Offshore wind speed real-time quality control method based on neural network, medium and system

Abstract

The invention provides an offshore wind speed real-time quality control method based on a neural network, a medium and a system, and belongs to the technical field of offshore wind speed quality control. The offshore wind speed real-time quality control method based on the neural network, the medium and the system comprise the steps of performing range detection on offshore wind speed data, identifying an abnormal peak value through peak detection, and determining the offshore wind speed according to the abnormal peak value. The method comprises the following steps: establishing a wind speed and air pressure multi-element correlation model, executing adjacent station detection, establishing a quality control character statistical model, setting buoy stations as nodes in a space-time diagram, and constructing a nine-layer space-time diagram convolutional neural network through natural connection edges and time connection edges; constructing a time coding layer comprising an adaptive delay conversion module and an autocorrelation time convolution module for modeling propagation delay and periodic characteristics of the wind speed data, and outputting quality control early warning information of the wind speed data and sensor fault alarm information; according to the invention, the weather change and the equipment fault can be distinguished, and the application universality of the system is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of offshore wind speed quality control, and in particular, relates to a neural network-based offshore wind speed real-time quality control method, medium and system. Background Art

[0002] Offshore wind speed monitoring is of great significance to marine meteorological forecasting, offshore wind energy development, offshore engineering construction and shipping safety. With the continuous expansion of the buoy observation network, data quality control has become a key link in ensuring the reliability of observation data. At present, offshore wind speed data quality control mainly uses statistical methods and empirical threshold judgments, including single-station extreme value tests, neighbor point consistency tests, and time continuity tests. These traditional methods perform well when processing wind speed data under normal meteorological conditions, but they have obvious shortcomings when faced with the influence of complex weather systems.

[0003] In the prior art, single-station extreme value tests mainly rely on thresholds obtained from historical data statistics for judgment. This method is too mechanical in judging data and cannot distinguish between sudden changes in wind speed caused by weather processes and abnormal data caused by equipment failures. For example, during a cold wave, the wind speed may increase significantly in a short period of time and exceed the statistical threshold, but this is a normal weather phenomenon and should not be judged as abnormal data. In the case of equipment failure, the measured wind speed value may be within the threshold range, but it cannot reflect the actual meteorological conditions.

[0004] The neighbor consistency test method judges the data quality by comparing the observation values ​​of adjacent stations. However, due to the significant non-uniformity of the spatial distribution of marine meteorological elements, the difference in wind speed between adjacent stations may be due to the influence of local weather systems. Existing test methods are difficult to accurately distinguish between such natural variations and observation anomalies, which can easily lead to misjudgment. Especially under the influence of weather systems such as typhoons and local convection, the reliability of neighbor consistency test is further reduced.

[0005] The temporal continuity test focuses on the temporal variation characteristics of wind speed at a single station. Traditional methods usually use a fixed rate of change threshold, which is too simplistic and cannot adapt to the diversity of wind speed changes under different weather systems. For example, when a frontal system passes, the wind speed may change rapidly, while under calm and stable weather conditions, the wind speed changes relatively slowly. The fixed threshold method is difficult to take into account both situations, often resulting in excessive rejection or omission of abnormal data.

[0006] In addition, existing technologies generally have the problem of single criterion and lack of coordinated analysis of multiple factors. Wind speed changes are often closely related to other meteorological factors such as air pressure and temperature. Relying solely on wind speed data for quality control ignores the physical connection between meteorological factors and reduces the scientificity and accuracy of quality control. For example, during severe convective weather, a sudden increase in wind speed is usually accompanied by a sharp change in air pressure. This synergistic relationship can serve as an important basis for judging the reliability of data. Summary of the invention

[0007] In view of this, the present invention provides a real-time quality control method, medium and system for offshore wind speed based on neural network, which can accurately distinguish between sudden weather changes and equipment failures, and adapt to wind speed changes under different weather systems.

[0008] The present invention is achieved in that: The first aspect of the present invention provides a real-time quality control method for offshore wind speed based on a neural network, a real-time quality control method for offshore wind speed based on a neural network, which includes the following steps: performing range detection on the input offshore wind speed data, identifying abnormal peaks through peak detection, establishing a multi-factor correlation model between wind speed and air pressure, performing adjacent station detection, establishing a quality control symbol statistical model, setting the buoy station as a node in the space-time graph and constructing a nine-layer space-time graph convolutional neural network through natural connection edges and time connection edges for extracting spatial correlation between stations and capturing time series change characteristics, designing an adaptive graph convolution module, constructing a time coding layer including an adaptive delay conversion module and an autocorrelation time convolution module for modeling the propagation delay and periodic characteristics of wind speed data, and outputting quality control warning information of wind speed data and sensor fault alarm information.

[0009] Furthermore, the range detection includes extreme range detection, empirical range detection and equipment range detection. The quality level is determined by calculating the pass rate of wind speed data, wherein the extreme range detection threshold is 40 meters per second, the empirical range detection sets different thresholds according to the season, the equipment range detection range is 0 to 60 meters per second, and when the pass rate is greater than 70%, the quality control symbol is set to 1, and when the pass rate is less than 70%, the quality control symbol is set to 3.

[0010] Furthermore, the peak detection calculates the average value of the wind speed data at adjacent time points before and after by setting a sliding time window as the peak reference value. When the difference between the current wind speed value and the peak reference value exceeds 3 times the standard deviation, the quality control symbol is set to 3, where the standard deviation is obtained by calculating the degree of discreteness of the wind speed data in the previous 12 hours.

[0011] Furthermore, the wind speed and air pressure multi-factor correlation model calculates the synchronous observation values ​​of wind speed data and air pressure data, and sets the quality control symbol to 3 when the wind speed change exceeds 2.5 times the wind speed standard deviation and the air pressure change is less than 2.5 times the air pressure standard deviation.

[0012] Furthermore, the adjacent station detection selects a buoy station within 200 kilometers as a reference station. When the adjacent moment wind speed change of the station to be detected exceeds 2 times of the first standard deviation and the adjacent moment wind speed change of the reference station is less than 2 times of the second standard deviation, the quality control symbol is set to 3.

[0013] Furthermore, the space-time graph convolutional neural network sets the buoy stations as nodes in the space-time graph, represents the physical connection relationship between adjacent stations through natural connection edges, represents the connection relationship of the same station at different time steps through time connection edges, and constructs a 9-layer space-time graph convolutional structure.

[0014] Furthermore, the adaptive graph convolution module introduces a learnable correlation matrix by constructing an initial adjacency matrix to represent the connection relationship between stations. The initial adjacency matrix is ​​used for the first 10 iterations in the early stage of training, and then the influence of the correlation matrix is ​​gradually introduced.

[0015] Furthermore, the autocorrelation time coding layer includes an adaptive delay conversion module and an autocorrelation time convolution module, which slices the historical data through a 24-hour sliding window, uses a dynamic time warping algorithm to perform sequence alignment and clustering, and converts the signal to the frequency domain through Fourier transform to calculate the autocorrelation features on different time scales.

[0016] A second aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores program instructions, and when the program instructions are run in a computer, they are used to execute the above-mentioned real-time quality control method for offshore wind speed based on a neural network.

[0017] The third aspect of the present invention provides a real-time offshore wind speed quality control system based on a neural network, which includes the above-mentioned computer-readable storage medium, the system is any one of a computer, a server, and a single-chip microcomputer, the computer-readable storage medium is arranged in the system, and the system is provided with a microprocessor for executing program instructions stored in the computer-readable storage medium.

[0018] The present invention effectively improves the accuracy of offshore wind speed data quality control by constructing a multi-level quality control system. First, an adaptive threshold system is used to dynamically adjust the judgment criteria according to seasonal characteristics and weather background, making quality control more in line with meteorological laws. During severe weather such as cold waves and typhoons, the system can automatically adjust the wind speed anomaly criterion according to changes in the air pressure field to avoid misjudging normal weather processes as data anomalies.

[0019] The introduction of a multi-factor collaborative analysis mechanism and the establishment of a correlation model between wind speed and air pressure changes provide a more reliable physical basis for the judgment of abnormal data. By analyzing the spatiotemporal consistency of wind speed changes and air pressure changes, the system can accurately identify wind speed changes caused by weather systems and distinguish them from abnormal data caused by equipment failures. This judgment method based on physical processes significantly improves the scientific nature of quality control.

[0020] The spatiotemporal graph convolutional neural network model designed by the present invention can adaptively learn the spatial correlation between sites, overcoming the limitation of the fixed distance threshold in the traditional neighbor consistency test. The model can accurately identify the influence range of the local weather system by capturing the dynamic correlation characteristics between sites, thus improving the accuracy of the spatial consistency test. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 is a flow chart of the method of the present invention; Figure 2 The average daily wind speed curve in Example 3; Figure 3 It is a scatter plot of the average wind speed for each month from 2021 to 2023 in Example 3, where a is the effective value range corresponding to different extreme values ​​in spring, b is the effective value range corresponding to different extreme values ​​in summer, c is the effective value range corresponding to different extreme values ​​in autumn, and d is the effective value range corresponding to different extreme values ​​in winter; Figure 4 The moon wind rose diagram in Example 3; Figure 5 It is the space-time diagram of adjacent stations in Example 4, where a is the space-time diagram of the station and b is the schematic diagram of the station; Figure 6 This is a diagram of the algorithm network structure in Example 4; Figure 7 This is a structural diagram of a dynamic delay conversion module in Example 4; Figure 8 This is the structure diagram of the autocorrelation time convolution module in Example 4 Fig. 9 This is the range inspection process in Example 2; Fig.10 This is the real-time quality control workflow of offshore wind speed in Example 2. DETAILED DESCRIPTION

[0022] In order to make the purpose, technical solution and advantages of the embodiments of the present invention more clear, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0023] like Figure 1 , Fig.10 FIG. 1 is a flowchart of a method for real-time quality control of offshore wind speed based on a neural network provided by the present invention. The method comprises the following steps: S10, performing range detection on the input offshore wind speed data, calculating the wind speed data pass rate according to the extreme value range, the experience range and the equipment range, and determining the wind speed data quality level based on the pass rate; S20, performing peak detection on the wind speed data, and identifying single-value peak data in the wind speed data by calculating the standard deviation of the wind speed data at adjacent moments; S30, establishing a multi-factor correlation model of wind speed and air pressure, calculating the absolute value of the change of wind speed data and air pressure data at adjacent moments, and determining whether the wind speed data is affected by the weather process; S40, performing adjacent station detection, identifying wind speed anomalies caused by gusts or human factors by comparing the standard deviation of the first wind speed sensor data with the standard deviation of the second wind speed sensor data; S50, establishing a quality control symbol statistical model based on the range detection, the peak detection, the multi-factor correlation model and the adjacent station detection results, and generating a quality control mark for wind speed data; S60, constructing a space-time graph convolutional neural network, setting the buoy station as a node in the space-time graph, and establishing a multi-layer space-time graph convolution structure based on natural connection edges and time connection edges; S70, design an adaptive graph convolution module to obtain long-range spatial dependencies between sites by learning the adjacency matrix; S80, constructing an autocorrelation time coding layer, including an adaptive delay conversion module and an autocorrelation time convolution module, for modeling the propagation delay and periodic characteristics of the wind speed data; S90, inputting the quality control identifier into the spatiotemporal graph convolutional neural network, and outputting quality control warning information of wind speed data and sensor failure alarm information.

[0024] The specific implementation methods of the above steps are described in detail below: The specific implementation method of step S10 is to first perform multi-level range detection on the received raw data of offshore wind speed, including extreme value range detection, experience range detection and equipment range detection. In the extreme value range detection, based on meteorological knowledge and long-term observation experience, the wind speed data is compared with the preset threshold value, and when the wind speed value is greater than 40 meters per second or less than 0 meters per second, it is determined to be abnormal data. In the experience range detection, by analyzing the statistical characteristics of historical data, a seasonal experience threshold model is established, and different wind speed thresholds are set for different seasons, such as setting the threshold value to 25 meters per second in spring, 30 meters per second in summer, 28 meters per second in autumn, and 35 meters per second in winter. In the equipment range detection, according to the factory technical specifications of the wind speed sensor, the range detection range is set to 0 to 60 meters per second. Subsequently, the three test results are weighted, wherein the extreme value range detection weight is 0.4, the experience range detection weight is 0.4, and the equipment range detection weight is 0.2, and the pass rate of the wind speed data is calculated. When the pass rate is greater than 70%, the quality control symbol is set to 1, indicating that the data is normal; when the pass rate is less than 70%, the quality control symbol is set to 3, indicating that the data is suspicious. The main purpose of this step is to quickly identify obviously abnormal wind speed data through multi-dimensional range detection, providing a basis for subsequent refined quality control.

[0025] The specific implementation method of step S20 is to realize peak detection through the sliding time window method, which is used to identify abnormal peaks in wind speed data. First, the length of the sliding time window is set to 3 time units. For the wind speed data at each time point, the average value of the wind speed data at the two adjacent time points before and after is calculated as the peak reference value. Then the difference between the wind speed value at the current moment and the peak reference value is calculated, and compared with 3 times the standard deviation. Among them, the standard deviation is obtained by calculating the discrete degree of wind speed data in the previous 12 hours. When the difference exceeds 3 times the standard deviation, the quality control symbol at that moment is set to 3, indicating that the data is suspicious. In practical applications, according to the wind speed change characteristics of different sea areas, the standard deviation multiple threshold can be dynamically adjusted, and it is generally recommended to be between 2.5 and 3.5 times. The main function of this step is to identify short-term abnormal data caused by factors such as equipment failure, sudden weather or human interference.

[0026] The specific implementation method of step S30 is to construct a multi-factor correlation model of wind speed and air pressure. First, the synchronous observation values ​​of wind speed data and air pressure data are obtained, and the changes at adjacent moments are calculated. For wind speed data, the absolute value of the difference in wind speed between the current moment and the previous moment is calculated; for air pressure data, the absolute value of the difference in air pressure at adjacent moments is also calculated. The standard deviations of the wind speed data sequence and the air pressure data sequence are calculated respectively, and recorded as the wind speed standard deviation and the air pressure standard deviation. When the change in wind speed exceeds 2.5 times the wind speed standard deviation, and the change in air pressure is less than 2.5 times the air pressure standard deviation, the quality control symbol at that moment is set to 3, indicating that the data is suspicious. This judgment method is based on meteorological principles, that is, in normal weather processes, significant changes in wind speed are usually accompanied by corresponding changes in air pressure. The purpose of this step is to improve the recognition accuracy of wind speed changes caused by weather processes through multi-factor joint analysis.

[0027] The specific implementation method of step S40 is to perform adjacent station detection. First, determine the adjacent reference station of the station to be detected, and generally select a buoy station within 200 kilometers as the reference station. Then obtain the wind speed observation data of the station to be detected and the reference station in the same time period, and the calculation time length is 12 hours. Calculate the wind speed standard deviation of the two stations respectively, and record it as the first standard deviation and the second standard deviation. When the change in wind speed at adjacent moments of the station to be detected exceeds 2 times of the first standard deviation, and the change in wind speed at adjacent moments of the reference station is less than 2 times of the second standard deviation, the quality control symbol of the station to be detected at that moment is set to 3, indicating that the data is suspicious. The main purpose of this step is to use the observation data of adjacent stations to verify each other and identify abnormal data caused by local gusts or human factors.

[0028] The specific implementation method of step S50 is to establish a quality control symbol statistical model. First, count the number of various quality control symbols at each time point, including quality control symbols generated by range detection, peak detection, multi-factor detection and adjacent station detection. Then, based on the combination of quality control symbols, generate the final quality control mark. When all tests show normal, the mark is 1; when only one test shows suspicious, the mark is 2; when two or more tests show suspicious, the mark is 3; when any test shows abnormal, the mark is 4. For data with 5 or more abnormal marks accumulated within 3 consecutive hours, it is judged as equipment failure. The main purpose of this step is to form the final data quality assessment result by comprehensively analyzing the results of various quality control tests.

[0029] The specific implementation method of step S60 is to construct a spatiotemporal graph convolutional neural network model. First, each buoy station is set as a node in the spatiotemporal graph, and the node features include wind speed value, quality control mark and other information. Then, based on the spatial relationship between the stations, two types of edge connections are established: the natural connection edge represents the physical connection relationship between adjacent stations, and the time connection edge represents the connection relationship of the same station at different time steps. On this basis, a 9-layer spatiotemporal graph convolution structure is constructed, each layer containing a graph convolution operation and a time convolution operation. The graph convolution operation is used to extract the spatial correlation between stations, and the time convolution operation is used to capture the characteristics of temporal changes. The main purpose of this step is to establish a deep learning framework that can simultaneously handle spatial relationships and temporal evolution.

[0030] The specific implementation method of step S70 is to realize the adaptive graph convolution module. First, construct the initial adjacency matrix, in which the matrix elements represent the connection relationship between the stations, and the value between the connected stations is 1, otherwise it is 0. Then introduce the learnable correlation matrix, and calculate the similarity between the stations through the Gaussian kernel function. In the early stage of training, the initial adjacency matrix is ​​used for the first 10 iterations, and then the influence of the correlation matrix is ​​gradually introduced. The initial value of the elements of the correlation matrix is ​​set to 1, and it is continuously updated by the gradient descent method during the training process. The main purpose of this step is to achieve adaptive learning of long-distance spatial dependencies between stations.

[0031] The specific implementation method of step S80 is to construct an autocorrelation time coding layer. In the adaptive delay conversion module, a sliding window with a length of 24 hours is used to slice the historical data, and a dynamic time warping algorithm is used to align and cluster the sequences to extract the typical wind speed change pattern. For the wind speed sequence at each moment, the similarity with the typical pattern is calculated through the attention mechanism, and weighted fusion is performed. In the autocorrelation time convolution module, the time feature is first extracted using a time convolution network, and then the signal is converted to the frequency domain through Fourier transform, and the autocorrelation features on different time scales are calculated. The features are converted back to the time domain through the inverse Fourier transform, and the most significant time correlation is selected for feature fusion. The main purpose of this step is to effectively model the propagation delay effect and periodic characteristics of wind speed data.

[0032] The specific implementation method of step S90 is to input the quality control mark into the trained spatiotemporal graph convolutional neural network to obtain the quality control warning information of the wind speed data and the sensor fault alarm information. For the warning information, the warning level is divided into three levels according to the degree of deviation between the predicted value output by the model and the measured value: when the predicted deviation is less than 1 meter per second, it is normal and no warning is required; when the predicted deviation is between 1 and 2 meters per second, it is a mild warning; when the predicted deviation is greater than 2 meters per second, it is a severe warning. For the fault alarm information, when a station has 5 or more abnormal marks accumulated within 3 consecutive hours, the sensor fault alarm is triggered. The main purpose of this step is to achieve real-time monitoring and early warning of wind speed data quality.

[0033] Through the implementation of the above steps, this method realizes all-round quality control of offshore wind speed data, can effectively identify and process various types of abnormal data, and ensure the reliability of wind speed data.

[0034] In the present invention, the calculation process involved is described in detail as follows: 1. The range detection pass rate calculation equation is specifically expressed as follows: ; In the formula, is the wind speed data passing rate; It is the result of the empirical range test, and takes the value of 0 or 1; It is the result of extreme value range detection, and takes the value of 0 or 1; It is the result of the equipment range test, and takes the value of 0 or 1; is the weight coefficient and satisfies ; 2. The peak detection calculation equation is specifically expressed as follows: ; ; ; In the formula, is the peak reference value; is the wind speed value at time t; is the standard deviation; is the time window length, the value is 12; is the wind speed deviation value at the current moment; 3. The calculation equation of wind speed and air pressure multi-factor correlation is specifically expressed as follows: ; ; ; ; In the formula, is the wind speed change; is the change in air pressure; is the standard deviation of wind speed variation; is the standard deviation of air pressure variation; is the time window length, the value is 24; 4. The calculation equation of the adaptive graph convolution model is specifically expressed as follows: ; ; In the formula, is the output feature; is the convolution kernel weight; is the input feature; is the adjacency matrix; is the learnable correlation matrix; is the distance between node i and node j; is the Gaussian kernel parameter; 5. The autocorrelation time coding calculation equation is specifically expressed as follows: ; ; In the formula, is the spectral density function; For time series Fourier transform of represents a conjugate complex number; is the autocorrelation function; is the time delay term; Parameter acquisition method description: 1. The weight coefficients were obtained through statistical analysis of historical data and were 0.4, 0.4, and 0.2; 2. Obtained through calculation of the latitude and longitude coordinates of the buoy station; 3. It is obtained by cross-validation method, and the value is generally taken as 1 / 3 of the minimum distance between stations; 4. It is calculated by the dynamic time warping algorithm, and its calculation steps are: Step 1: Calculate the distance matrix of two time series; Step 2: Use dynamic programming algorithm to find the optimal alignment path; Step 3: Calculate the time delay value according to the alignment path.

[0035] Explanation of the equation construction principle: 1. The range detection pass rate adopts the weighted summation form, taking into account the importance of different detection methods; 2. Peak detection uses the average value of adjacent points as a reference, which can effectively identify mutation points; 3. The standard deviation comparison method is used for the correlation of multiple factors, which reflects the synergy between wind speed and air pressure changes; 4. The adaptive graph convolution model uses a Gaussian kernel function to calculate spatial correlation, ensuring the physical property that correlation decreases with increasing distance; 5. Autocorrelation time coding uses spectrum analysis method, which can effectively capture periodic characteristics.

[0036] A second aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores program instructions, and when the program instructions are run in a computer, they are used to execute the above-mentioned real-time quality control method of offshore wind speed based on neural network.

[0037] A third aspect of the present invention provides a wind tunnel health monitoring data mining system, which includes the above-mentioned computer-readable storage medium, the system is any one of a computer, a server, and a single-chip microcomputer, the computer-readable storage medium is arranged in the system, and the system is provided with a microprocessor for executing program instructions stored in the computer-readable storage medium.

[0038] Specifically, the principle of the present invention is: based on the basic principles of meteorology and deep learning technology, the present invention constructs a complete set of offshore wind speed data quality control methods. In principle, the effectiveness of this method is mainly reflected in the following aspects: First, as a physical quantity describing the state of atmospheric motion, the change of wind speed must follow the basic laws of atmospheric motion. The present invention combines data quality control with atmospheric dynamics by establishing a multi-factor correlation model between wind speed and air pressure. When an abnormal wind speed occurs, the system will automatically check the corresponding air pressure change characteristics and judge the rationality of the data through the conservation relationship between physical quantities. This judgment method based on physical mechanisms essentially improves the scientific nature of quality control.

[0039] Secondly, the distribution of offshore meteorological elements has obvious spatial correlation, and adjacent areas are often affected by the same weather system. The spatiotemporal graph convolution structure designed by the present invention effectively describes the spatial relationship and temporal evolution characteristics between sites through natural connection edges and temporal connection edges. This network structure can automatically learn the spatial influence range of weather systems of different scales, making quality control more in line with the actual meteorological process.

[0040] Finally, the autocorrelation time coding technology used in the present invention can effectively capture the periodic characteristics and propagation delay effects of wind speed data. Through Fourier transform and dynamic time warping algorithm, the system can identify the typical pattern of wind speed change and dynamically adjust the judgment criteria according to the characteristics of different weather systems. This adaptive mechanism makes the quality control method more universal.

[0041] A specific embodiment 1 of the present invention is provided below. The specific method of each step in this embodiment 1 is described in detail as follows: The specific implementation method of step S10 is to establish a multi-level range detection process to conduct a comprehensive quality assessment of offshore wind speed data. First, based on meteorological theory and long-term observation experience, an extreme range detection threshold is established. When the wind speed measurement value is greater than 40 meters per second or less than 0 meters per second, it is determined to be abnormal data. At the same time, combined with the statistical characteristics of historical data, a seasonal empirical threshold model is constructed. The threshold is set to 25 meters per second in spring, 30 meters per second in summer, 28 meters per second in autumn, and 35 meters per second in winter. Differentiated detection is performed for the wind speed change characteristics in different seasons. According to the factory technical specifications of the wind speed sensor, the range detection range is set to 0 to 60 meters per second. The range detection pass rate is calculated using the weighted summation method, and the calculation formula is: ,in is the wind speed data passing rate, is the test result of the empirical range, is the extreme value range detection result, is the equipment range detection result, all three values ​​are 0 or 1, and the weight coefficient The values ​​are 0.4, 0.4, and 0.2 respectively. When the calculated pass rate is greater than 70%, the quality control symbol is set to 1, indicating that the data is normal; when the pass rate is less than 70%, the quality control symbol is set to 3, indicating that the data is suspicious. This step quickly identifies obviously abnormal wind speed data through multi-dimensional range detection, providing a basis for subsequent refined quality control.

[0042] The specific implementation of step S20 is to use the sliding time window method to perform peak detection, which is mainly used to identify abnormal peaks in wind speed data. First, the sliding time window length is set to 3 time units, and the peak reference value of the wind speed data at each time point is calculated. The calculation formula is: ,in is the peak reference value, is the wind speed value at time t. Then calculate the standard deviation of the wind speed data in the previous 12 hours, and the calculation formula is: ,in is the standard deviation, is the time window length. Calculate the wind speed deviation value at the current moment, and the calculation formula is: ,in is the wind speed deviation value at the current moment. Exceed When the data is detected, the quality control symbol at that moment is set to 3, indicating that the data is suspicious. This step can effectively identify short-term abnormal data caused by factors such as equipment failure, sudden weather or human interference.

[0043] The specific implementation of step S30 is to construct a multi-factor correlation model of wind speed and air pressure. First, the changes in wind speed and air pressure at adjacent moments are calculated, and the calculation formula is: and ,in is the wind speed change, is the change in air pressure. Then calculate the standard deviation of wind speed and air pressure changes, and the calculation formula is: and ,in is the standard deviation of wind speed variation, is the standard deviation of air pressure change. And the pressure change is less than When , the quality control symbol of this moment is set to 3, indicating that the data is suspicious. This step improves the recognition accuracy of wind speed changes caused by weather processes through multi-factor joint analysis.

[0044] The specific implementation method of step S40 is to perform adjacent station detection. First, determine the adjacent reference station of the station to be detected, and select the buoy station within 200 kilometers as the reference station. Calculate the standard deviation of the wind speed of the station to be detected and the reference station in the same time period, and the time length is 12 hours. When the change in wind speed at adjacent moments of the station to be detected exceeds 2 times of its standard deviation, and the change in wind speed at adjacent moments of the reference station is less than 2 times of its standard deviation, the quality control symbol of the station to be detected at that moment is set to 3, indicating that the data is suspicious. This step uses the observation data of adjacent stations to verify each other and identify abnormal data caused by local gusts or human factors.

[0045] The specific implementation method of step S50 is to establish a quality control symbol statistical model. First, count the number of various quality control symbols at each time point, including quality control symbols generated by range detection, peak detection, multi-factor detection and adjacent station detection. When all tests are normal, it is marked as 1; when only one test is suspicious, it is marked as 2; when two or more tests are suspicious, it is marked as 3; when any test is abnormal, it is marked as 4. For data with 5 or more abnormal marks within 3 consecutive hours, it is judged as equipment failure. This step forms the final data quality assessment result by comprehensively analyzing the results of various quality control.

[0046] The specific implementation of step S60 is to construct a spatiotemporal graph convolutional neural network model. Each buoy station is set as a node in the spatiotemporal graph, and the node features include wind speed value, quality control mark and other information. The calculation formula of the adaptive graph convolution model is: ,in is the output feature, is the convolution kernel weight, is the input feature, is the adjacency matrix, is the learnable correlation matrix. The spatial correlation calculation formula between sites is: ,in is the distance between node i and node j, is the Gaussian kernel parameter. This step establishes a deep learning framework that can handle both spatial relationships and temporal evolution.

[0047] The specific implementation of step S70 is to realize the adaptive graph convolution module. First, construct the initial adjacency matrix. The matrix elements represent the connection relationship between the stations. The value between the connected stations is 1, otherwise it is 0. Introduce a learnable correlation matrix, and calculate the similarity between the stations through the Gaussian kernel function. In the early stage of training, use the initial adjacency matrix for the first 10 iterations, and then gradually introduce the influence of the correlation matrix. This step realizes the adaptive learning of the long-distance spatial dependency relationship between stations.

[0048] The specific implementation of step S80 is to construct an autocorrelation time coding layer. The autocorrelation time coding calculation formula is: and ,in is the spectral density function, For time series The Fourier transform of represents the conjugate complex number, is the autocorrelation function, is the time delay term. The historical data is sliced ​​using a sliding window of 24 hours in length, and the dynamic time warping algorithm is used for sequence alignment and clustering to extract the typical wind speed change pattern. This step effectively models the propagation delay effect and periodic characteristics of wind speed data.

[0049] The specific implementation method of step S90 is to input the quality control mark into the trained spatiotemporal graph convolutional neural network to obtain the quality control warning information and sensor fault alarm information of the wind speed data. The warning level is divided into three levels: the prediction deviation is less than 1 meter per second, which is normal and no warning is required; the prediction deviation is between 1 and 2 meters per second, which is a mild warning; the prediction deviation is greater than 2 meters per second, which is a severe warning. When a station has 5 or more abnormal marks accumulated within 3 consecutive hours, the sensor fault alarm is triggered. This step realizes real-time monitoring and early warning of wind speed data quality.

[0050] In order to better understand and implement the present invention, the following provides an embodiment 2 of a specific application scenario of the present invention: the present invention has been actually applied to 15 offshore buoy stations arranged in the Yellow Sea and Bohai Sea, and the research area involves important sea areas such as Bohai Bay, Laizhou Bay, and Liaodong Bay. First, the quality control symbol classification standard used in wind speed data quality control is given as shown in Table 1: Table 1 Classification standard of quality control symbols for wind speed data quality control

[0051] The specific distribution information of the 15 buoy stations is shown in the following table: Table 2 Specific distribution information of buoy stations

[0052] The range detection determination method is shown in the following table: Table 3 Range detection determination method

[0053] in, is the wind speed at the nth moment, USER_MAX is the historical threshold calculated based on the historical experience of the buoy combined with the cumulative probability and wind direction, To set the pass rate, such as Fig. 9 As shown, The following conditions must be met: ; Among them, JZ is the extreme value range, Wdir*JY is the empirical wind direction, PJZ is the average value, and LC is the range.

[0054] The peak detection judgment conditions are shown in the following table: Table 4 Peak detection judgment conditions

[0055] Wherein, SPK_REF is the average value of adjacent data points n-2 and n as the peak reference, and SD is the standard deviation, satisfying the following conditions: ; In the above formula, TIM_DEV is the period for calculating SD. is the wind speed at any time, is the average wind speed during the period.

[0056] The multi-factor detection judgment conditions are shown in the following table: Table 5 Multi-factor detection judgment conditions

[0057] is the absolute value of the change in average wind speed at adjacent moments, is the absolute value of the air pressure change at the same moment, SD_WS is the standard deviation of wind speed, and SD_BP is the standard deviation of air pressure.

[0058] Adjacent station detection can be used for short-term and large changes in wind speed due to gusts, accidents or human factors in non-weather processes. Use wind data No. 1 (WS1) to compare with wind data No. 2 (Ws2) to calculate the standard deviation (SD1, SD2) of each device during this period (TIM_DEV).

[0059] The adjacent station detection judgment conditions are shown in the following table: Table 6 Adjacent station detection judgment conditions

[0060] in, is the wind speed of device No. 1 at the nth moment, is the wind speed of device No. 1 at the n-1th moment, is the wind speed of device No. 2 at the nth moment, is the wind speed of device No. 2 at the n-1th moment; SD1 is the standard deviation of the wind speed of device No. 1 during TIM_DEV, and SD2 is the standard deviation of the wind speed of device No. 2 during TIM_DEV.

[0061] From January 15 to 18, 2024, the Yellow Sea and Bohai Sea were affected by strong cold air, resulting in strong winds. The present invention performs quality control on the wind speed data during this period, and the specific implementation process is as follows: First, perform range detection. Considering the characteristics of winter wind speed thresholds, set empirical range thresholds for different sea areas: 35 meters per second for Bohai Bay stations (BH1-BH5), 32 meters per second for Laizhou Bay stations (LZ1-LZ5), and 38 meters per second for Liaodong Bay stations (LD1-LD5); the extreme range threshold is uniformly set to 40 meters per second, and the equipment range is 0 to 60 meters per second. The specific implementation process takes the BH1 station data as an example: The continuous observation data from 10:00 to 13:00 on January 15 are shown in the following table: Table 7 Continuous observation data of BH1 station from 10:00 to 13:00 on January 15

[0062] The pass rate calculation uses a weight system: the experience range detection weight is 0.4, the extreme range detection weight is 0.4, and the equipment range detection weight is 0.2. For example, the pass rate at 10:00 is calculated as: , since the passing rate is less than 70%, the quality control symbol is set to 3.

[0063] Then the peak detection is carried out. Take the data of BH2 station from 14:00 to 17:00 on January 15 as an example: Table 8 Continuous observation data of BH2 station from 14:00 to 17:00 on January 15

[0064] The analysis process at 16:00 is as follows: Calculate the peak reference value Meters per second, calculate the standard deviation of wind speed in the previous 12 hours Meters per second, wind speed deviation value at the current moment Meters per second, less than Meters per second, so the quality control symbol remains at 1.

[0065] When conducting multi-factor correlation detection, the synergy between wind speed changes and air pressure changes is analyzed. Take the observation data of BH3 station from 18:00 to 21:00 on January 15 as an example: Table 9 Continuous observation data of BH3 station from 18:00 to 21:00 on January 15

[0066] Calculate the standard deviation of wind speed variation within 24 hours Meters per second, standard deviation of pressure change Taking 18:00 as an example, the wind speed change is 6.7 meters per second, exceeding Meters per second, the pressure change is 3.7 hPa, less than hPa, so the quality control symbol is set to 3.

[0067] In the adjacent station detection, stations BH4 and BH3 were selected for analysis. The observation data from 20:00 to 23:00 on January 15 are as follows: Table 10 Continuous observation data of BH4 station from 20:00 to 23:00 on January 15

[0068] Taking 20:00 as an example, the wind speed change at BH4 station is Meters per second over Meters per second, and the wind speed change at BH3 station Meters per second less than Meters per second, so the quality control symbol of BH4 station is set to 3 at this moment.

[0069] In terms of quality control symbol statistics and equipment fault diagnosis, the data from BH5 station from 2:00 to 5:00 on January 16 was analyzed: Table 11 Continuous observation data of BH5 station from 2:00 to 5:00 on January 16

[0070] Since the quality control symbol was 3 for 10 consecutive times, it was determined that there was a fault in the equipment at the BH5 station, and the system automatically issued an equipment failure warning message.

[0071] The application results of the method of the present invention at 15 buoy stations are statistically shown as follows: Table 12 Statistical data of application results of 15 buoy stations

[0072] Through comparative analysis, the advantages of the present invention over traditional quality control methods are shown in the following table: Table 13 Advantages of the present invention over traditional quality control methods

[0073] The practical application of the present invention in the Yellow Sea and Bohai Sea shows that: 1. Through multi-level quality control processes, abnormal data can be effectively identified, and the quality control accuracy rate reaches 96%; 2. Using a multi-factor collaborative analysis method, we successfully distinguished between wind speed changes caused by weather processes and abnormal data caused by equipment failures; 3. The introduction of spatial correlation analysis improves the ability to identify local abnormal weather.

[0074] A specific embodiment 3 of step S10 of the present invention is provided below: The data standardization processing module continuously reads the database, automatically obtains the buoy wind data and related data, and then standardizes the data to ensure that the data from different manufacturers have a unified format for subsequent operations. The data will not be modified during the data quality control stage, and the quality control symbol records will be marked. The alarm processing module triggers different alarm prompts according to the quality control symbol records, including data abnormality alarms and equipment failure alarms.

[0075] In the visualization stage, different colors are set for different types of data, and data graphs are drawn in combination with the quality control symbol record table to intuitively display abnormal data, helping auditors to quickly grasp the overall situation and trend of the data. Auditors comprehensively analyze various situations and complete data audit work.

[0076] Data standardization processing module: Use advanced data acquisition tools to obtain real-time data streams from multiple different data sources, such as sensors from different manufacturers, and map them into a unified data structure framework. Use data cleaning technology to remove noise, redundant information, and incomplete data to ensure data consistency and comparability, providing a standardized basis for subsequent data analysis.

[0077] In this step, the time series analysis method and geographic information system technology are combined to identify and process the timestamp and spatial location information in the data stream, and record both the solar calendar and the lunar calendar dates. Combined with subsequent detection methods, the combination of data streams is dynamically adjusted to generate high-quality target real-time data streams and improve quality control efficiency.

[0078] Threshold determination method based on cumulative frequency method: The selection of threshold is crucial, which directly affects the accuracy of range detection. The wind threshold selection method proposed in this patent takes into account the influence of multiple factors, including seasonal changes and extreme weather events. The threshold obtained by this method can represent the normal wind speed in most cases. While excluding the influence of extreme weather, the obtained threshold will also give the main wind direction and probability of occurrence of extreme wind, making range detection more accurate.

[0079] Wind characteristic analysis: By analyzing the wind data, the characteristic distribution of the wind in the area is found. In order to better understand the seasonal changes of the ocean, the present invention adopts the ocean seasons. Taking station A as an example, the wind speed data is analyzed by daily average to find the characteristic changes of the wind speed at the station. Figure 2 As shown in the figure, the average daily wind speed is concentrated in the range of 2-10m / s. The average annual wind speed from 2021 to 2023 is 6.40m / s, and the wind speed distribution trend is relatively consistent. The wind speed gradually decreases in late winter and early spring, and drops to the lowest level of the year in summer. The wind speed gradually increases in autumn, and increases to the highest level of the year at the turn of autumn and winter. Therefore, the change of wind speed at station A has obvious seasonal characteristics, and a threshold can be set for each season.

[0080] Selection of threshold reference value: First, the maximum wind data of each month is counted, and combined with the results of scatter plot and seasonal analysis, a threshold that is neither too high nor too low is selected as the threshold reference value for each season.

[0081] The threshold reference value is obtained by using the theoretical analysis method. Then the cumulative frequency method can be used to verify the rationality of the reference value. First, set a range detection screening rate P, which means that the probability that the data cannot pass the range detection is 1-P. If the cumulative frequency corresponding to the reference value If it is basically consistent with 1-P, then this reference value is reasonable. If the two are significantly different, you need to change the reference value again and verify again to obtain the most appropriate threshold. The accumulation frequency can be set according to your needs.

[0082] ; in, is the wind speed threshold in season X, for The cumulative probability in season x over the years.

[0083] Next, the minimum value of the seasonal threshold is set to Hseaso-min, and the wind rose diagram is used to count the wind direction when H>Hseaso-min, to determine the main wind direction that exceeds the empirical range detection, and the probability of the main wind direction is set as the weight W of the empirical range detection. dir The process of threshold selection is demonstrated by taking the MF01002 station as an example.

[0084] (1) Reference value selection. As shown in Table 1 and Figure 3 As shown, the spring threshold is set to 17 m / s, summer 14 m / s, autumn 20 m / s, and winter 16 m / s.

[0085] Table 14 Statistics of monthly maximum average wind speed from 2021 to 2023

[0086] (2) Calculate the cumulative probability. Set P = 97%, then 1-P = 3%, that is, Px is more appropriate around 3%. The winter threshold is appropriate, and the rest need to be adjusted. Repeat this process until a suitable threshold is obtained.

[0087] Table 15 Seasonal threshold distribution probability statistics in the past three years

[0088] Assume Hseaso-min = 16, such as Figure 4 As shown in the figure, the wind rose diagram is used to calculate the wind direction when H>16. The main wind direction is northeast wind, and the probability of occurrence is =97%.

[0089] Embodiment 4: A specific embodiment 4 of steps S70 and S80 in the present invention is provided below. In view of the above problem, the present invention constructs a new short-term and medium-term wind speed prediction model, which can automatically capture the patterns embedded in the spatial structure of the observation station and their temporal dynamics, and predict the wind speed values ​​of multiple time steps in the future. Figure 5 As shown in Figure 1, when constructing the spatiotemporal graph, the buoy stations are regarded as nodes of the graph. There are two types of edges: natural edges, which are natural connections between nodes (orange); and time edges (blue), which connect the same nodes across consecutive time steps. Figure 5 As shown in Figure 2, a multi-layer spatiotemporal graph convolution is constructed on this basis to achieve the integration of information along the graph dimension and the time dimension.

[0090] A North Sea buoy dataset was constructed, on which the model was trained and tested, and the mean absolute error and mean square error were used to evaluate the model, providing a comparison curve between the predicted and true values. In addition, since the learnable adjacency matrix is ​​the only factor that determines the relationship between cities, the content learned by the network is visualized and some visual explanations are provided. The contributions of this invention are as follows: (1) A wind speed prediction model based on adaptive graph convolution and autocorrelation time coding is proposed to accurately predict wind speed in the North Sea, solving the problems of regionality, periodicity, long distance and time delay of complex wind data mentioned above; (2) An adaptive graph convolution module is proposed to capture the spatial dependencies between sites over long distances. An autocorrelation time encoding layer is designed, which includes two core components: an adaptive delay transformation module that can dynamically model the time delay during data propagation, and an autocorrelation time convolution module that is used to learn the inherent periodicity in the wind data series. (3) The method proposed in the present invention was experimented on a real data set collected. The experimental results showed that the mean absolute error (MAE) of the prediction for the next hour was 0.79, the mean square error (MSE) was 1.08, and the Pearson coefficient between the predicted value and the actual value was 96.5%.

[0091] like Figure 6 As shown in the figure, the network is composed of 9 AST-Blocks, each AST-Block includes an adaptive spatiotemporal graph convolution and an autocorrelation time coding layer. The adaptive graph convolution is performed before the autocorrelation time coding layer. A residual connection is added to the entire block. Each block contains its own, separate, learnable adjacency matrix in the spatial convolution.

[0092] The consistency of adjacent stations plays an important role in analyzing factors with poor internal consistency such as precipitation and wind. Researchers often use graph convolution to obtain this correlation. However, in reality, the distance between any two stations is hundreds of nautical miles, which means that the spatial dependency is long-distance. In addition, the GCN-based model has the problem of over-smoothing and is difficult to capture long-distance spatial dependencies. This paper proposes an adaptive graph convolution module to obtain long-distance spatial dependencies between stations.

[0093] ; ; in, It is the association matrix, indicating whether two sites are connected. is a learnable matrix that learns a unique graph for each sample and uses the classic Gaussian function to capture the similarity between sites. In formula 2, Indicates a fixed connection between sites. Represents the first 10 iterations ; In the subsequent iterations, A=P, the elements of P are initialized to 1, and the K-shape function is used to calculate the correlation between nodes, which is continuously updated during the training process. This use of prior knowledge makes it easier to train in the early stages, and the adaptive graph structure brings more flexibility.

[0094] The autocorrelation temporal encoding layer models the dynamic temporal correlation based on the self-attention mechanism and includes two core components: an adaptive delay transformation module that can dynamically model the time delay during data propagation, and an autocorrelation temporal convolution module that is used to learn the inherent periodicity in the wind data sequence.

[0095] The propagation of wind speed is affected by distance. At sea, the sites are sparsely arranged and far apart. It takes several minutes for the wind speed change at a site to affect the adjacent sites. The present invention proposes a dynamic delay feature conversion module, the goal of which is to optimize the network's perception capability by reducing the impact of delay.

[0096] like Figure 7 As shown in the figure, firstly, the historical data of the target station is sliced ​​using a sliding window of size S to obtain N wind speed sequences recorded as Next, we use the dynamic time warping (DTW) function to calculate each Perform clustering operation to obtain cluster centers Short-term wind speed patterns Then the historical data sequence of each site is compared with the extracted P, and similar pattern information is integrated into the sequence representation of each node. Specifically, for the sequence of site B at time t , first use 1X1Conv to get High dimensional representation of , and then use the weight matrix Convert each data sequence in the short-term wind speed pattern P into a memory vector ; Next, use the softmax function to get the similarity vector , and then use weighted summation to get the final representation of the historical sequence at that moment The weight matrix , It is randomly initialized first and then continuously updated during the training process.

[0097] Wind speed has the characteristics of daily periodicity and seasonality, which provides usable regular information for wind speed forecasting. The researchers proposed an autocorrelation time convolution module to obtain this characteristic.

[0098] like Figure 8 As shown, the input vector first passes through TCN to obtain the time dimension information, and then undergoes linear transformation to obtain Q, K, V. Q and K perform Fourier transform (FFT) to obtain the corresponding frequency domain signal Q FFT , K FFT , for K FFT Take the conjugate and add it to Q FFTElement-by-element multiplication is performed to obtain QK, the purpose of which is to combine the frequency domain information of Q and K. An inverse Fourier transform is performed on QK, which produces a time domain result that fuses the frequency domain information of Q and K. Next, the TopK important moments are selected from the results of the inverse Fourier transform to retain important information. The aggregated result passes through a linear layer and is added to the original V vector to supplement the underlying detail information. In the time domain, a SoftMax operation is performed on the results after different time delays to calculate the weights, and then these weights are used for weighted summation and finally fused to obtain the final result. The overall process realizes the correlation calculation of Q and K in the frequency domain through the operations of Fourier transform and inverse Fourier transform, and then combines the information of V through time delay aggregation and linear transformation to finally obtain the fused output result. This autocorrelation mechanism can more effectively capture the periodic characteristics of time series data.

[0099] ; ; It should be noted that the variables involved in the present invention are explained in detail as shown in Table 16 below.

[0100] Table 16 Variable explanation table

[0101] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A real-time quality control method for offshore wind speed based on neural network, characterized in that: The following steps are involved: The input offshore wind speed data is range-detected, abnormal peaks are identified through spike detection, a multi-factor correlation model of wind speed and air pressure is established, adjacent station detection is performed, a quality control symbol statistical model is established, the buoy stations are set as nodes in the space-time graph, and a nine-layer space-time graph convolutional neural network is constructed through natural connection edges and time connection edges to extract the spatial correlation between stations and capture the timing change characteristics. An adaptive graph convolution module is designed, and a time encoding layer including an adaptive delay conversion module and an autocorrelation time convolution module is constructed to model the propagation delay and periodic characteristics of wind speed data, and output quality control warning information of wind speed data and sensor failure alarm information.

2. The offshore wind speed real-time quality control method according to claim 1, characterized in that: The range detection includes extreme range detection, empirical range detection and equipment range detection. The quality level is determined by calculating the pass rate of wind speed data, wherein the extreme range detection threshold is 40 meters per second, the empirical range detection sets different thresholds according to the season, the equipment range detection range is 0 to 60 meters per second, and when the pass rate is greater than 70%, the quality control symbol is set to 1, and when the pass rate is less than 70%, the quality control symbol is set to 3.

3. The method for real-time quality control of offshore wind speed according to claim 2, characterized in that: The peak detection is performed by setting a sliding time window and calculating the average value of the wind speed data at adjacent time points as the peak reference value. When the difference between the current wind speed value and the peak reference value exceeds 3 times the standard deviation, the quality control symbol is set to 3, where the standard deviation is obtained by calculating the degree of dispersion of the wind speed data in the previous 12 hours.

4. The method for real-time quality control of offshore wind speed according to claim 3, characterized in that: The wind speed and air pressure multi-factor correlation model calculates the synchronous observation values ​​of wind speed data and air pressure data, and sets the quality control symbol to 3 when the wind speed change exceeds 2.5 times the wind speed standard deviation and the air pressure change is less than 2.5 times the air pressure standard deviation.

5. The method for real-time quality control of offshore wind speed according to claim 4, characterized in that: The adjacent station detection selects a buoy station within 200 kilometers as a reference station. When the adjacent moment wind speed change of the station to be detected exceeds 2 times of the first standard deviation and the adjacent moment wind speed change of the reference station is less than 2 times of the second standard deviation, the quality control symbol is set to 3.

6. The method for real-time quality control of offshore wind speed according to claim 5, characterized in that: The space-time graph convolutional neural network sets the buoy stations as nodes in the space-time graph, represents the physical connection relationship between adjacent stations through natural connection edges, represents the connection relationship of the same station at different time steps through time connection edges, and constructs a 9-layer space-time graph convolutional structure.

7. The method for real-time quality control of offshore wind speed according to claim 6, characterized in that: The adaptive graph convolution module constructs an initial adjacency matrix to represent the connection relationship between stations, introduces a learnable correlation matrix, uses the initial adjacency matrix for the first 10 iterations in the early stage of training, and then gradually introduces the influence of the correlation matrix.

8. The method for real-time quality control of offshore wind speed according to claim 7, characterized in that: The time coding layer of the autocorrelation time convolution module includes an adaptive delay conversion module and an autocorrelation time convolution module. It slices the historical data through a 24-hour sliding window, uses a dynamic time warping algorithm for sequence alignment and clustering, and converts the signal to the frequency domain through Fourier transform to calculate the autocorrelation features on different time scales.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores program instructions, and when the program instructions are executed in a computer, they are used to execute the real-time quality control method for offshore wind speed based on a neural network as described in any one of claims 1 to 8.

10. A real-time quality control system for offshore wind speed based on neural network, characterized in that: The system comprises the computer-readable storage medium as claimed in claim 9, wherein the system is any one of a computer, a server, and a single-chip microcomputer, the computer-readable storage medium is arranged in the system, and the system is provided with a microprocessor for executing program instructions stored in the computer-readable storage medium.

Images