A wind turbine fault diagnosis method and system based on SVD-SSA-LSTM

Abstract

The application relates to the technical field of fault diagnosis, in particular to a wind turbine fault diagnosis method and system based on SVD-SSA-LSTM. The method comprises the following steps: setting multiple monitoring points according to the equipment parameters of a wind turbine tower drum; obtaining vibration signals of each monitoring point according to a preset monitoring time node, and generating a fault feature data packet according to all the vibration signals; establishing a fault diagnosis model, generating a fault diagnosis result according to the fault diagnosis model and the fault feature data packet; decomposing the collected vibration signals by using singular value decomposition (SVD) noise reduction, removing the redundancy and noise components in the signals, and then diagnosing the faults of the wind turbine by using an SSA-LSTM fault diagnosis model; through the processing capacity of the LSTM network for time series data, the accuracy of the wind turbine fault diagnosis is improved, the diagnosis cycle is shortened, and the early warning efficiency for the fault risk of the wind turbine is improved.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of fault diagnosis, in particular to a wind turbine generator unit fault diagnosis method and system based on SVD-SSA-LSTM. BACKGROUND

[0002] As an important part of renewable energy, the reliability and maintenance efficiency of wind power generation are crucial to the stability and economy of energy supply. However, the gearbox in the wind turbine generator unit, as a key component of the transmission system, is prone to failure due to the influence of harsh operating environment.

[0003] Traditional fault diagnosis methods have problems such as long diagnosis cycle and low accuracy, which cannot meet the needs of modern wind power generation systems. SUMMARY

[0004] The purpose of the present application is to solve the above technical problems, and the present application provides a wind turbine generator unit fault diagnosis method and system based on SVD-SSA-LSTM, which aims to improve the diagnosis accuracy of wind turbine generator unit faults and shorten the diagnosis cycle.

[0005] In some embodiments of the present application, the collected vibration signals are decomposed using singular value decomposition (SVD) denoising to remove redundant and noise components in the signals, and then the SSA-LSTM fault diagnosis model is used to diagnose the faults of the wind turbine generator unit. Through the processing capability of the LSTM network for time series data, the accuracy of wind turbine generator unit fault diagnosis is improved, the diagnosis cycle is shortened, and the early warning efficiency of wind turbine generator unit fault risk is improved.

[0006] In some embodiments of the present application, by establishing multiple monitoring points, multiple sets of vibration signals are collected at a single monitoring time node, and all vibration signals are mutually corrected according to a preset calibration model, thereby improving the authenticity of the collected vibration signals and avoiding distortion of the collected vibration signals due to environmental disturbances, thereby affecting the fault diagnosis results.

[0007] In some embodiments of the present application, a wind turbine generator unit fault diagnosis method based on SVD-SSA-LSTM is provided, which comprises:

[0008] A plurality of monitoring points are set according to the equipment parameters of the wind turbine generator tower drum;

[0009] Vibration signals of each monitoring point are obtained according to a preset monitoring time node, and fault feature data packets are generated according to all vibration signals;

[0010] A fault diagnosis model is established, and a fault diagnosis result is generated according to the fault diagnosis model and the fault feature data packet;

[0011] When establishing multiple monitoring points, it comprises:

[0012] A sequence of monitoring points A is established, A=(a1, a2…ai…an), wherein ai is the i-th monitoring point; and n is the number of monitoring points.

[0013] In some embodiments of the present application, the fault feature data packet is generated according to the pretreatment model, comprising:

[0014] The vibration signals of each monitoring point at the current monitoring time node are acquired;

[0015] The all vibration signals are processed according to the preset calibration model;

[0016] The first-level vibration signal is generated according to the processing result;

[0017] The first-level vibration signal is decomposed according to the preset SVD model, and the fault time-frequency domain feature information is generated according to the decomposition result;

[0018] The fault feature data packet of the current monitoring time node is generated according to the fault time-frequency domain feature information.

[0019] In some embodiments of the present application, the first-level vibration signal is generated according to the processing result, comprising:

[0020] The vibration curve of each monitoring point is generated according to the all monitoring signals;

[0021] The standard comparison curve is generated according to the fusion result of the vibration curve;

[0022] The deviation evaluation value between each vibration curve and the standard comparison curve is generated in sequence;

[0023] A sequence of deviation evaluation values D is established, D=(d1, d2…di…dn), wherein di is the deviation evaluation value between the vibration curve of the i-th monitoring point and the standard comparison curve;

[0024] The correction evaluation value g is generated;

[0025] g=[ di];

[0026] The preset correction evaluation value threshold G1 is set;

[0027] If g

[0028] If g>G1, the second-level processing instruction is generated;

[0029] The first-level vibration signal is generated according to the processing instruction.

[0030] In some embodiments of the present application, when the fault diagnosis model is established, comprising:

[0031] The training set data is established according to the historical vibration data;

[0032] establishing an LSTM network and setting a plurality of parameter configuration strategies of the LSTM network;

[0033] establishing a sparrow individual sequence P, P=(p1, p2…pi…pm), wherein a single sparrow represents a parameter configuration strategy, and pi is the ith sparrow individual; and m is the number of parameter configuration strategies;

[0034] generating an initial fitness of each sparrow individual according to the training set data;

[0035] setting a sparrow position iteration strategy according to a preset SSA optimization model;

[0036] outputting a first-level fitness of each sparrow according to the iteration strategy;

[0037] establishing a first-level fitness sequence B, B=(b1, b2…bi…bm), wherein bi is the first-level fitness of the ith sparrow individual; and m is the number of sparrow individuals;

[0038] setting the sparrow individual corresponding to the maximum value bmax in the first-level fitness sequence B as a target sparrow individual;

[0039] setting a parameter configuration strategy corresponding to the target sparrow individual and an LSTM network to generate a fault diagnosis model.

[0040] In some embodiments of the present application, the sparrow position iteration strategy according to the preset SSA optimization model comprises:

[0041] dividing the sparrows in the sparrow individual sequence J into discoverers, joiners and sentinels;

[0042] establishing a discoverer iteration model, a joiner iteration model and a sentinel iteration model;

[0043] The discoverer iteration model is:

[0044]

[0045] wherein, is the jth dimension parameter of the ith sparrow in the tth iteration process, is the maximum iteration number, αm represents a random number between 0 and 1, is a safety factor, is an alarm value, m is a random number subject to normal distribution, Lm is a 1xd matrix with all elements being 1.

[0046] In some embodiments of the present application, the joiner iteration model comprises: ​

[0047] ;

[0048] wherein, Nm is the total number of sparrows, is the position of the sparrow with the best foraging state, is the position of the sparrow with the worst foraging state, A+ satisfies A + = A T ( AA T )-1, wherein A is a 1xd matrix consisting of random elements of 1 or -1.

[0049] In some embodiments of the application, the sentry iterative model comprises:

[0050] ;

[0051] wherein, is the central position of the entire sparrow population in the iteration, which is free from threats from natural enemies, β m is a compensation control parameter subject to a standard normal distribution, K m is a random number between -1 and 1, ε m is an infinitesimal number, is the fitness of the current sparrow, is the fitness of the sparrow currently in the best foraging position, is the fitness of the sparrow currently in the worst foraging position.

[0052] In some embodiments of the application, a wind turbine generator fault diagnosis system based on SVD-SSA-LSTM is provided, comprising:

[0053] a central control unit configured to set a plurality of monitoring points according to equipment parameters of a wind turbine tower drum;

[0054] a monitoring unit comprising a plurality of monitoring sub-modules, the monitoring sub-modules being configured to collect vibration signals of the respective monitoring points;

[0055] the central control unit comprises:

[0056] a first processing module configured to establish a monitoring point sequence A, A=(a1, a2…ai…an), wherein ai is the ith monitoring point; and n is the number of monitoring points;

[0057] a second processing module configured to obtain the vibration signals of the respective monitoring points according to a preset monitoring time node, and generate a fault feature data packet according to all the vibration signals;

[0058] The third processing module is configured to establish a fault diagnosis model, and generate a fault diagnosis result according to the fault diagnosis model and the fault feature data packet.

[0059] In some embodiments of the present application, the second processing module is further configured to:

[0060] obtain vibration signals of each monitoring point at the current monitoring time node;

[0061] process all the vibration signals according to a preset calibration model;

[0062] generate a primary vibration signal according to the processing result;

[0063] decompose the primary vibration signal according to a preset SVD model, and generate fault time-frequency domain feature information according to a decomposition result;

[0064] generate a fault feature data packet of the current monitoring time node according to the fault time-frequency domain feature information.

[0065] In some embodiments of the present application, the third processing module is further configured to:

[0066] establish a training set data according to historical vibration data;

[0067] establish an LSTM network, and set a plurality of parameter configuration strategies of the LSTM network;

[0068] establish a sparrow individual sequence P, P=(p1, p2…pi…pm), wherein a single sparrow represents a parameter configuration strategy, pi is an i-th sparrow individual, and m is a parameter configuration strategy quantity;

[0069] generate an initial fitness of each sparrow individual according to the training set data;

[0070] set a sparrow position iteration strategy according to a preset SSA optimization model;

[0071] output a primary fitness of each sparrow according to the iteration strategy;

[0072] establish a primary fitness sequence B, B=(b1, b2…bi…bm), wherein bi is a primary fitness of an i-th sparrow individual, and m is a sparrow individual quantity;

[0073] set a sparrow individual corresponding to a maximum value bmax in the primary fitness sequence B as a target sparrow individual;

[0074] set a parameter configuration strategy corresponding to the target sparrow individual and an LSTM network to generate a fault diagnosis model.

[0075] Compared with the prior art, the wind turbine fault diagnosis method and system based on SVD-SSA-LSTM has the beneficial effects that:

[0076] The collected vibration signals are decomposed by singular value decomposition (SVD) denoising to remove redundant and noise components in the signals, and then the SSA-LSTM fault diagnosis model is used to diagnose the faults of the wind turbine, the processing capability of the LSTM network for time series data is used to improve the accuracy of the wind turbine fault diagnosis, shorten the diagnosis cycle, and improve the early warning efficiency of the wind turbine fault risk.

[0077] By establishing multiple monitoring points, multiple groups of vibration signals are collected at a single monitoring time node, and all vibration signals are mutually corrected according to a preset calibration model, so as to improve the authenticity of the collected vibration signals, avoid distortion of the collected vibration signals due to environmental disturbance, and thus affect the fault diagnosis result. BRIEF DESCRIPTION OF DRAWINGS

[0078] Figure 1 is a flowchart of a wind turbine fault diagnosis method based on SVD-SSA-LSTM in a preferred embodiment of the present application. DETAILED DESCRIPTION

[0079] The specific embodiments of the present application will be further described in detail below in combination with the drawings and examples. The following examples are used to illustrate the present application, but not to limit the scope of the present application.

[0080] In the description of the present application, it should be understood that the terms "center", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer" and the like indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, and therefore cannot be understood as indicating or implying that the device or element referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as limiting the present application.

[0081] The terms "first", "second" are only for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the technical features indicated. Therefore, the features defined with "first", "second" can explicitly or implicitly include one or more of the features. In the description of the present application, unless otherwise specified, the meaning of "multiple" is two or more.

[0082] In the description of the present application, it should be noted that unless otherwise explicitly specified and limited, the terms "mounting", "connection", "connecting" should be understood broadly, for example, it can be fixed connection, or detachable connection, or integral connection; it can be mechanical connection, or electrical connection; it can be direct connection, or indirect connection through intermediate medium, or internal communication of two elements. For those skilled in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.

[0083] As shown in Figure 1 , a wind turbine fault diagnosis method based on SVD-SSA-LSTM according to the preferred embodiment of the present application, comprising:

[0084] According to the equipment parameters of the wind turbine tower, a plurality of monitoring points are set;

[0085] According to the preset monitoring time node, the vibration signals of each monitoring point are obtained, and the fault feature data packet is generated according to all the vibration signals;

[0086] A fault diagnosis model is established, and a fault diagnosis result is generated according to the fault diagnosis model and the fault feature data packet;

[0087] Among them, when establishing a plurality of monitoring points, it includes:

[0088] A monitoring point sequence A is established, A=(a1, a2…ai…an), wherein ai is the ith monitoring point; n is the number of monitoring points.

[0089] Specifically, according to the equipment parameters of the wind turbine tower, a plurality of monitoring points are set at the horizontal and vertical positions of the tower wall, and the vibration sensors are set at each monitoring point by magnetic adsorption.

[0090] Specifically, the fault feature data packet is generated according to the preprocessing model, which includes:

[0091] Obtain the vibration signals of each monitoring point at the current monitoring time node;

[0092] According to the preset calibration model, all the vibration signals are processed;

[0093] According to the processing result, a first vibration signal is generated;

[0094] According to the preset SVD model, the first vibration signal is decomposed, and the fault time-frequency domain feature information is generated according to the decomposition result;

[0095] According to the fault time-frequency domain feature information, the fault feature data packet of the current monitoring time node is generated.

[0096] Specifically, the fault time-frequency domain feature information includes but is not limited to maximum value, minimum value, peak value, variance, standard deviation, root mean square, kurtosis, frequency domain mean and standard deviation, etc.

[0097] Specifically, by constructing a trajectory matrix and performing singular value decomposition (SVD), the trend, periodic component and noise in the vibration signal are extracted, thereby obtaining the optimized time series data; further, the time series data is denoised, by selecting an appropriate threshold, the redundant and noise components in the signal are removed, and the fault time-frequency domain feature information is retained and extracted.

[0098] Specifically, when performing singular value decomposition, the SVD model adopts an adaptive hard threshold selection algorithm and a soft threshold method of unequal optimal weight shrinkage to improve the adaptability and accuracy of signal processing.

[0099] It can be understood that in the above embodiments, the singular value decomposition (SVD) denoising is used to decompose the collected vibration signal, remove the redundant and noise components in the signal, and then use the SSA-LSTM fault diagnosis model to diagnose the fault of the wind turbine, through the processing capability of the LSTM network for time series data, improve the accuracy of wind turbine fault diagnosis, shorten the diagnosis cycle, and improve the early warning efficiency of wind turbine fault risk.

[0100] In the preferred embodiment of the present application, the first-level vibration signal is generated according to the processing result, comprising:

[0101] Generating a vibration curve of each monitoring point according to all monitoring signals;

[0102] Generating a standard comparison curve according to the fusion result of the vibration curve;

[0103] Generating a deviation evaluation value between each vibration curve and the standard comparison curve in turn;

[0104] Establishing a deviation evaluation value sequence D, D=(d1, d2…di…dn), wherein di is the deviation evaluation value between the vibration curve of the i th monitoring point and the standard comparison curve;

[0105] Generating a correction evaluation value g;

[0106] g=[ di];

[0107] Presetting a correction evaluation value threshold G1;

[0108] If g

[0109] If g>G1, a second-level processing instruction is generated;

[0110] Generating a first-level vibration signal according to the processing instruction.

[0111] Specifically, by comparing each vibration curve, an average curve is generated as a standard comparison curve, and the greater the deviation evaluation value is, the lower the coincidence degree of the current vibration curve with the standard comparison curve is.

[0112] Specifically, the correction evaluation value threshold G1 can be set according to historical parameters.

[0113] The first processing instruction refers to that the current vibration signal is not affected by the environment, and the vibration signal of the monitoring point with the minimum deviation evaluation value in the horizontal direction and the vibration signal of the monitoring point with the minimum deviation evaluation value in the vertical direction are selected as the first vibration signal.

[0114] Specifically, the second processing instruction refers to that the current vibration signal is disturbed by the environment during the collection process, and the vibration signal of the monitoring point with a larger deviation evaluation value needs to be removed, and the first vibration signal is generated according to the removal result.

[0115] It can be understood that in the above embodiment, a plurality of monitoring points are established, a plurality of groups of vibration signals are collected at a single monitoring time node, and all vibration signals are mutually corrected according to a preset calibration model, so as to improve the authenticity of the collected vibration signal, avoid the distortion of the collected vibration signal caused by environmental disturbance, and thus affect the fault diagnosis result.

[0116] In the preferred embodiment of the present application, when establishing the fault diagnosis model, the following steps are included:

[0117] Establish a training set data according to historical vibration data;

[0118] Establish an LSTM network and set a plurality of parameter configuration strategies of the LSTM network;

[0119] Establish a sparrow individual sequence P, P=(p1, p2…pi…pm), wherein a single sparrow represents a group of parameter configuration strategies, pi is the i th sparrow individual, and m is the number of parameter configuration strategies;

[0120] Generate the initial fitness of each sparrow individual according to the training set data;

[0121] Set a sparrow position iteration strategy according to a preset SSA optimization model;

[0122] Output the first fitness of each sparrow according to the iteration strategy;

[0123] Establish a first fitness sequence B, B=(b1, b2…bi…bm), wherein bi is the first fitness of the i th sparrow individual, and m is the number of sparrow individuals;

[0124] Set the sparrow individual corresponding to the maximum value bmax in the first fitness sequence B as the target sparrow individual;

[0125] Set the parameter configuration strategy corresponding to the target sparrow individual and the LSTM network to generate a fault diagnosis model.

[0126] Specifically, the parameter configuration strategy includes the number of hidden layer nodes of the LSTM network, the learning rate and other related parameters.

[0127] Specifically, the training set data is obtained through simulation or digital twinning.

[0128] Specifically, each unit of the LSTM network includes a forgetting gate, an input gate and an output gate, and through the gating mechanism, further processing of historical information in the previous processing stage is realized.

[0129] Specifically, according to the preset SSA optimization model, the sparrow position iteration strategy is set, including:

[0130] The sparrows in the sparrow individual sequence J are divided into discoverers, joiners and alarmers;

[0131] The discoverer iteration model, the joiner iteration model and the alarmer iteration model are established;

[0132] Specifically, the SSA optimization model is initialized, and the related parameters of the SSA optimization model are set, such as population size, maximum iteration number, fitness function, etc.

[0133] Specifically, the discoverer iteration model is:

[0134] ;

[0135] Wherein, is the jthdimensional parameter of the ithsparrow in the tthiteration process, is the maximum iteration number, αm represents a random number between 0 and 1, is a safety factor, is an alarm value, m is a random number obeying normal distribution, Lm is a 1xdmatrix with all elements being 1.

[0136] Specifically, the joiner iteration model includes:

[0137] ;

[0138] Wherein, Nm is the total number of sparrows, is the optimal foraging state sparrow position, the worst foraging position of the sparrows, A+ satisfying A + = A T AA T -1, where A is a 1xd matrix consisting of random elements of 1 or -1.

[0139] In particular, the sentry iteration model comprises:

[0140] ;

[0141] wherein, is the center position of the entire sparrow population in the iteration, which is free from threats from natural enemies, β m is a compensation control parameter subject to a standard normal distribution, K m is a random number between -1 and 1, ε m is an infinitesimal number, is the fitness of the current sparrow, is the fitness of the sparrow currently in the best foraging position, is the fitness of the sparrow currently in the worst foraging position.

[0142] In particular, for sparrow individuals with higher fitness (producers), their positions are updated according to their current positions and parameters such as early warning values and safety values; for sparrow individuals with lower fitness (search individuals), a part of them monitor the producers and try to compete with them for food (i.e. update their own positions to approach the optimal position of the producers), and the other part chooses to fly to a more distant area to randomly search for food to increase the possibility of finding a better solution.

[0143] In particular, during the iteration optimization, the process of fitness evaluation and position updating is repeated until the maximum number of iterations is reached or other stopping conditions are met. The global optimal solution (i.e. the LSTM network parameter configuration represented by the sparrow individual with the highest fitness) is recorded and updated, and finally the global optimal solution is output as the optimized LSTM network parameter configuration.

[0144] Based on any one of the above preferred embodiments, another preferred embodiment of the SVD-SSA-LSTM-based wind turbine fault diagnosis method is provided, and in this preferred embodiment, an SVD-SSA-LSTM-based wind turbine fault diagnosis system is provided, comprising:

[0145] a central control unit configured to set a plurality of monitoring points according to the equipment parameters of the wind turbine tower;

[0146] ​The monitoring unit comprises a plurality of monitoring sub-modules, and the monitoring sub-modules are used to collect vibration signals of respective monitoring points;

[0147] The central control unit comprises:

[0148] The first processing module is used to establish a monitoring point sequence A, A=(a1, a2…ai…an), wherein ai is the ith monitoring point; and n is the number of monitoring points.

[0149] The second processing module is used to acquire vibration signals of respective monitoring points according to preset monitoring time nodes, and generate a fault feature data packet according to all vibration signals.

[0150] The third processing module is used to establish a fault diagnosis model, and generate a fault diagnosis result according to the fault diagnosis model and the fault feature data packet.

[0151] Specifically, the second processing module is further used to:

[0152] acquire vibration signals of respective monitoring points at a current monitoring time node;

[0153] process all vibration signals according to a preset calibration model;

[0154] generate a primary vibration signal according to a processing result;

[0155] decompose the primary vibration signal according to a preset SVD model, and generate fault time-frequency domain feature information according to a decomposition result;

[0156] generate a fault feature data packet of the current monitoring time node according to the fault time-frequency domain feature information.

[0157] In the preferred embodiment of the present application, the third processing module is further used to:

[0158] establish a training set data according to historical vibration data;

[0159] establish an LSTM network, and set a plurality of parameter configuration strategies of the LSTM network;

[0160] establish a sparrow individual sequence P, P=(p1, p2…pi…pm), wherein a single sparrow represents a group of parameter configuration strategies, pi is the ith sparrow individual; and m is the number of parameter configuration strategies.

[0161] generate an initial fitness of respective sparrow individuals according to the training set data;

[0162] set a sparrow position iteration strategy according to a preset SSA optimization model;

[0163] output a primary fitness of respective sparrows according to the iteration strategy;

[0164] A first-level fitness sequence B, B=(b1, b2…bi…bm), is established, wherein bi is the first-level fitness of the i-th sparrow individual; and m is the number of sparrow individuals.

[0165] The sparrow individual corresponding to the maximum value bmax in the first-level fitness sequence B is set as a target sparrow individual.

[0166] The parameter configuration strategy corresponding to the target sparrow individual and the LSTM network are set as a fault diagnosis model.

[0167] According to the first concept of the present application, singular value decomposition (SVD) is used to decompose the collected vibration signal, remove the redundant and noise components in the signal, and then use the SSA-LSTM fault diagnosis model to diagnose the fault of the wind turbine. Through the processing capacity of the LSTM network for time series data, the accuracy of the wind turbine fault diagnosis is improved, the diagnosis cycle is shortened, and the early warning efficiency of the wind turbine fault risk is improved.

[0168] According to the second concept of the present application, a plurality of monitoring points are established, a plurality of groups of vibration signals are collected at a single monitoring time node, and all vibration signals are mutually corrected according to a preset calibration model, thereby improving the authenticity of the collected vibration signals, avoiding distortion of the collected vibration signals due to environmental disturbance, and thereby affecting the fault diagnosis result.

[0169] The above only describes the preferred embodiments of the present application. It should be noted that, for those skilled in the art, without departing from the technical principles of the present application, a number of improvements and replacements can be made, which should also be considered as the protection scope of the present application.

Claims

1. A wind turbine generator fault diagnosis method based on SVD-SSA-LSTM, characterized in that, The method comprises the following steps: Setting multiple monitoring points according to the equipment parameters of the wind turbine tower; Obtaining the vibration signals of each monitoring point according to the preset monitoring time node, and generating a fault feature data packet according to all the vibration signals; Establishing a fault diagnosis model, and generating a fault diagnosis result according to the fault diagnosis model and the fault feature data packet; When establishing multiple monitoring points, the method comprises the following steps: Establishing a monitoring point sequence A, A=(a1, a2…ai…an), wherein ai is the ith monitoring point; n is the number of monitoring points; Generating a fault feature data packet according to a preprocessing model, which comprises the following steps: Obtaining the vibration signals of each monitoring point at the current monitoring time node; Processing all the vibration signals according to a preset calibration model; Generating a first-level vibration signal according to the processing result; Decomposing the first-level vibration signal according to a preset SVD model, and generating fault time-frequency domain feature information according to the decomposition result; Generating a fault feature data packet of the current monitoring time node according to the fault time-frequency domain feature information; When establishing a fault diagnosis model, the method comprises the following steps: Establishing a training set data according to historical vibration data; Establishing an LSTM network, and setting multiple parameter configuration strategies of the LSTM network; Establishing a sparrow individual sequence P, P=(p1, p2…pi…pm), wherein a single sparrow represents a group of parameter configuration strategies, pi is the ith sparrow individual; m is the number of parameter configuration strategies; Generating the initial fitness of each sparrow individual according to the training set data; Setting a sparrow position iteration strategy according to a preset SSA optimization model; Outputting the first-level fitness of each sparrow according to the iteration strategy; Establishing a first-level fitness sequence B, B=(b1, b2…bi…bm), wherein bi is the first-level fitness of the ith sparrow individual; m is the number of sparrow individuals; Setting the sparrow individual corresponding to the maximum value bmax in the first-level fitness sequence B as a target sparrow individual; Setting the parameter configuration strategy and the LSTM network corresponding to the target sparrow individual to generate the fault diagnosis model.

2. The SVD-SSA-LSTM-based wind turbine fault diagnosis method of claim 1, wherein, Generating a first-level vibration signal according to the processing result, which comprises the following steps: Generating the vibration curve of each monitoring point according to all the monitoring signals; Generating a standard comparison curve according to the fusion result of the vibration curve; Generating the deviation evaluation value between each vibration curve and the standard comparison curve in turn; Establishing a deviation evaluation value sequence D, D=(d1, d2…di…dn), wherein di is the deviation evaluation value between the vibration curve of the ith monitoring point and the standard comparison curve; Generating a correction evaluation value g; g=[ di]; Setting a preset correction evaluation value threshold G1; If g If g>G1, a second-level processing instruction is generated; Generating a first-level vibration signal according to the processing instruction.

3. The SVD-SSA-LSTM-based wind turbine fault diagnosis method of claim 2, wherein, Setting a sparrow position iteration strategy according to a preset SSA optimization model, which comprises the following steps: Dividing the sparrows in the sparrow individual sequence J into discoverers, joiners and sentinels; Establishing a discoverer iteration model, a joiner iteration model and a sentinel iteration model; The discoverer iteration model is as follows: ； wherein, is the jth dimension parameter of the ith sparrow in the tth iteration process, is the maximum number of iterations, represents a random number between 0 and 1, is a safety factor, is an alarm value, m is a random number subject to a normal distribution, is a 1xd matrix with all elements being 1.

4. The SVD-SSA-LSTM-based wind turbine generator fault diagnosis method of claim 3, wherein, The joiner iteration model comprises the following steps: ； wherein, is the total number of sparrows, is the position of the sparrow with the best foraging state, is the position of the sparrow with the worst foraging state, + satisfies A + = A T ( AA T )-1, where A is a 1 x d matrix of random elements of 1 or -1.

5. The SVD-SSA-LSTM-based wind turbine generator fault diagnosis method of claim 4, wherein, The sentinel iteration model comprises the following steps: ； wherein, is the center position of the entire sparrow population in the iteration, which is free from threats by natural enemies, β m is a compensation control parameter subject to a standard normal distribution, K m is a random number between -1 and 1, ε m is an infinitesimal number, is the fitness of the current sparrow, is the fitness of the sparrow currently in the best foraging position, is the fitness of the sparrow currently in the worst foraging position.

6. A wind turbine generator fault diagnosis system based on SVD-SSA-LSTM, adopting the wind turbine generator fault diagnosis method based on SVD-SSA-LSTM in any one of claims 1-5, characterized in that, The method comprises the following steps: The central control unit is used to set multiple monitoring points according to the equipment parameters of the wind turbine tower. The monitoring unit comprises a plurality of monitoring sub-modules for collecting vibration signals of respective monitoring points; The central control unit comprises: A first processing module for establishing a monitoring point sequence A, A=(a1, a2…ai…an), wherein ai is the ith monitoring point; and n is the number of monitoring points; A second processing module for obtaining vibration signals of respective monitoring points according to preset monitoring time nodes, and generating a fault feature data packet according to all the vibration signals; A third processing module for establishing a fault diagnosis model, and generating a fault diagnosis result according to the fault diagnosis model and the fault feature data packet.

7. The SVD-SSA-LSTM based wind turbine fault diagnosis system of claim 6, wherein, The second processing module is further configured to: Obtain vibration signals of respective monitoring points at a current monitoring time node; Process all the vibration signals according to a preset calibration model; Generate a primary vibration signal according to a processing result; Decompose the primary vibration signal according to a preset SVD model, and generate fault time-frequency domain feature information according to a decomposition result; Generate a fault feature data packet of the current monitoring time node according to the fault time-frequency domain feature information.

8. The SVD-SSA-LSTM based wind turbine fault diagnosis system of claim 7, wherein, The third processing module is further configured to: Establish a training set data according to historical vibration data; Establish an LSTM network, and set a plurality of parameter configuration strategies of the LSTM network; Establish a sparrow individual sequence P, P=(p1, p2…pi…pm), wherein a single sparrow represents a parameter configuration strategy, pi is the ith sparrow individual; and m is the number of parameter configuration strategies; Generate an initial fitness of respective sparrow individuals according to the training set data; Set a sparrow position iteration strategy according to a preset SSA optimization model; Output a primary fitness of respective sparrows according to the iteration strategy; Establish a primary fitness sequence B, B=(b1, b2…bi…bm), wherein bi is the primary fitness of the ith sparrow individual; and m is the number of sparrow individuals; Set a sparrow individual corresponding to a maximum value bmax in the primary fitness sequence B as a target sparrow individual; Set a parameter configuration strategy and an LSTM network corresponding to the target sparrow individual to generate a fault diagnosis model.

Images