Method, device and equipment for monitoring wind turbine cluster and storage medium

Abstract

The present disclosure relates to a wind turbine cluster monitoring method, device, equipment and storage medium. By obtaining historical operation monitoring data of each wind turbine, removing abnormal data in the historical operation monitoring data, extracting distribution features and correlation features of the historical operation monitoring data, clustering based on the distribution features and the correlation features, dividing into multiple unit clusters, determining the center unit of each unit cluster, training the state monitoring model based on the historical operation monitoring data of the center unit of each unit cluster, obtaining the trained state monitoring model corresponding to each unit cluster, and monitoring the state of the wind turbine of each unit cluster based on the trained state monitoring model corresponding to each unit cluster. Compared with the prior art, the embodiment of the present disclosure can reduce the number of models, has minimal demand for computing resources, is conducive to edge computing and model update iteration, while ensuring computing accuracy and meeting the demand of predictive maintenance.

Description

Technical Field

[0001] This disclosure relates to the field of wind turbine monitoring technology, and in particular to a monitoring method, device, equipment and storage medium for a wind turbine cluster. Background Technology

[0002] Wind farms typically employ corrective and preventative maintenance strategies. However, due to their inherent lag, these strategies often result in significant economic losses and even safety incidents caused by downtime. Predictive maintenance is a more ideal approach. With the development of artificial intelligence, predictive maintenance, based on condition monitoring and early fault detection, is rapidly evolving.

[0003] However, existing predictive maintenance methods typically focus on individual wind turbines rather than the entire wind farm, resulting in a "one turbine, one policy" approach. Although researchers have significantly improved the generalization performance of models, the current situation of one model corresponding to one turbine remains unresolved, leading to a massive number of models. Furthermore, as models develop further, the number of model parameters increases dramatically, drastically increasing the demand for computing resources. Due to limitations in edge computing capabilities at wind farms, these models are difficult to deploy and compute at the wind farm's edge. Summary of the Invention

[0004] To address the aforementioned technical issues, this disclosure provides a monitoring method, device, equipment, and storage medium for wind turbine clusters. This reduces the number of models and the number of model parameters, minimizing computational resource requirements. It facilitates edge computing and model updates and iterations while ensuring computational accuracy and meeting the needs of predictive maintenance.

[0005] In a first aspect, embodiments of this disclosure provide a method for monitoring a wind turbine cluster, the method comprising:

[0006] Obtain historical operation monitoring data for each wind turbine unit, and remove abnormal data from the historical operation monitoring data;

[0007] Extract the distribution and correlation characteristics of historical operation monitoring data for each wind turbine unit;

[0008] Clustering is performed based on the distribution characteristics and correlation characteristics to divide each wind turbine into multiple turbine clusters and determine the central turbine of each cluster.

[0009] The condition monitoring model is trained based on the historical operation monitoring data of the central unit of each unit cluster, so as to obtain the trained condition monitoring model corresponding to each unit cluster.

[0010] The status of the wind turbines in each cluster is monitored based on the pre-trained status monitoring model corresponding to each cluster.

[0011] In some embodiments, extracting the distribution characteristics and correlation characteristics of the historical operation monitoring data of each wind turbine includes:

[0012] First-order alignment is performed on the historical operation monitoring data of each wind turbine to obtain the alignment data of each wind turbine.

[0013] Feature extraction is performed on the alignment data of each wind turbine to obtain the distribution characteristics of the historical operation monitoring data of each wind turbine. The distribution characteristics include mean, variance, skewness and kurtosis.

[0014] The correlation coefficients between each input variable in the historical operation monitoring data of each wind turbine are calculated respectively, and the correlation coefficients between each input variable are used as the correlation characteristics of the historical operation monitoring data of each wind turbine.

[0015] In some embodiments, the step of clustering based on the distribution characteristics and the correlation characteristics to divide each wind turbine into multiple turbine clusters and determining the central turbine of each cluster includes:

[0016] Based on the distribution characteristics and the correlation characteristics, feature vectors for each wind turbine are constructed.

[0017] A pre-defined clustering algorithm is used to cluster the feature vectors of each wind turbine to obtain multiple turbine clusters, and the wind turbine closest to the cluster center in each cluster is determined as the central turbine.

[0018] In some embodiments, before training the condition monitoring model based on the historical operation monitoring data of the central unit of each unit cluster to obtain the trained condition monitoring model corresponding to each unit cluster, the method further includes:

[0019] A state monitoring model is constructed based on neural circuit strategies and spiking neural networks.

[0020] In some embodiments, the construction of the state monitoring model based on neural circuit strategies and spiking neural networks includes:

[0021] A state monitoring model is obtained by connecting spiking neurons in a spiking neural network using liquid time constant neurons based on a neural circuit strategy.

[0022] In some embodiments, training the condition monitoring model based on the historical operation monitoring data of the central unit of each unit cluster to obtain the trained condition monitoring model corresponding to each unit cluster includes:

[0023] For the central unit of each unit cluster, the historical operation monitoring data of the central unit is input into the status monitoring model, and the regression value of the target variable of the central unit is output through the status monitoring model.

[0024] Based on the measured values ​​of the target variable corresponding to the historical operation monitoring data of the central unit, the regression value of the target variable is calculated to obtain the loss calculation result;

[0025] The model parameters of the state monitoring model are updated based on the loss calculation results;

[0026] If the loss calculation results obtained based on the updated model parameters of the state monitoring model tend to converge, then the trained state monitoring model corresponding to each unit cluster is obtained.

[0027] In some embodiments, monitoring the state of the wind turbines in each turbine cluster based on the trained state monitoring model corresponding to each turbine cluster includes:

[0028] Obtain operational monitoring data for each wind turbine unit;

[0029] The operation monitoring data of each wind turbine are input into the trained state monitoring model corresponding to the turbine cluster to which each wind turbine belongs, so as to obtain the regression value of the target variable of each wind turbine.

[0030] Based on the regression values ​​of the target variables of each wind turbine and the current measured values ​​of the target variables of each wind turbine, the residuals of the target variables of each wind turbine are calculated.

[0031] The residuals of the target variables for each wind turbine are calculated using the exponentially weighted moving average method, and the target residuals of the target variables for each wind turbine are obtained.

[0032] For any wind turbine, if the target residual of the target variable of the wind turbine is greater than a preset threshold, a status warning will be issued for the wind turbine.

[0033] Secondly, embodiments of this disclosure provide a monitoring device for a wind turbine cluster, the device comprising:

[0034] The acquisition module is used to acquire historical operation monitoring data of each wind turbine and remove abnormal data from the historical operation monitoring data.

[0035] The extraction module is used to extract the distribution characteristics and correlation characteristics of historical operation monitoring data of each wind turbine.

[0036] The clustering module is used to perform clustering based on the distribution characteristics and the correlation characteristics, divide each wind turbine into multiple turbine clusters, and determine the central turbine of each turbine cluster.

[0037] The training module is used to train the status monitoring model based on the historical operation monitoring data of the central unit of each unit cluster, so as to obtain the trained status monitoring model corresponding to each unit cluster.

[0038] The monitoring module is used to monitor the status of the wind turbines in each turbine cluster based on the trained status monitoring model corresponding to each turbine cluster.

[0039] Thirdly, embodiments of this disclosure provide an electronic device, including:

[0040] Memory;

[0041] Processor; and

[0042] Computer programs;

[0043] The computer program is stored in the memory and configured to be executed by the processor to implement the method as described in the first aspect.

[0044] Fourthly, embodiments of this disclosure provide a computer-readable storage medium having a computer program stored thereon, the computer program being executed by a processor to implement the method as described in the first aspect.

[0045] Fifthly, embodiments of this disclosure also provide a computer program product comprising a computer program or instructions that, when executed by a processor, implement the method described in the first aspect.

[0046] The wind turbine cluster monitoring method, apparatus, equipment, and storage medium provided in this disclosure acquire historical operation monitoring data of each wind turbine, remove abnormal data from the historical operation monitoring data, extract the distribution characteristics and correlation characteristics of the historical operation monitoring data of each wind turbine, perform clustering based on the distribution characteristics and correlation characteristics to divide each wind turbine into multiple turbine clusters, determine the central turbine of each turbine cluster, train a state monitoring model based on the historical operation monitoring data of the central turbine of each turbine cluster, obtain a trained state monitoring model corresponding to each turbine cluster, and monitor the state of the wind turbines in each turbine cluster based on the trained state monitoring model corresponding to each turbine cluster. Compared with the prior art, this disclosure can reduce the number of models, has minimal computing resource requirements, is conducive to edge computing and model updates and iterations, and can ensure computing accuracy, thus meeting the needs of predictive maintenance. Attached Figure Description

[0047] The accompanying drawings, which are incorporated in and form a part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure.

[0048] To more clearly illustrate the technical solutions in the embodiments of this disclosure or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0049] Figure 1 A flowchart illustrating the monitoring method for a wind turbine cluster provided in this embodiment of the disclosure;

[0050] Figure 2 A flowchart of a monitoring method for a wind turbine cluster provided in another embodiment of this disclosure;

[0051] Figure 3 A flowchart of a monitoring method for a wind turbine cluster provided in another embodiment of this disclosure;

[0052] Figure 4 A schematic diagram of the structure of the monitoring device for a wind turbine cluster provided in an embodiment of this disclosure;

[0053] Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this disclosure. Detailed Implementation

[0054] To better understand the above-mentioned objectives, features, and advantages of this disclosure, the solutions disclosed herein will be further described below. It should be noted that, unless otherwise specified, the embodiments and features described herein can be combined with each other.

[0055] Numerous specific details are set forth in the following description in order to provide a full understanding of this disclosure, but this disclosure may also be implemented in other ways different from those described herein; obviously, the embodiments in the specification are only some, and not all, of the embodiments of this disclosure.

[0056] Wind farms typically employ corrective and preventative maintenance strategies. However, due to their inherent lag, these strategies often result in significant economic losses and even safety incidents caused by downtime. Predictive maintenance is a more ideal approach. With the development of artificial intelligence, predictive maintenance, based on condition monitoring and early fault detection, is rapidly evolving.

[0057] However, existing predictive maintenance methods typically focus on individual wind turbines rather than the entire wind farm, resulting in a "one turbine, one policy" approach. Although researchers have significantly improved the generalization performance of models, the current situation of one model corresponding to one turbine remains unresolved, leading to a massive number of models. Furthermore, as models develop further, the number of model parameters increases dramatically, drastically increasing the demand for computing resources. Due to limitations in edge computing capabilities at wind farms, these models are difficult to deploy and compute at the wind farm's edge.

[0058] To address this issue, this disclosure provides a method for monitoring wind turbine clusters, which will be described below with reference to specific embodiments.

[0059] Figure 1 This is a flowchart illustrating a monitoring method for a wind turbine cluster provided in this embodiment. The method can be applied to electronic devices, such as portable mobile devices like tablets and laptops, or fixed devices like personal computers and servers. The server can be a single server, a server cluster, a distributed cluster, or a centralized cluster. This method can be applied to scenarios involving the monitoring of wind turbine clusters. It reduces the number of models and model parameters, minimizing computational resource requirements, which is beneficial for edge computing and model updates, while ensuring computational accuracy and meeting the needs of predictive maintenance.

[0060] It is understood that the monitoring method for wind turbine clusters provided in this disclosure can also be applied to other scenarios.

[0061] The following is about Figure 1 The monitoring method for the wind turbine cluster shown is introduced, which includes the following steps:

[0062] S101. Obtain historical operation monitoring data for each wind turbine unit and remove abnormal data from the historical operation monitoring data.

[0063] In this step, the electronic equipment acquires historical operation monitoring data of each wind turbine in the wind farm, further filtering out abnormal data to obtain normal data for each wind turbine. Abnormal data includes, but is not limited to, shutdowns, power curtailment, and other abnormal power generation states. Optionally, the historical operation monitoring data can be Supervisory Control and Data Acquisition (SCADA) data, without limitation.

[0064] In some embodiments, after obtaining the historical operation monitoring data of each wind turbine and removing abnormal data from the historical operation monitoring data, the method further includes: normalizing the removed historical operation monitoring data of each wind turbine to obtain the target data of each wind turbine.

[0065] In this embodiment, in order to eliminate the influence of different dimensions between parameters, the input data is normalized according to the maximum-minimum normalization method to obtain the target data of each wind turbine, which facilitates subsequent data processing.

[0066] S102. Extract the distribution characteristics and correlation characteristics of the historical operation monitoring data of each wind turbine.

[0067] In this step, after acquiring historical operation monitoring data, the electronic equipment extracts the distribution and correlation characteristics of the historical operation monitoring data for each wind turbine. Optionally, distribution characteristics include, but are not limited to, mean, variance, skewness, and kurtosis. Correlation characteristics can be correlation coefficients, such as the Pearson correlation coefficient, without specific limitations.

[0068] In some embodiments, S102 may include, but is not limited to, S1021, S1022, and S1023:

[0069] S1021. Perform first-order alignment on the historical operation monitoring data of each wind turbine to obtain the alignment data of each wind turbine.

[0070] In this step, the historical operation monitoring data of each wind turbine are aligned in the first order to obtain the aligned data of each wind turbine. This can reduce the data distribution difference between the target domain and the source domain and improve the generalization ability of the model while preserving the original data volatility.

[0071] S1022. Extract features from the alignment data of each wind turbine to obtain the distribution features of the historical operation monitoring data of each wind turbine. The distribution features include mean, variance, skewness and kurtosis.

[0072] In this step, the electronic device will extract features from the alignment data of each wind turbine, analyze the mean, variance, skewness and kurtosis of the data, and extract the distribution characteristics of the historical operation monitoring data of each wind turbine.

[0073] S1023. Calculate the correlation coefficient between each input variable in the historical operation monitoring data of each wind turbine, and use the correlation coefficient between each input variable as the correlation feature of the historical operation monitoring data of each wind turbine.

[0074] To account for the mutual influence among the input variables, the correlation coefficients among the input variables in the historical operation monitoring data of each wind turbine are further calculated, and the correlation coefficients among the input variables are used as the correlation characteristics of the historical operation monitoring data of each wind turbine.

[0075] S103. Based on the distribution characteristics and the correlation characteristics, clustering is performed to divide each wind turbine into multiple turbine clusters, and the central turbine of each turbine cluster is determined.

[0076] In this step, the electronic device can cluster wind turbines based on the distribution and correlation characteristics, dividing each wind turbine into multiple clusters and identifying the central turbine in each cluster. Clustering wind turbines using distribution and correlation characteristics allows for the division of historical operational monitoring data, resulting in multiple clusters where each cluster shares a single model, minimizing the number of models required.

[0077] S104. Train the status monitoring model based on the historical operation monitoring data of the central unit of each unit cluster to obtain the trained status monitoring model corresponding to each unit cluster.

[0078] In this step, after identifying the central unit of each unit cluster, the electronic equipment can train the condition monitoring model based on the historical operation monitoring data of the central unit of each unit cluster, thereby obtaining the trained condition monitoring model corresponding to each unit cluster. Optionally, the condition monitoring model can be a spiking neural network driven by a neural circuit strategy, without specific limitations.

[0079] S105. Monitor the status of wind turbines in each turbine cluster based on the trained status monitoring model corresponding to each turbine cluster.

[0080] After obtaining the trained state monitoring model corresponding to each turbine cluster, the electronic equipment can monitor the state of the wind turbines in each turbine cluster based on the trained state monitoring model corresponding to each turbine cluster, so as to realize the state monitoring and predictive maintenance of each wind turbine in the wind farm.

[0081] This embodiment of the disclosure acquires historical operation monitoring data of each wind turbine, removes abnormal data from the historical operation monitoring data, extracts the distribution characteristics and correlation characteristics of the historical operation monitoring data of each wind turbine, performs clustering based on the distribution characteristics and correlation characteristics, divides each wind turbine into multiple turbine clusters, determines the central turbine of each cluster, and trains a state monitoring model based on the historical operation monitoring data of the central turbine of each cluster, obtaining a trained state monitoring model corresponding to each turbine cluster. The state of the wind turbines in each cluster is monitored based on the trained state monitoring model corresponding to each turbine cluster. Compared with the prior art, this embodiment of the disclosure can reduce the number of models, has minimal computing resource requirements, is conducive to edge computing and model updates and iterations, while ensuring computational accuracy and meeting the needs of predictive maintenance.

[0082] Figure 2 Here is a flowchart of a monitoring method for a wind turbine cluster provided in another embodiment of this disclosure, as shown below. Figure 2 As shown, the method includes the following steps:

[0083] S301. Obtain historical operation monitoring data for each wind turbine unit and remove abnormal data from the historical operation monitoring data.

[0084] Specifically, the implementation process and principle of S301 and S101 are the same, and will not be repeated here.

[0085] S302. Extract the distribution characteristics and correlation characteristics of the historical operation monitoring data of each wind turbine.

[0086] Specifically, the implementation process and principle of S302 and S102 are the same, and will not be repeated here.

[0087] S303. Based on the distribution characteristics and the correlation characteristics, construct the feature vector of each wind turbine.

[0088] In this step, the correlation characteristics of the historical operation monitoring data of each wind turbine are used as part of the data features, and together with the distribution characteristics of the historical operation monitoring data of each wind turbine, they constitute the feature vector of each wind turbine.

[0089] S304. The feature vectors of each wind turbine are clustered using a preset clustering algorithm to obtain multiple turbine clusters, and the wind turbine closest to the cluster center in each cluster is determined as the central turbine.

[0090] In this step, the electronic equipment uses a preset clustering algorithm to cluster the feature vectors of each wind turbine, resulting in multiple turbine clusters. The wind turbine closest to the cluster center is designated as the center of the cluster and thus the central turbine. Optionally, the preset clustering algorithm can be K-means clustering or other clustering algorithms; no specific limitation is made.

[0091] S305. Construct a state monitoring model based on neural circuit strategy and spiking neural network.

[0092] In this step, the electronic device constructs a state monitoring model based on a neural circuit strategy and a spiking neural network. Specifically, the electronic device drives a spiking neural network based on a neural circuit strategy to construct a state monitoring model. The constructed model has very few parameters and requires very few computing resources, which is beneficial for edge computing and model updates and iterations.

[0093] In some embodiments, a state monitoring model is obtained by connecting spiking neurons of a spiking neural network using liquid time constant neurons based on a neural circuit strategy.

[0094] In this embodiment, the electronic device connects the spiking neurons of the spiking neural network based on the liquid time constant neurons of the neural circuit strategy to obtain a state monitoring model. Optionally, the state monitoring model can be represented as follows: the first layer of spiking neurons and the second layer of spiking neurons are connected through the first layer of liquid time constant neurons, the second layer of spiking neurons and the third layer of spiking neurons are connected through the second layer of liquid time constant neurons, and so on.

[0095] Optionally, for the neural circuit strategy (NCP), there can be D layers of liquid time constant neurons, and the hidden state, i.e., the output, of each layer of liquid time constant neurons is determined by solving the following equation:

[0096]

[0097] Wherein, at time step t, x(t) defines the hidden state of each layer of liquid time constant neurons, denoted by x. D I(t) represents the hidden state of the D-th layer of liquid time constant neurons. I(t) is the exogenous input variable of the system, which can have m exogenous input variables, denoted by Id. m (t) represents m exogenous inputs. ω τ It is a time constant parameter vector, used This represents the time constant parameter vector of layer D, where A is the bias vector. D This represents the bias vector of the D-th layer. f() is... For a neural network with parameters, ⊙ is the Hadamard product.

[0098] Typically, an exact solution to the hidden state is obtained using an ordinary differential equation solver. However, considering the limited computational resources and time-sensitive requirements of wind farms, an approximate closed-form solution is obtained using a closed continuous-time network, as shown in the following equation:

[0099]

[0100] Spiking neural networks (SNNs) mimic the action pulses used by biological neurons to transmit information. This simulation method endows SNNs with superior performance in processing temporal information and analyzing temporal correlations. Furthermore, the sequence of 0 and 1 pulses can degenerate computation into an accumulation calculation, significantly reducing energy consumption and computational performance requirements.

[0101] In this embodiment, the basic unit of the spiking neural network is the spiking neuron, also known as a leaky integral discharge neuron. In practical applications, discrete difference equations are often used to approximate continuous differential equations. From the perspective of difference equations, the dynamic process of a spiking neuron is represented as follows:

[0102] V(t)=H(t-Δt)+I(t)

[0103]

[0104] H(t)=V reset S(t)+(1-S(t))βV(t)

[0105]

[0106] Where H(t) is the output at time step t. S(t) is the impulse defined as the Herveside step function. V reset It is a reset potential. In this disclosure, V is taken. reset =V rest =0. I(t) has a specific definition, representing x as input, The input is calculated by the function f1 with parameters at time step t.

[0107] When the threshold is reached, the spiking neuron will emit a pulse, and the output H(t) will be reset to V. reset Otherwise, the spiking neuron will not emit a pulse, and the membrane potential will decay at a rate β to H(t), as shown in the following equation:

[0108]

[0109] Since spiking neurons are not differentiable, backpropagation cannot be directly used to train spiking neural networks. Therefore, this disclosure uses the Sigmoid function as an alternative gradient. The impulse S(t) is expressed as follows:

[0110]

[0111] In the formula, α is a hyperparameter that controls the steepness of the Sigmoid function.

[0112] S306. For the central unit of each unit cluster, the historical operation monitoring data of the central unit is input into the status monitoring model, and the regression value of the target variable of the central unit is output through the status monitoring model.

[0113] In this step, for the central unit of each unit cluster, the historical operation monitoring data of the central unit is input into the condition monitoring model, and the condition monitoring model outputs the regression value of the target variable of the central unit. Based on the input data, the model obtains the regression value of the target variable.

[0114] S307. Based on the measured value of the target variable corresponding to the historical operation monitoring data of the central unit, the regression value of the target variable is calculated for loss, and the loss calculation result is obtained.

[0115] In this step, the electronic device can calculate the loss by comparing the measured value of the target variable with the regression value of the target variable, thus obtaining the loss calculation result. The closer the regression value is to the measured value, the smaller the loss calculation, and the better the model performance.

[0116] S308. Update the model parameters of the state monitoring model based on the loss calculation results.

[0117] After obtaining the loss calculation results, the electronic device can update the model parameters of the state monitoring model based on the loss calculation results.

[0118] S309. If the loss calculation results obtained based on the state monitoring model after updating the model parameters tend to converge, then the trained state monitoring model corresponding to each unit cluster is obtained.

[0119] In this step, if the loss calculation results obtained based on the state monitoring model after updating the model parameters tend to converge, that is, the loss calculation results tend to stabilize, it is determined that the state monitoring model training is complete, and the trained state monitoring model corresponding to each unit cluster is obtained.

[0120] S310. Monitor the status of wind turbines in each turbine cluster based on the trained status monitoring model corresponding to each turbine cluster.

[0121] Specifically, the implementation process and principle of S310 and S105 are the same, and will not be repeated here.

[0122] This embodiment of the disclosure acquires historical operation monitoring data of each wind turbine, removes abnormal data from the historical operation monitoring data, extracts the distribution characteristics and correlation characteristics of the historical operation monitoring data of each wind turbine, and constructs feature vectors for each wind turbine based on the distribution characteristics and correlation characteristics. Further, a preset clustering algorithm is used to cluster the feature vectors of each wind turbine, resulting in multiple turbine clusters. The wind turbine closest to the cluster center in each cluster is determined as the central turbine. A state monitoring model is constructed based on a neural circuit strategy and a spiking neural network. Next, for the central turbine of each turbine cluster, the historical operation monitoring data of the central turbine is input into the state monitoring model. The state monitoring model outputs the regression value of the target variable of the central turbine. Based on the measured value of the target variable corresponding to the historical operation monitoring data of the central turbine, a loss calculation is performed on the regression value of the target variable to obtain the loss calculation result. The model parameters of the state monitoring model are updated based on the loss calculation result. Then, if the loss calculation result obtained based on the state monitoring model with updated model parameters tends to converge, the trained state monitoring model corresponding to each turbine cluster is obtained. Furthermore, the status of the wind turbines in each turbine cluster is monitored based on the trained status monitoring model corresponding to each cluster. Through this method, the embodiments of this disclosure can reduce the number of models and the number of model parameters, minimizing the demand for computing resources. This is beneficial for edge computing and model updates and iterations, while ensuring computational accuracy and meeting the needs of predictive maintenance.

[0123] Figure 3 Here is a flowchart of a monitoring method for a wind turbine cluster provided in another embodiment of this disclosure, as shown below. Figure 3 As shown, the method includes the following steps:

[0124] S401. Obtain historical operation monitoring data for each wind turbine unit and remove abnormal data from the historical operation monitoring data.

[0125] Specifically, the implementation process and principle of S401 and S101 are the same, and will not be repeated here.

[0126] S402. Extract the distribution characteristics and correlation characteristics of the historical operation monitoring data of each wind turbine.

[0127] Specifically, the implementation process and principle of S402 and S102 are the same, and will not be repeated here.

[0128] S403. Based on the distribution characteristics and the correlation characteristics, clustering is performed to divide each wind turbine into multiple turbine clusters, and the central turbine of each turbine cluster is determined.

[0129] Specifically, the implementation process and principle of S403 and S103 are the same, and will not be repeated here.

[0130] S404. The status monitoring model is trained based on the historical operation monitoring data of the central unit of each unit cluster to obtain the trained status monitoring model corresponding to each unit cluster.

[0131] Specifically, the implementation process and principle of S404 and S104 are the same, and will not be repeated here.

[0132] S405. Obtain the operation monitoring data of each wind turbine unit.

[0133] In this step, the electronic equipment acquires real-time operational monitoring data from each wind turbine. This operational monitoring data can be SCADA data and is not limited to any particular type.

[0134] S406. Input the operation monitoring data of each wind turbine into the trained state monitoring model corresponding to the turbine cluster to which each wind turbine belongs, and obtain the regression value of the target variable of each wind turbine.

[0135] In this step, the operation monitoring data of each wind turbine is input into the pre-trained state monitoring model corresponding to the turbine cluster to which each wind turbine belongs. The pre-trained state monitoring model then outputs the regression value of the target variable for each wind turbine. The regression value represents the normal reference value of the target variable in the current state. The closer the measured value is to the regression value, the normal the state; a large deviation between the measured value and the regression value indicates an abnormal state.

[0136] S407. Based on the regression values ​​of the target variables of each wind turbine and the current measured values ​​of the target variables of each wind turbine, calculate the residuals of the target variables of each wind turbine.

[0137] In this step, the difference between the regression value of the target variable of each wind turbine output by the trained state monitoring model and the current measured value of the target variable of each wind turbine is determined as the residual of the target variable of each wind turbine.

[0138] S408. The residuals of the target variables of each wind turbine are calculated using the exponentially weighted moving average method to obtain the target residuals of the target variables of each wind turbine.

[0139] Furthermore, the residuals of the target variables for each wind turbine are calculated using the Exponentially Weighted Moving Average (EWMA) method, thus obtaining the target residuals of the target variables for each wind turbine.

[0140] S409. For any wind turbine, if the target residual of the target variable of the wind turbine is greater than the preset threshold, then a status warning is issued for the wind turbine.

[0141] In this step, a preset threshold is set. For any wind turbine, it is determined whether the target residual of the target variable of the wind turbine is greater than the preset threshold. If the target residual of the target variable of the wind turbine is greater than the preset threshold, a status warning is issued for the wind turbine.

[0142] This embodiment of the disclosure acquires historical operation monitoring data of each wind turbine, removes abnormal data from the historical operation monitoring data, extracts the distribution characteristics and correlation characteristics of the historical operation monitoring data of each wind turbine, and performs clustering based on the distribution characteristics and correlation characteristics to divide each wind turbine into multiple turbine clusters, determining the central turbine of each cluster. Next, the condition monitoring model is trained based on the historical operation monitoring data of the central turbine of each turbine cluster to obtain the trained condition monitoring model corresponding to each turbine cluster. The operation monitoring data of each wind turbine is acquired and input into the trained condition monitoring model corresponding to the turbine cluster to which each wind turbine belongs to obtain the regression value of the target variable of each wind turbine. Further, based on the regression value of the target variable of each wind turbine and the current measured value of the target variable of each wind turbine, the residual of the target variable of each wind turbine is calculated. The exponentially weighted moving average method is used to calculate the residual of the target variable of each wind turbine to obtain the target residual of the target variable of each wind turbine. Furthermore, for any wind turbine, if the target residual of the target variable of the wind turbine is greater than a preset threshold, a status warning is issued for the wind turbine. Through this method, the embodiments of this disclosure can reduce the number of models and the number of model parameters, minimizing the demand for computing resources, which is beneficial for edge computing and model updates and iterations, while ensuring computational accuracy and meeting the needs of predictive maintenance.

[0143] Figure 4 This is a schematic diagram of the structure of a monitoring device for a wind turbine cluster provided in an embodiment of this disclosure. The monitoring device for the wind turbine cluster can be an electronic device as described in the above embodiment, or it can be a component or assembly within that electronic device. The monitoring device for a wind turbine cluster provided in this embodiment can execute the processing flow provided in the embodiments of the monitoring method for wind turbine clusters, such as... Figure 4As shown, the monitoring device 50 for the wind turbine cluster includes: an acquisition module 51, an extraction module 52, a clustering module 53, a training module 54, and a monitoring module 55. The acquisition module 51 acquires historical operation monitoring data for each wind turbine and removes abnormal data from the historical operation monitoring data. The extraction module 52 extracts the distribution and correlation characteristics of the historical operation monitoring data for each wind turbine. The clustering module 53 performs clustering based on the distribution and correlation characteristics, dividing each wind turbine into multiple turbine clusters and determining the central turbine of each cluster. The training module 54 trains the status monitoring model based on the historical operation monitoring data of the central turbine of each turbine cluster, obtaining a trained status monitoring model for each turbine cluster. The monitoring module 55 monitors the status of the wind turbines in each turbine cluster based on the trained status monitoring model for each turbine cluster.

[0144] Optionally, when the extraction module 52 extracts the distribution features and correlation features of the historical operation monitoring data of each wind turbine, it is specifically used to: perform first-order alignment on the historical operation monitoring data of each wind turbine to obtain aligned data of each wind turbine; perform feature extraction on the aligned data of each wind turbine to obtain the distribution features of the historical operation monitoring data of each wind turbine, wherein the distribution features include mean, variance, skewness and kurtosis; calculate the correlation coefficient between each input variable in the historical operation monitoring data of each wind turbine, and use the correlation coefficient between each input variable as the correlation feature of the historical operation monitoring data of each wind turbine.

[0145] Optionally, when the clustering module 53 performs clustering based on the distribution features and the correlation features to divide each wind turbine into multiple turbine clusters and determines the central turbine of each cluster, it is specifically used to: construct feature vectors for each wind turbine based on the distribution features and the correlation features; cluster the feature vectors of each wind turbine using a preset clustering algorithm to obtain multiple turbine clusters, and determine the wind turbine closest to the cluster center in each turbine cluster as the central turbine.

[0146] Optionally, before training the state monitoring model based on the historical operation monitoring data of the central unit of each unit cluster to obtain the trained state monitoring model corresponding to each unit cluster, the monitoring device 50 of the wind turbine cluster further includes: a construction module 56, used to construct the state monitoring model based on the neural circuit strategy and the spiking neural network.

[0147] Optionally, when the construction module 56 constructs a state monitoring model based on the neural circuit strategy and the spiking neural network, it is specifically used to: connect the liquid time constant neurons based on the neural circuit strategy to the spiking neurons of the spiking neural network to obtain the state monitoring model.

[0148] Optionally, when the training module 54 trains the status monitoring model based on the historical operation monitoring data of the central unit of each unit cluster to obtain the trained status monitoring model corresponding to each unit cluster, it is specifically used for: for the central unit of each unit cluster, inputting the historical operation monitoring data of the central unit into the status monitoring model, and outputting the regression value of the target variable of the central unit through the status monitoring model; performing loss calculation on the regression value of the target variable based on the measured value of the target variable corresponding to the historical operation monitoring data of the central unit, and obtaining the loss calculation result; updating the model parameters of the status monitoring model based on the loss calculation result; if the loss calculation result obtained based on the status monitoring model after updating the model parameters tends to converge, then the trained status monitoring model corresponding to each unit cluster is obtained.

[0149] Optionally, when the monitoring module 55 monitors the status of the wind turbines in each turbine cluster based on the trained status monitoring model corresponding to each turbine cluster, it is specifically used for: acquiring the operation monitoring data of each wind turbine; inputting the operation monitoring data of each wind turbine into the trained status monitoring model corresponding to the turbine cluster to which each wind turbine belongs, to obtain the regression value of the target variable of each wind turbine; calculating the residual of the target variable of each wind turbine based on the regression value of the target variable of each wind turbine and the current measured value of the target variable of each wind turbine; calculating the residual of the target variable of each wind turbine using the exponential weighted moving average method, to obtain the target residual of the target variable of each wind turbine; and for any wind turbine, if the target residual of the target variable of the wind turbine is greater than a preset threshold, issuing a status warning for the wind turbine.

[0150] Figure 4 The monitoring device for the wind turbine cluster in the illustrated embodiment can be used to execute the technical solution of the above method embodiment. Its implementation principle and technical effect are similar, and will not be described again here.

[0151] Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this disclosure. The electronic device provided in this embodiment can execute the processing flow provided in the embodiment of the wind turbine cluster monitoring method, such as... Figure 5 As shown, the electronic device 80 includes: a memory 81, a processor 82, a computer program, and a communication interface 83; wherein the computer program is stored in the memory 81 and is configured to be executed by the processor 82 as described above for monitoring the wind turbine cluster.

[0152] In addition, this disclosure also provides a computer-readable storage medium having a computer program stored thereon, the computer program being executed by a processor to implement the wind turbine cluster monitoring method described in the above embodiments.

[0153] Furthermore, this disclosure also provides a computer program product, which includes a computer program or instructions that, when executed by a processor, implement the monitoring method for wind turbine clusters as described above.

[0154] It should be noted that the computer-readable medium described in this disclosure can be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this disclosure, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in connection with an instruction execution system, apparatus, or device. In this disclosure, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.

[0155] In some implementations, clients and servers can communicate using any currently known or future-developed network protocol, such as HTTP (Hypertext Transfer Protocol), and can interconnect with digital data communication (e.g., communication networks) of any form or medium. Examples of communication networks include local area networks (“LANs”), wide area networks (“WANs”), the Internet (e.g., the Internet of Things), and peer-to-peer networks (e.g., ad hoc peer-to-peer networks), as well as any currently known or future-developed networks.

[0156] The aforementioned computer-readable medium may be included in the aforementioned electronic device; or it may exist independently and not assembled into the electronic device.

[0157] The aforementioned computer-readable medium carries one or more programs that, when executed by the electronic device, cause the electronic device to:

[0158] Obtain historical operation monitoring data for each wind turbine unit, and remove abnormal data from the historical operation monitoring data;

[0159] Extract the distribution and correlation characteristics of historical operation monitoring data for each wind turbine unit;

[0160] Clustering is performed based on the distribution characteristics and correlation characteristics to divide each wind turbine into multiple turbine clusters and determine the central turbine of each cluster.

[0161] The condition monitoring model is trained based on the historical operation monitoring data of the central unit of each unit cluster, so as to obtain the trained condition monitoring model corresponding to each unit cluster.

[0162] The status of the wind turbines in each cluster is monitored based on the pre-trained status monitoring model corresponding to each cluster.

[0163] In addition, the electronic device can also perform other steps in the monitoring method for wind turbine clusters as described above.

[0164] Computer program code for performing the operations of this disclosure can be written in one or more programming languages ​​or a combination thereof, including but not limited to object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0165] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0166] The units described in the embodiments of this disclosure can be implemented in software or hardware. The names of the units are not, in some cases, intended to limit the specific unit.

[0167] The functions described above in this document can be performed, at least in part, by one or more hardware logic components. For example, exemplary types of hardware logic components that can be used, without limitation, include: Field Programmable Gate Arrays (FPGAs), Application-Specific Integrated Circuits (ASICs), Application Standard Products (ASSPs), System-on-Chip (SoCs), Complex Programmable Logic Devices (CPLDs), and so on.

[0168] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0169] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0170] The above description is merely a specific embodiment of this disclosure, enabling those skilled in the art to understand or implement it. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this disclosure. Therefore, this disclosure is not to be limited to the embodiments described herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A monitoring method for a cluster of wind turbine generators, characterized in that, The method includes: Obtain historical operation monitoring data for each wind turbine unit, and remove abnormal data from the historical operation monitoring data; Extract the distribution and correlation characteristics of historical operation monitoring data for each wind turbine unit; Clustering is performed based on the distribution characteristics and correlation characteristics to divide each wind turbine into multiple turbine clusters and determine the central turbine of each cluster. The condition monitoring model is trained based on the historical operation monitoring data of the central unit of each unit cluster, so as to obtain the trained condition monitoring model corresponding to each unit cluster. The status of the wind turbines in each turbine cluster is monitored based on the trained status monitoring model corresponding to each turbine cluster. The extraction of distribution and correlation characteristics of historical operation monitoring data for each wind turbine includes: First-order alignment is performed on the historical operation monitoring data of each wind turbine to obtain the alignment data of each wind turbine. Feature extraction is performed on the alignment data of each wind turbine to obtain the distribution characteristics of the historical operation monitoring data of each wind turbine. The distribution characteristics include mean, variance, skewness and kurtosis. Calculate the correlation coefficients between each input variable in the historical operation monitoring data of each wind turbine, and use the correlation coefficients between each input variable as the correlation characteristics of the historical operation monitoring data of each wind turbine. The clustering based on the distribution characteristics and correlation characteristics divides each wind turbine into multiple turbine clusters, and determines the central turbine of each cluster, including: Based on the distribution characteristics and the correlation characteristics, feature vectors for each wind turbine are constructed. A pre-defined clustering algorithm is used to cluster the feature vectors of each wind turbine to obtain multiple turbine clusters, and the wind turbine closest to the cluster center in each cluster is determined as the central turbine. The process of training the condition monitoring model based on the historical operation monitoring data of the central unit of each unit cluster to obtain the trained condition monitoring model corresponding to each unit cluster includes: For the central unit of each unit cluster, the historical operation monitoring data of the central unit is input into the status monitoring model, and the regression value of the target variable of the central unit is output through the status monitoring model. Based on the measured values ​​of the target variable corresponding to the historical operation monitoring data of the central unit, the regression value of the target variable is calculated to obtain the loss calculation result; The model parameters of the state monitoring model are updated based on the loss calculation results; If the loss calculation results obtained based on the state monitoring model after updating the model parameters tend to converge, then the trained state monitoring model corresponding to each unit cluster is obtained. The monitoring of the state of the wind turbines in each turbine cluster based on the trained state monitoring model corresponding to each turbine cluster includes: Obtain operational monitoring data for each wind turbine unit; The operation monitoring data of each wind turbine are input into the trained state monitoring model corresponding to the turbine cluster to which each wind turbine belongs, so as to obtain the regression value of the target variable of each wind turbine. Based on the regression values ​​of the target variables of each wind turbine and the current measured values ​​of the target variables of each wind turbine, the residuals of the target variables of each wind turbine are calculated. The residuals of the target variables for each wind turbine are calculated using the exponentially weighted moving average method, and the target residuals of the target variables for each wind turbine are obtained. For any wind turbine, if the target residual of the target variable of the wind turbine is greater than a preset threshold, a status warning will be issued for the wind turbine.

2. The monitoring method for a wind turbine cluster according to claim 1, characterized in that, Before training the condition monitoring model based on the historical operation monitoring data of the central unit of each unit cluster to obtain the trained condition monitoring model for each unit cluster, the method further includes: A state monitoring model is constructed based on neural circuit strategies and spiking neural networks.

3. The monitoring method for a wind turbine cluster according to claim 2, characterized in that, The state monitoring model constructed based on neural circuit strategies and spiking neural networks includes: A state monitoring model is obtained by connecting spiking neurons in a spiking neural network using liquid time constant neurons based on a neural circuit strategy.

4. A monitoring device for a wind turbine cluster, characterized in that, The device includes: The acquisition module is used to acquire historical operation monitoring data of each wind turbine and remove abnormal data from the historical operation monitoring data. The extraction module is used to extract the distribution characteristics and correlation characteristics of historical operation monitoring data of each wind turbine. The clustering module is used to perform clustering based on the distribution characteristics and the correlation characteristics, divide each wind turbine into multiple turbine clusters, and determine the central turbine of each turbine cluster. The training module is used to train the status monitoring model based on the historical operation monitoring data of the central unit of each unit cluster, so as to obtain the trained status monitoring model corresponding to each unit cluster. The monitoring module is used to monitor the status of the wind turbines in each turbine cluster based on the trained status monitoring model corresponding to each turbine cluster. When the extraction module extracts the distribution and correlation features of the historical operation monitoring data of each wind turbine, it specifically performs the following: first-order alignment of the historical operation monitoring data of each wind turbine to obtain aligned data for each wind turbine; feature extraction of the aligned data of each wind turbine to obtain the distribution features of the historical operation monitoring data of each wind turbine, wherein the distribution features include mean, variance, skewness, and kurtosis; and calculation of the correlation coefficients between each input variable in the historical operation monitoring data of each wind turbine, using the correlation coefficients between each input variable as the correlation features of the historical operation monitoring data of each wind turbine. The clustering module performs clustering based on the distribution characteristics and the correlation characteristics, dividing each wind turbine into multiple turbine clusters. When determining the central turbine of each cluster, it specifically performs the following: constructing feature vectors for each wind turbine based on the distribution characteristics and the correlation characteristics; using a preset clustering algorithm to cluster the feature vectors of each wind turbine to obtain multiple turbine clusters, and determining the wind turbine closest to the cluster center in each cluster as the central turbine. The training module trains the status monitoring model based on the historical operation monitoring data of the central unit of each unit cluster to obtain the trained status monitoring model for each unit cluster. Specifically, it is used for: inputting the historical operation monitoring data of the central unit into the status monitoring model for each unit cluster, and outputting the regression value of the target variable of the central unit through the status monitoring model; performing loss calculation on the regression value of the target variable based on the measured value of the target variable corresponding to the historical operation monitoring data of the central unit to obtain the loss calculation result; updating the model parameters of the status monitoring model based on the loss calculation result; if the loss calculation result obtained based on the status monitoring model after updating the model parameters tends to converge, then the trained status monitoring model for each unit cluster is obtained. When the monitoring module monitors the status of wind turbines in each turbine cluster based on the trained status monitoring model corresponding to each turbine cluster, it specifically performs the following functions: acquiring the operation monitoring data of each wind turbine; inputting the operation monitoring data of each wind turbine into the trained status monitoring model corresponding to the turbine cluster to which each wind turbine belongs, to obtain the regression value of the target variable of each wind turbine; calculating the residual of the target variable of each wind turbine based on the regression value of the target variable of each wind turbine and the current measured value of the target variable of each wind turbine; calculating the target residual of the target variable of each wind turbine using the exponentially weighted moving average method, to obtain the target residual of the target variable of each wind turbine; and for any wind turbine, if the target residual of the target variable of the wind turbine is greater than a preset threshold, issuing a status warning for the wind turbine.

5. An electronic device, characterized in that, include: Memory; processor; as well as Computer programs; The computer program is stored in the memory and configured to be executed by the processor to implement a monitoring method for a wind turbine cluster as described in any one of claims 1-3.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements a monitoring method for a wind turbine cluster as described in any one of claims 1-3.

Images