Offshore wind power cable fatigue assessment method, computer equipment and system

Abstract

The present invention discloses a fatigue assessment method, computer equipment, and system for offshore wind power cables, relating to the field of submarine cable monitoring technology. The method comprises: acquiring real-time electrical signals spatially distributed across the submarine cable; converting the real-time electrical signals into real-time optical phase information; converting the real-time optical phase information into real-time strain information; preprocessing the real-time strain information; extracting features from the preprocessed real-time strain information to generate fatigue feature vectors, and combining the fatigue feature vectors into a fatigue feature vector group, wherein the fatigue feature vectors include strain peak value, number of fatigue load cycles, load change rate, and statistical characteristics; inputting the fatigue feature vector group into a pretrained submarine cable fatigue prediction model to predict the submarine cable fatigue strain response value at the next moment; and calculating the current cumulative fatigue damage of the wind power submarine cable based on the submarine cable fatigue strain response value. The present invention significantly improves the accuracy and robustness of fatigue prediction.

Description

Technical Field

[0001] The present invention relates to the field of submarine cable monitoring technology, and in particular to a fatigue assessment method, computer equipment, and system for offshore wind power submarine cables. Background Art

[0002] In marine engineering applications, submarine cables used in offshore wind farms are exposed to complex sea conditions characterized by the coupling of wind, waves, currents, tides, and other factors. Consequently, they are inevitably subjected to dynamic fatigue loads from various directions, manifesting as continuous alternating loads in the axial, lateral, and vertical directions along the cables. These loads arise from the combined effects of factors such as the undulating seabed topography, tidal erosion, wind and wave action, and the periodic vibration of the wind turbine tower.

[0003] In recent years, with the expansion of wind farms into deeper waters, the length of submarine cables has continued to increase, creating a more complex dynamic environment. The fatigue loads on cables have become more frequent and greater in magnitude, making fatigue damage a key risk in the operation and maintenance of offshore wind power cables. Although existing submarine cable health monitoring systems and related technologies are gradually being applied during on-site operation and maintenance, they still rely primarily on periodic manual inspections and offline data analysis, failing to accurately and in real time reflect the fatigue state of submarine cables under complex and changing sea conditions.

[0004] Current submarine cable fatigue monitoring technologies primarily rely on field observations, intermittent data collection, and offline analysis using human expert systems. While these methods have improved fault detection and prediction capabilities to a certain extent, they are unable to capture the cable's response characteristics under complex dynamic loads in real time, resulting in significant lags in fatigue assessment results. Furthermore, existing methods rely heavily on expert experience to interpret and judge data. This subjective factor can easily lead to uncertainty in analysis results and increase fatigue damage prediction errors. Furthermore, due to the complex and ever-changing submarine cable installation environment, various external loads are continuously superimposed and difficult to clearly distinguish. Traditional methods are unable to effectively distinguish the contribution of different factors to fatigue damage, making it difficult to precisely control and optimize key factors. In actual engineering applications, to mitigate the adverse effects of these issues on the safe operation of submarine cables, a large number of experts are typically employed to conduct periodic fatigue assessments and analyses to compensate for the shortcomings of traditional methods in terms of accuracy and reliability. However, this approach is not only costly and labor-intensive, but also ill-suited to the future trend of large-scale, intelligent operation and maintenance of offshore wind power. Especially in the context of the continuous expansion of wind farms and the increasing complexity of submarine cable networks, the frequent introduction of expert systems for analysis not only wastes precious human and material resources, but also makes it difficult to effectively improve assessment efficiency and accuracy, ultimately resulting in a serious imbalance between resource input and benefit output.

[0005] Therefore, there is an urgent need to develop a real-time, intelligent and efficient submarine cable fatigue status assessment and health monitoring technology. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to provide a fatigue assessment method, computer equipment and system for offshore wind power cables, which can greatly improve the accuracy and robustness of fatigue prediction.

[0007] In order to solve the above technical problems, the present invention provides a fatigue assessment method for offshore wind power cables, comprising: acquiring real-time electrical signals of the spatial distribution of the cables in real time; converting the real-time electrical signals into real-time optical phase information; converting the real-time optical phase information into real-time strain information; preprocessing the real-time strain information; performing feature extraction on the preprocessed real-time strain information to generate fatigue feature vectors, and combining the fatigue feature vectors into a fatigue feature vector group, wherein the fatigue feature vectors include strain peak value, number of fatigue load cycles, load change rate and statistical characteristics; inputting the fatigue feature vector group into a pre-trained cable fatigue prediction model to predict the fatigue strain response value of the cable at the next moment; and calculating the fatigue cumulative damage of the current wind power cable based on the fatigue strain response value of the cable.

[0008] As an improvement to the above solution, the submarine cable fatigue prediction model includes a convolutional neural network and a long short-term memory neural network.

[0009] As an improvement to the above scheme, the training steps of the convolutional neural network and the long short-term memory neural network include: obtaining historical strain information; performing initial processing on the historical strain information to generate baseline strain information; converting the baseline strain information into a feature matrix; inputting the feature matrix into the convolutional neural network for feature extraction and spatial expansion to train the convolutional neural network and generate a spatial strain sequence; converting the spatial strain sequence into a temporal strain sequence; and inputting the temporal strain sequence into the long short-term memory neural network to train the long short-term memory neural network.

[0010] As an improvement to the above scheme, the step of preprocessing the real-time strain information includes: denoising the real-time strain information; removing outliers from the denoised real-time strain information; interpolating and smoothing the real-time strain information after the outlier removal; and normalizing the real-time strain information after the interpolating and smoothing.

[0011] As an improvement of the above scheme, the step of calculating the fatigue cumulative damage of the current wind power submarine cable based on the fatigue strain response value of the submarine cable includes: calculating the fatigue cumulative damage of the current wind power submarine cable through the fatigue strain response value of the submarine cable according to the Miner linear fatigue cumulative damage criterion.

[0012] As an improvement to the above solution, the step of converting the real-time electrical signal into real-time optical phase information includes: according to the formula , calculate the optical phase information; where, is the optical phase information, It is the in-phase component of the real-time electrical signal after passing through the low-pass filter. It is the orthogonal component of the real-time electrical signal after passing through the low-pass filter.

[0013] As an improvement of the above solution, the step of converting the real-time optical phase information into real-time strain information includes: according to the formula , calculate the strain information; among them, For real-time response information, is the wavelength of light, is the difference in real-time phase information, is the fiber refractive index, is the gauge length, is the Pockels coefficient of single-mode fiber glass.

[0014] As an improvement to the above solution, the offshore wind power cable fatigue assessment method further includes: comparing the fatigue cumulative damage amount with a preset damage threshold; and generating early warning information when the fatigue cumulative damage amount is greater than or equal to the damage threshold.

[0015] Accordingly, the present invention also provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, wherein the processor implements the steps of the above-mentioned offshore wind power cable fatigue assessment method when executing the computer program.

[0016] Correspondingly, the present invention also provides an offshore wind power cable fatigue assessment system, which includes interconnected PHI-OTDR optical fiber sensors and the above-mentioned computer equipment, and the PHI-OTDR optical fiber sensors obtain real-time electrical signals of the spatial distribution of the submarine cable in real time.

[0017] The implementation of the present invention has the following beneficial effects:

[0018] The present invention combines advanced sensing technology with intelligent data analysis algorithms to minimize reliance on expert experience and avoid uncertainty and errors caused by human factors. Furthermore, the invention accurately identifies the contribution of different dynamic loads to cable fatigue damage in real time, clearly distinguishes key load factors, and helps operations and maintenance teams understand the cable's operating status in advance.

[0019] Furthermore, the submarine cable fatigue prediction model employed in the present invention utilizes a convolutional neural network (CNN) and a long short-term memory neural network (LSTM) to construct a CNN-LSTM neural network. The present invention inputs historical strain information into the CNN and LSTM networks, respectively, and trains the CNN and LSTM networks to output a submarine cable risk event identification model.

[0020] In addition, the offshore wind power cable fatigue assessment method of the present invention can capture and accurately predict the fatigue damage status of the cable in real time, realize active maintenance and intelligent early warning, and provide effective technical guarantee and theoretical support for the safe and efficient operation of offshore wind power cables and reduce operation and maintenance costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 This is a flow chart of a first embodiment of the offshore wind power cable fatigue assessment method of the present invention;

[0022] Figure 2 Schematic diagram of the structure of the submarine cable fatigue prediction model in the present invention;

[0023] Figure 3 This is a flow chart of the second embodiment of the offshore wind power cable fatigue assessment method of the present invention. DETAILED DESCRIPTION

[0024] To make the objectives, technical solutions, and advantages of the present invention more apparent, the present invention will be further described in detail below with reference to the accompanying drawings. It is hereby stated that any directional terms such as "up," "down," "left," "right," "front," "back," "inside," and "outside" that appear or will appear herein are based solely on the accompanying drawings and are not intended to limit the present invention.

[0025] See also Figure 1 , Figure 1 The flowchart of the first embodiment of the offshore wind power cable fatigue assessment method of the present invention is shown, which includes:

[0026] S101, real-time acquisition of electrical signals of the spatial distribution of submarine cables;

[0027] The origin of the geodetic fixed coordinate system is defined as the seabed reference position (0,0,0). The x-axis is the direction of the cable axis, the z-axis is the direction perpendicular to the seabed and upward, and the y-axis is the direction perpendicular to the xz plane. The positive x-axis extends outward along the cable axis, the positive z-axis is perpendicular to the seabed and upward, and the positive y-axis is determined by the right-hand rule and points to the left, perpendicular to the cable axis.

[0028] The location coordinates of a specific monitoring section in a wind power submarine cable are expressed as (x1, y1, z1). The PHI-OTDR optical fiber sensor deployed inside the submarine cable can obtain the backscattered Rayleigh light signal of the submarine cable under complex sea conditions in real time and convert the backscattered Rayleigh light signal into a real-time electrical signal.

[0029] S102, converting the real-time electrical signal into real-time optical phase information;

[0030] By using heterodyne demodulation and the I / Q dual outputs of the PHI-OTDR optical fiber sensor, the real-time optical phase information of the spatial distribution of the submarine cable can be accurately extracted from the real-time electrical signal.

[0031] Accordingly, the optical phase information can be calculated according to the following formula:

[0032]

[0033] in:

[0034] is the optical phase information;

[0035] It is the in-phase component of the real-time electrical signal after it passes through the low-pass filter (LPF). It reflects the correlation between the real-time electrical signal and the cosine component of the local carrier and is the extraction of the real part signal during demodulation.

[0036] It is the quadrature component after the real-time electrical signal passes through the low-pass filter. It reflects the correlation between the real-time electrical signal and the local carrier sinusoidal component and is the extraction of the imaginary signal.

[0037] Therefore, when a submarine cable fails, the PHI-OTDR optical fiber sensor is subjected to forces in all directions. Therefore, the optical phase information can be effectively calculated through the effective elastic-optical coefficient of the PHI-OTDR optical fiber sensor.

[0038] S103, converting the real-time optical phase information into real-time strain information;

[0039] According to the standard linear conversion relationship, the real-time optical phase information can be converted into the real-time strain signal at the corresponding position to reflect the fatigue response state of the submarine cable.

[0040] Accordingly, the strain information can be calculated according to the following formula:

[0041]

[0042] in:

[0043] For real-time response information;

[0044] is the wavelength of light, generally 1550.12nm;

[0045] is the difference in real-time phase information, in rad;

[0046] is the fiber refractive index, which is generally 1.46;

[0047] is the gauge length, which is generally 4.8m;

[0048] is the Pockels coefficient of single-mode optical fiber glass, which is generally taken as 0.79.

[0049] S104, pre-processing the real-time strain information;

[0050] Accordingly, the steps of preprocessing the real-time strain information include:

[0051] (1) De-noising of real-time strain information;

[0052] (2) Eliminate outliers from the real-time strain information after denoising;

[0053] (3) Perform interpolation and smoothing processing on the real-time strain information after outlier removal;

[0054] (4) Perform data standardization on the real-time strain information after interpolation and smoothing.

[0055] Therefore, denoising, outlier removal, interpolation smoothing and data normalization are performed on the real-time strain information after phase conversion to ensure that the quality of the input data meets the requirements of the subsequent submarine cable fatigue prediction model.

[0056] S105, performing feature extraction on the preprocessed real-time strain information to generate fatigue feature vectors, and combining the fatigue feature vectors into a fatigue feature vector group;

[0057] Accordingly, the fatigue characteristic vector includes strain peak value, number of fatigue load cycles, load change rate and statistical characteristics, but is not limited thereto;

[0058] Therefore, fatigue sensitive features are extracted from the preprocessed real-time strain information and a real-time fatigue feature vector group is constructed.

[0059] S106, inputting the fatigue feature vector group into a pre-trained submarine cable fatigue prediction model to predict the fatigue strain response value of the submarine cable at the next moment;

[0060] It should be noted that the submarine cable fatigue prediction model is trained using historical strain data. By repeatedly adjusting the number of hidden layer neurons, activation functions and network parameters, a highly nonlinear mapping relationship between strain data and fatigue damage is established, enabling real-time prediction of the submarine cable fatigue strain response value at the next moment.

[0061] like Figure 2 As shown in Figure 2, the submarine cable fatigue prediction model includes a convolutional neural network (CNN) and a long short-term memory neural network (LSTM).

[0062] It should be noted that the submarine cable fatigue prediction model used in this invention utilizes a convolutional neural network (CNN) and a long short-term memory neural network (LSTM) to construct a CNN-LSTM neural network. This method inputs historical strain information into the CNN and LSTM, respectively, to train the CNN and LSTM.

[0063] Accordingly, the CNN-LSTM neural network in the present invention has the following particularities:

[0064] (1) Unlike traditional CNN-LSTM, which is mainly used for end-to-end sequence modeling, the model in this study explicitly uses CNN for spatial feature enhancement rather than directly participating in time series modeling. By inputting short-distance (10m) strain data obtained by the Φ-OTDR system into the CNN for local pattern learning and spatial expansion, the model effectively improves its sensitivity to spatial continuity and local anomalies, significantly enhancing its ability to identify multiple points along the submarine cable.

[0065] (2) The model adopts a serial mode of “spatial expansion → temporal modeling” in its structural design, rather than a simple parallel connection or feature splicing. This design fully utilizes the advantages of CNN in extracting rules in the spatial domain and the ability of LSTM in capturing trends in the temporal domain, ensuring that the deep information of spatial perception is retained and enhanced to the greatest extent before entering the LSTM, thereby improving the accuracy and stability of temporal prediction.

[0066] (3) Unlike the general CNN-LSTM method that directly processes raw spatiotemporal mixed data, this model increases structural interpretability and data controllability during the data flow process through an intermediate explicit data reconstruction process (for example, constructing a sliding window sequence from the CNN output and inputting it into the LSTM). This phased design facilitates the optimization of the parameters of each submodule and provides greater flexibility and maintainability in actual deployment.

[0067] (4) This method optimizes the physical characteristics of the Φ-OTDR system, taking into account its inherent limitations in spatial resolution and signal-to-noise ratio. By using CNN for spatial compensation and LSTM for trend prediction, combined with a final risk assessment module, this method not only improves model performance but also provides a more engineering-adaptable intelligent early warning solution for distributed fiber optic monitoring systems.

[0068] Furthermore, the training steps of convolutional neural networks and long short-term memory neural networks include:

[0069] (1) Obtain historical strain information;

[0070] The strain data with a duration of about 20 seconds and a corresponding spatial length of 10 meters were obtained.

[0071] (2) Preliminary processing of historical strain information to generate baseline strain information;

[0072] Perform preliminary processing on historical strain information to ensure that the data meets the quality and stability of convolutional neural network input. Specifically:

[0073] First, the original historical strain information is cleaned (distorted or missing points are removed); then, denoising is performed using an appropriate algorithm (for example, wavelet threshold denoising or denoising methods based on empirical mode decomposition); finally, the denoised historical strain signal is normalized (for example, the data is scaled to the interval [–1, 1] or [0, 1]) to eliminate dimensionality effects and accelerate model convergence.

[0074] (3) Convert the baseline strain information into a feature matrix;

[0075] Through data conversion, the 10m strain data is constructed into a feature matrix that conforms to the input format of the convolutional neural network (CNN). Specifically:

[0076] According to the pre-set spatial sampling interval (or window size), the baseline strain information of 10 consecutive meters is organized into the following feature matrix:

[0077]

[0078] Among them, the row dimension of the feature matrix corresponds to the spatial position, and the column dimension corresponds to the time sampling.

[0079] (4) Inputting the feature matrix into the convolutional neural network for feature extraction and spatial expansion to train the convolutional neural network and generate a spatial strain sequence;

[0080] The feature matrix is ​​input into a convolutional neural network (CNN) for local feature learning. Multi-layer convolution and pooling operations are used to extract spatial local correlation information. The original 10m spatial dimension is expanded to 80m through deconvolution (or deconvolution) layers or interpolation operations.

[0081] Therefore, by leveraging the powerful spatial feature expression and local feature learning capabilities of convolutional neural networks (CNNs), the local spatial features of the feature matrix are extracted and expanded. This allows the 10-meter feature matrix to be expanded into an 80-meter spatial strain sequence (i.e., the original 10-meter strain distribution can be mapped to a richer 80-meter feature map), thereby enriching the spatial information and enhancing the subsequent prediction model's perception of the far-field strain pattern.

[0082] (5) Convert the spatial strain series into a temporal strain series;

[0083] The expanded 80m spatial strain sequence is converted into a temporal strain sequence suitable for long short-term memory (LSTM) network input. Specifically:

[0084] The expanded spatial features can be divided into several time series segments according to the time step, forming the following 80m time strain series:

[0085]

[0086] (6) The time strain sequence is input into the long short-term memory neural network to train the long short-term memory neural network.

[0087] The time-strain sequence is input into an LSTM network. Recurrent units encode historical information and learn temporal dependencies, ultimately predicting strain trends for the next minute (for example, 60 seconds corresponds to multiple sampling moments). During training, a sliding window technique can be used to map several historical sequences of duration T to a future target sequence. The mean squared error (MSE) or other more robust metrics can be used as the loss function to optimize network parameters and improve prediction accuracy.

[0088] Therefore, the long short-term memory network (LSTM) is used to train and fit the prediction model of the time strain series, and finally output the strain value within the next minute to achieve short-term prediction and risk assessment of the strain situation of the submarine cable.

[0089] Accordingly, during the training process, the data is divided into training and test sets, and the convolutional neural network and long short-term memory neural network are trained and tested respectively, and the accuracy and robustness of the model in submarine cable strain prediction and risk identification are verified to ensure the reliability and effectiveness of the model in practical applications.

[0090] Furthermore, through the real-time rolling prediction update mechanism, the submarine cable fatigue prediction model can accurately and quickly capture the fatigue damage evolution trend of wind power submarine cables, effectively reducing the prediction error.

[0091] S107: Calculate the fatigue cumulative damage of the current wind power submarine cable according to the fatigue strain response value of the submarine cable.

[0092] Specifically, the fatigue cumulative damage of the current wind power submarine cable can be calculated based on the Miner linear fatigue cumulative damage criterion and the fatigue strain response value of the submarine cable.

[0093] As can be seen from the above, the present invention has the following beneficial effects:

[0094] 1. Efficient and accurate automated fatigue analysis surpasses the limitations of traditional methods. Traditional expert systems rely on preset rules, and manual calculations rely on empirical formulas, making it difficult to fully capture the complex characteristics of submarine cable fatigue damage. The submarine cable fatigue prediction model of this invention uses a data-driven approach to automatically extract spatial features and time series patterns, avoiding the limitations of manually set features and significantly improving the accuracy and robustness of fatigue prediction. Furthermore, the submarine cable fatigue prediction model can learn potential damage patterns from large-scale PhiOTDR monitoring data, reducing human intervention, improving analysis efficiency, and being applicable to different environmental conditions, with stronger generalization capabilities.

[0095] 2. Real-time monitoring and trend prediction capabilities enable intelligent early warning. Traditional methods primarily focus on post-event diagnosis, making real-time monitoring and preventative maintenance difficult to achieve. The invented submarine cable fatigue prediction model not only processes massive amounts of data and supports online monitoring, but also uses the sequence modeling capabilities of LSTM to predict strain changes up to the next minute, enabling trend prediction and early warning. Furthermore, combined with real-time sensor data, the submarine cable fatigue prediction model can dynamically adjust and proactively identify potential fatigue risks, providing intelligent decision-making support for operations and maintenance, effectively reducing the accident rate of submarine cables and extending their service life.

[0096] See also Figure 3 , Figure 3 A flow chart of a second embodiment of the offshore wind power cable fatigue assessment method of the present invention is shown, which includes:

[0097] S201, real-time acquisition of electrical signals of the spatial distribution of the submarine cable;

[0098] S202, converting the real-time electrical signal into real-time optical phase information;

[0099] S203, converting the real-time optical phase information into real-time strain information;

[0100] S204, pre-processing the real-time strain information;

[0101] S205 , performing feature extraction on the preprocessed real-time strain information to generate fatigue feature vectors, and combining the fatigue feature vectors into a fatigue feature vector group;

[0102] S206, inputting the fatigue feature vector group into a pre-trained submarine cable fatigue prediction model to predict the fatigue strain response value of the submarine cable at the next moment;

[0103] S207: Calculate the fatigue cumulative damage of the current wind power submarine cable according to the fatigue strain response value of the submarine cable.

[0104] S208, comparing the fatigue cumulative damage amount with a preset damage threshold;

[0105] S209: When the fatigue cumulative damage is greater than or equal to the damage threshold, a warning message is generated.

[0106] and Figure 1 Unlike the first embodiment shown, this embodiment compares the cumulative fatigue damage with a preset damage threshold. When the cumulative fatigue damage reaches the damage threshold, a warning signal is promptly issued to alert operation and maintenance personnel or the intelligent monitoring system to proactively intervene and implement targeted maintenance or protective measures, thereby effectively reducing safety hazards caused by fatigue.

[0107] Therefore, the fatigue assessment method for offshore wind power cables of the present invention is applicable to the actual operating conditions of submarine cables in wind farms. It can capture and accurately predict the fatigue damage status of submarine cables in real time, realize active maintenance and intelligent early warning, and provide effective technical guarantee and theoretical support for the safe and efficient operation of offshore wind power cables and the reduction of operation and maintenance costs.

[0108] Correspondingly, the present invention also discloses a computer device, including a memory and a processor, wherein the memory stores a computer program, wherein the processor implements the steps of the above-mentioned offshore wind power cable fatigue assessment method when executing the computer program.

[0109] In addition, the present invention also discloses an offshore wind power cable fatigue assessment system, including interconnected PHI-OTDR optical fiber sensors and the above-mentioned computer equipment, and the PHI-OTDR optical fiber sensors obtain real-time electrical signals of the spatial distribution of the submarine cable in real time.

[0110] From the above, it can be seen that the fatigue assessment method of offshore wind power cables of the present invention combines advanced sensing technology with intelligent data analysis algorithms, which can minimize the reliance on expert experience and avoid uncertainty and errors caused by human factors. At the same time, the present invention can accurately identify the contribution of different dynamic loads to the fatigue damage of submarine cables in real time, clearly distinguish key load factors, and help the operation and maintenance team to grasp the operating status of the submarine cables in advance and actively implement targeted maintenance and control measures. Therefore, the present invention not only greatly reduces the waste of human and material resources, but also significantly improves the level of safe operation of submarine cables and the accuracy of fatigue damage assessment, thereby promoting the innovation and development of offshore wind power operation and maintenance technology systems, and providing strong technical support and guarantee for the offshore wind power industry to achieve large-scale, efficient and intelligent development.

[0111] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A fatigue assessment method for offshore wind power cables, characterized in that: include: Real-time acquisition of electrical signals distributed in the space of submarine cables; Converting the real-time electrical signal into real-time optical phase information; converting the real-time optical phase information into real-time strain information; Preprocessing the real-time strain information; Performing feature extraction on the pre-processed real-time strain information to generate fatigue feature vectors, and combining the fatigue feature vectors into a fatigue feature vector group, wherein the fatigue feature vector includes strain peak value, number of fatigue load cycles, load change rate and statistical characteristics; Inputting the fatigue feature vector group into a pre-trained submarine cable fatigue prediction model to predict the fatigue strain response value of the submarine cable at the next moment; the submarine cable fatigue prediction model includes a convolutional neural network and a long short-term memory neural network; Calculating the fatigue cumulative damage of the current wind power submarine cable according to the fatigue strain response value of the submarine cable; The training steps of the convolutional neural network and the long short-term memory neural network include: obtaining historical strain information; performing preliminary processing on the historical strain information to generate baseline strain information; converting the baseline strain information into a feature matrix; inputting the feature matrix into the convolutional neural network for feature extraction and spatial expansion to train the convolutional neural network and generate a spatial strain sequence; converting the spatial strain sequence into a temporal strain sequence; and inputting the temporal strain sequence into the long short-term memory neural network to train the long short-term memory neural network.

2. The offshore wind power cable fatigue assessment method according to claim 1, characterized in that: The step of preprocessing the real-time strain information includes: performing denoising processing on the real-time strain information; performing outlier elimination processing on the real-time strain information after denoising; performing interpolation and smoothing processing on the real-time strain information after outlier elimination processing; Data standardization is performed on the real-time strain information after interpolation and smoothing.

3. The offshore wind power cable fatigue assessment method according to claim 1, characterized in that: The step of calculating the fatigue cumulative damage amount of the current wind power submarine cable according to the fatigue strain response value of the submarine cable includes: calculating the fatigue cumulative damage amount of the current wind power submarine cable according to the fatigue strain response value of the submarine cable according to the Miner linear fatigue cumulative damage criterion.

4. The offshore wind power cable fatigue assessment method according to claim 1, characterized in that: The step of converting the real-time electrical signal into real-time optical phase information comprises: According to the formula , calculate optical phase information; in, is the optical phase information, It is the in-phase component of the real-time electrical signal after passing through the low-pass filter. It is the orthogonal component of the real-time electrical signal after passing through the low-pass filter.

5. The offshore wind power cable fatigue assessment method according to claim 1, characterized in that: The step of converting the real-time optical phase information into real-time strain information comprises: According to the formula , calculate strain information; in, For real-time response information, is the wavelength of light, is the difference in real-time phase information, is the fiber refractive index, is the gauge length, is the Pockels coefficient of single-mode fiber glass.

6. The offshore wind power cable fatigue assessment method according to claim 1, characterized in that: Also includes: Comparing the fatigue cumulative damage amount with a preset damage threshold; When the fatigue cumulative damage amount is greater than or equal to the damage threshold, a warning message is generated.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the offshore wind power cable fatigue assessment method according to any one of claims 1 to 6 are implemented.

8. An offshore wind power cable fatigue assessment system, characterized in that: The device comprises a PHI-OTDR optical fiber sensor and the computer device according to claim 7, which are connected to each other, and the PHI-OTDR optical fiber sensor obtains real-time electrical signals of the spatial distribution of the submarine cable in real time.

Images