshold * w1); or although the outlier ratios r1, r2, r3 do not all exceed the corresponding third thresholds R1, R2, or R3, but all exceed the fifth threshold (Fifth threshold = Third threshold * w2), where both w1 and w2 are greater than 0 and less than 1. At this time, it can be determined that a fault has occurred in the pitch bearing of the wind turbine to be tested.

[0174] In addition, in cases other than the above, it is initially determined that the wind turbine has not failed. If there are some abnormal data, the on-site staff can be notified to conduct a patrol inspection for further confirmation.

[0175] As another alternative embodiment, in the above fault analysis stage, after calculating the outlier ratios r1, r2, r3, a comprehensive outlier value can also be calculated according to the weights of the pitch motor current, the pitch motor temperature rise, and the inter-axis blade angle difference when determining the pitch bearing fault.

[0176] Assume that the weights of the pitch motor current, the pitch motor temperature rise, and the inter-axis blade angle difference are k1, k2, and k3 respectively, then the comprehensive outlier value Q = k1 * r1 + k2 * r2 + k3 * r3.

[0177] Finally, the size relationship between the comprehensive outlier value Q and the historical comprehensive outlier value Q' detected in the previous detection cycle can be combined to determine whether a fault has occurred in the motor, and the severity of the fault can also be determined according to the difference between the two.

[0178] The pitch bearing detection method for a wind turbine provided by the present invention not only performs single-threshold judgment, but also considers the development trend of related indicators (such as the outlier ratio), effectively improving the detection accuracy.

[0179] Figure 6 It is a schematic structural diagram of a pitch bearing detection device for a wind turbine provided by the present invention, as Figure 6 shown, mainly including a data calling unit 61, a data arranging unit 62, a decision model calling unit 63, and a data comparing unit 64, where:

[0180] The data calling unit 61 is mainly used to obtain the time-series data of the wind turbine to be tested under the condition of shutdown and blade retraction, and construct a time-series data set of each target variable for pitch bearing detection;

[0181] The data arranging unit 62 is mainly used to construct a high-dimensional associated variable model related to any one of the target variables when any one of the target variables involves at least one associated variable, and determine the target variable detection values of each grid unit in the high-dimensional associated variable model according to the time-series data set of any one of the target variables.

[0182] The decision model calling unit 63 mainly calls the high-dimensional decision model related to any of the target variables and determines the target variable reference values corresponding to the detection values of each target variable in the high-dimensional decision model;

[0183] The data comparison unit 64 is mainly used to determine the detection result of the pitch bearing of the wind turbine to be tested according to all the detection values of the target variables and their corresponding target variable reference values.

[0184] Optionally, in the data comparison unit 64, the determination of the detection result is mainly achieved by performing the following steps:

[0185] First, determine the first ratio array and the first outlier proportion related to any of the target variables.

[0186] Among them, the first ratio array is composed of the ratios between the detection values of each target variable and their corresponding target variable reference values; the first outlier proportion is the proportion of the detection values of the target variables that are greater than their corresponding target variable reference values.

[0187] Then, the detection result of the pitch bearing of the wind turbine to be tested can be determined according to the first ratio array and the first outlier proportion related to all the target variables.

[0188] Assume that there are multiple target variables involving at least one associated variable, such as pitch motor current, pitch motor temperature rise, etc., then the first ratio array and the first outlier proportion related to each target variable can be obtained to digitally and intuitively express the operating state of the pitch bearing.

[0189] For example, if each ratio in the first ratio array related to all the target variables is greater than a threshold value, and the first outlier proportion related to all the target variables is greater than another threshold value, it can be considered that there is a serious fault in the pitch bearing of the wind turbine to be tested.

[0190] It should be noted that the pitch bearing detection device for wind turbines provided by the embodiments of the present invention can execute the pitch bearing detection method for wind turbines described in any of the above embodiments during specific operation, and this embodiment will not be elaborated here.

[0191] The pitch bearing detection device for wind turbines provided by the present invention forms a high-dimensional decision model by performing multivariate non-linear fitting on the associated variables of each target variable, and then comprehensively uses the target variable reference values in the high-dimensional decision model and the actually collected target variable detection values for fault detection, considering the comprehensive influence of multiple target variables and all the associated variables of each target variable, with higher accuracy and stronger interpretability based on the data characteristics of mechanism analysis.

[0192] The present invention further provides a wind power unit, including a wind turbine generator set body, wherein a detection processor is arranged in the wind turbine generator set body; it further includes a memory and a program or instruction stored on the memory and executable on the detection processor, and when the program or instruction is executed by the detection processor, it executes the steps of the motor set pitch bearing detection method provided in any of the above embodiments.

[0193] Figure 7 is a schematic structural diagram of the electronic device provided by the present invention, as Figure 7 shown, the electronic device may include: a processor 710, a communication interface 720, a memory 730, and a communication bus 740. Among them, the processor 710, the communication interface 720, and the memory 730 complete mutual communication through the communication bus 740. The processor 710 can call the logical instructions in the memory 730 to execute the wind turbine generator set pitch bearing detection method, and the method includes: obtaining the time series data of the wind turbine generator set to be tested under the condition of shutdown and blade retraction, and constructing a time series data set of each target variable for pitch bearing detection; in the case that any target variable involves at least one associated variable, constructing a high-dimensional associated variable model related to the any target variable, and determining the target variable detection values of each grid unit in the high-dimensional associated variable model according to the time series data set of the any target variable; calling a high-dimensional decision model related to the any target variable, and determining the target variable reference values corresponding to each target variable detection value in the high-dimensional decision model; determining the detection result of the pitch bearing of the wind turbine generator set to be tested according to all the target variable detection values and their corresponding target variable reference values.

[0194] In addition, when the logical instructions in the above-mentioned memory 730 can be implemented in the form of a software functional unit and sold or used as an independent product, they can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. And the foregoing storage medium includes: various media such as a USB flash drive, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk, or an optical disc that can store program codes.

[0195] On the other hand, the present invention also provides a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium. The computer program includes program instructions. When the program instructions are executed by a computer, the computer can execute the pitch bearing detection method provided by each of the above methods. The method includes: obtaining time-series data of a wind turbine unit to be tested under the condition of shutdown and pitch retraction, and constructing a time-series data set of each target variable for pitch bearing detection; when any target variable involves at least one associated variable, constructing a high-dimensional associated variable model related to the any target variable, and determining the target variable detection values of each grid cell in the high-dimensional associated variable model according to the time-series data set of the any target variable; calling a high-dimensional decision model related to the any target variable, and determining the target variable reference values corresponding to each target variable detection value in the high-dimensional decision model; and determining the detection result of the pitch bearing of the wind turbine unit to be tested according to all target variable detection values and their corresponding target variable reference values.

[0196] On another aspect, the present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it is configured to execute the pitch bearing detection method provided by each of the above embodiments. The method includes: obtaining time-series data of a wind turbine unit to be tested under the condition of shutdown and pitch retraction, and constructing a time-series data set of each target variable for pitch bearing detection; when any target variable involves at least one associated variable, constructing a high-dimensional associated variable model related to the any target variable, and determining the target variable detection values of each grid cell in the high-dimensional associated variable model according to the time-series data set of the any target variable; calling a high-dimensional decision model related to the any target variable, and determining the target variable reference values corresponding to each target variable detection value in the high-dimensional decision model; and determining the detection result of the pitch bearing of the wind turbine unit to be tested according to all target variable detection values and their corresponding target variable reference values.

[0197] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place, or may be distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. A person of ordinary skill in the art can understand and implement it without creative efforts.

[0198] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A detection method for a pitch bearing of a wind turbine, characterized in that, Including: Obtain the time-series data of the wind turbine to be tested under the condition of shutdown and blade retraction, and construct a time-series data set of each target variable for pitch bearing detection; When any target variable involves at least one associated variable, construct a high-dimensional associated variable model related to the any target variable, and determine the target variable detection values of each grid cell in the high-dimensional associated variable model according to the time-series data set of the any target variable; Call the high-dimensional decision model related to the any target variable, and determine the target variable reference value corresponding to each target variable detection value in the high-dimensional decision model; Determine the detection result of the pitch bearing of the wind turbine to be tested according to all target variable detection values and their corresponding target variable reference values; The method for constructing the high-dimensional associated variable model includes: screening out the target variable collected at each sampling moment and all its corresponding associated variables from the time-series data set; using the associated variables as one dimension respectively, performing binning discretization on the screened time-series data set to construct the high-dimensional associated variable model. Each time-series data in the time-series data set is represented by a data point in the high-dimensional associated variable model. A grid cell in the high-dimensional associated variable model refers to a spatial unit formed by the associated variables with preset intervals on their respective coordinate axes and the target variable from negative infinity to positive infinity; When the any target variable is the pitch motor current and the associated variables involved in the pitch motor current are the pitch angle and the blade azimuth angle, before calling the high-dimensional decision model related to the any target variable, it further includes: obtaining the historical time-series data of other wind turbines of the same type as the wind turbine to be tested under the condition of shutdown and blade retraction, and screening out the first time-series data related to the pitch motor current from the historical time-series data; dividing the blade azimuth angle interval into m first equal sub-intervals according to the first step size; dividing the pitch angle interval into n second equal sub-intervals; the n second equal sub-intervals are generated by dividing the pitch angle interval into multiple pitch angle sub-intervals and then performing equidistant division on each pitch angle sub-interval according to the corresponding pitch speed of each pitch angle sub-interval; the m first equal sub-intervals and the n second equal sub-intervals constitute the high-dimensional decision model related to the pitch motor current, and the high-dimensional decision model is a two-dimensional decision model; determining the target variable historical data of each grid cell located in the two-dimensional decision model according to the first time-series data; When any of the target variables includes the temperature rise of the pitch motor, and the associated variables involved in the temperature rise of the pitch motor are the pitch angle, the hub temperature, and the outdoor temperature, before calling the high-dimensional decision model related to any of the target variables, it further includes: obtaining the historical time-series data of other wind turbines of the same type as the wind turbine to be tested under the condition of shutdown and pitch retraction, and screening out the third time-series data related to the temperature rise of the pitch motor from the historical time-series data; constructing a high-dimensional decision model related to the temperature rise of the pitch motor, and the high-dimensional decision model is a three-dimensional decision model; determining the reference values of the target variables for each grid cell within the three-dimensional decision model according to the third time-series data.

2. The pitch bearing detection method for a wind turbine unit according to claim 1, characterized in that Determining the detection result of the pitch bearing of the wind turbine to be tested according to all the target variable detection values and their corresponding target variable reference values includes: Determining a first ratio array and a first outlier proportion related to any of the target variables; the first ratio array is composed of the ratios between each target variable detection value and its corresponding target variable reference value, and the first outlier proportion is the proportion of target variable detection values greater than their corresponding target variable reference values; Determining the detection result of the pitch bearing of the wind turbine to be tested according to the first ratio arrays and the first outlier proportions related to all the target variables.

3. The pitch bearing detection method for a wind turbine unit according to claim 2, wherein It further includes: When any of the target variables does not involve associated variables, calling the variable reference threshold related to any of the target variables; Determining a second ratio array and a second outlier proportion related to any of the target variables according to the time-series data set of any of the target variables and the variable reference threshold; The second ratio array is composed of the ratios between each time-series data and the variable reference threshold, and the second outlier proportion is the proportion of time-series data greater than the variable reference threshold; Re-determining the detection result of the pitch bearing of the wind turbine to be tested according to the first ratio arrays and the first outlier proportions related to all the target variables involving associated variables, and the second ratio arrays and the second outlier proportions related to all the target variables not involving associated variables.

4. The pitch bearing detection method for a wind turbine unit according to claim 1, wherein After determining the historical data of the target variables for each grid cell within the two-dimensional decision model according to the first time-series data, it further includes: When it is determined that the number of historical data of the target variables in each grid cell within the two-dimensional decision model is greater than the first threshold, determining the reference values of the target variables for each grid cell according to the historical data of the target variables in each grid cell within the two-dimensional decision model.

5. The pitch bearing detection method for a wind turbine according to claim 4, wherein Determining the reference values of the target variables for each grid cell according to the historical data of the target variables in each grid cell within the two-dimensional decision model includes: After arranging all the historical data of the target variables in any grid cell in ascending order of the pitch motor current, constructing a first box plot from all the historical data of the target variables and determining the first quartile and the third quartile of the first box plot; Determining the interquartile range of the first box plot according to the first quartile and the third quartile; Combine the interquartile range and the third quartile to determine the upper limit of the first box plot, and use the upper limit as the target variable reference value for any grid cell.

6. The pitch bearing detection method for a wind turbine according to claim 3, characterized in that When any of the target variables is the blade angle difference between shafts and the blade angle difference between shafts does not involve associated variables, before calling the variable reference threshold related to any of the target variables, it further includes: Obtain the historical time series data of other wind turbines of the same model as the wind turbine to be tested under the condition of shutdown and blade retraction, and screen out the second time series data related to the blade angle difference between shafts from the historical time series data; After arranging the second time series data in ascending order of the blade angle difference between shafts, construct a second box plot, and determine the first quartile and the third quartile of the second box plot; Determine the interquartile range of the second box plot according to the first quartile and the third quartile; Combine the interquartile range and the third quartile to determine the upper limit of the second box plot, and use the upper limit as the variable reference threshold related to the blade angle difference between shafts.

7. The pitch bearing detection method for a wind turbine unit according to claim 2, wherein Determine the detection result of the pitch bearing of the wind turbine to be tested according to the first ratio array and the first outlier ratio related to all target variables, including: If each ratio in the first ratio array of each target variable is greater than the second threshold related to each target variable, the comprehensive outlier of all target variables is greater than the third threshold related to each target variable and greater than the historical comprehensive outlier corresponding to all target variables in the previous detection period, it is determined that the pitch bearing of the wind turbine to be tested has a serious fault; the comprehensive outlier is determined according to the detection weight of each target variable and the first outlier ratio of each target variable; If each ratio in the first ratio array of each target variable is greater than the fourth threshold related to each target variable and not greater than the second threshold, and the comprehensive outlier of the target variable is greater than the fifth threshold related to each target variable and not greater than the third threshold, it is determined that the pitch bearing of the wind turbine to be tested has a minor fault; Otherwise, it is determined that the pitch bearing of the wind turbine to be tested is normal.

8. A pitch bearing detection device for a wind turbine, characterized in that, It includes: A data calling unit for obtaining the time series data of the wind turbine to be tested under the condition of shutdown and blade retraction, and constructing a time series data set of each target variable for pitch bearing detection; A data arrangement unit for constructing a high-dimensional associated variable model related to any of the target variables when any of the target variables involves at least one associated variable, and determining the target variable detection values of each grid cell in the high-dimensional associated variable model according to the time series data set of any of the target variables; A decision model calling unit for calling the high-dimensional decision model related to any of the target variables, and determining the target variable reference value corresponding to each target variable detection value in the high-dimensional decision model; A data comparison unit for determining the detection result of the pitch bearing of the wind turbine to be tested according to all target variable detection values and their corresponding target variable reference values; The method for constructing the high-dimensional correlation variable model includes: screening out the target variable collected at each sampling moment and all its corresponding correlation variables from the time-series data set; using the correlation variables as dimensions respectively, performing binning discretization on the screened time-series data set to construct the high-dimensional correlation variable model. Each time-series data in the time-series data set is represented by a data point in the high-dimensional correlation variable model. A grid cell in the high-dimensional correlation variable model refers to a spatial cell formed by the correlation variables with preset intervals on their respective coordinate axes and the target variable ranging from negative infinity to positive infinity. When any one of the target variables is the pitch motor current and the correlation variables involved in the pitch motor current are the pitch angle and the blade azimuth angle, before calling the high-dimensional decision model related to any one of the target variables, it further includes: obtaining the historical time-series data of other wind turbines of the same type as the wind turbine to be measured under the condition of shutdown and pitch retraction, and screening out the first time-series data related to the pitch motor current from the historical time-series data; dividing the blade azimuth angle interval into m first equal sub-intervals according to the first step size; dividing the pitch angle interval into n second equal sub-intervals; the n second equal sub-intervals are generated by dividing the pitch angle interval into multiple pitch angle sub-intervals and then performing equidistant division on each pitch angle sub-interval according to the corresponding pitch speed of each pitch angle sub-interval; the m first equal sub-intervals and the n second equal sub-intervals constitute the high-dimensional decision model related to the pitch motor current, and the high-dimensional decision model is a two-dimensional decision model; determining the historical data of the target variable in each grid cell located in the two-dimensional decision model according to the first time-series data. When any one of the target variables includes the pitch motor temperature rise and the correlation variables involved in the pitch motor temperature rise are the pitch angle, the hub temperature, and the outdoor temperature, before calling the high-dimensional decision model related to any one of the target variables, it further includes: obtaining the historical time-series data of other wind turbines of the same type as the wind turbine to be measured under the condition of shutdown and pitch retraction, and screening out the third time-series data related to the pitch motor temperature rise from the historical time-series data; constructing the high-dimensional decision model related to the pitch motor temperature rise, and the high-dimensional decision model is a three-dimensional decision model; determining the reference value of the target variable in each grid cell located in the three-dimensional decision model according to the third time-series data.

9. A wind turbine unit, characterized in that, It includes a wind turbine body, and a detection processor is provided in the wind turbine body; it further includes a memory and a program or instruction stored on the memory and executable on the detection processor. When the program or instruction is executed by the detection processor, it executes the steps of the motor group pitch bearing detection method according to any one of claims 1 to 7.

Images