Offshore wind turbine rotor heat dissipation system and method

Abstract

The application relates to the field of new energy, and particularly discloses a wind wheel heat dissipation system of an offshore wind turbine generator set and a method thereof. A deep neural network model is used as a feature extractor to extract dynamic change implicit features of solar heat load, hub heat load, convective heat transfer heat load and thermal radiation in a time dimension, the features are fused to obtain heat source dynamic change features, and then based on the dynamic change implicit features of the temperature of the wind wheel in a time sequence and the dynamic change implicit features of the heat source, the cooling water flow rate value of a water cooling system of the offshore wind turbine generator set is comprehensively and dynamically controlled in real time to optimize the heat dissipation effect of the wind wheel.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of new energy, and more particularly, to a wind wheel heat dissipation system of an offshore wind turbine generator set and a method thereof. BACKGROUND

[0002] The wind turbine generator set mainly consists of a tower, a nacelle and a wind wheel. In order to prevent the erosion of salt mist and rainwater, the wind wheel of the offshore wind turbine generator set is generally completely isolated from the outside world by a sealed structure. Since there are heat generating components such as variable pitch shaft control cabinets, super capacitor cabinets, variable pitch motors, main shafts and cables in the wind wheel, when the unit is running, the heat emitted from the surfaces of the heat generating components in the wind wheel will be discharged into the hub. Especially in the summer high temperature environment, the heat transferred from the outside to the inside of the wind wheel will also increase, thereby causing the temperature in the hub to be too high, affecting the normal operation of the unit, and even causing the unit to shut down due to over-temperature, resulting in loss of power generation.

[0003] The industry mainly adopts a water cooling system scheme for the heat dissipation of the wind wheel. However, the cooling effect of the existing water cooling scheme often cannot meet the application requirements, and the reason is that the control scheme of the water cooling system. Specifically, the existing water cooling system for offshore wind turbine generator sets usually transports cooling liquid at a predetermined flow rate for heat exchange, but since the working conditions of offshore wind turbine generator sets are different, the working environment is different, and the constant flow rate heat dissipation scheme cannot meet the heat dissipation requirements of the changing working conditions and environment.

[0004] Therefore, an optimized wind wheel heat dissipation scheme for offshore wind turbine generator sets is expected to dynamically control the flow rate of the water cooling system in real time, thereby optimizing the heat dissipation effect of the wind wheel. SUMMARY

[0005] In order to solve the above technical problems, the present application is proposed. The embodiments of the present application provide a wind wheel heat dissipation system of an offshore wind turbine generator set and a method thereof, which uses a deep neural network model as a feature extractor to extract the dynamic change implicit features of the solar heat load, the hub heat load, the convective heat transfer heat load and the thermal radiation in the time dimension, so as to mine the change feature information of the heat sources, that is, the various generating factors of the heat in the wind wheel. Then, based on the dynamic change implicit features of the temperature of the wind wheel and the heat sources in the time sequence, the cooling water flow rate value of the water cooling system of the offshore wind turbine generator set is dynamically controlled in real time to optimize the heat dissipation effect of the wind wheel.

[0006] Correspondingly, according to one aspect of the present application, a wind wheel heat dissipation system of an offshore wind turbine generator set is provided, which comprises:

[0007] a data monitoring and collecting module, configured to obtain solar heat load, hub heat load, convective heat transfer heat load and thermal radiation at a plurality of predetermined time points in a predetermined time period, temperature values of the wind wheel at the plurality of predetermined time points, and flow rate values of cooling water of the water cooling system at the plurality of predetermined time points;

[0008] a heat source data encoding module, configured to arrange the solar heat load, the hub heat load, the convective heat transfer heat load and the thermal radiation at the plurality of predetermined time points into input vectors respectively, and then input the input vectors into a time series encoder comprising a one-dimensional convolution layer to obtain first to fourth heat source feature vectors;

[0009] a heat source feature fusion module, configured to fuse the first to fourth heat source feature vectors to obtain a heat increase feature vector;

[0010] a heat feature correction module, configured to correct feature values at each position in the heat increase feature vector based on a mean value and a variance of the feature values at all positions in the heat increase feature vector to obtain a corrected heat increase feature vector;

[0011] a temperature and flow rate encoding module, configured to arrange the temperature values of the wind wheel at the plurality of predetermined time points and the flow rate values of the cooling water of the water cooling system at the plurality of predetermined time points into input vectors respectively, and then input the input vectors into the time series encoder comprising the one-dimensional convolution layer to obtain a temperature feature vector and a flow rate feature vector;

[0012] a Bayesian inference module, configured to fuse the flow rate feature vector, the temperature feature vector and the corrected heat increase feature vector using a Bayesian probability model to obtain a posterior probability vector; and

[0013] a heat dissipation control result generation module, configured to input the posterior probability vector into a classifier to obtain a classification result, the classification result being used to indicate whether the flow rate value of the cooling water at a current time point should be increased or decreased.

[0014] In the wind wheel heat dissipation system of the offshore wind turbine generator set, the heat source data encoding module comprises: an input vector construction unit, configured to arrange the solar heat load, the hub heat load, the convective heat transfer heat load and the thermal radiation at the plurality of predetermined time points into input vectors respectively according to a time dimension to obtain first to fourth input vectors; and a fully connected encoding unit, configured to use a fully connected layer of the time series encoder to perform fully connected encoding on the first to fourth input vectors respectively according to the following formula to extract high-dimensional hidden features of feature values at each position in the first to fourth input vectors respectively, wherein the formula is: wherein X is the first to fourth input vectors, Y is an output vector, W is a weight matrix, and B is a bias vector, represents matrix multiplication; and a one-dimensional convolution encoding unit configured to perform one-dimensional convolution encoding on the first to fourth input vectors respectively using one-dimensional convolution layers of the time series encoder to extract high-dimensional hidden correlation features between feature values at respective positions in the first to fourth input vectors respectively according to the following formula:

[0015]

[0016] wherein a is a width of the convolution kernel in the x direction, F(a) is a parameter vector of the convolution kernel, G(x-a) is a local vector matrix operated with the convolution kernel function, w is a size of the convolution kernel, and X represents the first to fourth input vectors.

[0017] In the wind wheel heat dissipation system of the offshore wind turbine generator set, the heat source feature fusion module comprises: a contribution degree evaluation unit configured to pass each of the first to fourth heat source feature vectors through a pre-classifier to obtain first to fourth probability values; and a fusion unit configured to calculate a position-weighted sum of the first to fourth heat source feature vectors using the first to fourth probability values as weights to obtain the heat increment feature vector.

[0018] In the wind wheel heat dissipation system of the offshore wind turbine generator set, the heat feature correction module is further configured to correct feature values at respective positions in the heat increment feature vector according to the following formula based on a mean and a variance of the feature values at all positions in the heat increment feature vector to obtain the corrected heat increment feature vector:

[0019]

[0020] wherein v i represents the feature values at respective positions in the heat increment feature vector, μ and σ respectively represent a mean and a variance of a set of the feature values at all positions in the heat increment feature vector, ReLU(·) represents a ReLU activation function, and exp(-σ) represents a natural exponential function value calculated with the negative number of the variance as the power.

[0021] In the wind wheel heat dissipation system of the offshore wind turbine generator set, the temperature and flow rate encoding module comprises: a time dimension arrangement unit configured to arrange temperature values of the wind wheel at the plurality of predetermined time points and flow rate values of cooling water of the water cooling system at the plurality of predetermined time points into input vectors according to a time dimension respectively to obtain a temperature input vector and a flow rate input vector; and a hidden feature extraction unit configured to perform full connection encoding on the temperature input vector and the flow rate input vector respectively using full connection layers of the time series encoder to extract high-dimensional hidden features of the feature values at respective positions in the temperature input vector and the flow rate input vector respectively according to the following formula: wherein X is the temperature input vector and the flow rate input vector, Y is an output vector, W is a weight matrix, and B is a bias vector, represents matrix multiplication; and a correlation feature extraction unit configured to use a one-dimensional convolution layer of the time series encoder to perform one-dimensional convolution encoding on the temperature input vector and the flow rate input vector, respectively, to extract high-dimensional correlation features between feature values at respective positions in the temperature input vector and the flow rate input vector, respectively, according to the following formula:

[0022]

[0023] wherein a is the width of a convolution kernel in the x direction, F(a) is a convolution kernel parameter vector, G(x-a) is a local vector matrix operated with the convolution kernel function, w is the size of the convolution kernel, and X represents the temperature input vector and the flow rate input vector.

[0024] In the wind wheel heat dissipation system of the offshore wind turbine generator set, the Bayesian inference module is further configured to fuse the flow rate feature vector, the temperature feature vector, and the corrected heat increase feature vector to obtain the posterior probability vector using a Bayesian probability model according to the following formula:

[0025] V = V1*V2 / V3

[0026] wherein V1 is a feature value at each position in the flow rate feature vector, V2 and V3 are feature values at each position in the temperature feature vector and the corrected heat increase feature vector, respectively, and V is a feature value at each position in the posterior probability vector.

[0027] In the wind wheel heat dissipation system of the offshore wind turbine generator set, the heat dissipation control result generation module is further configured to process the posterior probability vector to obtain the classification result using the classifier according to the following formula:

[0028] softmax{(W n ,B n ):…:(W1,B1)|X}

[0029] wherein W1 to W n are weight matrices, B1 to B n are bias vectors, and X is the posterior probability vector.

[0030] According to another aspect of the present application, a wind wheel heat dissipation method for an offshore wind turbine generator set is also provided, which comprises:

[0031] obtaining the solar heat load, the hub heat load, the convective heat transfer heat load and the thermal radiation at a plurality of predetermined time points in a predetermined time period, the temperature value of the wind wheel at the plurality of predetermined time points, and the flow rate value of the cooling water of the water cooling system at the plurality of predetermined time points;

[0032] arranging the solar heat load, the hub heat load, the convective heat transfer heat load and the thermal radiation at the plurality of predetermined time points into input vectors respectively, and inputting the input vectors into a time encoder comprising a one-dimensional convolution layer to obtain first to fourth heat source feature vectors;

[0033] fusing the first to fourth heat source feature vectors to obtain a heat increase feature vector;

[0034] correcting the feature values at each position in the heat increase feature vector based on the mean and variance of the feature values at all positions in the heat increase feature vector to obtain a corrected heat increase feature vector;

[0035] arranging the temperature value of the wind wheel at the plurality of predetermined time points and the flow rate value of the cooling water of the water cooling system at the plurality of predetermined time points into input vectors respectively, and inputting the input vectors into the time encoder comprising a one-dimensional convolution layer to obtain a temperature feature vector and a flow rate feature vector;

[0036] fusing the flow rate feature vector, the temperature feature vector and the corrected heat increase feature vector using a Bayesian probability model to obtain a posterior probability vector; and

[0037] inputting the posterior probability vector into a classifier to obtain a classification result, the classification result being used to indicate whether the flow rate value of the cooling water at a current time point should be increased or decreased.

[0038] In the wind wheel heat dissipation method of the offshore wind turbine generator set, the correction of the feature values at each position in the heat increase feature vector based on the mean and variance of the feature values at all positions in the heat increase feature vector to obtain a corrected heat increase feature vector comprises: correcting the feature values at each position in the heat increase feature vector based on the mean and variance of the feature values at all positions in the heat increase feature vector to obtain the corrected heat increase feature vector according to the following formula:

[0039]

[0040] wherein, v i represents the feature values at each position in the heat increase feature vector, μ and σ respectively represent the mean and variance of the set of feature values at all positions in the heat increase feature vector, ReLU(·) represents a ReLU activation function, and exp(-σ) represents the natural exponential function value calculated with the negative number of the variance as the power.

[0041] In the wind wheel heat dissipation method of the offshore wind turbine generator set, the step of fusing the flow velocity feature vector, the temperature feature vector and the corrected heat increase feature vector to obtain a posterior probability vector using a Bayesian probability model comprises: fusing the flow velocity feature vector, the temperature feature vector and the corrected heat increase feature vector to obtain the posterior probability vector using a Bayesian probability model according to the following formula:

[0042] The formula is:

[0043] V = V1*V2 / V3

[0044] Wherein, V1 is the feature value of each position in the flow velocity feature vector, V2 and V3 are the feature values of each position in the temperature feature vector and the corrected heat increase feature vector respectively, and V is the feature value of each position in the posterior probability vector.

[0045] Compared with the prior art, the offshore wind turbine generator set wind wheel heat dissipation system and method provided by the application use a deep neural network model as a feature extractor to extract dynamic change implicit features of solar heat load, hub heat load, convective heat transfer heat load and thermal radiation in the time dimension, so as to mine the change feature information of the heat source, that is, each factor of the heat generated in the wind wheel. Then, based on the dynamic change implicit features of the temperature of the wind wheel and the heat source in the time sequence, the cooling water flow rate value of the water cooling system of the offshore wind turbine generator set is dynamically controlled in real time to optimize the heat dissipation effect of the wind wheel. BRIEF DESCRIPTION OF DRAWINGS

[0046] The above and other objects, features and advantages of the present application will become more apparent from the following detailed description of embodiments of the present application, when taken in conjunction with the accompanying drawings. The drawings provided in the specification and the embodiments of the present application together serve to explain the present application and, therefore, to provide further understanding of the present application, and do not constitute a limitation of the present application. In the drawings, the same reference numerals refer to the same components or steps throughout the specification.

[0047] Figure 1 FIG. 1 illustrates a scene schematic diagram of an offshore wind turbine generator set wind wheel heat dissipation system according to an embodiment of the present application.

[0048] Figure 2 FIG. 2 illustrates a block diagram of an offshore wind turbine generator set wind wheel heat dissipation system according to an embodiment of the present application.

[0049] Figure 3 FIG. 3 illustrates an architecture schematic diagram of an offshore wind turbine generator set wind wheel heat dissipation system according to an embodiment of the present application.

[0050] Figure 4Fig. 1 is a block diagram of a heat source data encoding module in a wind wheel heat dissipation system of an offshore wind turbine generator set according to an embodiment of the present application.

[0051] Figure 5 Fig. 2 is a block diagram of a heat source feature fusion module in a wind wheel heat dissipation system of an offshore wind turbine generator set according to an embodiment of the present application.

[0052] Figure 6 Fig. 3 is a block diagram of a temperature and flow rate encoding module in a wind wheel heat dissipation system of an offshore wind turbine generator set according to an embodiment of the present application.

[0053] Figure 7 Fig. 4 is a flowchart of a wind wheel heat dissipation method of an offshore wind turbine generator set according to an embodiment of the present application. DETAILED DESCRIPTION

[0054] Hereinafter, example embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only a part but not all of the embodiments of the present application, and it should be understood that the present application is not limited to the described example embodiments.

[0055] SUMMARY

[0056] Since the heat inside the wind wheel of the offshore wind turbine generator set mainly comes from solar heat load, hub heat load, convective heat transfer heat load and thermal radiation, the present application expects to use a deep neural network model as a feature extractor to extract the dynamic change implicit features of these data in the time dimension, so as to mine the various generating factors of the heat inside the wind wheel, that is, the change feature information of the heat source. Then, based on the dynamic change implicit features of the temperature of the wind wheel and the dynamic change implicit features of the heat source, the cooling water flow rate value of the water cooling system of the offshore wind turbine generator set is comprehensively and dynamically controlled in real time to optimize the heat dissipation effect of the wind wheel.

[0057] Specifically, in the technical solution of the present application, first, the solar heat load, the hub heat load, the convective heat transfer heat load and the thermal radiation at a plurality of predetermined time points in a predetermined time period, the temperature values of the wind wheel at the plurality of predetermined time points, and the flow rate values of the cooling water of the water cooling system at the plurality of predetermined time points are obtained by the sensors.

[0058] It should be understood that, considering the dynamic variation of the solar heat load, the hub heat load, the convective heat transfer heat load and the thermal radiation in the heat source in the time dimension, in order to control the cooling water flow rate of the water cooling system in real time and dynamically to optimize the heat dissipation effect of the wind wheel, it is necessary to deeply mine the dynamic implicit variation characteristics of the heat source data in the time dimension. That is, the solar heat load, the hub heat load, the convective heat transfer heat load and the thermal radiation at the plurality of predetermined time points are arranged as input vectors respectively to integrate the heat source data at each time point, and the obtained input vectors are encoded through a time sequence encoder containing a one-dimensional convolution layer to obtain first to fourth heat source feature vectors. In a specific example of the present application, the time sequence encoder is composed of fully connected layers and one-dimensional convolution layers arranged alternately, which respectively extract the correlation of each heat source data in the time dimension through one-dimensional convolution encoding and the high-dimensional implicit features of each heat source data through fully connected encoding.

[0059] Then, the first to fourth heat source feature vectors are fused to fuse the dynamic implicit correlation characteristics of the solar heat load, the hub heat load, the convective heat transfer heat load and the thermal radiation to obtain a heat increase feature vector corresponding to the heat source. Specifically, it should be understood that, considering that the conversion rate of heat generated by different heat sources into wind wheel heat is different, which can be represented by the pre-classification idea, therefore, a pre-classifier is used to process each heat source feature vector in the first to fourth heat source feature vectors to obtain first to fourth probability values of the contribution of each heat source to the wind wheel heat, and further the probability values can be used as weights to calculate the position weighted sum of the first to fourth heat source feature vectors to obtain the heat increase feature vector.

[0060] It should be understood that, in the calculation of the position weighted sum of the first to fourth heat source feature vectors to obtain the heat increase feature vector using the first to fourth probability values as weights, due to the inter-class difference of the source data corresponding to the first to fourth heat source feature vectors, the feature distribution may have distribution difference when passing through the time sequence encoder, thereby causing the heat increase feature vector to have poor expression consistency in the direction of the vector when performing position weighted sum.

[0061] Therefore, the heat increase feature vector, for example, denoted as V, is subjected to directional recursive squeezing-activation optimization, represented as:

[0062]

[0063] μ and σ are the mean and variance of the feature set v i ∈V.

[0064] That is, the directional recursion of the feature distribution is activated based on the statistical characteristics of the feature set along the vector direction, so as to infer the directional distribution of the features at each sampling position of each feature vector, and the channel squeezing-activation mechanism composed of the ReLU-Sigmoid function is adopted to obtain the attention-enhanced confidence value of the vector direction, so as to enhance the consistency of the feature expression in the length dimension of the vector, thereby improving the accuracy of classification. In this way, the heat dissipation effect of the wind wheel can be optimized.

[0065] Then, for the temperature values of the wind wheel at the plurality of predetermined time points and the flow rate values of the cooling water of the water cooling system, similarly, the implicit feature association in the time dimension is also considered, so the temperature values of the wind wheel at the plurality of predetermined time points and the flow rate values of the cooling water of the water cooling system at the plurality of predetermined time points are respectively arranged as input vectors and then encoded through the time sequence encoder containing the one-dimensional convolution layer, to respectively extract the dynamic change feature information of the temperature values of the wind wheel and the flow rate values of the cooling water of the water cooling system in time sequence, thereby obtaining a temperature feature vector and a flow rate feature vector.

[0066] Further, when the solar heat load, hub heat load, convective heat transfer heat load and thermal radiation in the heat source change, the heat dissipation of the wind wheel is performed by dynamically controlling the cooling water flow rate value of the water cooling system of the offshore wind turbine generator set in real time. Therefore, in the technical solution of the present application, the flow rate feature vector is used as the prior probability, and the purpose of the technical solution of the present application is to update the prior probability to obtain the posterior probability when new evidence, i.e. when the solar heat load, hub heat load, convective heat transfer heat load and thermal radiation in the heat source change. Then according to the Bayes formula, the posterior probability is the prior probability multiplied by the event probability divided by the evidence probability, therefore, in the technical solution of the present application, the Bayesian probability model is used to fuse the flow rate feature vector, the temperature feature vector and the corrected heat increase feature vector to obtain a posterior probability vector, wherein the flow rate feature vector is the prior, the temperature feature vector is the event, and the corrected heat increase feature vector is the evidence. Then, the posterior probability vector can be used to obtain a classification result for representing whether the flow rate value of the cooling water at the current time point should be increased or decreased. In this way, the cooling water flow rate value of the water cooling system of the offshore wind turbine generator set can be dynamically and accurately controlled to optimize the heat dissipation effect of the wind wheel.

[0067] Based on this, the offshore wind turbine cooling system provided by the present application comprises: a data monitoring and acquisition module, which is used to obtain solar heat load, hub heat load, convective heat transfer heat load and thermal radiation in a predetermined time period, temperature values of the wind wheel at a plurality of predetermined time points, and flow rate values of the cooling water of the water cooling system at the plurality of predetermined time points; a heat source data encoding module, which is used to arrange the solar heat load, the hub heat load, the convective heat transfer heat load and the thermal radiation at the plurality of predetermined time points into input vectors respectively, and then obtain first to fourth heat source feature vectors through a time sequence encoder comprising a one-dimensional convolution layer; a heat source feature fusion module, which is used to fuse the first to fourth heat source feature vectors to obtain a heat increase feature vector; a heat feature correction module, which is used to correct the feature values at each position in the heat increase feature vector based on the mean and variance of the feature values at all positions in the heat increase feature vector to obtain a corrected heat increase feature vector; a temperature and flow rate encoding module, which is used to arrange the temperature values of the wind wheel at the plurality of predetermined time points and the flow rate values of the cooling water of the water cooling system at the plurality of predetermined time points into input vectors respectively, and then obtain a temperature feature vector and a flow rate feature vector through the time sequence encoder comprising the one-dimensional convolution layer; a Bayesian inference module, which is used to fuse the flow rate feature vector, the temperature feature vector and the corrected heat increase feature vector using a Bayesian probability model to obtain a posterior probability vector; and a cooling control result generation module, which is used to pass the posterior probability vector through a classifier to obtain a classification result, wherein the classification result is used to indicate that the flow rate value of the cooling water at the current time point should be increased or should be decreased.

[0068] Figure 1 The offshore wind turbine cooling system according to the embodiment of the present application is illustrated in the scene schematic diagram of the offshore wind turbine cooling system. As shown in the application scenario of the offshore wind turbine cooling system, first, the solar heat load, the hub heat load and the convective heat transfer heat load at a plurality of predetermined time points in a predetermined time period are obtained through a heat load sensor (for example, F1 as shown in the embodiment of the present application), the thermal radiation at the plurality of predetermined time points in the predetermined time period is obtained through a thermal radiation sensor (for example, F2 as shown in the embodiment of the present application), the temperature values of the wind wheel at the plurality of predetermined time points are obtained through a temperature sensor (for example, F3 as shown in the embodiment of the present application), and the flow rate values of the cooling water of the water cooling system at the plurality of predetermined time points are obtained through a flow rate sensor (for example, F4 as shown in the embodiment of the present application). Figure 1 Figure 1 Figure 1 Figure 1 Figure 1 Figure 1 Figure 1 Figure 1 ​​​​​​​The flow rate of the cooling water in C) shown in the diagram. Then, the solar heat load, hub heat load, convective heat transfer load, thermal radiation, rotor temperature, and cooling water flow rate of the water-cooling system are input to the server of the wind turbine cooling algorithm (e.g., such as...) Figure 1 In the illustrated S), the server is able to process the solar heat load, hub heat load, convective heat transfer heat load, thermal radiation, wind turbine temperature value, and cooling water flow rate value of the water cooling system using the wind turbine heat dissipation algorithm of the offshore wind turbine to obtain a control result indicating whether the cooling water flow rate value should be increased or decreased at the current time point.

[0069] After introducing the basic principles of this application, various non-limiting embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0070] Exemplary System

[0071] Figure 2 The diagram illustrates a block diagram of a wind turbine cooling system for an offshore wind turbine generator according to an embodiment of this application. Figure 2 As shown, the wind turbine cooling system 100 of an offshore wind turbine generator according to an embodiment of this application includes: a data monitoring and acquisition module 110, used to acquire solar heat load, hub heat load, convective heat transfer heat load and thermal radiation at multiple predetermined time points within a predetermined time period, the temperature value of the wind turbine at the multiple predetermined time points, and the flow rate value of the cooling water of the water cooling system at the multiple predetermined time points; a heat source data encoding module 120, used to arrange the solar heat load, hub heat load, convective heat transfer heat load and thermal radiation at the multiple predetermined time points into input vectors and then pass them through a time-series encoder containing a one-dimensional convolutional layer to obtain first to fourth heat source feature vectors; a heat source feature fusion module 130, used to fuse the first to fourth heat source feature vectors to obtain a heat increase feature vector; and a heat feature correction module 140, used to correct the heat increase feature vector based on the heat increase feature vector. The heat gain feature vector is corrected by taking the mean and variance of the feature values ​​at each position. A temperature and flow rate encoding module 150 is used to arrange the temperature values ​​of the wind turbine and the flow rate values ​​of the cooling water in the water-cooling system at multiple predetermined time points into input vectors, which are then passed through a time-series encoder containing a one-dimensional convolutional layer to obtain temperature and flow rate feature vectors. A Bayesian inference module 160 is used to fuse the flow rate feature vector, the temperature feature vector, and the corrected heat gain feature vector using a Bayesian probability model to obtain a posterior probability vector. Finally, a heat dissipation control result generation module 170 is used to pass the posterior probability vector through a classifier to obtain a classification result, which indicates whether the flow rate value of the cooling water at the current time point should increase or decrease.

[0072] Figure 3 The illustration shows a schematic diagram of the wind turbine cooling system of an offshore wind turbine generator according to an embodiment of this application. Figure 3 As shown, firstly, the solar heat load, hub heat load, convective heat transfer heat load, and thermal radiation at multiple predetermined time points within a predetermined time period are acquired, along with the wind turbine temperature value at each predetermined time point and the cooling water flow rate value of the water-cooling system at each predetermined time point. Next, the solar heat load, hub heat load, convective heat transfer heat load, and thermal radiation at each predetermined time point are arranged into input vectors and passed through a time-series encoder containing a one-dimensional convolutional layer to obtain first to fourth heat source feature vectors. Then, the first to fourth heat source feature vectors are fused to obtain a heat increase feature vector. Next, based on the mean and variance of the feature values ​​at all positions in the heat increase feature vector, the feature values ​​at each position in the heat increase feature vector are corrected to obtain a corrected heat increase feature vector. Finally, the wind turbine temperature value and the cooling water flow rate value of the water-cooling system at each predetermined time point are arranged into input vectors and passed through the time-series encoder containing a one-dimensional convolutional layer to obtain a temperature feature vector and a flow rate feature vector. Next, a Bayesian probability model is used to fuse the flow velocity feature vector, the temperature feature vector, and the corrected heat gain feature vector to obtain a posterior probability vector. Then, the posterior probability vector is passed through a classifier to obtain a classification result, which indicates whether the cooling water flow velocity at the current time point should increase or decrease.

[0073] In the aforementioned offshore wind turbine rotor cooling system 100, the data monitoring and acquisition module 110 is used to acquire solar heat load, hub heat load, convective heat transfer heat load, and thermal radiation at multiple predetermined time points within a predetermined time period, the rotor temperature at each of the multiple predetermined time points, and the cooling water flow rate of the water-cooling system at each of the multiple predetermined time points. The solar heat load, hub heat load, convective heat transfer heat load, and thermal radiation at the multiple predetermined time points within the predetermined time period, the rotor temperature at each of the multiple predetermined time points, and the cooling water flow rate of the water-cooling system at each of the multiple predetermined time points are acquired by corresponding sensors.

[0074] In the wind wheel heat dissipation system 100 of the offshore wind turbine generator set, the heat source data encoding module 120 is configured to arrange the solar heat load, the hub heat load, the convective heat transfer heat load and the thermal radiation at the plurality of predetermined time points into input vectors respectively, and then encode the input vectors through a time series encoder including a one-dimensional convolution layer to obtain first to fourth heat source feature vectors. It should be understood that, considering the dynamic change rules of the solar heat load, the hub heat load, the convective heat transfer heat load and the thermal radiation in the heat source in the time dimension, in order to dynamically control the cooling water flow rate value of the water cooling system in real time to optimize the heat dissipation effect of the wind wheel, it is necessary to deeply mine the dynamic implicit change characteristics of the heat source data in the time dimension. That is, the solar heat load, the hub heat load, the convective heat transfer heat load and the thermal radiation at the plurality of predetermined time points are arranged into input vectors respectively to integrate the heat source data at each time point, and then the obtained input vectors are encoded through the time series encoder including the one-dimensional convolution layer to obtain the first to fourth heat source feature vectors.

[0075] Figure 4 FIG. 1 illustrates a block diagram of a heat source data encoding module in a wind wheel heat dissipation system of an offshore wind turbine generator set according to an embodiment of the present application. As shown in the figure, Figure 4 In one example, in the wind wheel heat dissipation system 100 of the offshore wind turbine generator set, the heat source data encoding module 120 includes: an input vector construction unit 121 configured to arrange the solar heat load, the hub heat load, the convective heat transfer heat load and the thermal radiation at the plurality of predetermined time points into input vectors respectively according to the time dimension to obtain first to fourth input vectors; a fully connected encoding unit 122 configured to use a fully connected layer of the time series encoder to perform fully connected encoding on the first to fourth input vectors respectively according to the following formula to extract high-dimensional implicit features of feature values at each position in the first to fourth input vectors respectively, wherein the formula is: wherein X is the first to fourth input vectors, Y is an output vector, W is a weight matrix, B is a bias vector, represents matrix multiplication; and a one-dimensional convolution encoding unit 123 configured to use a one-dimensional convolution layer of the time series encoder to perform one-dimensional convolution encoding on the first to fourth input vectors respectively according to the following formula to extract high-dimensional implicit correlation features between feature values at each position in the first to fourth input vectors respectively, wherein the formula is:

[0076]

[0077] wherein a is the width of the convolution kernel in the x direction, F(a) is the convolution kernel parameter vector, G(x-a) is the local vector matrix operated with the convolution kernel function, w is the size of the convolution kernel, and X represents the first to fourth input vectors.

[0078] In the wind wheel heat dissipation system 100 of the offshore wind turbine generator set, the heat source feature fusion module 130 is configured to fuse the first to fourth heat source feature vectors to obtain a heat increase feature vector. That is, the dynamic implicit correlation features of the solar heat load, the hub heat load, the convective heat transfer heat load, and the heat radiation are fused to obtain the heat increase feature vector corresponding to the heat source. Specifically, it should be understood that the conversion rates of the heat generated by different heat sources into the wind wheel heat are different, which can be characterized by the pre-classification idea. Therefore, the pre-classifier is used to process each of the first to fourth heat source feature vectors to obtain the first to fourth probability values of the contribution of each of the heat sources to the wind wheel heat. Further, the first to fourth probability values can be used as weights to calculate the position-wise weighted sum of the first to fourth heat source feature vectors to obtain the heat increase feature vector.

[0079] Figure 5 FIG. 7 illustrates a block diagram of a heat source feature fusion module in a wind wheel heat dissipation system of an offshore wind turbine generator set according to an embodiment of the present application. As shown in the figure, in one example, in the wind wheel heat dissipation system 100 of the offshore wind turbine generator set, the heat source feature fusion module 130 includes a contribution evaluation unit 131 configured to pass each of the first to fourth heat source feature vectors through a pre-classifier to obtain first to fourth probability values, and a fusion unit 132 configured to calculate a position-wise weighted sum of the first to fourth heat source feature vectors using the first to fourth probability values as weights to obtain the heat increase feature vector. Figure 5

[0080] In the wind wheel heat dissipation system 100 of the offshore wind turbine generator set, the heat feature correction module 140 is configured to correct the feature values of each position in the heat increase feature vector based on the mean and variance of the feature values of all positions in the heat increase feature vector to obtain a corrected heat increase feature vector. It should be understood that when the position-wise weighted sum of the first to fourth heat source feature vectors is calculated using the first to fourth probability values as weights to obtain the heat increase feature vector, due to the inter-class differences in the source data corresponding to the first to fourth heat source feature vectors, the feature distribution of the heat increase feature vector may have distribution differences when passing through the time encoder, resulting in poor consistency in the expression of the heat increase feature vector in its vector direction when performing position-wise weighted sum. Therefore, the heat increase feature vector, denoted as V, is subjected to directional recursive squeeze-and-excitation optimization.

[0081] ​In one example, in the wind wheel heat dissipation system 100 of the offshore wind turbine generator set described above, the thermal feature correction module 140 is further configured to correct the feature values of each position in the thermal increase feature vector based on the mean and variance of the feature values of all positions in the thermal increase feature vector to obtain the corrected thermal increase feature vector according to the following formula:

[0082]

[0083] wherein v i represents the feature values of each position in the thermal increase feature vector, μ and σ respectively represent the mean and variance of the feature value set of all positions in the thermal increase feature vector, ReLU(·) represents the ReLU activation function, and exp(-σ) represents the natural exponential function value calculated with the negative number of the variance as the power.

[0084] That is, the directional recursion of the feature distribution is activated based on the statistical characteristics of the feature set along the vector direction, so as to infer the directional distribution of the feature at each sampling position of each feature vector, and the channel squeezing-activation mechanism composed of the ReLU-Sigmoid function is adopted to obtain the confidence value of the attention enhancement of the vector direction, so as to enhance the consistency of the feature expression in the length dimension of the vector, thereby improving the accuracy of classification. In this way, the heat dissipation effect of the wind wheel can be optimized.

[0085] In the wind wheel heat dissipation system 100 of the offshore wind turbine generator set described above, the temperature and flow rate encoding module 150 is configured to arrange the temperature values of the wind wheel at the plurality of predetermined time points and the flow rate values of the cooling water of the water cooling system at the plurality of predetermined time points into input vectors respectively, and then pass them through the time sequence encoder comprising a one-dimensional convolution layer to obtain a temperature feature vector and a flow rate feature vector. For the temperature values of the wind wheel at the plurality of predetermined time points and the flow rate values of the cooling water of the water cooling system at the plurality of predetermined time points, similarly, the dynamic feature association in the time dimension is also considered, therefore, the temperature values of the wind wheel at the plurality of predetermined time points and the flow rate values of the cooling water of the water cooling system at the plurality of predetermined time points are arranged into input vectors respectively, and then passed through the time sequence encoder comprising a one-dimensional convolution layer to encode, so as to extract the dynamic change feature information of the temperature values of the wind wheel and the flow rate values of the cooling water of the water cooling system in time sequence respectively, thereby obtaining the temperature feature vector and the flow rate feature vector.

[0086] Figure 6 Fig. 3 shows a block diagram of a temperature and flow rate encoding module in a wind wheel heat dissipation system of an offshore wind turbine generator set according to an embodiment of the present application. As shown in Fig. 3, the temperature and flow rate encoding module 150 is configured to arrange the temperature values of the wind wheel at the plurality of predetermined time points and the flow rate values of the cooling water of the water cooling system at the plurality of predetermined time points into input vectors respectively, and then pass them through the time sequence encoder comprising a one-dimensional convolution layer to obtain a temperature feature vector and a flow rate feature vector. Figure 6As shown, in one example, in the wind wheel heat dissipation system 100 of the offshore wind turbine generator set described above, the temperature and flow rate coding module 150 comprises: a time dimension arrangement unit 151, configured to arrange the temperature values of the wind wheel at the plurality of predetermined time points and the flow rate values of the cooling water of the water cooling system at the plurality of predetermined time points into input vectors according to the time dimension respectively to obtain a temperature input vector and a flow rate input vector; an implicit feature extraction unit 152, configured to use the full connection layer of the time series encoder to perform full connection coding on the temperature input vector and the flow rate input vector respectively according to the following formula to extract high-dimensional implicit features of the feature values at each position in the temperature input vector and the flow rate input vector respectively, wherein the formula is: wherein X is the temperature input vector and the flow rate input vector, Y is an output vector, W is a weight matrix, and B is a bias vector, represents matrix multiplication; and a correlation feature extraction unit 153, configured to use the one-dimensional convolution layer of the time series encoder to perform one-dimensional convolution coding on the temperature input vector and the flow rate input vector respectively according to the following formula to extract high-dimensional implicit correlation features between the feature values at each position in the temperature input vector and the flow rate input vector respectively, wherein the formula is:

[0087]

[0088] wherein a is the width of the convolution kernel in the x direction, F(a) is the convolution kernel parameter vector, G(x-a) is the local vector matrix operated with the convolution kernel function, w is the size of the convolution kernel, and X represents the temperature input vector and the flow rate input vector.

[0089] In the wind wheel heat dissipation system 100 of the offshore wind turbine generator set, the Bayesian inference module 160 is configured to fuse the flow rate feature vector, the temperature feature vector and the corrected heat increment feature vector to obtain a posterior probability vector by using a Bayesian probability model. Further, when the solar heat load, the hub heat load, the convective heat transfer heat load and the thermal radiation in the heat source change, the heat dissipation of the wind wheel is performed by dynamically controlling the cooling water flow rate value of the water cooling system of the offshore wind turbine generator set in real time. Therefore, in the technical solution of the present application, the flow rate feature vector is used as the prior probability, and the purpose of the technical solution of the present application is to update the prior probability to obtain the posterior probability when new evidence, i.e., when the solar heat load, the hub heat load, the convective heat transfer heat load and the thermal radiation in the heat source change. Then according to the Bayesian formula, the posterior probability is the prior probability multiplied by the event probability divided by the evidence probability. Therefore, in the technical solution of the present application, the Bayesian probability model is used to fuse the flow rate feature vector, the temperature feature vector and the corrected heat increment feature vector to obtain the posterior probability vector, wherein the flow rate feature vector is used as the prior probability, the temperature feature vector is used as the event, and the corrected heat increment feature vector is used as the evidence.

[0090] In one example, in the wind wheel heat dissipation system 100 of the offshore wind turbine generator set, the Bayesian inference module 160 is further configured to fuse the flow rate feature vector, the temperature feature vector and the corrected heat increment feature vector to obtain the posterior probability vector by using a Bayesian probability model according to the following formula:

[0091] The formula is:

[0092] V = V1*V2 / V3

[0093] Wherein, V1 is the feature value of each position in the flow rate feature vector, V2 and V3 are the feature values of each position in the temperature feature vector and the corrected heat increment feature vector respectively, and V is the feature value of each position in the posterior probability vector.

[0094] In the wind wheel heat dissipation system 100 of the offshore wind turbine generator set, the heat dissipation control result generation module 170 is configured to obtain a classification result by classifying the posterior probability vector, and the classification result is used to indicate whether the cooling water flow rate value at the current time point should be increased or decreased. In this way, the cooling water flow rate value of the water cooling system of the offshore wind turbine generator set can be dynamically and accurately controlled to optimize the heat dissipation effect of the wind wheel.

[0095] In one example, in the wind wheel heat dissipation system 100 of the offshore wind turbine generator set described above, the heat dissipation control result generation module 170 is further configured to: use the classifier to process the posterior probability vector according to the following formula to obtain the classification result, wherein the formula is:

[0096] softmax(W1, B1)T X n n softmax(W1, B1)T X

[0097] wherein W1 to W n are weight matrices, B1 to B n are bias vectors, and X is the posterior probability vector.

[0098] In summary, the wind wheel heat dissipation system 100 of the offshore wind turbine generator set based on the embodiments of the present application is illustrated, which uses a deep neural network model as a feature extractor to extract the dynamic change implicit features of the solar heat load, the hub heat load, the convective heat transfer heat load and the thermal radiation in the time dimension, so as to mine the change characteristic information of the heat sources, i.e., the various generation factors of the heat in the wind wheel. Then, based on the dynamic change implicit features of the temperature of the wind wheel in the time sequence and the dynamic change implicit features of the heat sources, the cooling water flow rate value of the water cooling system of the offshore wind turbine generator set is dynamically controlled in real time to optimize the heat dissipation effect of the wind wheel.

[0099] As described above, the wind wheel heat dissipation system 100 of the offshore wind turbine generator set according to the embodiments of the present application can be implemented in various terminal devices, such as a server for the heat dissipation of the wind wheel of the offshore wind turbine generator set, etc. In one example, the wind wheel heat dissipation system 100 of the offshore wind turbine generator set according to the embodiments of the present application can be integrated into a terminal device as a software module and / or a hardware module. For example, the wind wheel heat dissipation system 100 of the offshore wind turbine generator set can be a software module in the operating system of the terminal device, or can be an application program developed for the terminal device; of course, the wind wheel heat dissipation system 100 of the offshore wind turbine generator set can also be one of the many hardware modules of the terminal device.

[0100] Alternatively, in another example, the wind wheel heat dissipation system 100 of the offshore wind turbine generator set and the terminal device can also be separate devices, and the wind wheel heat dissipation system 100 of the offshore wind turbine generator set can be connected to the terminal device through a wired and / or wireless network, and transmit interactive information in a conventional data format.

[0101] Exemplary Method

[0102] Figure 5 ​A flowchart of a wind wheel heat dissipation method of an offshore wind turbine generator system according to an embodiment of the present application. As shown in Figure 5 The wind wheel heat dissipation method of the offshore wind turbine generator system according to the embodiment of the present application includes: S110, acquiring solar heat load, hub heat load, convective heat transfer heat load and thermal radiation at a plurality of predetermined time points in a predetermined time period, temperature values of the wind wheel at the plurality of predetermined time points, and flow rate values of cooling water of the water cooling system at the plurality of predetermined time points; S120, arranging the solar heat load, the hub heat load, the convective heat transfer heat load and the thermal radiation at the plurality of predetermined time points into input vectors respectively, and then passing through a time sequence encoder containing a one-dimensional convolution layer to obtain first to fourth heat source feature vectors; S130, fusing the first to fourth heat source feature vectors to obtain a heat increase feature vector; S140, based on the mean and variance of the feature values at all positions in the heat increase feature vector, correcting the feature values at each position in the heat increase feature vector to obtain a corrected heat increase feature vector; S150, arranging the temperature values of the wind wheel at the plurality of predetermined time points and the flow rate values of the cooling water of the water cooling system at the plurality of predetermined time points into input vectors respectively, and then passing through the time sequence encoder containing the one-dimensional convolution layer to obtain a temperature feature vector and a flow rate feature vector; S160, using a Bayesian probability model to fuse the flow rate feature vector, the temperature feature vector and the corrected heat increase feature vector to obtain a posterior probability vector; and S170, passing the posterior probability vector through a classifier to obtain a classification result, the classification result being used to indicate that the flow rate value of the cooling water at the current time point should be increased or should be decreased.

[0103] In one example, in the wind wheel heat dissipation method of the offshore wind turbine generator system described above, the correction of the feature values at each position in the heat increase feature vector based on the mean and variance of the feature values at all positions in the heat increase feature vector to obtain the corrected heat increase feature vector includes: based on the mean and variance of the feature values at all positions in the heat increase feature vector, correcting the feature values at each position in the heat increase feature vector to obtain the corrected heat increase feature vector according to the following formula:

[0104]

[0105] wherein v i represents the feature values at each position in the heat increase feature vector, μ and σ respectively represent the mean and variance of the set of feature values at all positions in the heat increase feature vector, ReLU(·) represents a ReLU activation function, and exp(-σ) represents the natural exponential function value calculated with the negative number of the variance as the power.

[0106] In one example, in the wind wheel heat dissipation method of the offshore wind turbine generator set, the step of using the Bayesian probability model to fuse the flow velocity feature vector, the temperature feature vector and the corrected heat increase feature vector to obtain a posterior probability vector comprises: using the Bayesian probability model to fuse the flow velocity feature vector, the temperature feature vector and the corrected heat increase feature vector to obtain the posterior probability vector according to the following formula: V = V1*V2 / V3.

[0107] V = V1*V2 / V3

[0108] wherein V1 is a feature value of each position in the flow velocity feature vector, V2 and V3 are feature values of each position in the temperature feature vector and the corrected heat increase feature vector respectively, and V is a feature value of each position in the posterior probability vector.

[0109] In summary, the wind wheel heat dissipation method of the offshore wind turbine generator set is illustrated, which uses a deep neural network model as a feature extractor to extract the dynamic change implicit features of the solar heat load, the hub heat load, the convective heat transfer heat load and the thermal radiation in the time dimension, so as to mine the change feature information of the heat source, i.e. the various generation factors of the heat in the wind wheel. Then, based on the dynamic change implicit features of the temperature of the wind wheel and the dynamic change implicit features of the heat source, the cooling water flow velocity value of the water cooling system of the offshore wind turbine generator set is dynamically controlled in real time to optimize the heat dissipation effect of the wind wheel.

[0110] The basic principles of the application are described above in combination with specific embodiments, but it should be pointed out that the advantages, advantages, effects and the like mentioned in the application are only examples and are not limiting, and these advantages, advantages, effects and the like cannot be considered as the application must have. In addition, the above specific details are only for the purpose of example and for the purpose of understanding, and are not limited to the application which must use the above specific details to realize.

[0111] The block diagrams of the devices, apparatuses, equipment, systems involved in the application are only illustrative examples and are not intended to require or imply that the connection, arrangement, configuration shown in the block diagram must be connected, arranged, configured. As those skilled in the art will recognize, these devices, apparatuses, equipment, systems can be connected, arranged, configured in any way. Words such as "include", "contain", "have" and the like are open-ended words, mean "include but not limited to", and can be used interchangeably. The words "or" and "and" used herein mean the word "and / or", and can be used interchangeably unless the context clearly indicates otherwise. The word "such as" used herein means the phrase "such as but not limited to", and can be used interchangeably.

[0112] It is also important to note that the devices, apparatuses and methods described in the present application can be embodied in a variety of other forms, modi fications and alt ernatives, some of which have been discussed above and some of which are gathe red as wi ll be apparent to those reasonably skilled in the art. The described aspects and embodiments are to be considered in a descriptive sense only and not for purposes of limitation. Therefore, the scope of the present application is not to be determined strictly by the description in the specification but by the appended claims, and all changes that come within the meaning and range of equivalency of the claims are intended to be embraced therein.

[0113] The above description of disclosed aspects is intended to be illustrative, and not restrictive. Other modifications used to illustrate and describe the present aspects will be apparent to those of ordinary skill in the art upon reviewing the above description, and it will be apparent that changes can be made in the form, implementations, and implementations of the present aspects without departing from the spirit and scope of the application. Accordingly, the disclosure is not intended to be limited to the described aspects and embodiments, but is to be commensurate with the widest scope embodied by the principles and novel features described herein.

[0114] The above description has been presented for the purpose of illustration and description. Furthermore, this description is not intended to limit the embodiments of the present application to forms disclosed herein. Although various example aspects and embodiments have been discussed above, those of ordinary skill in the art will appreciate a variety of modifications, alternative constructions, adaptations, and equivalents.

Claims

1. A wind turbine cooling system for an offshore wind turbine generator set, characterized in that, include: The data monitoring and acquisition module is used to acquire solar heat load, hub heat load, convective heat transfer heat load and thermal radiation at multiple predetermined time points within a predetermined time period, the temperature value of the wind turbine at the multiple predetermined time points, and the flow rate value of the cooling water of the water cooling system at the multiple predetermined time points. The heat source data encoding module is used to arrange the solar heat load, hub heat load, convective heat transfer heat load and thermal radiation at the multiple predetermined time points into input vectors, and then pass them through a time encoder containing a one-dimensional convolutional layer to obtain the first to fourth heat source feature vectors. A heat source feature fusion module is used to fuse the feature vectors of the first to fourth heat sources to obtain a heat increase feature vector. The thermal feature correction module is used to correct the feature values ​​at each position in the thermal increase feature vector based on the mean and variance of the feature values ​​at all positions in the thermal increase feature vector to obtain the corrected thermal increase feature vector. The temperature and flow rate encoding module is used to arrange the temperature values ​​of the wind turbine and the flow rate values ​​of the cooling water of the water cooling system at the multiple predetermined time points into input vectors, and then pass them through the time encoder containing a one-dimensional convolutional layer to obtain temperature feature vectors and flow rate feature vectors. The Bayesian inference module is used to fuse the flow velocity feature vector, the temperature feature vector, and the corrected thermal increase feature vector using a Bayesian probability model to obtain a posterior probability vector; and The heat dissipation control result generation module is used to pass the posterior probability vector through a classifier to obtain a classification result, which is used to indicate whether the flow rate of the cooling water at the current time point should be increased or decreased.

2. The wind turbine cooling system for offshore wind turbine generators according to claim 1, characterized in that, The heat source data encoding module includes: An input vector construction unit is used to arrange the solar heat load, hub heat load, convective heat transfer heat load and thermal radiation at the multiple predetermined time points into input vectors according to the time dimension to obtain the first to fourth input vectors. A fully connected encoding unit is used to perform fully connected encoding on the first to fourth input vectors using the fully connected layers of the temporal encoder, respectively, according to the following formula, to extract high-dimensional latent features of the feature values ​​at each position in the first to fourth input vectors, wherein the formula is: Where X is the first to fourth input vectors, Y is the output vector, W is the weight matrix, and B is the bias vector. Represents matrix multiplication; and A one-dimensional convolutional coding unit is used to perform one-dimensional convolutional coding on the first to fourth input vectors using the one-dimensional convolutional layers of the temporal encoder, respectively, according to the following formula, to extract high-dimensional implicit correlation features between feature values ​​at each position in the first to fourth input vectors, wherein the formula is: Where a is the width of the convolution kernel in the x direction, F(a) is the convolution kernel parameter vector, G(xa) is the local vector matrix operated with the convolution kernel function, w is the size of the convolution kernel, and X represents the first to fourth input vectors.

3. The wind turbine cooling system for offshore wind turbine generators according to claim 2, characterized in that, The heat source feature fusion module includes: The contribution evaluation unit is used to pass each heat source feature vector in the first to fourth heat source feature vectors through a pre-classifier to obtain the first to fourth probability values; and The fusion unit is used to calculate the position-weighted sum of the first to fourth heat source feature vectors using the first to fourth probability values ​​as weights to obtain the heat increase feature vector.

4. The wind turbine cooling system for offshore wind turbine generators according to claim 3, characterized in that, The thermal feature correction module is further configured to: based on the mean and variance of the feature values ​​at all positions in the thermal increase feature vector, correct the feature values ​​at each position in the thermal increase feature vector using the following formula to obtain the corrected thermal increase feature vector; Among them, v i Let μ and σ represent the feature values ​​at each position in the thermally amplified feature vector, respectively, and let ReLU(·) represent the ReLU activation function. Let exp(-σ) represent the natural exponential function value calculated by raising the variance to the negative power of the variance.

5. The wind turbine cooling system for offshore wind turbine generators according to claim 4, characterized in that, The temperature and flow rate encoding module includes: The time dimension arrangement unit is used to arrange the temperature values ​​of the wind turbines and the flow rate values ​​of the cooling water of the water cooling system at the multiple predetermined time points into input vectors according to the time dimension to obtain temperature input vector and flow rate input vector respectively. The hidden feature extraction unit is used to perform fully connected encoding on the temperature input vector and the flow velocity input vector using the fully connected layer of the temporal encoder according to the following formula to extract high-dimensional hidden features of the feature values ​​at each position in the temperature input vector and the flow velocity input vector, respectively, wherein the formula is: Where X is the temperature input vector and the flow rate input vector, Y is the output vector, W is the weight matrix, and B is the bias vector. Represents matrix multiplication; and The correlation feature extraction unit is used to perform one-dimensional convolutional encoding on the temperature input vector and the flow velocity input vector using the one-dimensional convolutional layer of the temporal encoder, respectively, according to the following formula, to extract high-dimensional implicit correlation features between feature values ​​at each position in the temperature input vector and the flow velocity input vector, respectively, wherein the formula is: Where a is the width of the convolution kernel in the x direction, F(a) is the convolution kernel parameter vector, G(xa) is the local vector matrix operated with the convolution kernel function, w is the size of the convolution kernel, and X represents the temperature input vector and the flow rate input vector.

6. The wind turbine cooling system for offshore wind turbine generators according to claim 5, characterized in that, The Bayesian inference module is further configured to: use a Bayesian probability model to fuse the flow velocity feature vector, the temperature feature vector, and the corrected thermal increase feature vector using the following formula to obtain the posterior probability vector; The formula is as follows: V = V1 * V2 / V3 Wherein, V1 is the feature value at each position in the flow velocity feature vector, V2 and V3 are the feature values ​​at each position in the temperature feature vector and the corrected heat increase feature vector, respectively, and V is the feature value at each position in the posterior probability vector.

7. The wind turbine cooling system for offshore wind turbine generators according to claim 6, characterized in that, The heat dissipation control result generation module is further configured to: process the posterior probability vector using the classifier according to the following formula to obtain the classification result, wherein the formula is: softmax{(W n B n ):…:(W1,B1)|X}, where W1 to W n The weight matrix is ​​B1 to B1. n Let X be the bias vector, and let X be the posterior probability vector.

8. A method for cooling the rotor of an offshore wind turbine, characterized in that, include: The solar heat load, hub heat load, convective heat transfer heat load and thermal radiation at multiple predetermined time points within a predetermined time period are obtained, along with the wind turbine temperature values ​​at the multiple predetermined time points and the cooling water flow rate values ​​of the water cooling system at the multiple predetermined time points. The solar heat load, hub heat load, convective heat transfer heat load and thermal radiation at the multiple predetermined time points are arranged into input vectors and then passed through a time encoder containing a one-dimensional convolutional layer to obtain the first to fourth heat source feature vectors. The feature vectors of the first to fourth heat sources are fused to obtain the heat increase feature vector; Based on the mean and variance of the eigenvalues ​​at all positions in the thermal increase feature vector, the eigenvalues ​​at each position in the thermal increase feature vector are corrected to obtain the corrected thermal increase feature vector. The temperature values ​​of the wind turbine at the multiple predetermined time points and the flow rate values ​​of the cooling water in the water cooling system at the multiple predetermined time points are arranged into input vectors and then passed through the time encoder containing a one-dimensional convolutional layer to obtain temperature feature vectors and flow rate feature vectors. A Bayesian probabilistic model is used to fuse the flow velocity feature vector, the temperature feature vector, and the corrected thermal amplification feature vector to obtain a posterior probability vector; and The posterior probability vector is passed through a classifier to obtain a classification result, which is used to indicate whether the flow rate of the cooling water at the current time point should increase or decrease.

9. The method for cooling the rotor of an offshore wind turbine generator set according to claim 8, characterized in that, The step of correcting the feature values ​​at each position in the heat increase feature vector based on the mean and variance of the feature values ​​at all positions in the heat increase feature vector to obtain the corrected heat increase feature vector includes: correcting the feature values ​​at each position in the heat increase feature vector based on the mean and variance of the feature values ​​at all positions in the heat increase feature vector using the following formula to obtain the corrected heat increase feature vector. Among them, v i Let μ and σ represent the feature values ​​at each position in the thermally amplified feature vector, respectively, and let ReLU(·) represent the ReLU activation function. Let exp(-σ) represent the natural exponential function value calculated by raising the variance to the negative power of the variance.

10. The method for cooling the rotor of an offshore wind turbine generator set according to claim 9, characterized in that, The step of using a Bayesian probability model to fuse the flow velocity feature vector, the temperature feature vector, and the corrected thermal amplification feature vector to obtain a posterior probability vector includes: using a Bayesian probability model to fuse the flow velocity feature vector, the temperature feature vector, and the corrected thermal amplification feature vector using the following formula to obtain the posterior probability vector; The formula is as follows: V = V1 * V2 / V3 Wherein, V1 is the feature value at each position in the flow velocity feature vector, V2 and V3 are the feature values ​​at each position in the temperature feature vector and the corrected heat increase feature vector, respectively, and V is the feature value at each position in the posterior probability vector.

Images