Floating type wind power mooring tension monitoring method and system based on multi-source data fusion

Abstract

The invention belongs to the technical field of offshore wind power. According to the floating type wind power mooring tension monitoring method and system based on multi-source data fusion, feature extraction is carried out on preprocessed mooring cable body dynamic data, floating body platform global motion data and environment context data; dynamic features of the mooring cable body, global motion features of the floating body platform and environmental context features are obtained; performing cross-modal feature fusion on the dynamic features of the mooring cable body, the global motion features of the floating body platform and the environmental context features to obtain fusion features; and a time convolution network is adopted to extract time sequence features in the fusion features, a global feature vector is obtained according to the time sequence features, and a mooring tension prediction result is obtained according to the global feature vector. According to the method, static assumption and data dimension limitation of an existing method are overcome, and accurate and reliable monitoring of the mooring tension of the catenary of the semi-submersible platform of the floating type offshore wind turbine generator is achieved.

Description

Technical Field

[0001] This invention relates to the field of offshore wind power technology, specifically to a method and system for monitoring the mooring tension of floating wind turbines based on multi-source data fusion. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] The mooring system is the core component ensuring the stability and safety of floating offshore wind turbines, and real-time, accurate monitoring of mooring tension is crucial for assessing their safe operation. Currently, most floating offshore wind turbines in operation employ a technical approach that combines semi-submersible platforms with catenary mooring systems.

[0004] The mainstream methods for mooring tension monitoring are divided into two categories: direct measurement and indirect measurement. Direct measurement is achieved by deploying tension sensors at the top of the mooring cable. However, these sensors are exposed to harsh marine environments for extended periods, resulting in issues such as susceptibility to damage, low reliability, high maintenance costs, and short service life. In extreme conditions, they may even fail, failing to provide early warning of mooring breakage and posing a serious threat to the overall safety of the floating wind turbine. Indirect measurement, through numerical calculation and inversion of mooring tension, can effectively improve the reliability of real-time mooring tension monitoring. Existing research has yielded several mooring tension inversion methods, but they all generally have certain limitations. Traditional mooring tension calculation methods are mainly based on catenary theory, establishing an analytical relationship between mooring cable attitude and tension under quasi-static assumptions. Because they completely ignore the inertial forces and hydrodynamic effects generated by the mooring cable under complex marine environmental loads, they fail to accurately reflect the high-frequency, nonlinear dynamic responses in actual engineering, leading to a significant decrease in prediction accuracy. Especially when capturing extreme tension peaks crucial for safety assessment, there is a large error between the actual tension value and the actual value, making it difficult to meet the requirements of high-precision real-time monitoring.

[0005] In recent years, with the development of artificial intelligence technology, data-driven machine learning methods, especially recurrent neural networks (RNN, LSTM, etc.), have been gradually applied to the field of indirect measurement of mooring tension. Existing technical solutions mainly fall into two categories: First, tension inversion is performed by measuring the local inclination angle at key locations of the mooring cable. While this method is relatively simple and convenient to deploy sensors, it is limited by the non-unique mapping relationship between inclination angle and tension. That is, the same inclination angle may correspond to mooring tension under different motion states, leading to significant uncertainty in the tension inversion results based solely on a single inclination angle parameter. Second, mooring tension is inverted based on the six-degree-of-freedom motion data of the floating platform. This method can capture the nonlinear dynamics of the mooring cable to some extent, but because the input data lacks an explicit physical correlation with the mooring motion response and does not fully consider the local dynamic response of the mooring cable itself, when the input data exhibits nonlinearity and non-stationarity, the tension value is prone to overfitting or getting trapped in local minima, thus limiting its generalization ability and accuracy. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a method and system for monitoring the mooring tension of floating wind turbines based on multi-source data fusion. This overcomes the static assumptions and data dimension limitations of existing methods, and achieves accurate and reliable monitoring of the catenary mooring tension of semi-submersible platforms for floating offshore wind turbines.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides a method for monitoring the mooring tension of floating wind turbines based on multi-source data fusion.

[0008] A method for monitoring the mooring tension of floating wind turbines based on multi-source data fusion includes the following steps: The acquired dynamic data of the mooring cable body, global motion data of the floating platform, and environmental context data are preprocessed. Feature extraction was performed on the preprocessed dynamic data of the mooring cable body, global motion data of the floating platform, and environmental context data to obtain the dynamic features of the mooring cable body, the global motion features of the floating platform, and the environmental context features. Cross-modal feature fusion is performed on the dynamic characteristics of the mooring cable itself, the global motion characteristics of the floating platform, and the environmental context characteristics to obtain fused features; Temporal features are extracted from the fused features using a temporal convolutional network. A global feature vector is obtained based on the temporal features, and the mooring tension prediction result is obtained based on the global feature vector.

[0009] Secondly, the present invention provides a floating wind turbine mooring tension monitoring system based on multi-source data fusion.

[0010] A floating wind turbine mooring tension monitoring system based on multi-source data fusion includes: The data preprocessing unit is configured to preprocess the acquired dynamic data of the mooring cable body, the global motion data of the floating platform, and the environmental context data. The feature extraction unit is configured to extract features from the preprocessed dynamic data of the mooring cable body, the global motion data of the floating platform, and the environmental context data to obtain the dynamic features of the mooring cable body, the global motion features of the floating platform, and the environmental context features. The feature fusion unit is configured to perform cross-modal feature fusion on the dynamic features of the mooring cable body, the global motion features of the floating platform, and the environmental context features to obtain fused features. The tension prediction unit is configured to: extract temporal features from the fused features using a temporal convolutional network, obtain a global feature vector based on the temporal features, and obtain the mooring tension prediction result based on the global feature vector.

[0011] Thirdly, the present invention provides a computer device, comprising: a processor and a computer-readable storage medium; A processor, adapted to execute computer programs; A computer-readable storage medium storing a computer program, which, when executed by a processor, implements the floating wind turbine mooring tension monitoring method based on multi-source data fusion, which is the first aspect of the present invention.

[0012] Fourthly, the present invention provides a computer-readable storage medium storing a computer program adapted to be loaded by a processor and executed by the method for monitoring the mooring tension of floating wind turbines based on multi-source data fusion, which is a first aspect of the present invention.

[0013] Fifthly, the present invention provides a computer program product, which includes a computer program. When the computer program is executed by a processor, it implements the floating wind turbine mooring tension monitoring method based on multi-source data fusion according to the first aspect of the present invention.

[0014] Compared with the prior art, the beneficial effects of the present invention are: This invention innovatively integrates dynamic data of the mooring cable itself (acceleration and tilt angle), global platform motion data, and environmental context data to construct a comprehensive dynamic response information system for the mooring system. In particular, the introduction of acceleration data fundamentally solves the technical challenge of the uncertain mapping relationship between tilt angle and tension in traditional methods, enabling the model to distinguish tension differences corresponding to different dynamic states at the same tilt angle. Compared to the significant error (typically exceeding 15%) in predicting extreme tension peaks using traditional catenary methods, this invention reduces the prediction error to less than 10%, greatly improving monitoring accuracy.

[0015] Unlike traditional catenary methods based on quasi-static assumptions, the deep learning architecture employed in this invention can fully capture the nonlinear dynamic response of mooring systems in complex marine environments. Through the synergistic effect of temporal convolutional networks (TCN) and bidirectional long short-term memory networks (BiLSTM), it effectively models the high-frequency fluctuations and low-frequency evolution characteristics of mooring systems, overcoming the inherent defects of traditional methods in handling inertial forces and hydrodynamic effects.

[0016] To address the limitations of single-data-source inversion in existing machine learning methods, this invention proposes a multi-source temporal feature hierarchical fusion network (MTHF-Net). Through a cross-attention mechanism, it achieves deep interaction between features of different modalities. The innovative design, with platform motion as the query and mooring cable dynamics as the key, conforms to the physical law that "platform motion is the cause and cable response is the effect," establishing a more accurate and reliable mapping relationship and significantly reducing prediction uncertainty.

[0017] This invention innovatively integrates a Monte Carlo uncertainty estimation module, capable of simultaneously outputting tension prediction values ​​and their confidence levels. This function enables the system to identify abnormal operating conditions not fully covered by training data, and to issue timely reliability alerts when prediction uncertainty is high, providing dual protection for operation and maintenance decisions and greatly enhancing the system's engineering practicality.

[0018] This invention significantly reduces sensor failure risk and maintenance costs by employing non-contact motion measurement instead of easily damaged direct tension measurement. Simultaneously, the optimized model can be deployed on embedded edge devices, achieving real-time inference performance of less than 100 milliseconds, fully meeting the real-time requirements for online mooring tension monitoring in engineering projects.

[0019] The multi-branch neural network architecture designed in this invention can adaptively learn the dynamic characteristics of mooring systems under different sea conditions. By injecting environmental context information, the model can automatically adjust its prediction strategy according to environmental conditions such as wind, waves, and currents, maintaining excellent prediction performance under various operating conditions, and providing an all-weather, all-condition safety monitoring solution for floating wind power mooring systems.

[0020] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0021] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0022] Figure 1A flowchart illustrating a floating wind turbine mooring tension monitoring method based on multi-source data fusion, provided as an exemplary embodiment of the present invention; Figure 2 A schematic diagram of the principle of a floating wind power mooring tension monitoring system based on multi-source data fusion, provided as an exemplary embodiment of the present invention; Figure 3 A schematic diagram of a computer device provided for an exemplary embodiment of the present invention. Detailed Implementation

[0023] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0024] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0025] This implementation proposes a floating wind turbine mooring tension monitoring method based on multi-source data fusion. Through the innovative combination of multi-source information fusion and deep learning technology, it fundamentally solves the technical bottlenecks of existing mooring tension monitoring methods in terms of accuracy, reliability, and applicability, and provides advanced and effective technical support for the safe operation of floating offshore wind turbine mooring systems.

[0026] Specifically, such as Figure 1 As shown, the process includes the following: Step S101: Construct a multi-layered data acquisition system to simultaneously acquire three types of heterogeneous sensor data: 1) Dynamic data of the mooring cable body, which is measured in real time by using triaxial accelerometers and tilt sensors installed near the cable guide hole or at specific locations on the mooring cable to measure the acceleration components and biaxial spatial tilt angle of the mooring cable in three orthogonal directions; 2) Global motion data of the floating platform, which is acquired by using the inertial measurement unit on the platform and the Beidou satellite navigation system receiver to obtain the motion time sequence data of the platform's six degrees of freedom; 3) Environmental context data, which is collected by using environmental sensors such as wind, waves, and current to collect environmental parameters such as wind speed, wave height, wave period, and current velocity.

[0027] Step S102: Preprocess and feature engineer the aforementioned multi-source heterogeneous data to eliminate noise interference, resolve data inconsistency issues, and standardize data scale, providing high-quality, well-organized input data for subsequent deep learning models. The specific processing flow includes the following steps: (1) Filtering and noise reduction: To address the high-frequency noise and abnormal fluctuations in the original time series data, a Butterworth low-pass filter is used for smoothing, which can suppress high-frequency noise interference and abnormal fluctuations, and retain the low-frequency signal and effective frequency band signal that reflect the essential dynamic characteristics of the mooring system.

[0028] (2) Timestamp alignment and multi-source data synchronization: Due to the inherent delay in data acquisition and transmission from different sensors, the Dynamic Time Warping (DTW) algorithm is used to accurately match and align the timestamps of the multi-source data streams, ensuring that acceleration, tilt angle, platform motion and environmental data are strictly synchronized in the time dimension, and eliminating phase errors caused by inconsistent sampling times.

[0029] (3) Missing value imputation: To address data loss or outlier removal that may occur during data transmission, cubic spline interpolation is used to imput the missing data, ensuring that the imputed data curve is continuous and smooth, and maximizing the restoration of the real physical process. For segments with consecutive missing values ​​exceeding a preset threshold (e.g., 1 second), they are marked and excluded from the training set to ensure data reliability.

[0030] (4) Data Standardization: To eliminate the huge differences in the numerical range of various physical quantities due to different dimensions, and to avoid large-scale features dominating the gradient descent direction during model training, a robust scaling method is adopted. The standardization is calculated based on the statistical quantiles of the data. The median and interquartile range are statistically robust indicators, less susceptible to extreme outliers (such as outliers introduced by splashing waves, communication interruptions, or occasional sensor malfunctions). This method can effectively suppress the interference of outliers, generate a more stable and reliable data distribution, and to a certain extent, play a slight regularization role, helping to suppress overfitting, accelerate the model convergence process, and improve training generalization ability and stability. The calculation formula is as follows: (1); (2); For any training set data , The median. Interquartile range, It is the 25th percentile. It is at the 75th percentile.

[0031] Through the above systematic preprocessing process, a clean, synchronous, complete and scale-uniform multi-source time series dataset is finally obtained, laying a solid foundation for feature extraction and fusion of subsequent deep learning models.

[0032] Step S103: Construct a deep learning model of Multi-source Temporalfeature extraction and Hierarchical Fusion Network (MTHF-Net). This model adopts a parallel encoding-cross fusion-joint decoding architecture, specifically including: (1) Feature extraction branch one: mooring cable dynamic feature encoding branch.

[0033] This branch aims to extract physically meaningful dynamic features directly from the raw sensor signals of the mooring cable itself and solve the fundamental technical problem of the uncertain mapping relationship between inclination angle and tension. Under the same inclination angle, the mooring cable may be in different dynamic states such as acceleration, deceleration, or uniform motion, resulting in drastically different inertial forces and significant differences in mooring tension. To address the uncertainty and physical ambiguity of tension mapping from a single inclination angle, acceleration and inclination angle data are fused into the input to uniquely determine the dynamic state of the mooring cable. Based on this, a convolutional neural network (CNN) is used to extract key local dynamic features from the acceleration and inclination angle signals, and a squeeze-excitation (SE) attention mechanism is introduced to adaptively weight different physical quantity channels, thereby obtaining the key local dynamic features of the mooring cable.

[0034] 1) Input data and definitions.

[0035] Input the preprocessed dynamic timing data of the mooring cables, represented as a matrix. ,in For time step, The number of channels (corresponding to triaxial acceleration and biaxial spatial tilt).

[0036] 2) This branch adopts a modular architecture combining convolutional neural networks and a squeeze-excitation attention mechanism. The specific steps are as follows: Phase 1: Data Feature Extraction.

[0037] Time series data The input consists of a module with two one-dimensional convolutional layers. The first layer uses a larger kernel to capture dynamic patterns and trends over a longer period; the second layer uses a smaller kernel to focus on finer short-term feature changes. The convolutional kernel slides along the time axis to extract features. (3); Specifically, for the first... The first channel, the first Perform the following operations at each time point to extract data features: (4); in, For convolution kernel weights, For bias parameters, For the convolution kernel time dimension, To output the number of feature channels, To modify the activation function of the linear unit.

[0038] Phase Two: Squeezing-incentivizing attention mechanism weighting.

[0039] By introducing an attention mechanism, the model can dynamically allocate computational resources and focus on key information, thereby achieving more efficient and flexible processing capabilities when dealing with complex tasks. The computational steps are as follows: Squeezing operation: Applying compression to the data features of the convolution output. Global average pooling is performed to compress the global spatiotemporal information of each channel into a scalar. The calculation formula is as follows: (5); Obtain the channel statistics descriptor: (6); Activation operation: The non-linear interaction relationships between channels are learned through a two-layer fully connected network to generate the weights for each channel. (7); in, and As weight, and For bias, For the Sigmoid function, This is the ReLU activation function.

[0040] Feature recalibration: adjusting the activation weights Applied to original features The final calibrated output features are obtained. : (8); This step can effectively increase the contribution of key channels with high signal-to-noise ratio (such as acceleration or tilt angle in a specific direction that is sensitive to changes in tension) and suppress noise channels or redundant information.

[0041] (2) Feature extraction branch two: platform motion feature encoding branch.

[0042] By employing a stacked bidirectional long short-term memory network (Stacked BiLSTM), we can simultaneously utilize past and future contextual information to gain a more comprehensive understanding of the platform's motion state and capture the long-term temporal dependencies and low-frequency evolution patterns in the platform's six degrees of freedom motion. 1) Input data and definitions.

[0043] Input the platform's six-degree-of-freedom motion timing data preprocessed by the system. The six dimensions correspond to sway, sway, heave, roll, pitch and yaw respectively.

[0044] 2) This branch adopts an L-layer stacked bidirectional LSTM architecture, with each layer containing The specific steps for creating a hidden unit are as follows: Phase 1: Forward propagation computation.

[0045] For the The forward LSTM unit of the layer mainly defines the input gate. Forgotten Gate and output gate In the current cell state Discarding, adding, or exporting information at time steps The calculation formula is as follows: (9); in, The weight matrix is ​​the input to the gate. The weight matrix is ​​the hidden state to the gate. This is the bias vector.

[0046] Phase Two: Backpropagation computation.

[0047] For the The backward LSTM unit of the layer mainly defines the input gate. Forgotten Gate and output gate In the current cell state Discarding, adding, or exporting information at time steps The calculation formula is as follows: (10); Phase 3: Output splicing.

[0048] No. Layer in time step The final output is a concatenation of the forward and backward hidden states: (11); Phase 4: Output Features.

[0049] go through The final output of the stacked BiLSTM process contains a comprehensive temporal feature representation of the platform motion sequence: (12); (3) Feature extraction branch three: environment context encoding branch.

[0050] A multilayer perceptron (MLP) with two hidden layers is used to encode environmental parameters, aiming to transform the input environmental parameters through deep nonlinear transformation. This is mapped to a high-dimensional feature that can effectively represent the complex sea state context. The MLP in this branch consists of the following three sequentially connected fully connected layers: 1) First hidden layer: basic nonlinear mapping.

[0051] This layer performs the first nonlinear transformation from the original input space to the hidden space, capturing the initial interactions between environmental parameters: (13); 2) Second hidden layer: high-level feature abstraction.

[0052] This layer, based on the primary features, further synthesizes and abstracts more complex higher-order features directly related to mooring tension dynamics: (14); 3) Output layer: Feature extraction.

[0053] This layer further extracts high-level features and expands the encoded environmental feature vector into temporal features, aligning it with the temporal dimension of other branches, and outputs a context feature matrix. Specifically, it can be expressed as: (15); (16); Step S104: Cross-modal feature fusion.

[0054] The core innovation lies in the cross-modal feature fusion layer, which enables information interaction and alignment between mooring cable dynamics, platform motion, and environmental context, establishing a nonlinear mapping relationship between multi-source data. This is achieved using platform motion features... To query sequences based on the dynamic characteristics of mooring cables A cross-attention mechanism for key and value sequences enables the model to proactively uncover the deep nonlinear correlation between the platform's global motion and the mooring cable's local dynamic response. Simultaneously, environmental features are incorporated. The outputs of cross-attention are added point-by-point to each other as a global bias injection. Through the innovative design of the cross-modal feature fusion layer, this invention effectively solves the problem of insufficient information utilization caused by simple concatenation or independent processing of multi-source data in traditional methods, and achieves true deep feature interaction and alignment. The specific steps are as follows: 1) Input feature specification.

[0055] This fusion layer receives the outputs from three feature extraction branches as input, which are the dynamic feature matrices of the mooring cable. Platform motion feature matrix and environmental context feature matrix .in, To ensure a uniform time step, , , These represent the output feature dimensions of each branch.

[0056] 2) Feature alignment and dimensional uniformity.

[0057] Perform dimension alignment processing on the feature data of each feature extraction branch: (17); (18); (19); in, To query the transformation weight matrix, The weight matrix is ​​a key transformation matrix. The weight matrix represents the value changes. , For a unified feature dimension.

[0058] 3) Cross-attention mechanism.

[0059] A cross-attention mechanism is employed, using platform motion features as the query and mooring cable dynamic features as the key and value. The core idea is to effectively extract relevant information by calculating the similarity between input sequences and weighting different parts of the input. Its mathematical expression is as follows: (20); (twenty one).

[0060] 4) Environment context injection.

[0061] Inject environmental context features as a global bias into the fused features: (twenty two); 5) Residual connectivity and layer normalization.

[0062] To maintain gradient flow and training stability, residual connections and layer normalization are introduced: (twenty three); in, This is the final output feature matrix after deep fusion. Transform the weight matrix for the residual link.

[0063] Step S105: Spatiotemporal context modeling layer and output layer.

[0064] A Temporal Convolutional Network (TCN) is used to extract and refine the temporal features of the fused feature sequence, and finally, a fully connected layer outputs the predicted value of mooring tension. The TCN employs a dilated causal convolutional architecture to ensure that no future information is introduced during the modeling process, meeting the requirements of real-time prediction. Its core operations are as follows: (twenty four); in, This represents the feature vector of the l-th layer network at time step t, where l is the number of network layers and t is a time step in the time series. This represents the linear rectified activation function, used to introduce nonlinear characteristics into the network; This indicates a batch normalization operation; This represents the number of feature channels in the (l-1)th layer; This represents the convolution weights of the l-th layer; the (l-1)-th layer at time step... , the characteristic value at channel c; This represents the bias term of the l-th layer.

[0065] The output layer is responsible for mapping the spatiotemporal context features extracted by the temporal convolutional network to the final mooring tension prediction value. First, for the last layer... The output is then subjected to global temporal pooling to aggregate information from all time steps, forming a global feature vector: (25); Subsequently, the global feature vector is mapped to a scalar output, i.e., the predicted mooring tension value, through a fully connected layer, as expressed by: (26); in, and The learned weights and biases for the output layer.

[0066] Optionally, to quantify the reliability of the prediction results, Monte Carlo sampling is enabled during the inference phase to output the mean and standard deviation of the predicted values, thereby achieving tension prediction and uncertainty estimation. After the training phase, the Dropout layers in the network (including the aforementioned feature extraction network, fusion network, and spatiotemporal context modeling network) are not turned off. During prediction, the same input samples are used... implement Each independent forward propagation yields a set of predicted values. Based on this sampling result, the final prediction output and its uncertainty are calculated: (27); (28); in, As the final predicted value of mooring tension, This serves as a measure of the uncertainty of the predicted value. High uncertainty indicates that the model has a lower awareness of the input pattern, which can provide additional decision-making basis for system early warning.

[0067] Step S106: Model deployment and system implementation.

[0068] The trained and validated MTHF-Net model is deployed in an embedded edge computing device or central processing unit, connected to a real-time data stream, processed according to the aforementioned preprocessing procedure, and then input into the optimized model. This enables online, continuous, and high-precision inversion and monitoring of mooring tension, and provides data support for system status assessment and early warning decision-making.

[0069] Figure 2 A floating wind turbine mooring tension monitoring system based on multi-source data fusion is shown, comprising: The data preprocessing unit 201 is configured to preprocess the acquired dynamic data of the mooring cable body, the global motion data of the floating platform, and the environmental context data. The feature extraction unit 202 is configured to extract features from the preprocessed dynamic data of the mooring cable body, the global motion data of the floating platform, and the environmental context data to obtain the dynamic features of the mooring cable body, the global motion features of the floating platform, and the environmental context features. The feature fusion unit 203 is configured to perform cross-modal feature fusion on the dynamic features of the mooring cable body, the global motion features of the floating platform, and the environmental context features to obtain fused features. Tension prediction unit 204 is configured to: extract temporal features from the fused features using a temporal convolutional network, obtain a global feature vector based on the temporal features, and obtain the mooring tension prediction result based on the global feature vector.

[0070] It is understood that the aforementioned units can be individually or entirely merged into one or more other units, or some of the units can be further divided into multiple functionally smaller units. This achieves the same operation without affecting the technical effects of the embodiments of this application. The aforementioned units are based on logical functional division. In practical applications, the function of one unit can be implemented by multiple units, or the function of multiple units can be implemented by one unit. In other embodiments of this application, the system may also include other units. In practical applications, these functions can also be implemented with the assistance of other units, and can be implemented collaboratively by multiple units.

[0071] According to another embodiment of this application, the system of this embodiment can be constructed by running a computer program (including program code) capable of performing the steps involved in the corresponding method of the present invention on a general-purpose computing device, such as a computer, which includes processing elements and storage elements such as a central processing unit (CPU), random access memory (RAM), and read-only memory (ROM). The computer program can be recorded on, for example, a computer-readable recording medium, loaded into the aforementioned computing device through the computer-readable recording medium, and run therein.

[0072] Figure 3 A computer device is shown, which includes a processor 301, a communication interface 302, and a computer-readable storage medium 303. The processor 301, communication interface 302, and computer-readable storage medium 303 can be connected via a bus or other means.

[0073] The communication interface 302 is used to receive and send data. The computer-readable storage medium 303 can be stored in the memory of the electronic device. The computer-readable storage medium 303 is used to store computer programs, which include program instructions. The processor 301 is used to execute the program instructions stored in the computer-readable storage medium 303.

[0074] The processor 301 is the computing and control core of the electronic device. It is suitable for implementing one or more instructions, specifically for loading and executing one or more instructions to achieve the corresponding method flow or corresponding function.

[0075] Processor 301 is configured to perform the following procedure: The acquired dynamic data of the mooring cable body, global motion data of the floating platform, and environmental context data are preprocessed. Feature extraction was performed on the preprocessed dynamic data of the mooring cable body, global motion data of the floating platform, and environmental context data to obtain the dynamic features of the mooring cable body, the global motion features of the floating platform, and the environmental context features. Cross-modal feature fusion is performed on the dynamic characteristics of the mooring cable itself, the global motion characteristics of the floating platform, and the environmental context characteristics to obtain fused features; Temporal features are extracted from the fused features using a temporal convolutional network. A global feature vector is obtained based on the temporal features, and the mooring tension prediction result is obtained based on the global feature vector.

[0076] This invention also provides a computer-readable storage medium, which is a memory device in an electronic device for storing programs and data. It is understood that the computer-readable storage medium here may include both built-in storage media in the electronic device and extended storage media supported by the electronic device. The computer-readable storage medium provides storage space for storing the processing system of the electronic device.

[0077] Furthermore, this storage space also contains one or more instructions suitable for loading and execution by the processor. These instructions can be one or more computer programs (including program code). It should be noted that the computer-readable storage medium here can be a high-speed RAM memory; alternatively, it can also be at least one computer-readable storage medium located remotely from the aforementioned processor.

[0078] In one embodiment, the computer-readable storage medium stores one or more instructions; the processor loads and executes the one or more instructions stored in the computer-readable storage medium to perform the following process: The acquired dynamic data of the mooring cable body, global motion data of the floating platform, and environmental context data are preprocessed. Feature extraction was performed on the preprocessed dynamic data of the mooring cable body, global motion data of the floating platform, and environmental context data to obtain the dynamic features of the mooring cable body, the global motion features of the floating platform, and the environmental context features. Cross-modal feature fusion is performed on the dynamic characteristics of the mooring cable itself, the global motion characteristics of the floating platform, and the environmental context characteristics to obtain fused features; Temporal features are extracted from the fused features using a temporal convolutional network. A global feature vector is obtained based on the temporal features, and the mooring tension prediction result is obtained based on the global feature vector.

[0079] The present invention also provides a computer program product or computer program comprising computer instructions stored in a computer-readable storage medium. A processor of an electronic device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the electronic device to perform the following process: The acquired dynamic data of the mooring cable body, global motion data of the floating platform, and environmental context data are preprocessed. Feature extraction was performed on the preprocessed dynamic data of the mooring cable body, global motion data of the floating platform, and environmental context data to obtain the dynamic features of the mooring cable body, the global motion features of the floating platform, and the environmental context features. Cross-modal feature fusion is performed on the dynamic characteristics of the mooring cable itself, the global motion characteristics of the floating platform, and the environmental context characteristics to obtain fused features; Temporal features are extracted from the fused features using a temporal convolutional network. A global feature vector is obtained based on the temporal features, and the mooring tension prediction result is obtained based on the global feature vector.

[0080] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed in this application can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0081] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the flow or function according to the embodiments of this application is generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in or transmitted through a computer-readable storage medium. The computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic cable, digital cable) or wireless (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that a computer can access or a data processing device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state drive), etc.

[0082] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for monitoring the mooring tension of floating wind turbines based on multi-source data fusion, characterized in that, The process includes the following: The acquired dynamic data of the mooring cable body, global motion data of the floating platform, and environmental context data are preprocessed. Feature extraction was performed on the preprocessed dynamic data of the mooring cable body, global motion data of the floating platform, and environmental context data to obtain the dynamic features of the mooring cable body, the global motion features of the floating platform, and the environmental context features. Cross-modal feature fusion is performed on the dynamic characteristics of the mooring cable body, the global motion characteristics of the floating platform, and the environmental context characteristics to obtain fused features; Temporal convolutional networks are used to extract temporal features from the fused features, and global features are obtained based on the temporal features. The vector is used to obtain the mooring tension prediction result based on the global feature vector.

2. The floating wind turbine mooring tension monitoring method based on multi-source data fusion as described in claim 1, characterized in that, Dynamic data of the mooring cable itself, including: acceleration components of the mooring cable in three orthogonal directions and biaxial spatial tilt angle; Global motion data of the floating platform, including: motion timing data of the six degrees of freedom of the floating platform; Environmental context data, including: wind speed, wave height, wave period, and current velocity.

3. The floating wind turbine mooring tension monitoring method based on multi-source data fusion as described in claim 1, characterized in that, Feature extraction is performed on the preprocessed dynamic data of the mooring cable body, including: The preprocessed mooring cable dynamic time series data The convolutional module performs feature extraction, and for the first feature map in the output feature map... The first channel, the first Perform the following operations at each time point to extract data features. : ； in, For convolution kernel weights, For bias parameters, For the convolution kernel time dimension, To output the number of feature channels, To modify the activation function of the linear unit; Data features of convolution output Perform global average pooling to compress the global spatiotemporal information of each channel into a scalar: ; Obtain the channel statistics descriptor: ; The non-linear interaction relationships between channels are learned through a two-layer fully connected network to generate the weights of each channel: ,in, and As weight, and For bias, For the Sigmoid function, It is the ReLU activation function; Incentive weight Applied to original features The final calibrated output features are obtained. : .

4. The floating wind turbine mooring tension monitoring method based on multi-source data fusion as described in claim 1, characterized in that, A multi-layer stacked bidirectional LSTM architecture is used for feature extraction of global motion data of the floating platform; A multilayer perceptron with two hidden layers is used to encode the environmental context data to obtain environmental context features.

5. The floating wind turbine mooring tension monitoring method based on multi-source data fusion as described in claim 1, characterized in that, Dynamic characteristics of the mooring cable Global motion characteristics of floating platforms and environmental context features Cross-modal feature fusion is performed to obtain fused features, including: Perform dimension alignment processing on the feature data of each feature extraction branch: , , , To query the transformation weight matrix, The weight matrix is ​​a key transformation matrix. The weight matrix represents the value changes. and To unify feature dimensions, To ensure a uniform time step; Preliminary fusion features were obtained using a cross-attention mechanism, including: , ; Contextual features are injected as global biases into the initial fused features: ,in, This is the final fusion feature.

6. The floating wind turbine mooring tension monitoring method based on multi-source data fusion as described in claim 1, characterized in that, Global features are obtained based on temporal features. The vector, based on the global feature vector, yields the mooring tension prediction result, including: Global temporal pooling is performed on the output of the last layer of the temporal convolutional network to aggregate information from all time steps, forming a global feature vector. This global feature vector is then mapped to a scalar output through a fully connected layer, and this scalar output is used as the mooring tension prediction result.

7. A floating wind turbine mooring tension monitoring system based on multi-source data fusion, characterized in that, include: The data preprocessing unit is configured to preprocess the acquired dynamic data of the mooring cable body, the global motion data of the floating platform, and the environmental context data. The feature extraction unit is configured to extract features from the preprocessed dynamic data of the mooring cable body, the global motion data of the floating platform, and the environmental context data to obtain the dynamic features of the mooring cable body, the global motion features of the floating platform, and the environmental context features. The feature fusion unit is configured to perform cross-modal feature fusion on the dynamic features of the mooring cable body, the global motion features of the floating platform, and the environmental context features to obtain fused features. The tension prediction unit is configured to: extract temporal features from the fused features using a temporal convolutional network, obtain a global feature vector based on the temporal features, and obtain the mooring tension prediction result based on the global feature vector.

8. A computer device, characterized in that, include: Processor and computer-readable storage media; A processor, adapted to execute computer programs; A computer-readable storage medium storing a computer program, which, when executed by the processor, implements the floating wind turbine mooring tension monitoring method based on multi-source data fusion as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program adapted to be loaded by a processor and executed as described in any one of claims 1 to 6: the floating wind turbine mooring tension monitoring method based on multi-source data fusion.

10. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the floating wind turbine mooring tension monitoring method based on multi-source data fusion as described in any one of claims 1 to 6.

Images