CN115357065A

CN115357065A - Remote intelligent dehumidification control system and method for offshore wind turbine

Info

Publication number: CN115357065A
Application number: CN202211036246.5A
Authority: CN
Inventors: 黄力哲; 王力军; 陈思; 徐峰; 王洪兴; 刘荣波; 郝蛟蛟; 吕胜波
Original assignee: Huaneng Renewables Corp Ltd; Beijing Huaneng Xinrui Control Technology Co Ltd
Current assignee: Huaneng Renewables Corp Ltd; Beijing Huaneng Xinrui Control Technology Co Ltd
Priority date: 2022-08-27
Filing date: 2022-08-27
Publication date: 2022-11-18
Anticipated expiration: 2042-08-27
Also published as: CN115357065B

Abstract

The application relates to the field of intelligent dehumidification control of offshore wind turbines, and particularly discloses a remote intelligent dehumidification control system and a method of the remote intelligent dehumidification control system. In this way, the dynamic moisture change characteristics brought by the wind data are introduced into the dehumidifier power control model, so that the power control of the dehumidifier can be prospective and predictive, and the humidity of the offshore wind turbine in a future time period can be kept within a preset reasonable range.

Description

Remote intelligent dehumidification control system and method for offshore wind turbine

Technical Field

The invention relates to the field of intelligent dehumidification control of offshore wind turbines, in particular to a remote intelligent dehumidification control system and a remote intelligent dehumidification control method for an offshore wind turbine.

Background

Offshore wind turbines work in high humidity, high salt environments for long periods of time. Metal equipment and facilities inside the fan are easy to corrode. Generally, the critical point of relative humidity of metal corrosion is 45% -50%, corrosion occurs when the relative humidity exceeds the limit, the corrosion speed is accelerated along with the increase of the relative humidity, and once dew condensation occurs on steel, the corrosion process is quicker.

At present, each fan is provided with a dehumidifier, in the working mode of the existing dehumidifier, the dehumidifier is operated in a fixed power mode and a preset working time, and the control mode not only can cause the dehumidifier to generate a large amount of ineffective work (for example, when the dryness of the fan is very low, the fan still works), but also can expect that the dehumidifier can carry out self-adaptive adjustment on the working power according to specific conditions rather than still working at the preset power if the humidity in the offshore fan is suddenly increased.

Therefore, an optimized intelligent dehumidification control scheme is desired.

Disclosure of Invention

The present application is proposed to solve the above-mentioned technical problems. The embodiment of the application provides a remote intelligent dehumidification control system and a method thereof for an offshore wind turbine, which are characterized in that an artificial intelligent control technology is adopted, a multi-scale neighborhood feature extraction module is utilized to dynamically and implicitly extract the humidity values at a plurality of time points, wind data and the working power of a dehumidifier respectively, and Bayes is utilized to perform data feature fusion to perform real-time dynamic regulation and control on the power of the dehumidifier. In this way, the dynamic moisture change characteristics brought by the wind data are introduced into the dehumidifier power control model, so that the power control of the dehumidifier can be prospective and predictive, and the humidity of the offshore wind turbine in a future time period can be kept within a preset reasonable range.

According to an aspect of the present application, there is provided a remote intelligent dehumidification control system of an offshore wind turbine, including:

the system comprises an environment data acquisition module, a data processing module and a data processing module, wherein the environment data acquisition module is used for acquiring humidity values of a plurality of preset time points in a preset time period acquired by a hygrometer and wind data of the plurality of preset time points acquired by a wind speed station, the wind data comprises a wind speed value and a wind direction, and the wind direction is represented by a sine value or a cosine value of an angle;

the working state data acquisition module is used for acquiring the working power of the dehumidifier at the plurality of preset time points;

the single-point wind data coding module is used for enabling wind data of each preset time point in the wind data of the preset time points to pass through a full connection layer so as to obtain a plurality of wind characteristic vectors corresponding to the preset time points;

the wind data multi-scale neighborhood coding module is used for arranging the plurality of wind characteristic vectors corresponding to each preset time point into one-dimensional characteristic vectors and then obtaining the multi-scale wind characteristic vectors through the first multi-scale neighborhood characteristic extraction module;

the humidity data multi-scale neighborhood coding module is used for enabling the humidity values of the plurality of preset time points to pass through the second multi-scale neighborhood feature extraction module to obtain a multi-scale humidity feature vector;

the power data multi-scale neighborhood coding module is used for enabling the working power of the dehumidifier at the plurality of preset time points to pass through the third multi-scale neighborhood feature extraction module so as to obtain a multi-scale power feature vector;

a Bayesian inference module for fusing the multi-scale power feature vector, the multi-scale humidity feature vector, and the multi-scale wind feature vector using a Bayesian probability model to obtain a posterior probability feature vector, wherein the multi-scale power feature vector is a prior probability vector, the multi-scale humidity feature vector is an event probability vector, and the multi-scale wind feature vector is an evidence probability vector;

the posterior distribution correction module is used for correcting the characteristic value of each position in the posterior probability characteristic vector based on the reciprocal of the mean value of the characteristic values of all the positions in the posterior probability characteristic vector to obtain a corrected posterior probability characteristic vector; and

and the control result generation module is used for enabling the corrected posterior probability characteristic vector to pass through a classifier to obtain a classification result, and the classification result is used for indicating that the power value of the dehumidifier at the current time point should be increased, kept or reduced.

In the above remote intelligent dehumidification control system of the offshore wind turbine, the single-point wind data encoding module is further configured to: arranging the wind data of each preset time point in the wind data of the preset time points into an input vector according to a time dimension; and performing full-connection coding on the input vector by using the full-connection layer according to the following formula to extract high-dimensional implicit features of feature values of all positions in the input vector so as to obtain a plurality of wind feature vectors corresponding to all preset time points, wherein the formula is as follows:

wherein X is the input vector, Y is the output vector, W is the weight matrix, B is the offset vector,

representing a matrix multiplication.

In the above remote intelligent dehumidification control system of an offshore wind turbine, the wind data multi-scale neighborhood coding module includes: a first convolution unit, configured to input the one-dimensional feature vector into a first convolution layer of the first multi-scale neighborhood feature extraction module to obtain a first neighborhood scale wind feature vector, where the first convolution layer has a first one-dimensional convolution kernel of a first length; a second convolution unit, configured to input the one-dimensional feature vector into a second convolution layer of the first multi-scale neighborhood feature extraction module to obtain a second neighborhood scale wind feature vector, where the second convolution layer has a second one-dimensional convolution kernel with a second length, and the first length is different from the second length; and the cascading unit is used for cascading the first neighborhood scale wind characteristic vector and the second neighborhood scale wind characteristic vector to obtain the multi-scale wind characteristic vector.

In the remote intelligent dehumidification control system of the offshore wind turbine, the first convolution unit is further used for; performing one-dimensional convolution coding on the one-dimensional feature vector by using a first convolution layer of the first multi-scale neighborhood feature extraction module according to the following formula to obtain a first neighborhood scale wind feature vector;

wherein the formula is:

wherein, a is the width of the first convolution kernel in the X direction, F is a first convolution kernel parameter vector, G is a local vector matrix operated with the convolution kernel function, w is the size of the first convolution kernel, and X represents a tension input vector.

In the above remote intelligent dehumidification control system of an offshore wind turbine, the second convolution unit is further configured to: performing one-dimensional convolution coding on the one-dimensional feature vector by using a second convolution layer of the first multi-scale neighborhood feature extraction module according to the following formula to obtain a second neighborhood scale wind feature vector;

wherein the formula is:

wherein b is the width of the second convolution kernel in the X direction, F is a parameter vector of the second convolution kernel, G is a local vector matrix operated with the convolution kernel function, m is the size of the second convolution kernel, and X represents a tension input vector.

In the above remote intelligent dehumidification control system of an offshore wind turbine, the bayesian inference module is further configured to: fusing the multi-scale power feature vector, the multi-scale humidity feature vector and the multi-scale wind feature vector by using a Bayesian probability model to obtain the posterior probability feature vector;

wherein the formula is:

qi＝pi*ai/bi

where pi is the eigenvalue of each location in the multi-scale power eigenvector, ai and bi are the eigenvalues of each location in the multi-scale humidity eigenvector and the multi-scale wind eigenvector, respectively, and qi is the eigenvalue of each location in the posterior probability eigenvector.

In the remote intelligent dehumidification control system of above-mentioned offshore wind turbine, posterior distribution correction module is further used for: based on the reciprocal of the mean value of the feature values of all the positions in the posterior probability feature vector, correcting the feature values of all the positions in the posterior probability feature vector by the following formula to obtain the corrected posterior probability feature vector;

wherein the formula is:

wherein V represents the posterior probability feature vector,

an inverse number indicating the mean value of the feature values of all the positions in the a posteriori probability feature vector, indicates a dot-by-position multiplication.

In the above remote intelligent dehumidification control system of the offshore wind turbine, the control result generation module is further configured to: processing the corrected a posteriori probability feature vectors using the classifier to obtain the classification result, which isWherein the formula is: softmax { (W) _n ,B _n ):…:(W ₁ ,B ₁ ) I X }, wherein W ₁ To W _n As a weight matrix, B ₁ To B _n And X is the corrected posterior probability feature vector.

According to another aspect of the present application, a remote intelligent dehumidification control method of an offshore wind turbine includes:

acquiring humidity values of a plurality of preset time points in a preset time period collected by a hygrometer and wind data of the plurality of preset time points collected by a wind speed station, wherein the wind data comprise a wind speed value and a wind direction, and the wind direction is represented by a sine value or a cosine value of an angle;

obtaining the working power of the dehumidifier at the plurality of preset time points;

wind data of each preset time point in the wind data of the preset time points pass through a full connection layer to obtain a plurality of wind characteristic vectors corresponding to the preset time points;

arranging the plurality of wind characteristic vectors corresponding to each preset time point into one-dimensional characteristic vectors, and then obtaining multi-scale wind characteristic vectors through a first multi-scale neighborhood characteristic extraction module;

enabling the humidity values of the plurality of preset time points to pass through a second multi-scale neighborhood characteristic extraction module to obtain a multi-scale humidity characteristic vector;

the working power of the dehumidifier at the preset time points is processed by a third multi-scale neighborhood characteristic extraction module to obtain a multi-scale power characteristic vector;

fusing the multi-scale power feature vector, the multi-scale humidity feature vector and the multi-scale wind feature vector by using a Bayesian probability model to obtain a posterior probability feature vector, wherein the multi-scale power feature vector is a prior probability vector, the multi-scale humidity feature vector is an event probability vector, and the multi-scale wind feature vector is an evidence probability vector;

correcting the characteristic value of each position in the posterior probability characteristic vector based on the reciprocal of the mean value of the characteristic values of all the positions in the posterior probability characteristic vector to obtain a corrected posterior probability characteristic vector; and

and passing the corrected posterior probability feature vector through a classifier to obtain a classification result, wherein the classification result is used for indicating that the power value of the dehumidifier at the current time point should be increased, kept or reduced.

In the above method for remotely and intelligently controlling dehumidification of an offshore wind turbine, passing the wind data at each predetermined time point in the wind data at the plurality of predetermined time points through a full connection layer to obtain a plurality of wind eigenvectors corresponding to each predetermined time point, the method includes: arranging wind data of each preset time point in the wind data of the preset time points into an input vector according to a time dimension; and performing full-connection coding on the input vector by using the full-connection layer according to the following formula to extract high-dimensional implicit features of feature values of all positions in the input vector so as to obtain a plurality of wind feature vectors corresponding to all preset time points, wherein the formula is as follows:

representing a matrix multiplication.

In the above remote intelligent dehumidification control method of an offshore wind turbine, the step of arranging the plurality of wind eigenvectors corresponding to each predetermined time point as one-dimensional eigenvectors and then obtaining multi-scale wind eigenvectors by a first multi-scale neighborhood feature extraction module includes: inputting the one-dimensional feature vector into a first convolution layer of the first multi-scale neighborhood feature extraction module to obtain a first neighborhood scale wind feature vector, wherein the first convolution layer has a first one-dimensional convolution kernel with a first length; inputting the one-dimensional feature vector into a second convolution layer of the first multi-scale neighborhood feature extraction module to obtain a second neighborhood scale wind feature vector, wherein the second convolution layer has a second one-dimensional convolution kernel with a second length, and the first length is different from the second length; and cascading the first neighborhood scale wind feature vector and the second neighborhood scale wind feature vector to obtain the multi-scale wind feature vector.

In the above remote intelligent dehumidification control method of an offshore wind turbine, inputting the one-dimensional feature vector into the first convolution layer of the first multi-scale neighborhood feature extraction module to obtain a first neighborhood scale wind feature vector, includes: performing one-dimensional convolution coding on the one-dimensional feature vector by using a first convolution layer of the first multi-scale neighborhood feature extraction module according to the following formula to obtain a first neighborhood scale wind feature vector;

wherein the formula is:

In the above remote intelligent dehumidification control method of an offshore wind turbine, inputting the one-dimensional feature vector into the second convolution layer of the first multi-scale neighborhood feature extraction module to obtain a second neighborhood scale wind feature vector, includes: performing one-dimensional convolution coding on the one-dimensional feature vector by using a second convolution layer of the first multi-scale neighborhood feature extraction module according to the following formula to obtain a second neighborhood scale wind feature vector;

wherein the formula is:

In the above remote intelligent dehumidification control method of an offshore wind turbine, a bayesian probability model is used to fuse the multi-scale power feature vector, the multi-scale humidity feature vector and the multi-scale wind feature vector to obtain a posterior probability feature vector, where the multi-scale power feature vector is a prior probability vector, the multi-scale humidity feature vector is an event probability vector, and the multi-scale wind feature vector is an evidence probability vector, and the method includes: fusing the multi-scale power feature vector, the multi-scale humidity feature vector and the multi-scale wind feature vector by using a Bayesian probability model to obtain the posterior probability feature vector;

wherein the formula is:

qi＝pi*ai/bi

where pi is the eigenvalue of each location in the multi-scale power eigenvector, ai and bi are the eigenvalues of each location in the multi-scale humidity eigenvector and the multi-scale wind eigenvector, respectively, and qi is the eigenvalue of each location in the a posteriori probability eigenvector.

In the above remote intelligent dehumidification control method of an offshore wind turbine, based on a reciprocal of a mean value of feature values of all positions in the posterior probability feature vector, correcting the feature values of each position in the posterior probability feature vector to obtain a corrected posterior probability feature vector, includes: based on the reciprocal of the mean value of the characteristic values of all the positions in the posterior probability characteristic vector, correcting the characteristic value of each position in the posterior probability characteristic vector by the following formula to obtain the corrected posterior probability characteristic vector;

wherein the formula is:

wherein V represents the posterior probability feature vector,

In the above remote intelligent dehumidification control method of an offshore wind turbine, passing the corrected posterior probability feature vector through a classifier to obtain a classification result, where the classification result is used to indicate that the power value of the dehumidifier at the current time point should be increased, maintained, or decreased, and includes: processing the corrected posterior probability feature vector using the classifier to obtain the classification result with the following formula: softmax { (W) _n ,B _n ):…:(W ₁ ,B ₁ ) | X }, wherein W ₁ To W _n As a weight matrix, B ₁ To B _n Is a bias vector, and X is the corrected posterior probability feature vector.

Compared with the prior art, the remote intelligent dehumidification control system and method of the offshore wind turbine provided by the application have the advantages that by adopting an artificial intelligent control technology, the humidity values, the wind data and the working power of the dehumidifier at multiple time points are respectively subjected to dynamic implicit characteristic extraction by using the multi-scale neighborhood characteristic extraction module, and the real-time dynamic regulation and control of the power of the dehumidifier are carried out by fusing data characteristics by using Bayes. In this way, the dynamic change characteristics of moisture brought by wind data are introduced into the dehumidifier power control model, so that the power control of the dehumidifier has foresight and predictability, and the humidity of the offshore wind turbine in a future time period can be kept within a preset reasonable range.

Drawings

The above and other objects, features and advantages of the present application will become more apparent by describing in more detail embodiments of the present application with reference to the attached drawings. The accompanying drawings are included to provide a further understanding of the embodiments of the application and are incorporated in and constitute a part of this specification, illustrate embodiments of the application and together with the description serve to explain the principles of the application. In the drawings, like reference numbers generally represent like parts or steps.

Fig. 1 is an application scenario diagram of a remote intelligent dehumidification control system of an offshore wind turbine according to an embodiment of the application.

FIG. 2 is a block diagram of a remote intelligent dehumidification control system of an offshore wind turbine according to an embodiment of the application.

Fig. 3 is a flowchart of a remote intelligent dehumidification control method of an offshore wind turbine according to an embodiment of the application.

Fig. 4 is a schematic structural diagram of a remote intelligent dehumidification control method for an offshore wind turbine according to an embodiment of the application.

Detailed Description

Hereinafter, example embodiments according to the present application will be described in detail with reference to the accompanying drawings. It should be understood that the described embodiments are only some embodiments of the present application and not all embodiments of the present application, and that the present application is not limited by the example embodiments described herein.

Overview of scenes

As previously mentioned, offshore wind turbines operate in high humidity, high salt environments for long periods of time. Metal equipment and facilities inside the fan are easy to corrode. Generally, the critical point of the relative humidity of metal corrosion is 45% -50%, the corrosion phenomenon occurs when the relative humidity exceeds the limit, the corrosion speed is accelerated along with the increase of the relative humidity, and once dew condensation occurs on steel, the corrosion process is quicker.

At present, each fan is provided with a dehumidifier, in the working mode of the existing dehumidifier, the dehumidifier is operated in a fixed power mode and a preset working time, the control mode not only can cause the dehumidifier to generate a large amount of ineffective work (for example, when the dryness of the fan is very low, the dehumidifier still works), but also, if the humidity in the offshore fan suddenly increases, the dehumidifier is expected to carry out self-adaptive adjustment on the working power according to specific conditions instead of still working at the preset power. Therefore, an optimized intelligent dehumidification control scheme is desired.

Accordingly, the inventor considers that the humidity in the offshore wind turbine at each time point has implicit characteristic associated information, and the change of the humidity also causes the change of the working power of the dehumidifier, so if the power of the dehumidifier is dynamically regulated in real time according to the actual condition of the humidity in the wind turbine, the implicit associated characteristic mining needs to be performed on the humidity value at each time point and the working power value of the dehumidifier. And the inventor also considers that wind brings more moisture, therefore, the moisture dynamic change characteristic brought by the wind data is introduced into the dehumidifier power control model, so that the power control of the dehumidifier has foresight and predictability, and the humidity of the offshore wind turbine in the future time period can be kept within a preset reasonable range.

Specifically, in the technical solution of the present application, first, for the collection of environmental data, humidity values of a plurality of predetermined time points in a predetermined time period are collected by a hygrometer disposed in a wind turbine and wind data of the plurality of predetermined time points are collected by a wind speed station, where the wind data includes a wind speed value and a wind direction, and the wind direction is represented by a sine value or a cosine value of an angle. And then, acquiring the working power of the dehumidifier at the plurality of preset time points through a power detector for the collection of the working state data of the dehumidifier.

Further, aiming at the wind data of the plurality of preset time points, the wind data of each preset time point in the wind data of the plurality of preset time points is processed through a full connection layer, so as to extract the high-dimensional hidden features of the wind data of each preset time point, and thus, a plurality of wind feature vectors corresponding to each preset time point are obtained.

Then, the wind eigenvectors corresponding to the preset time points are arranged into one-dimensional eigenvectors according to the time dimension so as to facilitate subsequent feature mining. It should be understood that convolutional neural networks were originally models applied in the image domain, but the idea of local feature extraction can be applied to time series data analysis as well. For example, a time-series convolution structure with a convolution kernel size of 3, the convolution kernel is moved along the time dimension in the form of a sliding window for time-series data input, and outputs a weighted sum of the data within each time-series segment. Each convolution unit stacks a plurality of convolution kernels to output a multi-dimensional feature. The large convolution kernel can extract features from a large-scale time sequence neighborhood, wherein the influence of each item of numerical value in the neighborhood is smaller, so that the fluctuation of input data is weakened, and the influence of noise points on output features is reduced. However, the difference of numerical value changes is weakened by the large-scale convolution kernel, and the problem of excessive smoothness is easily caused, so that the output characteristics lose the discrimination capability. In contrast, small scale convolution kernels are better able to retain information in the input data, but are also more susceptible to interference from noise therein. Therefore, the convolution units with different sizes are combined to extract the features of different time sequence scales in consideration of the characteristics of convolution with different scales. And then completing feature fusion by adopting a feature splicing mode, thereby obtaining the multi-scale neighborhood features.

That is, specifically, in the technical solution of the present application, further, the convolution layers of the one-dimensional convolution kernels with different scales of the first multi-scale neighborhood feature extraction module are used to perform one-dimensional convolution coding on the one-dimensional feature vectors respectively, and then the obtained feature vectors corresponding to the two one-dimensional convolution kernels with different scales are cascaded to obtain the multi-scale wind feature vector. Particularly, by the method, the characteristic information implicit in the dynamic change of the wind data can be extracted, the output characteristics comprise the smoothed characteristics and the originally input characteristics, the information loss is avoided, and the accuracy of subsequent classification is improved. In other examples of the present application, the multi-scale neighborhood feature extraction module may further include a greater number of one-dimensional convolution layers, which perform neighborhood-related feature extraction of different scales using one-dimensional convolution kernels of different lengths, which is not limited in this application.

Similarly, for the humidity values of the plurality of predetermined time points and the working power of the dehumidifier, after the humidity values of the plurality of predetermined time points are arranged into humidity input vectors and the working power of the dehumidifier of the plurality of predetermined time points is arranged into power input vectors, one-dimensional convolution coding is performed in the convolution layer of the multi-scale neighborhood feature extraction module with the one-dimensional convolution kernels of different scales, and then the obtained feature vectors corresponding to the two one-dimensional convolution kernels of different scales are cascaded to obtain the multi-scale humidity feature vector and the multi-scale power feature vector respectively. Therefore, the dynamic change characteristic information of the humidity value and the working power of the dehumidifier on the time dimension can be extracted.

It should be appreciated that in view of using the multi-scale power feature vector as a prior probability, the technical solution in the present application aims to update the prior probability to get a posterior probability in case of new evidence, i.e. moisture brought by new said wind data. Then, according to a bayesian formula, a posterior probability is a prior probability multiplied by an event probability divided by an evidence probability, and therefore, in the technical scheme of the application, a bayesian probability model is used to fuse the multi-scale power feature vector, the multi-scale humidity feature vector and the multi-scale wind feature vector to obtain the posterior probability feature vector, wherein the multi-scale power feature vector is the prior probability vector, the multi-scale humidity feature vector is the event probability vector, and the multi-scale wind feature vector is the evidence probability vector. In this way, the moisture dynamic change characteristics brought by the wind data can be introduced into the dehumidifier power control model, so that the power control of the dehumidifier is prospective and predictive, and the humidity of the offshore wind turbine in the future period can be kept within a preset reasonable range. And then, the posterior probability feature vector is passed through a classifier to obtain a classification result which is used for indicating that the power value of the dehumidifier at the current time point should be increased, maintained or reduced.

However, the fusion of the multi-scale power feature vector, the multi-scale humidity feature vector and the multi-scale wind feature vector using the bayesian probability model is a calculation by location, that is, each location of the posterior probability feature vector corresponds to the posterior probability under different time sequence features, and has a certain phase attribute. However, the classification of the classifier is usually location-independent, which results in possible bias of the classification results.

Therefore, in the technical solution of the present application, the vectors for performing phase sensing on the posterior probability feature vectors are further aggregated according to location, that is:

wherein V represents the posterior probability feature vector,

The optimized representation of the phase perception of the posterior probability feature vector V' introduces a class real value-virtual value representation of amplitude-phase, the vectors are spliced and expanded according to positions based on the principle of Euler formula, and position aggregation is carried out in a multi-layer phase perception mode, so that induction deviation possibly caused when a real value classification task without position attribute is carried out on the posterior probability feature vector V is compensated, and the accuracy of a classification result is improved.

Based on this, this application provides a remote intelligent dehumidification control system of offshore wind turbine, and it includes: the system comprises an environment data acquisition module, a data processing module and a data processing module, wherein the environment data acquisition module is used for acquiring humidity values of a plurality of preset time points in a preset time period acquired by a hygrometer and wind data of the plurality of preset time points acquired by a wind speed station, the wind data comprises a wind speed value and a wind direction, and the wind direction is represented by a sine value or a cosine value of an angle; the working state data acquisition module is used for acquiring the working power of the dehumidifier at the plurality of preset time points; the single-point wind data coding module is used for enabling wind data of each preset time point in the wind data of the preset time points to pass through a full connection layer so as to obtain a plurality of wind characteristic vectors corresponding to the preset time points; the wind data multi-scale neighborhood coding module is used for arranging the plurality of wind characteristic vectors corresponding to the preset time points into one-dimensional characteristic vectors and then obtaining the multi-scale wind characteristic vectors through the first multi-scale neighborhood characteristic extraction module; the humidity data multi-scale neighborhood coding module is used for enabling the humidity values of the plurality of preset time points to pass through the second multi-scale neighborhood feature extraction module to obtain a multi-scale humidity feature vector; the power data multi-scale neighborhood coding module is used for enabling the working power of the dehumidifier at the plurality of preset time points to pass through the third multi-scale neighborhood feature extraction module so as to obtain a multi-scale power feature vector; a Bayesian inference module for fusing the multi-scale power feature vector, the multi-scale humidity feature vector and the multi-scale wind feature vector to obtain a posterior probability feature vector by using a Bayesian probability model, wherein the multi-scale power feature vector is a prior probability vector, the multi-scale humidity feature vector is an event probability vector, and the multi-scale wind feature vector is an evidence probability vector; the posterior distribution correction module is used for correcting the characteristic values of all the positions in the posterior probability characteristic vector based on the reciprocal of the mean value of the characteristic values of all the positions in the posterior probability characteristic vector to obtain a corrected posterior probability characteristic vector; and the control result generation module is used for enabling the corrected posterior probability characteristic vector to pass through a classifier to obtain a classification result, and the classification result is used for indicating that the power value of the dehumidifier at the current time point should be increased, kept or reduced.

Fig. 1 illustrates an application scenario diagram of a remote intelligent dehumidification control system of an offshore wind turbine according to an embodiment of the application. As shown in fig. 1, in this application scenario, first, humidity values at a plurality of predetermined time points within a predetermined time period are collected by a hygrometer (e.g., H as illustrated in fig. 1) disposed in a fan (e.g., F as illustrated in fig. 1) and wind data at the plurality of predetermined time points are collected by a wind speed station (e.g., W as illustrated in fig. 1), the wind data including a wind speed value and a wind direction, the wind direction being represented by a sine value or a cosine value of an angle, and an operating power of a dehumidifier (e.g., T as illustrated in fig. 1) at the plurality of predetermined time points is obtained by a power detector (e.g., P as illustrated in fig. 1). Then, the acquired humidity values, wind data and operating power of the dehumidifier at the plurality of predetermined time points in the predetermined time period are input into a server (for example, a cloud server S as illustrated in fig. 1) in which a remote intelligent dehumidification control algorithm of an offshore wind turbine is deployed, wherein the server can process the humidity values, wind data and operating power of the dehumidifier at the plurality of predetermined time points in the predetermined time period by using the remote intelligent dehumidification control algorithm of the offshore wind turbine to generate a classification result indicating that the power value of the dehumidifier at the current time point should be increased, maintained or decreased.

Having described the general principles of the present application, various non-limiting embodiments of the present application will now be described with reference to the accompanying drawings.

Exemplary System

FIG. 2 illustrates a block diagram of a remote intelligent dehumidification control system of an offshore wind turbine in accordance with an embodiment of the present application. As shown in fig. 2, the remote intelligent dehumidification control system 200 of an offshore wind turbine according to an embodiment of the present application includes: an environment data acquisition module 210, configured to acquire humidity values at a plurality of predetermined time points within a predetermined time period acquired by a hygrometer and wind data at the plurality of predetermined time points acquired by a wind speed station, where the wind data includes a wind speed value and a wind direction, and the wind direction is represented by a sine value or a cosine value of an angle; the working state data acquisition module 220 is configured to acquire the working powers of the dehumidifier at the plurality of predetermined time points; a single-point wind data encoding module 230, configured to pass wind data at each predetermined time point in the wind data at the plurality of predetermined time points through a full connection layer to obtain a plurality of wind feature vectors corresponding to each predetermined time point; a wind data multi-scale neighborhood coding module 240, configured to arrange the multiple wind feature vectors corresponding to each predetermined time point into one-dimensional feature vectors, and then obtain multi-scale wind feature vectors through a first multi-scale neighborhood feature extraction module; a humidity data multi-scale neighborhood coding module 250, configured to pass the humidity values at the multiple predetermined time points through a second multi-scale neighborhood feature extraction module to obtain a multi-scale humidity feature vector; the power data multi-scale neighborhood coding module 260 is used for enabling the working power of the dehumidifier at the plurality of preset time points to pass through a third multi-scale neighborhood feature extraction module so as to obtain a multi-scale power feature vector; a bayesian inference module 270 configured to fuse the multi-scale power feature vector, the multi-scale humidity feature vector, and the multi-scale wind feature vector using a bayesian probability model to obtain a posterior probability feature vector, where the multi-scale power feature vector is a prior probability vector, the multi-scale humidity feature vector is an event probability vector, and the multi-scale wind feature vector is an evidence probability vector; a posterior distribution correction module 280, configured to correct the feature values of each position in the posterior probability feature vector based on a reciprocal of a mean value of the feature values of all positions in the posterior probability feature vector to obtain a corrected posterior probability feature vector; and a control result generating module 290, configured to pass the corrected a posteriori probability feature vector through a classifier to obtain a classification result, where the classification result is used to indicate that the power value of the dehumidifier at the current time point should be increased, maintained, or decreased.

Specifically, in this embodiment of the present application, the environment data collecting module 210 and the operating state data collecting module 220 are configured to obtain humidity values at a plurality of predetermined time points in a predetermined time period collected by a hygrometer and wind data at the plurality of predetermined time points collected by a wind speed station, where the wind data includes a wind speed value and a wind direction, and the wind direction is represented by a sine value or a cosine value of an angle, and obtain operating power of the dehumidifier at the plurality of predetermined time points. As described above, in the technical solution of the present application, since the humidity in the offshore wind turbine at each time point has implicit characteristic related information and the change of the humidity also causes the change of the operating power of the dehumidifier, if the power of the dehumidifier is to be dynamically adjusted in real time according to the actual condition of the humidity in the wind turbine, it is necessary to perform deep implicit related characteristic mining on the humidity value at each time point and the operating power value of the dehumidifier. And also considering that wind brings more moisture, in the technical scheme of the application, the dynamic moisture change characteristics brought by the wind data are introduced into the dehumidifier power control model, so that the power control of the dehumidifier has foresight and predictability, and the humidity of the offshore wind turbine in the future time period can be kept within a preset reasonable range.

That is, specifically, in the technical solution of the present application, first, for the collection of the environmental data, humidity values at a plurality of predetermined time points in a predetermined time period are collected by a hygrometer disposed in a wind turbine, and wind data at the plurality of predetermined time points are collected by a wind speed station, the wind data including a wind speed value and a wind direction, the wind direction being represented by a sine value or a cosine value of an angle. And then, acquiring the working power of the dehumidifier at a plurality of preset time points through a power detector for the collection of the working state data of the dehumidifier.

Specifically, in this embodiment of the present application, the single-point wind data encoding module 230 is configured to pass wind data at each predetermined time point in the wind data at the multiple predetermined time points through a full connection layer to obtain multiple wind feature vectors corresponding to each predetermined time point. That is, in the technical solution of the present application, further, for the wind data at the plurality of predetermined time points, the wind data at each predetermined time point in the wind data at the plurality of predetermined time points is processed through a full connection layer to extract high-dimensional implicit features of the wind data at each predetermined time point, so as to obtain a plurality of wind feature vectors corresponding to each predetermined time point.

More specifically, in this embodiment of the present application, the single-point wind data encoding module is further configured to: arranging wind data of each preset time point in the wind data of the preset time points into an input vector according to a time dimension; and performing full-connection coding on the input vector by using the full-connection layer according to the following formula to extract high-dimensional implicit features of feature values of all positions in the input vector so as to obtain a plurality of wind feature vectors corresponding to all preset time points, wherein the formula is as follows:

representing a matrix multiplication.

Specifically, in this embodiment of the present application, the wind data multi-scale neighborhood coding module 240 is configured to arrange the plurality of wind feature vectors corresponding to each predetermined time point into one-dimensional feature vectors, and then obtain the multi-scale wind feature vectors through the first multi-scale neighborhood feature extraction module. That is, in the technical solution of the present application, the multiple wind eigenvectors corresponding to each predetermined time point are further arranged as a one-dimensional eigenvector according to a time dimension, so as to facilitate subsequent feature mining. It should be understood that convolutional neural networks were originally models applied in the image domain, but the idea of local feature extraction can be applied to time series data analysis as well. For example, a time-series convolution structure with a convolution kernel size of 3, the convolution kernel is moved along the time dimension in the form of a sliding window for time-series data input, and outputs a weighted sum of the data within each time-series segment. Each convolution unit stacks a plurality of convolution kernels to output a multi-dimensional feature. The large convolution kernel can extract features from a large-scale time sequence neighborhood, wherein the influence of each item of numerical value in the neighborhood is smaller, so that the fluctuation of input data is weakened, and the influence of noise points on output features is reduced. However, the difference of numerical value changes is weakened by the large-scale convolution kernel, and the problem of excessive smoothness is easily caused, so that the output characteristics lose the discrimination capability. In contrast, small scale convolution kernels are better able to retain information in the input data, but are also more susceptible to interference from noise therein. Therefore, the convolution units with different sizes are combined to extract the features of different time sequence scales in consideration of the characteristics of convolution with different scales. And then completing feature fusion by adopting a feature splicing mode, thereby obtaining the multi-scale neighborhood features.

More specifically, in this embodiment of the present application, the wind data multi-scale neighborhood coding module includes: and the first convolution unit is used for inputting the one-dimensional feature vector into a first convolution layer of the first multi-scale neighborhood feature extraction module to obtain a first neighborhood scale wind feature vector, wherein the first convolution layer has a first one-dimensional convolution kernel with a first length. Accordingly, in one specific example, the one-dimensional feature vector is one-dimensionally convolution encoded using a first convolution layer of the first multi-scale neighborhood feature extraction module to obtain the first neighborhood scale wind feature vector;

wherein the formula is:

wherein, a is the width of the first convolution kernel in the X direction, F is a first convolution kernel parameter vector, G is a local vector matrix operated with the convolution kernel function, w is the size of the first convolution kernel, and X represents a tension input vector. A second convolution unit, configured to input the one-dimensional feature vector into a second convolution layer of the first multi-scale neighborhood feature extraction module to obtain a second neighborhood scale wind feature vector, where the second convolution layer has a second one-dimensional convolution kernel with a second length, and the first length is different from the second length. Accordingly, in one particular example, the one-dimensional feature vector is one-dimensionally convolution encoded using a second convolution layer of the first multi-scale neighborhood feature extraction module to obtain the second neighborhood scale wind feature vector;

wherein the formula is:

wherein b is the width of the second convolution kernel in the X direction, F is a parameter vector of the second convolution kernel, G is a local vector matrix operated with the convolution kernel function, m is the size of the second convolution kernel, and X represents a tension input vector. The cascade unit is used for cascading the first neighborhood scale wind characteristic vector and the second neighborhood scale wind characteristic vector to obtain the multi-scale wind characteristic vector.

Specifically, in this embodiment of the present application, the humidity data multi-scale neighborhood coding module 250 and the power data multi-scale neighborhood coding module 260 are configured to pass humidity values of the multiple predetermined time points through a second multi-scale neighborhood feature extraction module to obtain a multi-scale humidity feature vector, and pass working powers of the dehumidifier of the multiple predetermined time points through a third multi-scale neighborhood feature extraction module to obtain a multi-scale power feature vector. It should be understood that, in the technical solution of the present application, similarly, after arranging the humidity values of the plurality of predetermined time points and the operating power of the dehumidifier into humidity input vectors, and arranging the operating power of the dehumidifier of the plurality of predetermined time points into power input vectors, performing one-dimensional convolution coding in the convolution layer having one-dimensional convolution kernels of different scales through the multi-scale neighborhood feature extraction module, and then cascading the obtained feature vectors corresponding to two one-dimensional convolution kernels of different scales to obtain the multi-scale humidity feature vector and the multi-scale power feature vector respectively. Therefore, the dynamic change characteristic information of the humidity value and the working power of the dehumidifier on the time dimension can be extracted.

Specifically, in this embodiment of the present application, the bayesian inference module 270 is configured to use a bayesian probability model to fuse the multi-scale power feature vector, the multi-scale humidity feature vector, and the multi-scale wind feature vector to obtain a posterior probability feature vector, where the multi-scale power feature vector is a prior probability vector, the multi-scale humidity feature vector is an event probability vector, and the multi-scale wind feature vector is an evidence probability vector. It should be appreciated that in view of using the multi-scale power feature vector as a prior probability, the technical solution in the present application aims to update the prior probability to get a posterior probability in case of new evidence, i.e. moisture brought by new said wind data. Then, according to a bayesian formula, a posterior probability is a prior probability multiplied by an event probability divided by an evidence probability, and therefore, in the technical scheme of the application, a bayesian probability model is used to fuse the multi-scale power feature vector, the multi-scale humidity feature vector and the multi-scale wind feature vector to obtain the posterior probability feature vector, wherein the multi-scale power feature vector is the prior probability vector, the multi-scale humidity feature vector is the event probability vector, and the multi-scale wind feature vector is the evidence probability vector. In this way, the moisture dynamic change characteristics brought by the wind data can be introduced into the dehumidifier power control model, so that the power control of the dehumidifier is prospective and predictive, and the humidity of the offshore wind turbine in the future period can be kept within a preset reasonable range.

More specifically, in this embodiment of the present application, the bayesian inference module is further configured to: the Bayesian inference module further to: fusing the multi-scale power feature vector, the multi-scale humidity feature vector and the multi-scale wind feature vector by using a Bayesian probability model to obtain the posterior probability feature vector;

wherein the formula is:

qi＝pi*ai/bi

Specifically, in this embodiment of the present application, the posterior distribution correction module 280 is configured to correct the feature values of each position in the posterior probability feature vector based on a reciprocal of a mean of the feature values of all positions in the posterior probability feature vector to obtain a corrected posterior probability feature vector. It should be understood that, in the technical solution of the present application, after obtaining the posterior probability feature vector, the posterior probability feature vector is passed through a classifier to obtain a classification result indicating that the power value of the dehumidifier at the current time point should be increased, maintained, or decreased. However, the fusion of the multi-scale power feature vector, the multi-scale humidity feature vector and the multi-scale wind feature vector using the bayesian probability model is a calculation by location, that is, each location of the posterior probability feature vector corresponds to the posterior probability under different time sequence features, and has a certain phase attribute. However, the classification of the classifier is usually location-independent, which results in possible bias of the classification results. Therefore, in the technical solution of the present application, the vectors for phase sensing of the posterior probability feature vectors are further aggregated according to location.

More specifically, in this embodiment of the present application, the posterior distribution correcting module is further configured to: based on the reciprocal of the mean value of the feature values of all the positions in the posterior probability feature vector, correcting the feature values of all the positions in the posterior probability feature vector by the following formula to obtain the corrected posterior probability feature vector;

wherein the formula is:

wherein V represents the posterior probability feature vector,

an inverse number indicating the mean value of the feature values of all the positions in the a posteriori probability feature vector, indicates a dot-by-position multiplication. It should be understood that, here,the optimized representation of the phase perception of the posterior probability feature vector V' introduces the analog real value-virtual value representation of amplitude-phase, the vectors are spliced and expanded according to positions by the principle based on the Euler formula, and position aggregation is carried out in a multi-layer phase perception mode, so that induction deviation possibly caused when a real value classification task without position attribute is carried out on the posterior probability feature vector V is compensated, and the accuracy of a classification result is improved.

Specifically, in this embodiment of the present application, the control result generating module 290 is configured to pass the corrected posterior probability feature vector through a classifier to obtain a classification result, where the classification result is used to indicate that the power value of the dehumidifier at the current time point should be increased, maintained, or decreased. Accordingly, in one specific example, the corrected a posteriori probability feature vectors are processed using the classifier to obtain the classification result with the following formula: softmax { (W) _n ,B _n ):…:(W ₁ ,B ₁ ) | X }, wherein W ₁ To W _n As a weight matrix, B ₁ To B _n Is a bias vector, and X is the corrected posterior probability feature vector.

In summary, the remote intelligent dehumidification control system 200 of the offshore wind turbine according to the embodiment of the present application is illustrated, and the system dynamically extracts implicit characteristics of humidity values, wind data and working power of the dehumidifier at a plurality of time points by using a multi-scale neighborhood characteristic extraction module by using an artificial intelligent control technology, and performs real-time dynamic regulation and control of the power of the dehumidifier by using bayes to perform fusion of data characteristics. In this way, the dynamic change characteristics of moisture brought by wind data are introduced into the dehumidifier power control model, so that the power control of the dehumidifier has foresight and predictability, and the humidity of the offshore wind turbine in a future time period can be kept within a preset reasonable range.

As described above, the remote intelligent dehumidification control system 200 of an offshore wind turbine according to the embodiment of the present application may be implemented in various terminal devices, such as a server of a remote intelligent dehumidification control algorithm of an offshore wind turbine. In one example, the remote intelligent dehumidification control system 200 of the offshore wind turbine according to the embodiment of the present application may be integrated into a terminal device as a software module and/or a hardware module. For example, the remote intelligent dehumidification control system 200 of the offshore wind turbine may be a software module in the operating system of the terminal device, or may be an application developed for the terminal device; of course, the remote intelligent dehumidification control system 200 of the offshore wind turbine may also be one of many hardware modules of the terminal device.

Alternatively, in another example, the remote intelligent dehumidification control system 200 of the offshore wind turbine and the terminal device may be separate devices, and the remote intelligent dehumidification control system 200 of the offshore wind turbine may be connected to the terminal device through a wired and/or wireless network and transmit the interaction information according to an agreed data format.

Exemplary method

Fig. 3 illustrates a flowchart of a remote intelligent dehumidification control method of an offshore wind turbine. As shown in fig. 3, the remote intelligent dehumidification control method for an offshore wind turbine according to the embodiment of the application includes the steps of: s110, acquiring humidity values of a plurality of preset time points in a preset time period collected by a hygrometer and wind data of the plurality of preset time points collected by a wind speed station, wherein the wind data comprise a wind speed value and a wind direction, and the wind direction is represented by a sine value or a cosine value of an angle; s120, obtaining the working power of the dehumidifier at the preset time points; s130, enabling the wind data of each preset time point in the wind data of the preset time points to pass through a full-connection layer to obtain a plurality of wind characteristic vectors corresponding to the preset time points; s140, arranging the plurality of wind characteristic vectors corresponding to each preset time point into one-dimensional characteristic vectors, and then obtaining multi-scale wind characteristic vectors through a first multi-scale neighborhood characteristic extraction module; s150, enabling the humidity values of the plurality of preset time points to pass through a second multi-scale neighborhood feature extraction module to obtain a multi-scale humidity feature vector; s160, enabling the working power of the dehumidifier at the plurality of preset time points to pass through a third multi-scale neighborhood characteristic extraction module to obtain a multi-scale power characteristic vector; s170, fusing the multi-scale power feature vector, the multi-scale humidity feature vector and the multi-scale wind feature vector by using a Bayesian probability model to obtain a posterior probability feature vector, wherein the multi-scale power feature vector is a prior probability vector, the multi-scale humidity feature vector is an event probability vector, and the multi-scale wind feature vector is an evidence probability vector; s180, correcting the characteristic value of each position in the posterior probability characteristic vector based on the reciprocal of the mean value of the characteristic values of all the positions in the posterior probability characteristic vector to obtain a corrected posterior probability characteristic vector; and S190, the corrected posterior probability feature vector is processed by a classifier to obtain a classification result, and the classification result is used for indicating that the power value of the dehumidifier at the current time point should be increased, kept or reduced.

Fig. 4 illustrates an architecture diagram of a remote intelligent dehumidification control method of an offshore wind turbine according to an embodiment of the application. As shown in fig. 4, in the network architecture of the remote intelligent dehumidification control method of the offshore wind turbine, first, wind data at each of the plurality of predetermined time points (e.g., P1 as illustrated in fig. 4) obtained is passed through a full connection layer (e.g., FC as illustrated in fig. 4) to obtain a plurality of wind feature vectors (e.g., V1 as illustrated in fig. 4) corresponding to each of the predetermined time points; then, arranging the plurality of wind feature vectors corresponding to the respective predetermined time points into a one-dimensional feature vector (e.g., V2 as illustrated in fig. 4), and then passing through a first multi-scale neighborhood feature extraction module (e.g., MS1 as illustrated in fig. 4) to obtain a multi-scale wind feature vector (e.g., VF1 as illustrated in fig. 4); then, passing the humidity values (e.g., P2 as illustrated in fig. 4) of the plurality of predetermined time points through a second multi-scale neighborhood feature extraction module (e.g., MS2 as illustrated in fig. 4) to obtain a multi-scale humidity feature vector (e.g., VF2 as illustrated in fig. 4); next, passing the operating power (e.g., P3 as illustrated in fig. 4) of the dehumidifier at the plurality of predetermined time points through a third multi-scale neighborhood feature extraction module (e.g., MS3 as illustrated in fig. 4) to obtain a multi-scale power feature vector (e.g., VF3 as illustrated in fig. 4); then, fusing the multi-scale power feature vector, the multi-scale humidity feature vector and the multi-scale wind feature vector using a bayesian probabilistic model (e.g., BPM as illustrated in fig. 4) to obtain a posterior probability feature vector (e.g., VP1 as illustrated in fig. 4); then, based on the reciprocal of the mean of the feature values of all the positions in the posterior probability feature vector, correcting the feature values of each position in the posterior probability feature vector to obtain a corrected posterior probability feature vector (e.g., VP2 as illustrated in fig. 4); and, finally, passing the corrected a posteriori probability feature vector through a classifier (e.g., circle S as illustrated in fig. 4) to obtain a classification result indicating that the power value of the dehumidifier at the current point in time should be increased, should be maintained, or should be decreased.

More specifically, in step S110 and step S120, humidity values at a plurality of predetermined time points in a predetermined time period collected by a hygrometer and wind data at the plurality of predetermined time points collected by a wind speed station are obtained, the wind data including a wind speed value and a wind direction, the wind direction being represented by a sine value or a cosine value of an angle, and operating power of the dehumidifier at the plurality of predetermined time points is obtained. That is, in the technical solution of the present application, considering that the humidity in the offshore wind turbine at each time point has implicit characteristic related information and that the change of the humidity also causes the change of the working power of the dehumidifier, if the power of the dehumidifier is to be dynamically regulated in real time according to the actual condition of the humidity in the wind turbine, deep implicit related characteristic mining needs to be performed on the humidity value at each time point and the working power value of the dehumidifier. And also considering that wind brings more moisture, in the technical scheme of the application, the dynamic moisture change characteristics brought by the wind data are introduced into the dehumidifier power control model, so that the power control of the dehumidifier has foresight and predictability, and the humidity of the offshore wind turbine in the future time period can be kept within a preset reasonable range.

That is, specifically, in the technical solution of the present application, first, for the collection of the environmental data, humidity values at a plurality of predetermined time points within a predetermined time period are collected by a hygrometer disposed in a wind turbine and wind data at the plurality of predetermined time points are collected by a wind speed station, where the wind data includes a wind speed value and a wind direction, and the wind direction is represented by a sine value or a cosine value of an angle. And then, acquiring the working power of the dehumidifier at the plurality of preset time points through a power detector for the collection of the working state data of the dehumidifier.

More specifically, in step S130 and step S140, the wind data at each predetermined time point in the wind data at the predetermined time points is passed through a full-connected layer to obtain a plurality of wind feature vectors corresponding to each predetermined time point, and the plurality of wind feature vectors corresponding to each predetermined time point are arranged as a one-dimensional feature vector and then passed through a first multi-scale neighborhood feature extraction module to obtain a multi-scale wind feature vector. That is, in the technical solution of the present application, further, for the wind data at the plurality of predetermined time points, the wind data at each predetermined time point in the wind data at the plurality of predetermined time points is processed through a full connection layer to extract high-dimensional implicit features of the wind data at each predetermined time point, so as to obtain a plurality of wind feature vectors corresponding to each predetermined time point.

Further, the wind eigenvectors corresponding to the preset time points are arranged into a one-dimensional eigenvector according to the time dimension so as to facilitate subsequent feature mining. It should be understood that convolutional neural networks were originally models applied in the image domain, but the idea of local feature extraction can be applied to time series data analysis as well. For example, a time-series convolution structure with a convolution kernel size of 3, the convolution kernel is moved along the time dimension in the form of a sliding window for time-series data input, and outputs a weighted sum of the data within each time-series segment. Each convolution unit stacks a plurality of convolution kernels to output a multi-dimensional feature. The large convolution kernel can extract features from a large-scale time sequence neighborhood, wherein the influence of each item of numerical value in the neighborhood is smaller, so that the fluctuation of input data is weakened, and the influence of noise points on output features is reduced. However, the difference of numerical value changes is weakened by the large-scale convolution kernel, and the problem of excessive smoothness is easily caused, so that the output characteristics lose the discrimination capability. In contrast, small scale convolution kernels are better able to retain information in the input data, but are also more susceptible to interference from noise therein. Therefore, the convolution units with different sizes are combined to extract the features of different time sequence scales in consideration of the characteristics of convolution with different scales. And then completing feature fusion by adopting a feature splicing mode, thereby obtaining the multi-scale neighborhood features.

That is to say, specifically, in the technical solution of the present application, further, the convolution layers of the one-dimensional convolution kernels with different scales of the first multi-scale neighborhood feature extraction module are used to respectively perform one-dimensional convolution encoding on the one-dimensional feature vectors, and then the obtained feature vectors corresponding to the two one-dimensional convolution kernels with different scales are cascaded to obtain the multi-scale wind feature vector. Particularly, through the method, the characteristic information implied by the dynamic change of the wind data can be extracted, the output characteristics not only comprise the smoothed characteristics, but also keep the originally input characteristics, the information loss is avoided, and the accuracy of the subsequent classification is improved. In other examples of the present application, the multi-scale neighborhood feature extraction module may further include a greater number of one-dimensional convolution layers, which perform neighborhood-related feature extraction of different scales using one-dimensional convolution kernels of different lengths, which is not limited in this application.

More specifically, in step S150 and step S160, the humidity values of the plurality of predetermined time points are passed through a second multi-scale neighborhood feature extraction module to obtain a multi-scale humidity feature vector, and the working powers of the dehumidifier of the plurality of predetermined time points are passed through a third multi-scale neighborhood feature extraction module to obtain a multi-scale power feature vector. It should be understood that, in the technical solution of the present application, similarly, after arranging the humidity values of the plurality of predetermined time points and the operating power of the dehumidifier into humidity input vectors, and arranging the operating power of the dehumidifier of the plurality of predetermined time points into power input vectors, performing one-dimensional convolution coding in the convolution layer having one-dimensional convolution kernels of different scales through the multi-scale neighborhood feature extraction module, and then cascading the obtained feature vectors corresponding to two one-dimensional convolution kernels of different scales to obtain the multi-scale humidity feature vector and the multi-scale power feature vector respectively. Therefore, the dynamic change characteristic information of the humidity value and the working power of the dehumidifier on the time dimension can be extracted.

More specifically, in step S170, a bayesian probability model is used to fuse the multi-scale power feature vector, the multi-scale humidity feature vector and the multi-scale wind feature vector to obtain a posterior probability feature vector, where the multi-scale power feature vector is a prior probability vector, the multi-scale humidity feature vector is an event probability vector, and the multi-scale wind feature vector is an evidence probability vector. It should be understood that, considering the use of the multi-scale power eigenvector as a prior probability, the purpose of the solution in the present application is to update the prior probability to get a posterior probability in case of new evidence, i.e. moisture brought by new said wind data. Then, according to a bayesian formula, a posterior probability is a prior probability multiplied by an event probability divided by an evidence probability, and therefore, in the technical scheme of the application, a bayesian probability model is used to fuse the multi-scale power feature vector, the multi-scale humidity feature vector and the multi-scale wind feature vector to obtain the posterior probability feature vector, wherein the multi-scale power feature vector is the prior probability vector, the multi-scale humidity feature vector is the event probability vector, and the multi-scale wind feature vector is the evidence probability vector. In this way, the moisture dynamic change characteristics brought by the wind data can be introduced into the dehumidifier power control model, so that the power control of the dehumidifier is prospective and predictive, and the humidity of the offshore wind turbine in the future period can be kept within a preset reasonable range. More specifically, in step S180, the feature values of the respective positions in the posterior probability feature vector are corrected based on the reciprocal of the mean of the feature values of all the positions in the posterior probability feature vector to obtain a corrected posterior probability feature vector. It should be understood that, in the technical solution of the present application, after obtaining the posterior probability feature vector, the posterior probability feature vector is passed through a classifier to obtain a classification result indicating that the power value of the dehumidifier at the current time point should be increased, maintained, or decreased. However, the fusion of the multi-scale power feature vector, the multi-scale humidity feature vector and the multi-scale wind feature vector using the bayesian probability model is a calculation by location, that is, each location of the posterior probability feature vector corresponds to the posterior probability under different time sequence features, and has a certain phase attribute. However, the classification of the classifier is usually location-independent, which results in possible bias of the classification results. Therefore, in the technical solution of the present application, the vectors for phase sensing are further aggregated by location for the posterior probability feature vectors.

More specifically, in step S190, the corrected a posteriori probability feature vector is passed through a classifier to obtain a classification result, which is used to indicate that the power value of the dehumidifier at the current time point should be increased, maintained or decreased.

In summary, the remote intelligent dehumidification control method for the offshore wind turbine based on the embodiment of the application is clarified, and the method adopts an artificial intelligent control technology, utilizes a multi-scale neighborhood feature extraction module to dynamically extract implicit features of humidity values, wind data and working power of a dehumidifier at multiple time points, and utilizes bayes to fuse data features to dynamically regulate and control the power of the dehumidifier in real time. In this way, the dynamic change characteristics of moisture brought by wind data are introduced into the dehumidifier power control model, so that the power control of the dehumidifier has foresight and predictability, and the humidity of the offshore wind turbine in a future time period can be kept within a preset reasonable range.

The foregoing describes the general principles of the present application in conjunction with specific embodiments, however, it is noted that the advantages, effects, etc. mentioned in the present application are merely examples and are not limiting, and they should not be considered essential to the various embodiments of the present application. Furthermore, the foregoing disclosure of specific details is provided for purposes of illustration and understanding only, and is not intended to limit the application to the details which are set forth in order to provide a thorough understanding of the present application.

The block diagrams of devices, apparatuses, systems referred to in this application are only given as illustrative examples and are not intended to require or imply that the connections, arrangements, configurations, etc. must be made in the manner shown in the block diagrams. These devices, apparatuses, devices, systems may be connected, arranged, configured in any manner, as will be appreciated by those skilled in the art. Words such as "including," "comprising," "having," and the like are open-ended words that mean "including, but not limited to," and are used interchangeably therewith. The words "or" and "as used herein mean, and are used interchangeably with, the word" and/or, "unless the context clearly dictates otherwise. The word "such as" is used herein to mean, and is used interchangeably with, the phrase "such as but not limited to".

It should also be noted that in the devices, apparatuses, and methods of the present application, each component or step can be decomposed and/or re-combined. These decompositions and/or recombinations are to be considered as equivalents of the present application.

The previous description of the disclosed aspects is provided to enable any person skilled in the art to make or use the present application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other aspects without departing from the scope of the application. Thus, the present application is not intended to be limited to the aspects shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

The foregoing description has been presented for purposes of illustration and description. Furthermore, the description is not intended to limit embodiments of the application to the form disclosed herein. While a number of example aspects and embodiments have been discussed above, those of skill in the art will recognize certain variations, modifications, alterations, additions and sub-combinations thereof.

Claims

1. The utility model provides a remote intelligent dehumidification control system of offshore wind turbine which characterized in that includes:

the wind data multi-scale neighborhood coding module is used for arranging the plurality of wind characteristic vectors corresponding to the preset time points into one-dimensional characteristic vectors and then obtaining the multi-scale wind characteristic vectors through the first multi-scale neighborhood characteristic extraction module;

2. The remote intelligent dehumidification control system of an offshore wind turbine of claim 1, wherein the single point wind data encoding module is further configured to: arranging wind data of each preset time point in the wind data of the preset time points into an input vector according to a time dimension; and performing full-connection coding on the input vector by using the full-connection layer according to the following formula to extract high-dimensional implicit features of feature values of all positions in the input vector so as to obtain a plurality of wind feature vectors corresponding to all preset time points, wherein the formula is as follows:

representing a matrix multiplication.

3. The remote intelligent dehumidification control system of an offshore wind turbine of claim 2, wherein the wind data multi-scale neighborhood coding module comprises:

a first convolution unit, configured to input the one-dimensional feature vector into a first convolution layer of the first multi-scale neighborhood feature extraction module to obtain a first neighborhood scale wind feature vector, where the first convolution layer has a first one-dimensional convolution kernel of a first length;

a second convolution unit, configured to input the one-dimensional feature vector into a second convolution layer of the first multi-scale neighborhood feature extraction module to obtain a second neighborhood scale wind feature vector, where the second convolution layer has a second one-dimensional convolution kernel with a second length, and the first length is different from the second length; and

the cascade unit is used for cascading the first neighborhood scale wind feature vector and the second neighborhood scale wind feature vector to obtain the multi-scale wind feature vector.

4. The remote intelligent dehumidification control system of an offshore wind turbine of claim 3, wherein the first volume unit is further configured to; performing one-dimensional convolution coding on the one-dimensional feature vector by using a first convolution layer of the first multi-scale neighborhood feature extraction module according to the following formula to obtain a first neighborhood scale wind feature vector;

wherein the formula is:

5. The remote intelligent dehumidification control system of an offshore wind turbine of claim 4, wherein the second convolution unit is further configured to: performing one-dimensional convolution coding on the one-dimensional feature vector by using a second convolution layer of the first multi-scale neighborhood feature extraction module according to the following formula to obtain a second neighborhood scale wind feature vector;

wherein the formula is:

6. The remote intelligent dehumidification control system of an offshore wind turbine of claim 5, wherein the Bayesian inference module is further configured to: fusing the multi-scale power feature vector, the multi-scale humidity feature vector and the multi-scale wind feature vector by using a Bayesian probability model to obtain the posterior probability feature vector;

wherein the formula is:

qi＝pi*ai/bi

7. The remote intelligent dehumidification control system of an offshore wind turbine of claim 6, wherein the posterior distribution correction module is further configured to: based on the reciprocal of the mean value of the feature values of all the positions in the posterior probability feature vector, correcting the feature values of all the positions in the posterior probability feature vector by the following formula to obtain the corrected posterior probability feature vector;

wherein the formula is:

wherein V represents the posterior probability feature vector,

representing the posterior probability characteristicThe inverse of the mean of the feature values of all the positions in the eigenvector, indicates a dot-by-position multiplication.

8. The remote intelligent dehumidification control system of an offshore wind turbine of claim 7, wherein the control result generation module is further configured to: processing the corrected posterior probability feature vector using the classifier to obtain the classification result with the following formula: softmax { (W) _n ,B _n ):…:(W ₁ ,B ₁ ) I X }, wherein W ₁ To W _n As a weight matrix, B ₁ To B _n Is a bias vector, and X is the corrected posterior probability feature vector.

9. A remote intelligent dehumidification control method of an offshore wind turbine is characterized by comprising the following steps:

acquiring humidity values of a plurality of preset time points in a preset time period collected by a hygrometer and wind data of the plurality of preset time points collected by a wind speed station, wherein the wind data comprises a wind speed value and a wind direction, and the wind direction is represented by a sine value or a cosine value of an angle;

the working power of the dehumidifier at the plurality of preset time points passes through a third multi-scale neighborhood feature extraction module to obtain a multi-scale power feature vector;

10. The remote intelligent dehumidification control method of an offshore wind turbine as recited in claim 9, wherein the passing wind data at each of the plurality of predetermined time points through a full connection layer to obtain a plurality of wind eigenvectors corresponding to each of the predetermined time points comprises:

arranging wind data of each preset time point in the wind data of the preset time points into an input vector according to a time dimension; and

performing full-connection coding on the input vector by using the full-connection layer according to the following formula so as to extract high-dimensional implicit features of feature values of all positions in the input vector to obtain the plurality of wind feature vectors corresponding to all preset time points, wherein the formula is as follows:

representing a matrix multiplication.