Wind driven generator slip ring fault early warning method

Abstract

The invention relates to the technical field of wind driven generators, in particular to a wind driven generator slip ring fault early warning method, which comprises the following steps: acquiring slip ring operation data and historical slip ring operation data of a wind driven generator, and constructing a slip ring fault convolutional network model, historical slip ring operation data of the wind driven generator is divided into a training set and a test set, the slip ring fault convolutional network model is trained through the training set to obtain a slip ring fault prediction result, a loss value between the test set and the slip ring fault prediction result is calculated, parameters of the slip ring fault convolutional network model are updated through loss value iteration, and the slip ring fault prediction result is obtained. And after the maximum number of iterations is reached, outputting the trained slip ring fault convolutional network model. And calculating slip ring operation data of the wind driven generator through the trained slip ring fault convolutional network model to obtain a slip ring normal state probability and a slip ring fault state probability, judging whether the slip ring fault state probability reaches a threshold value or not, and if yes, executing early warning operation.

Description

Technical Field

[0001] The invention relates to the technical field of wind turbine generators, and in particular to a slip ring fault early warning method for a wind turbine generator. Background Art

[0002] Wind turbines are key equipment for renewable energy utilization, and their slip rings are important components for transmitting electrical energy and signals. The normal operation of slip rings directly affects the performance and life of wind turbines. At present, traditional slip ring fault diagnosis methods mainly rely on expert experience and simple statistical analysis, and have the following disadvantages: 1. Manual diagnosis is inefficient and difficult to achieve real-time monitoring; 2. Simple threshold judgment methods have poor adaptability to complex working conditions and are prone to misjudgment or missed judgments; 3. It is difficult to adapt to the dynamic characteristics changes caused by equipment aging and environmental changes, and the fault can often be discovered only when it develops to a serious stage, which seriously affects the operation and maintenance efficiency and reliability of wind turbines. Based on this, in response to the above problems, we designed a wind turbine slip ring fault early warning method. Summary of the invention

[0003] The purpose of the present invention is to provide a wind turbine slip ring fault early warning method, which, on the one hand, automatically extracts data features of multiple sensors through a slip ring fault convolutional network model to achieve accurate prediction of slip ring faults, has the advantages of strong adaptability and high prediction accuracy, and can significantly improve the operation and maintenance efficiency of wind turbines; on the other hand, by timely warning of potential faults, preventive maintenance can be achieved, maintenance costs can be effectively reduced, equipment life can be extended, and the overall reliability and economic benefits of wind turbines can be improved.

[0004] The embodiments of the present invention are implemented by the following technical solutions:

[0005] A wind turbine slip ring fault early warning method, the method comprising the steps of:

[0006] For the slip ring device of the wind turbine, various sensors are deployed and data is collected at a set collection frequency to obtain the slip ring operation data of the wind turbine;

[0007] Obtain the historical slip ring operation data of the wind turbine, build a slip ring fault convolutional network model, divide the historical slip ring operation data of the wind turbine into a training set and a test set, train the slip ring fault convolutional network model through the training set, obtain the slip ring fault prediction result, and calculate the loss value between the test set and the slip ring fault prediction result, iteratively update the parameters of the slip ring fault convolutional network model through the loss value until the maximum number of iterations is reached, and then output the trained slip ring fault convolutional network model;

[0008] The slip ring operation data of the wind turbine is calculated by the trained slip ring fault convolutional network model to obtain the slip ring normal state probability and slip ring fault state probability, and determine whether the slip ring fault state probability reaches the threshold. If so, an early warning operation is performed.

[0009] Optionally, the multiple sensors include: temperature sensor, vibration sensor, current sensor, voltage sensor and speed sensor; the temperature sensors are respectively arranged on the body of the wind turbine slip ring device, near the bearing and near the contact surface of the carbon brush and the slip ring to obtain the slip ring temperature operating data of the wind turbine; the vibration sensor is arranged on the shell of the wind turbine slip ring device to obtain the slip ring vibration acceleration operating data of the wind turbine; the current sensor and voltage sensor are respectively arranged on the wires at the input and output ends of the wind turbine slip ring device to obtain the slip ring current and voltage operating data of the wind turbine; the speed sensor is arranged on the shaft of the wind turbine slip ring device to obtain the slip ring speed operating data of the wind turbine, and the wind turbine slip ring temperature operating data, the wind turbine slip ring vibration acceleration operating data, the wind turbine slip ring current and voltage operating data and the wind turbine slip ring speed operating data constitute the wind turbine slip ring operating data.

[0010] Optionally, the slip ring operation data of the wind turbine and the historical slip ring operation data of the wind turbine are obtained, and are processed by wavelet transform to remove noise, and each index in the slip ring operation data of the wind turbine and the historical slip ring operation data of the wind turbine is normalized to the interval [0,1].

[0011] Optionally, the slip ring fault convolutional network model is specifically composed of an input layer, a convolutional layer, an activation layer, a pooling layer, a fully connected layer and an output layer, wherein the input data in the input layer is characterized as: Among them, X is the input data, H is the height, W is the width, and C is the number of channels, which represents different types of sensors; the convolution layer is:

[0012]

[0013] Among them, Y i,j,k is the output of the kth filter at the (i, j) position of the output feature, K is the convolution kernel weight, b k is the bias term of the kth filter, M and N are the height and width of the convolution kernel respectively;

[0014] The activation layer adopts the Relu activation function;

[0015] The calculation formula of the pooling layer is:

[0016]

[0017] Among them, Y i,j is the value of the (i, j) position of the output feature, P and Q are the sizes of the pooling window, and s is the step size;

[0018] The calculation formula of the fully connected layer is:

[0019]

[0020] Among them, Y i is the i-th element of the output vector, X j is the jth element of the input vector, W ij is the element of the weight matrix, b i is bias;

[0021] The output layer is characterized as: Among them, Y k is the probability of the kth category, including the probability of the normal state of the slip ring and the probability of the faulty state of the slip ring; K is the total number of categories, z k is the output of the kth neuron in the output layer, z j is the linear output of the fully connected layer.

[0022] Optionally, the parameters of the slip ring fault convolutional network model are iteratively updated through the loss value, and the specific calculation formula is:

[0023]

[0024] Among them, L is the loss function, y i,k is the true label, Prediction output of the convolutional network model for slip ring faults.

[0025] Optionally, during the training process of the slip ring fault convolutional network model, the parameter updating process is as follows:

[0026] Solve the gradient of the loss function L for each parameter in the slip ring fault convolutional network model:

[0027]

[0028] Where L is the loss function, θ is the parameter of the slip ring fault convolutional network model;

[0029] The parameters of the slip ring fault convolutional network model are updated by the stochastic gradient descent optimization algorithm:

[0030]

[0031] Among them, η is the learning rate, t is the current number of iterations;

[0032]

[0033] θ t+1 =θ t -v t+1

[0034] Among them, γ is the momentum parameter,

[0035]

[0036] in, and is a hyperparameter and ε is a constant.

[0037] Optionally, the slip ring operation data of the wind turbine is calculated by the trained slip ring fault convolutional network model, which is specifically:

[0038] Y(normal),Y(fault)=CNN(X new )

[0039] Among them, Y(normal) and Y(fault) are the probability of normal state and fault state of slip ring respectively, X new Slip ring operating data for wind turbines.

[0040] Optionally, the determining whether the slip ring fault state probability reaches a threshold value is specifically as follows:

[0041] Set the threshold to τ;

[0042] Determine whether Y(fault) is greater than or equal to τ. If so, perform a warning operation; if not, cyclically obtain the slip ring operation data of the wind turbine.

[0043] The technical solution of the embodiment of the present invention has at least the following advantages and beneficial effects:

[0044] On the one hand, the embodiment of the present invention automatically extracts data features of multiple sensors through the slip ring fault convolutional network model to achieve accurate prediction of slip ring faults, which has the advantages of strong adaptability and high prediction accuracy, and can significantly improve the operation and maintenance efficiency of wind turbines; on the other hand, by timely warning of potential faults, preventive maintenance can be achieved, maintenance costs can be effectively reduced, equipment life can be extended, and the overall reliability and economic benefits of wind turbines can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 A schematic flow chart of a wind turbine slip ring fault early warning method provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0046] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0047] like Figure 1 As shown, the present invention provides one embodiment: a wind turbine slip ring fault early warning method, the method comprising the steps of:

[0048] For the slip ring device of the wind turbine, various sensors are deployed and data is collected at a set collection frequency to obtain the slip ring operation data of the wind turbine;

[0049] Obtain the historical slip ring operation data of the wind turbine, build a slip ring fault convolutional network model, divide the historical slip ring operation data of the wind turbine into a training set and a test set, train the slip ring fault convolutional network model through the training set, obtain the slip ring fault prediction result, and calculate the loss value between the test set and the slip ring fault prediction result, iteratively update the parameters of the slip ring fault convolutional network model through the loss value until the maximum number of iterations is reached, and then output the trained slip ring fault convolutional network model;

[0050] The slip ring operation data of the wind turbine is calculated by the trained slip ring fault convolutional network model to obtain the slip ring normal state probability and slip ring fault state probability, and determine whether the slip ring fault state probability reaches the threshold. If so, an early warning operation is performed.

[0051] In the implementation, the acquisition frequency is specifically set to 1KHZ, and a set of data is formed at an interval of 10 minutes. Specifically, the various sensors arranged for the slip ring device of the wind turbine in this embodiment include: temperature sensor, vibration sensor, current sensor, voltage sensor and speed sensor; the temperature sensor is arranged on the body of the wind turbine slip ring device, near the bearing and near the contact surface of the carbon brush and the slip ring, so as to obtain the slip ring temperature operation data of the wind turbine; the vibration sensor is arranged on the shell of the wind turbine slip ring device to obtain the slip ring vibration acceleration operation data of the wind turbine; the current sensor and the voltage sensor are arranged on the wires at the input and output ends of the wind turbine slip ring device, so as to obtain the slip ring current and voltage operation data of the wind turbine; the speed sensor is arranged on the shaft of the wind turbine slip ring device to obtain the slip ring speed operation data of the wind turbine, and the slip ring temperature operation data of the wind turbine, the slip ring vibration acceleration operation data of the wind turbine, the slip ring current and voltage operation data of the wind turbine and the slip ring speed operation data of the wind turbine constitute the slip ring operation data of the wind turbine.

[0052] In the specific implementation, the temperature sensor of this embodiment adopts a thermocouple, the vibration sensor adopts a piezoelectric acceleration sensor, the current sensor adopts a Hall effect current sensor, and the speed sensor adopts a photoelectric encoder. It can be understood that other sensors of the same type applicable to the slip ring device of the wind turbine can be selected in this embodiment to realize the corresponding collection function.

[0053] Furthermore, the slip ring operation data of the wind turbine and the historical slip ring operation data of the wind turbine are obtained, and are processed by wavelet transform to remove noise, and each index in the slip ring operation data of the wind turbine and the historical slip ring operation data of the wind turbine is normalized to the interval [0,1].

[0054] In the specific implementation of this embodiment, the slip ring fault convolutional network model is specifically composed of an input layer, a convolutional layer, an activation layer, a pooling layer, a fully connected layer and an output layer, wherein the input data in the input layer is characterized as follows: Among them, X is the input data, H is the height, W is the width, and C is the number of channels, which represents different types of sensors; the convolution layer is:

[0055]

[0056] Among them, Y i,j,k is the output of the kth filter at the (i, j) position of the output feature, K is the convolution kernel weight, b k is the bias term of the kth filter, M and N are the height and width of the convolution kernel respectively;

[0057] The activation layer adopts the Relu activation function;

[0058] The calculation formula of the pooling layer is:

[0059]

[0060] Among them, Y i,j is the value of the (i, j) position of the output feature, P and Q are the sizes of the pooling window, and s is the step size;

[0061] The calculation formula of the fully connected layer is:

[0062]

[0063] Among them, Y i is the i-th element of the output vector, X j is the jth element of the input vector, W ij is the element of the weight matrix, b i is bias;

[0064] The output layer is characterized as: Among them, Y k is the probability of the kth category, including the probability of the normal state of the slip ring and the probability of the faulty state of the slip ring; K is the total number of categories, z k is the output of the kth neuron in the output layer, z j is the linear output of the fully connected layer.

[0065] The above formula of this embodiment describes the forward propagation process of the historical slip ring operation data of the input wind turbine from input to output in the convolutional neural network. Among them, the input data is transformed specifically from the convolutional layer, activation layer, pooling layer, fully connected layer, etc., and finally the prediction result is generated in the output layer.

[0066] In a further implementation of this embodiment, the parameters of the slip ring fault convolutional network model are iteratively updated through the loss value, and the specific calculation formula is:

[0067]

[0068] Among them, L is the loss function, y i,k is the true label, Prediction output of the convolutional network model for slip ring faults.

[0069] After the aforementioned forward propagation is completed, the gradient is calculated through the loss function, and the gradient is passed back to each layer of the convolutional neural network through back propagation.

[0070] Furthermore, this embodiment updates the parameters of the convolutional neural network according to the calculated gradient through the following optimization algorithm, specifically:

[0071] During the training process of the slip ring fault convolutional network model, its parameter updating process is as follows:

[0072] Solve the gradient of the loss function L for each parameter in the slip ring fault convolutional network model:

[0073]

[0074] Where L is the loss function, θ is the parameter of the slip ring fault convolutional network model;

[0075] The parameters of the slip ring fault convolutional network model are updated by the stochastic gradient descent optimization algorithm:

[0076]

[0077] Among them, η is the learning rate, t is the current number of iterations;

[0078]

[0079] θt+1 =θ t -v t+1

[0080] Among them, γ is the momentum parameter,

[0081]

[0082] in, and is a hyperparameter and ε is a constant.

[0083] In implementation, the slip ring fault convolutional network model that has been trained calculates the slip ring operation data of the wind turbine, which is specifically:

[0084] Y(normal),Y(fault)=CNN(X new )

[0085] Among them, Y(normal) and Y(fault) are the probability of normal state and fault state of slip ring respectively, X new Slip ring operating data for wind turbines.

[0086] Specifically, the determination of whether the slip ring fault state probability reaches a threshold value is specifically as follows:

[0087] Set the threshold to τ;

[0088] Determine whether Y(fault) is greater than or equal to τ. If so, perform a warning operation; if not, cyclically obtain the slip ring operation data of the wind turbine.

[0089] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A wind turbine slip ring fault early warning method, characterized in that: The steps of the method include: For the slip ring device of the wind turbine, various sensors are deployed and data is collected at a set collection frequency to obtain the slip ring operation data of the wind turbine; Obtain the historical slip ring operation data of the wind turbine, build a slip ring fault convolutional network model, divide the historical slip ring operation data of the wind turbine into a training set and a test set, train the slip ring fault convolutional network model through the training set, obtain the slip ring fault prediction result, and calculate the loss value between the test set and the slip ring fault prediction result, iteratively update the parameters of the slip ring fault convolutional network model through the loss value until the maximum number of iterations is reached, and then output the trained slip ring fault convolutional network model; The slip ring operation data of the wind turbine is calculated by the trained slip ring fault convolutional network model to obtain the slip ring normal state probability and slip ring fault state probability, and determine whether the slip ring fault state probability reaches the threshold. If so, an early warning operation is performed.

2. The wind turbine generator slip ring fault early warning method according to claim 1, characterized in that: The multiple sensors include: temperature sensor, vibration sensor, current sensor, voltage sensor and speed sensor; the temperature sensor is respectively arranged on the body of the wind turbine slip ring device, near the bearing and near the contact surface between the carbon brush and the slip ring to obtain the slip ring temperature operation data of the wind turbine; the vibration sensor is arranged on the shell of the wind turbine slip ring device to obtain the slip ring vibration acceleration operation data of the wind turbine; the current sensor and voltage sensor are respectively arranged on the wires at the input end and the output end of the wind turbine slip ring device to obtain the slip ring current and voltage operation data of the wind turbine; the speed sensor is arranged on the shaft of the wind turbine slip ring device to obtain the slip ring speed operation data of the wind turbine, and the slip ring temperature operation data of the wind turbine, the slip ring vibration acceleration operation data of the wind turbine, the slip ring current and voltage operation data of the wind turbine and the slip ring speed operation data of the wind turbine constitute the slip ring operation data of the wind turbine.

3. The wind turbine generator slip ring fault early warning method according to claim 2, characterized in that: The slip ring operation data of the wind turbine and the historical slip ring operation data of the wind turbine are obtained, and are processed by wavelet transform to remove noise, and each index in the slip ring operation data of the wind turbine and the historical slip ring operation data of the wind turbine is normalized to the interval [0,1].

4. The wind turbine generator slip ring fault early warning method according to claim 3 is characterized in that: The slip ring fault convolutional network model is specifically composed of an input layer, a convolutional layer, an activation layer, a pooling layer, a fully connected layer and an output layer, wherein the input data in the input layer is characterized as follows: Among them, X is the input data, H is the height, W is the width, and C is the number of channels, which represents different types of sensors; the convolution layer is: Among them, Y i,j,k is the output of the kth filter at the (i, j) position of the output feature, K is the convolution kernel weight, b k is the bias term of the kth filter, M and N are the height and width of the convolution kernel respectively; The activation layer adopts the Relu activation function; The calculation formula of the pooling layer is: Among them, Y i,j is the value of the (i, j) position of the output feature, P and Q are the sizes of the pooling window, and s is the step size; The calculation formula of the fully connected layer is: Among them, Y i is the i-th element of the output vector, X j is the jth element of the input vector, W ij is the element of the weight matrix, b i is bias; The output layer is characterized as: Among them, Y k is the probability of the kth category, including the probability of the normal state of the slip ring and the probability of the faulty state of the slip ring; K is the total number of categories, z k is the output of the kth neuron in the output layer, z j is the linear output of the fully connected layer.

5. The wind turbine generator slip ring fault early warning method according to claim 4, characterized in that: The parameters of the slip ring fault convolutional network model are iteratively updated through the loss value, and the specific calculation formula is: Among them, L is the loss function, y i,k is the true label, Prediction output of the convolutional network model for slip ring faults.

6. The wind turbine slip ring fault early warning method according to claim 5, characterized in that: During the training process of the slip ring fault convolutional network model, its parameter updating process is as follows: Solve the gradient of the loss function L for each parameter in the slip ring fault convolutional network model: Where L is the loss function, θ is the parameter of the slip ring fault convolutional network model; The parameters of the slip ring fault convolutional network model are updated by the stochastic gradient descent optimization algorithm: Among them, η is the learning rate, t is the current number of iterations; i t+1 =θ t -v t+1 Among them, γ is the momentum parameter, in, and is a hyperparameter and ε is a constant.

7. The wind turbine generator slip ring fault early warning method according to claim 6, characterized in that: The slip ring fault convolutional network model trained to calculate the slip ring operation data of the wind turbine is specifically: Y(normal),Y(fault)=CNN(X new ) Among them, Y(normal) and Y(fault) are the probability of normal state and fault state of slip ring respectively, X new Slip ring operating data for wind turbines.

8. The wind turbine generator slip ring fault early warning method according to claim 7, characterized in that: The determination of whether the slip ring fault state probability reaches a threshold value is specifically as follows: Set the threshold to τ; Determine whether Y(fault) is greater than or equal to τ. If so, perform a warning operation; if not, cyclically obtain the slip ring operation data of the wind turbine.

Images