Fault detection method, device and equipment for wind generating set and storage medium

Abstract

The invention relates to the technical field of wind power generation detection, in particular to a fault detection method, device and equipment of a wind generating set and a storage medium. The method comprises the steps that a current monitoring time sequence sample set of the wind generating set is acquired; detecting the current monitoring data fragment according to the parameter standard feature set with the corresponding length, and determining whether the current monitoring data fragment is abnormal or not; obtaining target state data and target environment data according to the abnormal current monitoring data fragment; obtaining a fault prediction result; by means of the mode, the parameter standard feature sets which represent the normal change rule of the state parameters and have different lengths are mined based on the historical data, and the current monitoring data fragments which are not matched with the parameter standard feature sets can be analyzed and obtained in time; according to the invention, fault detection is carried out on various state parameters in the timestamp range corresponding to the current monitoring data fragment, the monitoring parameter change abnormity is detected before the state parameters exceed the corresponding threshold values, fault prediction can be carried out in advance, and the fault detection efficiency is improved.

Description

Technical Field

[0001] The present application relates to the field of wind power generation detection technology, and in particular to a fault detection method, device, equipment and storage medium for a wind power generator set. Background Art

[0002] A wind turbine is a device that realizes two energy conversion processes: from wind energy to mechanical energy and from mechanical energy to electrical energy. The wind wheel system realizes the energy conversion from wind energy to mechanical energy, while the generator and control system realize the energy conversion process from mechanical energy to electrical energy. Due to differences in operating environment, equipment operation, maintenance and factory time, wind turbines generally have different faults during operation.

[0003] In the background technology, various state parameters of a wind turbine are monitored in real time. When a state parameter exceeds a corresponding state parameter threshold, it is determined that the wind turbine has an operating fault corresponding to the state parameter. When the fault is found, component damage may have occurred. The above-mentioned fault detection method does not fully mine the normal operating characteristics of the wind turbine, and cannot detect abnormal changes in the monitoring parameters before the state parameter exceeds the corresponding threshold, which is not conducive to improving the efficiency of fault detection. Summary of the invention

[0004] The embodiments of the present application provide a fault detection method, device, equipment and storage medium for a wind turbine generator set, which can improve the efficiency of fault detection.

[0005] In a first aspect, an embodiment of the present application provides a fault detection method for a wind turbine generator set, comprising:

[0006] Acquire current monitoring data of at least one monitoring parameter of the wind turbine generator set, and acquire multiple current monitoring time series sample sets of the monitoring parameter according to the current monitoring data, wherein each of the current monitoring time series sample sets includes multiple current monitoring data segments of the same length, and the lengths of the current monitoring data segments in different current monitoring time series sample sets are different;

[0007] The current monitoring data segment in the current monitoring time series sample set is detected according to the parameter standard feature set of the corresponding length of the monitoring parameter to determine whether the current monitoring data segment is abnormal; the parameter standard feature set is used to characterize the normal change law of the monitoring parameter, and the parameter standard feature set includes multiple similar historical data segments of corresponding lengths;

[0008] Acquire the target state data and target environment data of the wind turbine generator set according to the abnormal current monitoring data segment; the timestamp range corresponding to the target state data and the timestamp range corresponding to the environment data are respectively consistent with the timestamp range corresponding to the abnormal current monitoring data segment;

[0009] A fault prediction result of the wind turbine generator set is obtained according to the target state data and the target environment data.

[0010] Optionally, acquiring a fault prediction result of the wind turbine generator set according to the target state data and the target environment data includes:

[0011] The target state data and the target environment data are input into a pre-trained fault prediction model to obtain the fault prediction result output by the fault prediction model, wherein the fault prediction result includes at least one fault type and the fault occurrence probability corresponding to the fault type; wherein the fault detection model is trained based on historical state data and historical environment data within a first time range before the fault occurs.

[0012] Optionally, the step of training the fault detection model includes:

[0013] Acquire a training set, the training set comprising a plurality of data vectors and fault classification labels corresponding to the data vectors, the data vectors comprising a plurality of data sequences, the plurality of data sequences comprising a time series of at least one state parameter and a time series of at least one environmental indicator;

[0014] Inputting the data vector into a fault detection model to obtain a fault prediction result output by the fault detection model, wherein the fault prediction result includes a predicted fault type and a fault occurrence probability corresponding to the preset fault type;

[0015] The training loss of the fault detection model is calculated according to the fault prediction result, and the parameters of the fault detection model are adjusted according to the training loss until the fault detection model converges.

[0016] Optionally, before acquiring current monitoring data of at least one monitoring parameter of the wind turbine generator set, the method includes:

[0017] Acquire historical data of at least one monitoring parameter of a wind turbine generator set, and acquire multiple historical time series sample sets of the monitoring parameter according to the historical data, wherein each of the historical time series sample sets includes multiple historical data segments of the same length, and the lengths of the historical data segments in different historical time series sample sets are different;

[0018] A parameter standard feature set of corresponding length is constructed according to each of the historical time series sample sets.

[0019] Optionally, constructing the parameter standard feature set of corresponding length according to each of the historical time series sample sets includes:

[0020] Generate a historical data segment curve according to the time series data fitting of each of the historical data segments;

[0021] Obtaining the similarity between every two historical data segment curves;

[0022] When the similarity is greater than or equal to a first preset similarity threshold, the two historical data segments corresponding to the similarity are put into the parameter standard feature set.

[0023] Optionally, the detecting the current monitoring data segment in the current monitoring time series sample set according to the parameter standard feature set of corresponding length to determine whether the current monitoring data segment is abnormal includes:

[0024] Generate a historical data segment curve according to the time series data fitting of each historical data segment in the parameter standard feature set;

[0025] Generate a current monitoring data segment curve according to the time series data fitting of each current monitoring data segment in the current monitoring time series sample set;

[0026] Obtaining the similarity between the current monitoring data segment curve and each of the historical data segment curves, and obtaining a similarity average of the current monitoring data segment according to the multiple similarities;

[0027] When the similarity mean is less than a second preset similarity threshold, it is determined that the corresponding current monitoring data segment is abnormal.

[0028] Optionally, constructing the parameter standard feature set of corresponding length according to each of the historical time series sample sets includes:

[0029] According to the Euclidean distance between every two of the historical data fragments, a similarity portrait of each of the historical time series sample sets and a portrait index of the similarity portrait are obtained; the similarity portrait includes a minimum distance value between each of the historical data fragments and each of the remaining historical data fragments in the historical time series sample set, the minimum distance value is the distance between the historical data fragment and the most similar historical data fragment, and the portrait index is used to record the timestamp range information of the most similar historical data fragment of each of the historical data fragments in the historical data;

[0030] A minimum distance value that is less than or equal to a preset distance threshold is selected in the similarity portrait, and the corresponding historical data segment is obtained according to the selected minimum distance value and the portrait index. The parameter standard feature set of the corresponding length is constructed according to the obtained historical data segment, and the preset distance threshold is determined according to the minimum distance value of each historical data segment.

[0031] In a second aspect, an embodiment of the present application provides a fault detection device for a wind turbine generator set, comprising:

[0032] A first data module is used to obtain current monitoring data of at least one monitoring parameter of the wind turbine generator set, and obtain multiple current monitoring time series sample sets of the monitoring parameter according to the current monitoring data, each of the current monitoring time series sample sets includes multiple current monitoring data segments of the same length, and the lengths of the current monitoring data segments in different current monitoring time series sample sets are different;

[0033] An abnormal warning module is used to detect the current monitoring data segment in the current monitoring time series sample set according to the parameter standard feature set of the corresponding length of the monitoring parameter to determine whether the current monitoring data segment is abnormal; the parameter standard feature set is used to characterize the normal change law of the monitoring parameter, and the parameter standard feature set includes multiple similar historical data segments of corresponding lengths;

[0034] A second data module is used to obtain target state data and target environment data of the wind turbine generator set according to the abnormal current monitoring data segment; the timestamp range corresponding to the target state data and the timestamp range corresponding to the environment data are respectively consistent with the timestamp range corresponding to the abnormal current monitoring data segment;

[0035] A fault detection module is used to obtain a fault prediction result of the wind turbine generator set according to the target state data and the target environment data.

[0036] In a third aspect, an embodiment of the present application provides a computer device, the device comprising a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other via the communication bus;

[0037] Memory, used to store computer programs;

[0038] The processor is used to implement the steps of the above-mentioned fault detection method for wind turbine generator set when executing the program stored in the memory.

[0039] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the above-mentioned wind turbine generator set fault detection method are implemented.

[0040] The embodiments of the present application provide a fault detection method, apparatus, device and storage medium for a wind turbine generator set, which obtain current monitoring data of at least one monitoring parameter of the wind turbine generator set, and obtain multiple current monitoring time series sample sets of the monitoring parameter according to the current monitoring data; detect the current monitoring data segment in the current monitoring time series sample set according to the parameter standard feature set of the corresponding length of the monitoring parameter to determine whether the current monitoring data segment is abnormal; obtain the target state data and target environment data of the wind turbine generator set according to the abnormal current monitoring data segment; obtain the fault prediction result of the wind turbine generator set according to the target state data and the target environment data; through the above-mentioned method, based on historical data mining of parameter standard feature sets of different lengths that characterize the normal change law of state parameters, it is possible to timely analyze and obtain the current monitoring data segments that do not match the parameter standard feature sets, and perform fault detection on various state parameters in the timestamp range corresponding to the current monitoring data segment, thereby realizing the detection of abnormal changes in monitoring parameters before the state parameters exceed the corresponding threshold values, and being able to perform fault prediction in advance, thereby improving the efficiency of fault detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0042] Figure 1 A schematic diagram of an application environment of an embodiment of the present application.

[0043] Figure 2 Schematic diagram of a flow chart of a fault detection method for a wind turbine generator set according to an embodiment of the present application.

[0044] Figure 3 Schematic diagram of the structure of a fault prediction model in a fault detection method for a wind turbine generator set according to an embodiment of the present application.

[0045] Figure 4 Schematic diagram of the structure of a fault detection device for a wind turbine generator set according to an embodiment of the present application.

[0046] Figure 5 A schematic diagram of the structure of a computer device according to an embodiment of the present application.

[0047] Figure 6 A schematic diagram of the structure of a computer-readable storage medium according to an embodiment of the present application. DETAILED DESCRIPTION

[0048] The embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present application, and cannot be understood as limiting the present application.

[0049] In order to enable those skilled in the art to better understand the solutions of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without making creative work are within the scope of protection of the present application.

[0050] In the embodiments of the present application, it should be noted that, in this article, relational terms such as first and second, etc. are merely used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations.

[0051] Moreover, the terms "comprises," "comprising," or any other variation thereof are intended to encompass non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus that includes the element.

[0052] In the description of the embodiments of the present application, words such as "example" or "for example" are used to indicate examples, illustrations or descriptions. Any embodiment or design described as "example" or "for example" in the embodiments of the present application is not to be interpreted as being more preferred or having more advantages than another embodiment or design. The use of words such as "example" or "for example" is intended to present relative concepts in a clear manner.

[0053] In addition, the "plurality" in the embodiments of the present application refers to two or more than two. In view of this, in the embodiments of the present application, "plurality" can also be understood as "at least two". "At least one" can be understood as one or more, for example, one, two or more. For example, including at least one means including one, two or more, and there is no limit on which ones are included. For example, including at least one of A, B and C, then A, B, C, A and B, A and C, B and C, or A, B and C can be included.

[0054] See also Figure 1 , Figure 1 It is a schematic diagram of an application environment provided by an embodiment of the present application, which illustrates a schematic diagram of real-time monitoring of a wind turbine generator set. The schematic diagram includes a wind turbine generator set 100 and a parameter monitoring device 110. The parameter monitoring device 110 includes multiple signal sensors 1101, multiple signal processing modules 1102 arranged in a one-to-one correspondence with the multiple signal sensors 1101, and a fault detection module 1103 respectively connected to the multiple signal processing modules 1102. Some of the signal sensors 1101 are used to collect physical signals of monitoring parameters of the wind turbine generator set 100, and some of the signal sensors 1101 are used to collect physical signals of environmental parameters, such as temperature signals, current signals, voltage signals, vibration signals, etc. The signal processing module 1102 is used to process the collected physical signals to obtain the measurement values ​​of the corresponding parameters, and the fault detection module 1103 is used to perform fault detection according to each monitoring parameter and each environmental parameter.

[0055] An embodiment of the present application provides a fault detection method for a wind turbine generator set. Figure 2 It is a flow chart of a fault detection method for a wind turbine generator set provided in an embodiment of the present application. This specification provides method operation steps such as the embodiment or flow chart, but may include more or fewer operation steps based on conventional or non-creative labor. The order of steps listed in the embodiment is only one way of executing the steps among many orders, and does not represent the only order of execution. When the actual system or server product is executed, it can be executed in the order shown in the embodiment or the drawings or in parallel (for example, in a parallel processor or multi-threaded processing environment). Specifically, Figure 2 As shown, the fault detection method of the wind turbine generator set may include the following steps:

[0056] Step S101: obtaining current monitoring data of at least one monitoring parameter of a wind turbine generator set, and obtaining a plurality of current monitoring time series sample sets of the monitoring parameter according to the current monitoring data.

[0057] Each current monitoring time series sample set includes multiple current monitoring data segments of the same length, the lengths of the current monitoring data segments in different current monitoring time series sample sets are different, and the current monitoring data segments are intercepted from the current monitoring data of the monitoring parameters.

[0058] The monitoring parameters may be, for example, stator winding temperature, stator voltage, stator current, rotor current, generator output power, rotor speed, stator amplitude, bearing amplitude, etc.

[0059] Due to changes in environmental data such as wind speed and temperature, the parameters of wind turbines change. For example, when the wind speed changes, the aerodynamic force acting on the blades of the wind turbine will also change, and the lift and drag on the blades will increase or decrease with the increase or decrease of wind speed; since the blades are connected to the hub, main shaft and other components, these changing forces will be transmitted to the generator through the transmission system. Changes in wind speed will also cause changes in the speed of the wind turbine rotor. According to the power characteristic curve of the wind turbine, when the wind speed increases within a certain range, in order to capture more wind energy, the speed of the wind wheel will increase accordingly. Therefore, changes in wind speed will cause changes in various monitoring parameters. The current monitoring data of a certain monitoring parameter is time series data, which is a time series generated based on the sampling values ​​of the monitoring parameter at each acquisition time within a certain monitoring timestamp range. The sampling values ​​are arranged in chronological order, and the sampling values ​​are generally real values.

[0060] The corresponding length of the monitoring timestamp range may be, for example, 24 hours, 12 hours, 6 hours, 2 hours or 1 hour.

[0061] The current monitoring data segment is a time series data of a predetermined length intercepted from the current monitoring data of the corresponding monitoring parameter, and the length can be 6 hours, 3 hours, 2 hours, 30 minutes, 10 minutes, etc. A plurality of current monitoring time series sample sets are constructed according to the current monitoring data. For example, the number of current monitoring time series sample sets is k, and the k current monitoring time series sample sets are respectively a current monitoring time series sample set with a length of m1, a current monitoring time series sample set with a length of m2, a current monitoring time series sample set with a length of m3, ..., a current monitoring time series sample set with a length of m k Specifically, in the current monitoring time series sample set with a length of m1, the current monitoring data segment is represented by T i，m1 , which represents a continuous segment of length m1 starting from the i-th position of the current monitoring data T. The current monitoring data segment T i，m1 =t i , t i+1 ,…,t i+m1-1 , where t l is the lth sampling value of the current monitoring data T, i≤l≤i+m1-1, 1≤i≤n-m1+1.

[0062] Exemplarily, a current monitoring time series sample set of length m1 can be constructed in the following manner: construct a sliding window of length m1, slide the sliding window on the current monitoring data, and each time a segment within the sliding window is intercepted as a current monitoring data segment, and the sliding step length of the sliding window can be a first preset sliding step length. A current monitoring time series sample set of length m2 can be constructed in the following manner: construct a sliding window of length m2, slide the sliding window on the current monitoring data, and each time a segment within the sliding window is intercepted as a current monitoring data segment, and the sliding step length of the sliding window can be a second preset sliding step length. There is an overlapping portion between the two current monitoring data segments respectively intercepted by two adjacent sliding windows, and the amount of the overlapping portion is related to the sliding step length of the sliding window.

[0063] Step S102: Detect the current monitoring data segment in the current monitoring time series sample set according to the parameter standard feature set of the corresponding length of the monitoring parameter to determine whether the current monitoring data segment is abnormal.

[0064] The parameter standard feature set is used to characterize the normal variation pattern of the monitoring parameters, and the parameter standard feature set includes multiple similar historical data segments of corresponding lengths.

[0065] Among them, for each monitoring parameter, the lengths of multiple current monitoring time series sample sets correspond to the lengths of multiple parameter standard feature sets. For example, the number of parameter standard feature sets is also k, and the k parameter standard feature sets are respectively a parameter standard feature set with a length of m1, a parameter standard feature set with a length of m2, a parameter standard feature set with a length of m3, ..., and a parameter standard feature set with a length of m k The standard feature set of parameters.

[0066] The parameter standard feature set of each length is used to characterize the normal change characteristics of the monitoring parameter. The different historical data segments contained in the parameter standard feature set of each length have the same length and are highly similar to each other, which can represent the normal change law under this length. The parameter standard feature sets of different lengths correspond to the normal change law patterns of the monitoring parameters.

[0067] The parameter standard feature set is constructed based on the historical data of the corresponding monitoring parameters. Similar to the current monitoring data, the historical data of the corresponding monitoring parameters is also time series data, which is a time series generated based on the sampling values ​​of the monitoring parameters at each collection time within a certain historical timestamp range. The sampling values ​​are arranged in chronological order, and the sampling values ​​are generally real values. Compared with the monitoring timestamp range, the length of the historical timestamp range can be longer, for example, the length of the historical timestamp range can be one month, two months, six months, twelve months, twenty-four months, or thirty-six months.

[0068] For each current monitoring data segment in the current monitoring time series sample set of a certain length of the monitoring parameter, similarity calculation is performed between the current monitoring data segment and each historical data segment in the parameter standard feature set of the same length, and based on the similarity calculation result, it is determined whether the normal change pattern represented by the current monitoring data segment and the parameter standard feature set is consistent, so as to detect whether the current monitoring data segment is abnormal.

[0069] The abnormal reflection of the current monitoring data segment indicates that the normal change pattern represented by the standard feature set of parameters of the corresponding length is inconsistent with that of the current monitoring data segment. There is no known necessary correlation between whether the monitoring parameter corresponding to the current monitoring data segment is abnormal or whether it exceeds the parameter threshold. Generally, the abnormality of the current monitoring data segment occurs before the corresponding monitoring parameter becomes abnormal or exceeds the parameter threshold. The abnormality of the current monitoring data segment can play a certain role in early warning. After the abnormality of the current monitoring data segment occurs, the wind turbine generator set can be promptly troubleshooted and the operating environment improved to avoid component damage.

[0070] Step S103: acquiring target state data and target environment data of the wind turbine generator set according to the abnormal current monitoring data segment.

[0071] The timestamp range corresponding to the target state data and the timestamp range corresponding to the environment data are respectively consistent with the timestamp range corresponding to the abnormal current monitoring data segment.

[0072] The target timestamp range is determined according to the timestamp range corresponding to the abnormal current monitoring data segment, and the corresponding monitoring state data segments are respectively intercepted from the time series data of multiple monitoring parameters according to the target timestamp range, and the monitoring state data segments of multiple monitoring parameters form the target state data; the corresponding environmental data segment is intercepted from the time series data of at least one environmental parameter according to the target timestamp range, and the at least one environmental data segment forms the target environmental data. The environmental parameter may be, for example, wind speed, temperature, or air pressure.

[0073] Exemplarily, the target state data includes data segments of generator stator winding temperature, data segments of stator voltage, data segments of stator current, data segments of rotor current, data segments of generator output power, and data segments of rotor speed.

[0074] Step S104: Obtaining a fault prediction result of the wind turbine generator set according to the target state data and the target environment data.

[0075] Among them, there is an abnormal current monitoring data segment within the target timestamp range, and fault prediction is performed on the data of each monitoring parameter and at least one environmental parameter within the target timestamp range, which can improve the accuracy of fault prediction.

[0076] The target state data and the target environment data can be analyzed, and the fault prediction result of the wind turbine generator set can be determined according to the analysis result. Alternatively, a convolutional neural network can be used to perform fault prediction on the target state data and the target environment data to obtain the fault prediction result of the wind turbine generator set.

[0077] In this embodiment, by mining historical data to obtain parameter standard feature sets of different lengths that characterize the normal change patterns of state parameters, it is possible to timely analyze and obtain current monitoring data segments that do not match the parameter standard feature sets, and perform fault detection on various state parameters within the timestamp range corresponding to the current monitoring data segments. This enables detection of abnormal changes in monitoring parameters before the state parameters exceed the corresponding thresholds, enables fault prediction in advance, and improves fault detection efficiency.

[0078] As an implementation mode, in step S104, the target state data and the target environment data are input into a pre-trained fault prediction model to obtain a fault prediction result output by the fault prediction model, wherein the fault prediction result includes at least one fault type and the fault occurrence probability corresponding to the fault type; wherein the fault detection model is trained based on historical state data and historical environment data within a first time range before the fault occurs.

[0079] The fault prediction model may be a BP (back propagation) neural network, and the BP neural network includes an input layer, at least one hidden layer and an output layer.

[0080] like Figure 3 As shown, the input layer includes multiple input nodes, and the target state data includes monitoring state data fragments X1, monitoring state data fragments X2, ..., monitoring state data fragments X n The target environment data includes environment data segment Y1, environment data segment Y2, ..., environment data segment Y m , at least one monitoring state data and at least one environmental data segment correspond to an input node respectively. The input layer transmits the input data to the hidden layer, the hidden layer performs nonlinear transformation on each input data, and transmits the processed data to the output layer, the output layer is used to output the fault prediction result, and the output layer includes an output node.

[0081] The training data used to train the fault detection model is the data of a period of time before the fault occurs. For example, the first period is t a , the timestamp of the fault occurrence t x , then the timestamp ranges corresponding to the historical state data and historical environment data are both (t x -t a -t0~t x-t0), 0≤t0≤t, t is the preset lead time.

[0082] In this embodiment, the data before the fault occurs is used for training, so that the fault detection model can predict the fault before the fault occurs. The fault detection model is used to detect the target state data and target environment data determined by the abnormal current monitoring data segment, which is beneficial to further improve the efficiency of fault detection.

[0083] In some implementations, the step of training the fault detection model includes:

[0084] S201, obtaining a training set, the training set including multiple data vectors and fault classification labels corresponding to the data vectors, the data vectors including multiple data sequences, the multiple data sequences including at least one time series of a state parameter and at least one time series of an environmental indicator.

[0085] The time stamp range (t x -t a -t0~t x -t0), 0≤t0≤t, t is the preset lead time, t x is the timestamp of the fault occurrence, t a For example, the fault classification label may be a mechanical fault or an electrical fault; or, the fault classification label may be a bearing vibration fault, a stator vibration fault, a turn-to-turn short circuit fault, a phase-to-phase short circuit fault, a layer-to-layer short circuit fault, or a stator winding fault.

[0086] S202, inputting the data vector into a fault detection model to obtain a fault prediction result output by the fault detection model, wherein the fault prediction result includes a predicted fault type and a fault occurrence probability corresponding to a preset fault type.

[0087] S203, calculating the training loss of the fault detection model according to the fault prediction result, and adjusting the parameters of the fault detection model according to the training loss until the fault detection model converges.

[0088] The training loss may be, for example, a cross entropy loss.

[0089] In this embodiment, the data before the fault occurs is used for training, so that the fault detection model can learn the changing patterns of the state parameters and environmental parameters monitored before the faults of various types of faults occur, and can predict the faults before they occur.

[0090] As an implementation manner, the following steps are also included before step S101:

[0091] S301, obtaining historical data of at least one monitoring parameter of the wind turbine generator set, and obtaining multiple historical time series sample sets of the monitoring parameter based on the historical data, each historical time series sample set includes multiple historical data segments of the same length, and the lengths of the historical data segments in different historical time series sample sets are different, and the historical data segments are intercepted from the historical data of the monitoring parameter.

[0092] Among them, the number of historical time series sample sets is k, and the k historical time series sample sets are respectively a historical time series sample set with a length of m1, a historical time series sample set with a length of m2, a historical time series sample set with a length of m3, ..., a historical time series sample set with a length of m k Specifically, in the historical time series sample set with a length of m1, the historical data segment is represented as T′ i，m1 , which represents a continuous segment of length m1 starting from the i-th position of the historical data T′, the historical data segment T′ i，m1 =t′ i , t′ i+1 ,…,t′ i+m1-1 , where t′ l is the lth sampling value of the historical data T′, i≤l≤i+m1-1, 1≤i≤n-m1+1.

[0093] Exemplarily, a historical time series sample set with a length of m1 can be constructed in the following manner: construct a sliding window with a length of m1, slide the sliding window on the historical data, and each time a segment within the sliding window is intercepted as a historical data segment, and the sliding step length of the sliding window can be a first preset sliding step length. A historical time series sample set with a length of m2 can be constructed in the following manner: construct a sliding window with a length of m2, slide the sliding window on the historical data, and each time a segment within the sliding window is intercepted as a historical data segment, and the sliding step length of the sliding window can be a second preset sliding step length. There is an overlapping portion between two historical data segments respectively intercepted by two adjacent sliding windows, and the amount of the overlapping portion is related to the sliding step length of the sliding window.

[0094] S302, constructing a parameter standard feature set of corresponding length according to each historical time series sample set, the parameter standard feature set is used to characterize the normal change law of the monitoring parameter, and the parameter standard feature set includes multiple similar historical data fragments.

[0095] Among them, taking the parameter standard feature set with a length of m1 as an example, its construction process is the process of selecting multiple historical data fragments with extremely high similarity in the historical time series sample set with a length of m1.

[0096] In this embodiment, by constructing historical time series sample sets of different lengths and constructing parameter standard feature sets of corresponding lengths based on the historical time series sample sets, it is beneficial to fully mine historical data.

[0097] In some implementations, the similarity between the time series data may be characterized based on the similarity between the fitting curves of the time series data. Specifically, step S302 includes the following steps:

[0098] S401, generating a historical data segment curve according to the time series data fitting of each historical data segment.

[0099] S402, obtaining the similarity between every two historical data segment curves.

[0100] Exemplarily, the calculation formula (1) for the similarity of two historical data segment curves is as follows:

[0101]

[0102] S is the similarity of the two historical data segment curves, X i is the i-th data extracted from the first historical data segment curve, Y i is the i-th data extracted from the second historical data segment curve, N is the number of data extracted from the first historical data segment curve, and N is also the number of data extracted from the first historical data segment curve.

[0103] S403: When the similarity is greater than or equal to a first preset similarity threshold, two historical data segments corresponding to the similarity are added into a parameter standard feature set.

[0104] In the process of putting the selected historical data segments into the parameter standard feature set, there may be at least two identical historical data segments in the parameter standard feature set, and a deduplication operation can be performed subsequently.

[0105] Exemplarily, the first preset similarity threshold may be determined in the following manner: multiple similarities obtained in step S402 are arranged in ascending order, the sum of the first-ranked similarity and the similarity adjustment value is obtained, and the sum of the first-ranked similarity and the similarity adjustment value is used as the first preset similarity threshold. The similarity adjustment value may be determined based on an empirical value.

[0106] In this embodiment, the similarity between the time series data is characterized based on the similarity between the fitting curves of the time series data, which is conducive to improving the calculation accuracy.

[0107] In some implementations, in the process of using the parameter standard feature set to perform anomaly detection on the current monitoring data segment, the similarity between the fitting curves of the time series data can also be used to characterize the similarity between the time series data. Specifically, step S102 includes the following steps:

[0108] S501, generating a historical data segment curve according to the time series data fitting of each historical data segment in the parameter standard feature set.

[0109] S502, generating a current monitoring data segment curve according to the time series data fitting of each current monitoring data segment in the current monitoring time series sample set.

[0110] S503, obtaining the similarity between the current monitoring data segment curve and each historical data segment curve, and obtaining the average similarity of the current monitoring data segment according to the multiple similarities.

[0111] The similarity between the current monitoring data segment curve and the historical data segment curve can be calculated according to formula (1), and the mean similarity of the current monitoring data segment is the quotient of the sum of the similarities between the current monitoring data segment curve and different historical data segment curves and the number of similarities.

[0112] S504: When the similarity mean is less than a second preset similarity threshold, determine that the corresponding current monitoring data segment is abnormal.

[0113] Among them, the larger the similarity mean is, the higher the consistency between the normal change pattern represented by the standard feature set of parameters of the corresponding length; if the similarity mean is too low, it means that there is a lack of consistency between the normal change pattern represented by the standard feature set of parameters of the corresponding length, and the current monitoring data segment does not conform to the normal change pattern of the corresponding length and is an abnormal segment.

[0114] In this embodiment, the similarity between the time series data is characterized based on the similarity between the fitting curves of the time series data, which is conducive to improving the calculation accuracy.

[0115] As an implementation method, a matrix portrait algorithm may be used to mine the normal change rules of historical data. Specifically, step S302 includes the following steps:

[0116] S601, based on the Euclidean distance between every two historical data fragments, obtain the similarity portrait of each historical time series sample set and the portrait index of the similarity portrait; the similarity portrait includes the minimum distance value between each historical data fragment and each of the remaining historical data fragments in the historical time series sample set, the minimum distance value is the distance between the historical data fragment and the most similar historical data fragment, and the portrait index is used to record the timestamp range information of the most similar historical data fragment of each historical data fragment in the historical data.

[0117] First, for each length of the historical time series sample set, the Euclidean distance between every two historical data fragments in the historical time series sample set is obtained.

[0118] The smaller the Euclidean distance between two historical data segments, the higher the similarity. Assume that the length of the historical time series sample set is m, and the number of historical data segments in the historical time series sample set is n. For the historical data segment T i，m , which represents a continuous segment of length m starting from the i-th position of the historical data T.

[0119] Historical data fragment T i，m The distance set from the rest of the historical data segments is D i，m =[d i，1 , d i，2 , …, d i，n-1 ], where d i，j is the historical data segment T i，m With the rest of the historical data fragment T j，m The Euclidean distance between them, where 1≤j≤n-1.

[0120] For example, the Euclidean distance can be calculated according to the following formula (2).

[0121]

[0122] Where m represents the length of the historical data segment, μ i represents the mean of all historical data fragments in the historical time series sample set, δ i Represents the standard deviation of all historical data fragments in the historical time series sample set, μ j represents the mean of all the remaining historical data fragments in the historical time series sample set, δ j Represents the standard deviation of all remaining historical data fragments in the historical time series sample set, Represents historical data segment T i，m With the rest of the historical data fragment T j，m The dot product between the historical data segments T i，m =ta1,ta2,…,ta m , the remaining historical data fragments T j，m =tb1,tb2,…,tb m .

[0123] Then, from the historical data segment T i，m The minimum distance value is obtained from the distances of the remaining historical data segments, and the remaining historical data segments corresponding to the minimum distance value are historical data segments T i，m The most similar historical data fragment is obtained, and the timestamp range information of the most similar historical data fragment in the historical data is obtained.

[0124] Specifically, the distance minimum MP i，m is d i，1 , d i，2 , …, di，n-1 The minimum value in the MP i，m is the historical data segment T i，m With the rest of the historical data fragment T k，m The distance between the remaining historical data fragments T k，m is the historical data segment T i，m The most similar historical data segment, the timestamp range information may include the most similar historical data segment T k，m The start and end timestamps of the historical data.

[0125] The obtained similarity portrait is [MP 1，m , M.P. 2，m , …, MP n，m ], portrait index [I 1，m , I 2，m ,…,I n，m ], I i，m is the historical data segment T i，m The timestamp range information of the most similar historical data fragments.

[0126] S602, select a minimum distance value that is less than or equal to a preset distance threshold in the similarity portrait, obtain the corresponding historical data segment based on the selected minimum distance value and the portrait index, and construct a parameter standard feature set of corresponding length based on the obtained historical data segment. The preset distance threshold is determined based on the minimum distance value of each historical data segment.

[0127] Among them, the minimum value of multiple distance minima in the similarity portrait is the minimum distance value MP of length m m , that is, the minimum distance MP m MP 1，m , M.P. 2，m , …, MP n，m The preset distance threshold is the minimum distance MP m The sum of the first adjustment value and the second adjustment value, the second adjustment value is, for example, 0.1 to 0.4.

[0128] When the minimum distance value is less than or equal to the preset distance threshold, the two historical data segments corresponding to the minimum distance value are placed in the parameter standard feature set. In the process of placing the selected historical data segments into the parameter standard feature set, the parameter standard feature set may contain at least two identical historical data segments, and subsequent deduplication operations can be performed.

[0129] In this embodiment, a parameter standard feature set is constructed through a matrix profiling algorithm to fully explore the normal change patterns of historical data segments of corresponding lengths, which is beneficial to improving the efficiency and accuracy of fault detection.

[0130] In some implementations, when using the parameter standard feature set to perform anomaly detection on the current monitoring data segment, the Euclidean distance between two time series data may also be used to characterize the similarity between the time series data. Specifically, step S102 includes the following steps:

[0131] S601, obtaining the Euclidean distance between the current monitoring data segment and each historical data segment in the parameter standard feature set, and obtaining the mean value of the Euclidean distance of the current monitoring data segment based on multiple Euclidean distances.

[0132] The Euclidean distance between the current monitoring data segment and the historical data segment can be calculated according to formula (2). The mean Euclidean distance of the current monitoring data segment is the quotient of the sum of the Euclidean distances between the current monitoring data segment and different historical data segments and the number of Euclidean distances.

[0133] S602: When the mean Euclidean distance is greater than a second preset distance threshold, it is determined that the corresponding current monitoring data segment is abnormal.

[0134] Among them, the smaller the mean value of the Euclidean distance is, the higher the consistency is between the normal change pattern represented by the standard feature set of parameters of the corresponding length; if the mean value of the Euclidean distance is too high, it means that there is a lack of consistency between the normal change pattern represented by the standard feature set of parameters of the corresponding length, and the current monitoring data segment does not conform to the normal change pattern of the corresponding length and is an abnormal segment.

[0135] In this embodiment, the Euclidean distance between two time series data is used to characterize the similarity between the time series data, which is beneficial to improving the calculation accuracy.

[0136] An embodiment of the present application provides a fault detection device for a wind turbine generator set. Figure 4As shown, the fault detection device 40 of the wind turbine generator set includes: a first data module 41, an abnormal warning module 42, a second data module 43 and a fault detection module 44. Among them, the first data module 41 is used to obtain the current monitoring data of at least one monitoring parameter of the wind turbine generator set, and obtain multiple current monitoring time series sample sets of the monitoring parameters according to the current monitoring data. Each current monitoring time series sample set includes multiple current monitoring data segments of the same length. The lengths of the current monitoring data segments in different current monitoring time series sample sets are different, and the current monitoring data segments are intercepted from the current monitoring data of the monitoring parameters. The abnormal warning module 42 is used to detect the current monitoring data segment in the current monitoring time series sample set according to the parameter standard feature set of the corresponding length of the monitoring parameter to determine whether the current monitoring data segment is abnormal; the parameter standard feature set is used to characterize the normal change law of the monitoring parameter, and the parameter standard feature set includes multiple similar historical data segments of corresponding lengths. The second data module 43 is used to obtain the target state data and target environment data of the wind turbine generator set according to the abnormal current monitoring data segment; the timestamp range corresponding to the target state data and the timestamp range corresponding to the environment data are respectively consistent with the timestamp range corresponding to the abnormal current monitoring data segment. The fault detection module 44 is used to obtain the fault prediction result of the wind turbine generator set according to the target state data and the target environment data.

[0137] As an implementation mode, the fault detection module 44 is used to input the target state data and the target environment data into a pre-trained fault prediction model to obtain a fault prediction result output by the fault prediction model, wherein the fault prediction result includes at least one fault type and the fault occurrence probability corresponding to the fault type; wherein the fault detection model is trained based on the historical state data and the historical environment data within a first time range before the fault occurs.

[0138] As an implementation mode, the fault detection module 44 is used to obtain a training set, the training set includes multiple data vectors and fault classification labels corresponding to the data vectors, the data vectors include multiple data sequences, the multiple data sequences include at least one time series of a state parameter and at least one time series of an environmental indicator; the data vectors are input into a fault detection model to obtain a fault prediction result output by the fault detection model, the fault prediction result includes a predicted fault type and a fault probability corresponding to a preset fault type; the training loss of the fault detection model is calculated according to the fault prediction result, and the parameters of the fault detection model are adjusted according to the training loss until the fault detection model converges.

[0139] As an implementation mode, the first data module 41 is used to obtain historical data of at least one monitoring parameter of the wind turbine generator set, and obtain multiple historical time series sample sets of the monitoring parameters based on the historical data, each historical time series sample set includes multiple historical data segments of the same length, and the lengths of the historical data segments in different historical time series sample sets are different, and the historical data segments are intercepted from the historical data of the monitoring parameters; a parameter standard feature set of corresponding length is constructed based on each historical time series sample set, and the parameter standard feature set is used to characterize the normal change law of the monitoring parameter, and the parameter standard feature set includes multiple similar historical data segments.

[0140] As an implementation mode, the first data module 41 is used to generate a historical data segment curve based on the time series data fitting of each historical data segment; obtain the similarity between every two historical data segment curves; when the similarity is greater than or equal to a first preset similarity threshold, put the two historical data segments corresponding to the similarity into a parameter standard feature set.

[0141] As an implementation mode, the second data module 43 is used to generate a historical data segment curve based on the time series data fitting of each historical data segment in the parameter standard feature set; generate a current monitoring data segment curve based on the time series data fitting of each current monitoring data segment in the current monitoring time series sample set; obtain the similarity between the current monitoring data segment curve and each historical data segment curve, and obtain the average similarity of the current monitoring data segment based on multiple similarities; when the similarity average is less than a second preset similarity threshold, determine that the corresponding current monitoring data segment is abnormal.

[0142] As an implementation mode, the first data module 41 is used to obtain a similarity portrait and a portrait index of the similarity portrait for each historical time series sample set based on the Euclidean distance between every two historical data fragments; the similarity portrait includes a minimum distance value between each historical data fragment and each of the remaining historical data fragments in the historical time series sample set, the minimum distance value is the distance between the historical data fragment and the most similar historical data fragment, and the portrait index is used to record the timestamp range information of the most similar historical data fragment of each historical data fragment in the historical data; a minimum distance value less than or equal to a preset distance threshold is selected in the similarity portrait, and the corresponding historical data fragment is obtained according to the selected minimum distance value and the portrait index, and a parameter standard feature set of corresponding length is constructed according to the obtained historical data fragment, and the preset distance threshold is determined according to the minimum distance value of each historical data fragment.

[0143] An embodiment of the present application provides a computer device, see Figure 5As shown, the computer device 50 includes a processor 51, a communication interface 52, a memory 53 and a communication bus 54, wherein the processor 51, the communication interface 52 and the memory 53 communicate with each other via the communication bus 54; the memory 53 is used to store computer programs; the processor 51 is used to implement the steps of the above-mentioned wind turbine fault detection method when executing the program stored in the memory 53.

[0144] The processor 51 may also be a CPU (Central Processing Unit). The processor 51 may also be an integrated circuit chip having signal processing capabilities. The processor 51 may also be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0145] The memory 53 may be, for example, a read-only memory (ROM), and the processor 51 and the memory 53 are connected to each other via a communication bus 54. The communication interface 52 is also connected to the communication bus 54.

[0146] See also Figure 6 , Figure 6 The schematic diagram of the structure of the computer-readable storage medium of an embodiment of the present application. The computer-readable storage medium 60 of the embodiment of the present application stores a computer program 61 capable of implementing all the above methods, and the computer program 61 is a computer executable instruction or a computer program, wherein the computer program 61 can be stored in the above computer-readable storage medium in the form of a software product, and can include several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor (processor) to perform all or part of the steps of the method described in each embodiment of the present application. The computer-readable storage medium 60 can be a non-volatile storage medium or a volatile storage medium, and the aforementioned computer-readable storage medium 60 can include: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes, or terminal devices such as computers, servers, mobile phones, and tablets.

[0147] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0148] In addition, each functional unit in each embodiment of the present application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The above integrated unit can be implemented in the form of hardware or in the form of software functional units. The above is only an implementation method of the present application, and does not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made by using the content of the specification and drawings of this application, or directly or indirectly used in other related technical fields, is also included in the patent protection scope of the present application.

[0149] The above is only an implementation method of the present application. It should be pointed out that a person skilled in the art can make improvements without departing from the inventive concept of the present application, but these improvements are within the scope of protection of the present application.

Claims

1. A fault detection method for a wind turbine generator set, characterized in that: include: Acquire current monitoring data of at least one monitoring parameter of the wind turbine generator set, and acquire multiple current monitoring time series sample sets of the monitoring parameter according to the current monitoring data, wherein each of the current monitoring time series sample sets includes multiple current monitoring data segments of the same length, and the lengths of the current monitoring data segments in different current monitoring time series sample sets are different; The current monitoring data segment in the current monitoring time series sample set is detected according to the parameter standard feature set of the corresponding length of the monitoring parameter to determine whether the current monitoring data segment is abnormal; the parameter standard feature set is used to characterize the normal change law of the monitoring parameter, and the parameter standard feature set includes multiple similar historical data segments of corresponding lengths; Acquire the target state data and target environment data of the wind turbine generator set according to the abnormal current monitoring data segment; the timestamp range corresponding to the target state data and the timestamp range corresponding to the environment data are respectively consistent with the timestamp range corresponding to the abnormal current monitoring data segment; A fault prediction result of the wind turbine generator set is obtained according to the target state data and the target environment data.

2. The fault detection method for a wind turbine generator set according to claim 1, characterized in that: The obtaining of the fault prediction result of the wind turbine generator set according to the target state data and the target environment data comprises: The target state data and the target environment data are input into a pre-trained fault prediction model to obtain the fault prediction result output by the fault prediction model, wherein the fault prediction result includes at least one fault type and the fault occurrence probability corresponding to the fault type; wherein the fault detection model is trained based on historical state data and historical environment data within a first time range before the fault occurs.

3. The fault detection method for a wind turbine generator set according to claim 2, characterized in that: The training step of the fault detection model includes: Acquire a training set, the training set comprising a plurality of data vectors and fault classification labels corresponding to the data vectors, the data vectors comprising a plurality of data sequences, the plurality of data sequences comprising a time series of at least one state parameter and a time series of at least one environmental indicator; Inputting the data vector into a fault detection model to obtain a fault prediction result output by the fault detection model, wherein the fault prediction result includes a predicted fault type and a fault occurrence probability corresponding to the preset fault type; The training loss of the fault detection model is calculated according to the fault prediction result, and the parameters of the fault detection model are adjusted according to the training loss until the fault detection model converges.

4. The fault detection method for a wind turbine generator set according to claim 1, characterized in that: Before obtaining the current monitoring data of at least one monitoring parameter of the wind turbine generator set, the method includes: Acquire historical data of at least one monitoring parameter of a wind turbine generator set, and acquire multiple historical time series sample sets of the monitoring parameter according to the historical data, wherein each of the historical time series sample sets includes multiple historical data segments of the same length, and the lengths of the historical data segments in different historical time series sample sets are different; A parameter standard feature set of corresponding length is constructed according to each of the historical time series sample sets.

5. The fault detection method of a wind turbine generator set according to claim 4, characterized in that: The step of constructing the parameter standard feature set of corresponding length according to each of the historical time series sample sets includes: Generate a historical data segment curve according to the time series data fitting of each of the historical data segments; Obtaining the similarity between every two historical data segment curves; When the similarity is greater than or equal to a first preset similarity threshold, the two historical data segments corresponding to the similarity are put into the parameter standard feature set.

6. The fault detection method of a wind turbine generator set according to claim 5, characterized in that: The detecting the current monitoring data segment in the current monitoring time series sample set according to the parameter standard feature set of corresponding length to determine whether the current monitoring data segment is abnormal includes: Generate a historical data segment curve according to the time series data fitting of each historical data segment in the parameter standard feature set; Generate a current monitoring data segment curve according to the time series data fitting of each current monitoring data segment in the current monitoring time series sample set; Obtaining the similarity between the current monitoring data segment curve and each of the historical data segment curves, and obtaining a similarity average of the current monitoring data segment according to the multiple similarities; When the similarity mean is less than a second preset similarity threshold, it is determined that the corresponding current monitoring data segment is abnormal.

7. The fault detection method of a wind turbine generator set according to claim 4, characterized in that: The step of constructing the parameter standard feature set of corresponding length according to each of the historical time series sample sets includes: According to the Euclidean distance between every two of the historical data fragments, a similarity portrait of each of the historical time series sample sets and a portrait index of the similarity portrait are obtained; the similarity portrait includes a minimum distance value between each of the historical data fragments and each of the remaining historical data fragments in the historical time series sample set, the minimum distance value is the distance between the historical data fragment and the most similar historical data fragment, and the portrait index is used to record the timestamp range information of the most similar historical data fragment of each of the historical data fragments in the historical data; A minimum distance value that is less than or equal to a preset distance threshold is selected in the similarity portrait, and the corresponding historical data segment is obtained according to the selected minimum distance value and the portrait index. The parameter standard feature set of the corresponding length is constructed according to the obtained historical data segment, and the preset distance threshold is determined according to the minimum distance value of each historical data segment.

8. A fault detection device for a wind turbine generator set, characterized in that: include: A first data module is used to obtain current monitoring data of at least one monitoring parameter of the wind turbine generator set, and obtain multiple current monitoring time series sample sets of the monitoring parameter according to the current monitoring data, each of the current monitoring time series sample sets includes multiple current monitoring data segments of the same length, and the lengths of the current monitoring data segments in different current monitoring time series sample sets are different; An abnormal warning module is used to detect the current monitoring data segment in the current monitoring time series sample set according to the parameter standard feature set of the corresponding length of the monitoring parameter to determine whether the current monitoring data segment is abnormal; the parameter standard feature set is used to characterize the normal change law of the monitoring parameter, and the parameter standard feature set includes multiple similar historical data segments of corresponding lengths; A second data module, used for acquiring target state data and target environment data of the wind turbine generator set according to the abnormal current monitoring data segment; The timestamp range corresponding to the target state data and the timestamp range corresponding to the environment data are respectively consistent with the timestamp range corresponding to the abnormal current monitoring data segment; A fault detection module is used to obtain a fault prediction result of the wind turbine generator set according to the target state data and the target environment data.

9. A computer device, characterized in that: The device includes a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other via the communication bus; Memory, used to store computer programs; The processor is used to implement the steps of the fault detection method for a wind turbine generator set according to any one of claims 1 to 7 when executing the program stored in the memory.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the fault detection method for a wind turbine generator set according to any one of claims 1 to 7 are implemented.

Images