Offshore wind plant cooperative operation and maintenance method based on dynamic simulation of atmospheric marine environment

Abstract

The invention relates to an offshore wind plant collaborative operation and maintenance method based on atmospheric marine environment dynamic simulation, and is suitable for the field of new energy. The method comprises the following steps: S1, performing clock synchronization, coordinate conversion, exception elimination, missing value repair and dimension standardization processing on acquired multi-source heterogeneous data to generate a standardized data set with a unified format; s2, constructing an atmospheric marine environment field dynamic model and an equipment state evolution model, performing coupling training on the two types of models through a multi-scale space-time fusion algorithm, generating an environment change trend prediction result and an equipment operation state prediction result, and dynamically correcting model parameters according to prediction errors; and S3, carrying out quantitative grading on equipment fault early warning and environmental risks, carrying out matching scheduling in combination with a real-time operation and maintenance resource state, generating a collaborative operation and maintenance plan through a space-time optimization algorithm, executing the collaborative operation and maintenance plan and feeding back an execution state in real time, and realizing iterative optimization of the operation and maintenance plan.

Description

Technical Field

[0001] This invention relates to a collaborative operation and maintenance method for offshore wind farms based on dynamic simulation of the atmospheric and marine environment. It is applicable to the field of new energy. Background Technology

[0002] As offshore wind power is developed on a large scale in deep-sea areas, operation and maintenance (O&M) operations face three major challenges: strong dynamic coupling between the atmospheric and marine environments, slow response to equipment failures, and low efficiency of multi-task collaboration. The spatiotemporal abrupt changes in environmental fields (wind, waves, and currents) are strongly coupled with the degradation of the operating status of equipment such as wind turbines and cables. However, traditional O&M models rely solely on static threshold early warning and discrete decision-making mechanisms, which are insufficient to meet the real-time requirements of resource scheduling in complex marine environments. Summary of the Invention

[0003] The technical problem to be solved by this invention is to provide a collaborative operation and maintenance method for offshore wind farms based on dynamic simulation of the atmospheric and marine environment, in order to address the above-mentioned problems.

[0004] The technical solution adopted in this invention is: a collaborative operation and maintenance method for offshore wind farms based on dynamic simulation of the atmospheric and marine environment, comprising: S1. Collect atmospheric and marine environmental data, wind farm equipment operation status data, and operation and maintenance resource status data of offshore wind farm areas through multiple types of monitoring points. Perform clock synchronization, coordinate transformation, anomaly removal, missing value repair, and dimensional standardization on the collected multi-source heterogeneous data to generate a standardized dataset in a unified format. S2. Based on the standardized dataset, construct a dynamic model of the atmospheric and marine environmental field and a model of equipment status evolution. Couple the two models through a multi-scale spatiotemporal fusion algorithm to generate prediction results of environmental change trends and equipment operating status. Dynamically correct the model parameters according to the prediction error. S3. Based on the predicted results of environmental change trends, the predicted results of equipment operation status, and the pre-stored operation and maintenance rules, the equipment fault warning and environmental risks are quantitatively classified and graded. The real-time operation and maintenance resource status is matched and scheduled. A collaborative operation and maintenance plan is generated through a spatiotemporal optimization algorithm. The collaborative operation and maintenance plan is executed and the execution status is fed back in real time to realize the iterative optimization of the operation and maintenance plan.

[0005] The atmospheric and marine environmental data includes wind field parameters, wave field parameters, and flow field parameters, which are collected collaboratively by monitoring equipment deployed on the sea surface, underwater observation platforms on the seabed, and onshore wind measurement towers. The equipment operating status data includes the wind turbine's rotational speed, vibration amplitude, temperature, and the insulation resistance of the submarine cable, which are collected by sensors deployed in key parts of the equipment. The operation and maintenance resource status data includes the location of operation and maintenance personnel, equipment availability, and material inventory, which are collected through GPS positioning devices and inventory management systems.

[0006] In step S1, clock synchronization is achieved through a satellite time synchronization system to control timestamp errors; coordinate transformation converts the latitude and longitude of the monitoring point into planar coordinates through the UTM projected coordinate system.

[0007] In step S1, the anomaly removal uses the 3σ principle to identify abnormal data points and removes impulse noise through sliding median filtering; the missing value repair uses linear interpolation, and data segments that are continuously missing for more than 2 hours are marked as invalid; the dimensional standardization uses Min-Max standardization to transform the data to the [0,1] interval.

[0008] In step S2, the atmospheric and oceanic environmental field dynamic model includes a wind speed field model, a wave field model, and an ocean current field model. The wind speed field model is constructed by integrating the WRF meteorological model and the CFD method, the wave field model is constructed based on the SWAN model, and the ocean current field model is constructed based on the FVCOM model. The equipment state evolution model extracts the equipment fault characteristic frequency and combines it with the exponential degradation model to quantify the evolution law of equipment health status.

[0009] In step S2, the multi-scale spatiotemporal fusion algorithm includes the following sub-steps: S21. Synchronize the standardized dataset by timestamp and resample using a sliding time window to generate a unified time series; S22. Perform a 3-level Daubechies-4 wavelet decomposition on the unified time series to obtain the low-frequency approximate components. With high-frequency detail components Calculate the energy percentage of each high-frequency component. ; S23. Generate a grid covering the offshore wind farm area, and construct a Gaussian semivariogram based on the Kriging interpolation method. Combine the location of each monitoring point and the collected atmospheric and marine environmental data to generate the environmental field matrix of each grid cell. S24. The splicing of time decomposition components and the environmental field matrix forms a spatiotemporal feature tensor, which is then combined with the energy proportion. Input spatiotemporal convolutional neural network, using The model is trained using a weighted mixed loss function, and the prediction results are output.

[0010] In step S2, the prediction error is obtained through continuous... The average relative error of each time window Quantification; when When the value exceeds the preset value, extract historical data within the most recent preset time period to construct an updated training set, freeze the parameters of the front 3 layers of the spatiotemporal convolutional neural network, fine-tune the weights of the last fully connected layer, and dynamically correct the model parameters.

[0011] In step S3, the quantitative classification of equipment fault warnings and environmental risks based on the environmental change trend prediction results, equipment operating status prediction results, and pre-stored operation and maintenance rules includes: Construct a device fault feature vector, including fault location weight, parameter deviation, and fault development rate. The parameter deviation is determined based on the difference between the device operating status prediction result and the device operating status threshold corresponding to the environmental change trend prediction result. The fault development rate is the development rate of the device operating status prediction result after it deviates from the threshold. Based on the equipment fault feature vector, the fault level value is calculated, and the equipment fault warning level is determined based on the fault level value. The environmental risk level is determined based on the duration of deviation from the safe threshold in the environmental change trend prediction results.

[0012] In step S3, the resource matching includes: selecting suitable maintenance personnel based on task skill requirements, planning the operation paths of maintenance vessels and drones based on environmental prediction results, checking material inventory, and triggering the shortage material replenishment process.

[0013] In step S3, the spatiotemporal optimization algorithm is a spatiotemporal collaborative genetic algorithm, which uses "task-resource-time" as chromosome encoding, and the fitness function weightedly considers task priority, environmental risk period and spatial clustering efficiency.

[0014] A collaborative operation and maintenance device for offshore wind farms based on dynamic simulation of the atmospheric and marine environment includes: The data acquisition module is used to collect atmospheric and marine environmental data, wind farm equipment operation status data, and operation and maintenance resource status data of offshore wind farm areas through multiple types of monitoring points. It performs clock synchronization, coordinate transformation, anomaly removal, missing value repair, and dimensional standardization on the collected multi-source heterogeneous data to generate a standardized dataset in a unified format. The dynamic simulation module is used to construct a dynamic model of the atmospheric and marine environmental field and a model of equipment status evolution based on the standardized dataset. The two models are coupled and trained through a multi-scale spatiotemporal fusion algorithm to generate prediction results of environmental change trends and equipment operating status. The model parameters are dynamically corrected according to the prediction error. The operation and maintenance decision module is used to quantify and classify equipment fault warnings and environmental risks based on the predicted results of environmental change trends, equipment operating status, and pre-stored operation and maintenance rules. It also performs matching and scheduling in combination with real-time operation and maintenance resource status, generates collaborative operation and maintenance plans through spatiotemporal optimization algorithms, executes the collaborative operation and maintenance plans, and provides real-time feedback on the execution status to achieve iterative optimization of the operation and maintenance plans.

[0015] A storage medium storing a computer program executable by a processor, wherein the computer program, when executed, implements the steps of the collaborative operation and maintenance method for offshore wind farms.

[0016] A collaborative operation and maintenance device for offshore wind farms includes a memory and a processor. The memory stores a computer program that can be executed by the processor. When the computer program is executed, it implements the steps of the collaborative operation and maintenance method for offshore wind farms.

[0017] The beneficial effects of this invention are: This invention constructs a coupling system between an atmospheric and marine environmental field dynamic model (wind speed field, wave field, and ocean current field model) and an equipment state evolution model, and combines a multi-scale spatiotemporal fusion algorithm to achieve deep correlation between the two types of data.

[0018] This invention uses a multi-scale spatiotemporal fusion algorithm to capture the coupling relationship between transient environmental changes and equipment trend degradation through time decomposition (3-layer Daubechies-4 wavelet decomposition) and spatial reconstruction (Kriging interpolation of the environmental field matrix), which significantly improves prediction accuracy.

[0019] This invention enhances the robustness of extreme environment error penalty and equipment status prediction by using a weighted hybrid loss function (MSE+MAE), enabling the model to accurately identify equipment failure risks caused by strong winds and high waves in advance.

[0020] This invention constructs a coupled model of atmospheric and marine environmental fields and equipment status through a dynamic simulation module, and combines a multi-scale spatiotemporal fusion algorithm to achieve collaborative prediction of environmental change trends and equipment failure risks. It solves the problem of the lack of correlation analysis between environment and equipment status in existing technologies, enabling operation and maintenance decisions to identify the chain risk of "environmental change - equipment failure" in advance, changing the lag of traditional static threshold early warning, and realizing the transformation from "post-failure response" to "pre-risk prevention".

[0021] This invention constructs a feature vector based on fault location weight, parameter deviation, and fault development rate, and combines it with the quantitative classification of environmental risk duration to make task priority determination more accurate and avoid resource misallocation.

[0022] This invention integrates multi-dimensional data such as the skills of maintenance personnel, the environmental resistance of equipment, and material inventory in the resource matching process, and plans the operation path in combination with the environmental field matrix to avoid water depth exceeding the limit and high-risk areas, thereby shortening the average round-trip voyage of maintenance vessels.

[0023] This invention employs a spatiotemporal collaborative genetic algorithm, using "task-resource-time" as the encoding. The fitness function takes into account priority, environmental time period, and spatial clustering, reducing the multi-task conflict rate, shortening the average maintenance time of a single device, and reducing annual operation and maintenance costs.

[0024] This invention integrates the status of operation and maintenance resources with marine environmental constraints, and uses a spatiotemporal collaborative genetic algorithm to arrange operation and maintenance plans according to task priority, environmental risk period, and wind turbine geographic clustering. This solves the problems of resource conflicts and path redundancy caused by single data dimension and static scheduling in the prior art, realizes dynamic matching of multiple tasks and multiple resources, and improves the orderliness of operation and maintenance task execution and resource utilization efficiency.

[0025] This invention quantifies prediction accuracy based on the average relative error over n consecutive time windows. When the error exceeds a preset value, the model is corrected through a lightweight update strategy of "freezing front-end parameters + fine-tuning the fully connected layer" to reduce prediction errors in scenarios of sudden environmental changes.

[0026] The dynamic adaptability of the model and operation and maintenance plan in this invention enables the solution to flexibly cope with complex environments such as typhoons and sudden ocean currents, completely solving the problem of "plan generation-execution" break and ensuring the continuity and stability of deep-sea wind power operation and maintenance.

[0027] This invention forms a closed-loop mechanism of "plan generation - execution - feedback - optimization" by collecting real-time data on equipment parameter adjustment results, resource execution status, and environmental risk evolution. Combined with the online model correction capability of the dynamic simulation module, it solves the problems of broken control processes and poor model adaptability in existing technologies, enabling operation and maintenance strategies to adapt to sudden changes in the deep-sea environment in real time, and ensuring the continuous effectiveness and environmental adaptability of operation and maintenance plans.

[0028] In this invention, environment-device coupled prediction provides a data foundation for accurate scheduling, multi-task collaborative scheduling improves resource utilization efficiency, and dynamic adaptive mechanism ensures adaptability to complex scenarios, ultimately achieving the operation and maintenance goal of "accurate prediction, excellent scheduling, and stable execution". Attached Figure Description

[0029] Figure 1 The flowchart is for an example. Detailed Implementation

[0030] The embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention. The step numbers in the following embodiments are set only for ease of explanation, and there is no limitation on the order between the steps. The execution order of each step in the embodiments can be adaptively adjusted according to the understanding of those skilled in the art.

[0031] In the description of this invention, "multiple" means two or more. The use of "first" and "second" is for distinguishing technical features only and should not be construed as indicating or implying relative importance, or implicitly indicating the number of indicated technical features, or the order of the indicated technical features. Furthermore, unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art.

[0032] Example 1: This example is a collaborative operation and maintenance method for offshore wind farms based on dynamic simulation of the atmospheric and marine environment, including:

[0033] S1. Collect atmospheric and marine environmental data, wind farm equipment operation status data, and operation and maintenance resource status data of offshore wind farm areas through multiple types of monitoring points. Perform clock synchronization, coordinate transformation, anomaly removal, missing value repair, and dimensional standardization on the collected multi-source heterogeneous data to generate a standardized dataset in a unified format.

[0034] In this embodiment, the various types of monitoring points include atmospheric and marine environmental monitoring points, equipment operation status monitoring points, and operation and maintenance resource status monitoring points.

[0035] In this example, the atmospheric and marine environmental data collected by the atmospheric and marine environmental monitoring point includes wind field parameters, wave field parameters, and current field parameters, specifically including data such as wind speed, wind direction, wave height, ocean current speed, and seawater salinity.

[0036] Atmospheric and marine environmental monitoring stations include sub-monitoring stations for sea surface meteorology, seawater parameters, and atmospheric environment, among which: The marine meteorological monitoring sub-monitoring points are integrated into the marine meteorological buoys, and collect wind speed and wind direction data through ultrasonic anemometers and sea wave height data through pressure wave height meters. As a further explanation of this step, this embodiment can use a spherical marine meteorological buoy with a diameter of 3 meters (such as a NOMAD-type buoy), integrating an ultrasonic anemometer and a pressure wave height meter. The buoy is fixed to the central area of ​​the wind farm by anchor chains, in water depths of 30-50 meters. Wind speed and direction data are transmitted in real time to a shore-based server via a 4G / 5G communication module.

[0037] The seawater parameter monitoring sub-monitoring points are integrated into the underwater observation platform. They collect ocean current velocity data through an electromagnetic current meter and seawater temperature and salinity data through a temperature, salinity and depth meter. As a further explanation of this step, the underwater observation platform in this embodiment can adopt a bottom-mounted structure, deployed near the wind turbine foundation, and equipped with an electromagnetic current meter and a temperature, salinity, and depth (TDM) meter. Current velocity data is transmitted via an underwater acoustic communication link, and the TDM data sampling interval is 10 seconds.

[0038] The atmospheric environment monitoring sub-monitoring points are integrated into the onshore wind measurement tower. They collect wind speed and wind direction profile data at different altitudes using lidar anemometers, and collect atmospheric temperature, humidity, and air pressure data through meteorological stations.

[0039] As a further explanation of this step, the onshore wind measurement tower in this embodiment can be constructed by installing a lidar anemometer and a weather station in layers. The lidar periodically generates wind speed profile data, while the weather station collects temperature, humidity, and air pressure parameters in real time, which are then transmitted to a data center via a fiber optic network.

[0040] The equipment operation status data collected by the equipment operation status monitoring points include the wind turbine's speed, vibration amplitude, temperature, and the insulation resistance of the submarine cable.

[0041] The equipment status monitoring system uses vibration sensors, speed encoders, and temperature sensors deployed on the wind turbines, insulation resistance testers for submarine cables, and GPS positioning equipment on maintenance vessels to collect data such as wind turbine speed, generator temperature, and cable insulation resistance in real time.

[0042] As a further explanation of this step, in this embodiment, an incremental speed encoder can be installed at the main shaft of the wind turbine to monitor the impeller speed; a thermal resistance sensor can be embedded in the generator stator winding to collect the generator temperature and connected to the data acquisition device through a three-wire connection; and a vibration sensor can be deployed in the gearbox bearing housing to collect vibration data according to a set sampling rate.

[0043] In this embodiment, the operation and maintenance resource status data includes the location of operation and maintenance personnel, equipment availability, and material inventory, which are collected through GPS positioning devices and an inventory management system.

[0044] In this example, a satellite timing system is used to synchronize the clock of the monitoring equipment with multi-source heterogeneous data to control timestamp errors, convert the geographical coordinates of different monitoring points to a preset coordinate system to control conversion errors, and cross-validate the multi-sensor data of the same monitoring object. When the data deviation exceeds a preset threshold, a sensor fault warning is triggered.

[0045] In this embodiment, a satellite time receiver is used to provide a time reference for all monitoring equipment, and the clocks of the shore-based server and the field equipment are synchronized periodically through network protocols. At the same time, the wind farm area adopts the UTM projection coordinate system, and the latitude and longitude of the monitoring points are converted into planar coordinates according to parameters such as the central meridian longitude and scale factor. In addition, the multi-sensor data of the same monitoring object are cross-validated. When the data deviation exceeds the preset threshold, a sensor fault warning is triggered, the backup sensor is automatically switched and a maintenance work order is generated.

[0046] Geographical location data for different monitoring points are converted into planar coordinates using the UTM projection coordinate system. The conversion formula is: ; in, The latitude and longitude of the monitoring point; This is the Earth's semi-major axis; Longitude of the central meridian; The Earth's eccentricity; It is a scaling factor; This represents the offset of the origin of the projection zone.

[0047] This embodiment cleans and standardizes atmospheric and marine environmental data and wind farm equipment operation status data, including removing abnormal fluctuation data points and repairing missing data, aligning timestamps and unifying sampling frequencies for multi-source heterogeneous data, and converting numerical data of different dimensions into standard scales to eliminate the influence of dimensions. As a further explanation of this step, this embodiment can use the 3σ principle to identify outliers in data such as wind speed and vibration (e.g., deviations from the mean by more than 3 standard deviations), and remove impulse noise through sliding median filtering; missing data is repaired using linear interpolation between adjacent time points, and data segments with consecutive missing values ​​exceeding 2 hours are marked as invalid; simultaneously, for data from different sources (e.g., 10-second sampling from meteorological buoys, 1kHz sampling from wind turbine vibration), a fixed time window resampling is used, and the timestamps are aligned using the nearest neighbor interpolation method to generate a uniform time series with equal intervals; in addition, for data of different dimensions such as wind speed and temperature, Min-Max standardization is used to transform them to the [0,1] interval, as shown in the formula: ; in, The original monitoring data is to be standardized; This represents the lower physical boundary of this data type. This represents the physical boundary of the data type. This is the standardized data.

[0048] S2. Based on the standardized dataset, construct a dynamic model of the atmospheric and marine environmental field and a model of equipment status evolution. Couple the two models through a multi-scale spatiotemporal fusion algorithm to generate prediction results of environmental change trends and equipment operating status. Dynamically correct the model parameters according to the prediction error.

[0049] This embodiment is based on the principles of fluid dynamics, and combines the topography, coastline features and historical meteorological data of the wind farm area to construct a spatiotemporal distribution model of wind speed field, wave field and ocean current field, and constructs an equipment health status evolution model based on the physical parameters and historical operating data of the wind farm equipment.

[0050] As a further explanation of this step, the construction of the atmospheric and marine environmental dynamic model in this embodiment is based on fluid dynamics theory, and a multi-model coupling strategy is used to simulate the spatiotemporal distribution of the wind farm area, specifically including: Wind speed field modeling: By integrating the large-scale wind field output of the WRF meteorological model and combining computational fluid dynamics (CFD) methods, a three-dimensional wind speed shear model is constructed for the wind farm terrain (such as coastline tortuosity and wind turbine density). The spatiotemporal changes of local wind speed and direction are analyzed by meshing (such as 500m×500m resolution). Wave field modeling: Based on the SWAN model (SimulatingWavesNearshore), wind speed field, water depth topography and tidal data are input, wave action equations are solved, and the evolution law of parameters such as effective wave height and wave direction is output. Ocean current field modeling: Based on FVCOM (FiniteVolumeCommunityOceanModel), considering tidal forces, wind-driven currents and seabed topography constraints, the ocean fluid motion equations are numerically solved to characterize the spatiotemporal distribution characteristics of surface and bottom ocean current velocities.

[0051] Furthermore, taking a wind turbine gearbox as an example, health status prediction is achieved through vibration spectrum analysis and degradation modeling: Signal acquisition and processing: High-frequency vibration signals (sampling rate ≥10kHz) are collected using vibration sensors deployed in the bearing housing (such as IEPE type), converted into frequency domain data by Fast Fourier Transform (FFT), and fault characteristic frequencies are extracted (such as the inner ring fault frequency can be derived from parameters such as the number of rolling elements and rotational speed); Feature-life mapping: Based on historical fault data, the correlation between the characteristic frequency amplitude and the remaining life is fitted, and an exponential degradation model is used to quantitatively describe the evolution process of the equipment's health status.

[0052] In this embodiment, a multi-scale spatiotemporal fusion algorithm is used to couple and train two types of models to generate prediction results of environmental change trends and equipment operating status, including: S21. Synchronize the standardized dataset by timestamp and use a sliding time window to resample and generate a unified time series.

[0053] Atmospheric and marine environmental data are synchronized with wind farm equipment operation status data by timestamp, and sliding time window resampling is used to generate a unified sequence with equal time intervals. ,like This aligns data streams of different frequencies along the time dimension, where: Wind speed; Wave height; The amplitude of the vibration; For temperature.

[0054] S22. Perform a 3-level Daubechies-4 wavelet decomposition on the unified time series to obtain the low-frequency approximate components. With high-frequency detail components Calculate the energy percentage of each high-frequency component. .

[0055] ; Calculate the energy percentage at each scale : ; in, For the first The energy percentage of each component.

[0056] S23. Generate a grid covering the offshore wind farm area, and construct a Gaussian semivariogram based on the Kriging interpolation method. Combine the location of each monitoring point and the collected atmospheric and marine environmental data to generate the environmental field matrix of each grid cell.

[0057] Based on the Kriging interpolation method, spatial interpolation is performed on environmental field data to construct a Gaussian semivariogram. : ; in, The Euclidean distance between monitoring points; For the nugget effect; This is the base value; For variable range;

[0058] Weights are determined by solving the Kriging equations. Generate continuous spatial features The output environmental field matrix has a grid resolution of 500 meters. ; in, For monitoring points At any moment The observed values.

[0059] S24. The splicing of time decomposition components and the environmental field matrix forms a spatiotemporal feature tensor, which is then combined with the energy proportion. Input spatiotemporal convolutional neural network, using The model is trained using a weighted mixed loss function, and the prediction results are output.

[0060] Low-frequency approximation components of time decomposition High-frequency detail components , , Environmental field matrix reconstructed with space spliced ​​into a spatiotemporal feature tensor : ;

[0061] A spatiotemporal convolutional neural network is constructed, consisting of spatiotemporal convolutional layers, long short-term memory layers, and an output layer. The output layer performs a linear transformation to generate the prediction results. Predicted duration Configuration based on the wind farm operation and maintenance decision-making cycle; among which, To predict wind speed; To predict wave height; To predict vibration amplitude; To predict temperature; The spatiotemporal convolutional neural network is trained using the Adam optimizer to adaptively update parameters, and the loss function is... The batch size and number of training epochs, among other hyperparameters, are adaptively adjusted based on the data scale and hardware performance. Let the mean squared error loss function be used. This is the mean absolute error loss function.

[0062] This invention compares the prediction results with the actual monitoring data, and triggers the model parameter optimization process when the prediction deviation exceeds a preset threshold for multiple consecutive periods.

[0063] As a further explanation of this step, the reliability of the model can be ensured in this embodiment through the following quantitative threshold judgment and process optimization: On the one hand, based on engineering experience in offshore wind power operation and maintenance, the following verification indicators are set: Wind speed prediction: If the average relative error of 5 consecutive prediction periods (e.g., 1 hour / period) exceeds 15%, the model is deemed to need optimization; Wave height prediction: If the root mean square error (RMSE) exceeds the reasonable deviation range of the wave height in the engineering design (e.g., 0.5 meters), parameter adjustment is triggered; On the other hand, when the model deviation is detected to exceed the limit, the closed-loop optimization process is initiated: Scenario-based data preparation: Extract historical monitoring data from the most recent 72 hours (covering normal operating conditions, extreme weather, and other scenarios) to construct a training set—this duration can fully cover a typical weather process (such as a typhoon), ensuring data compatibility with complex scenarios; Intelligent parameter tuning: Using a Bayesian optimization algorithm, key parameters such as the WRF wind shear index and the SWAN spectral attenuation coefficient are iteratively adjusted to minimize prediction error and achieve dynamic updating of model parameters. Dynamic verification and update: The optimized model is verified by 12 hours of real-time monitoring data. Once the verification is successful, the online running model is replaced, forming a closed-loop control of "prediction-verification-optimization" to continuously ensure the accuracy of the model.

[0064] The following example illustrates the parameter optimization process: Real-time computing continuous The average relative error of each time window : ; in, For the first The center moment of a time window; For a moment The actual monitoring value; For a moment The model's predicted values;

[0065] when When the error rate exceeds 15%, initiate online correction: Extract historical data from the most recent few hours (typically 72 hours, which can be dynamically adjusted according to the wind farm environment) to build a model update training set. Freeze the kernel parameters of the front-end layers of the spatiotemporal convolutional neural network (typically the first 3 layers, used to extract stable spatiotemporal features) and only fine-tune the weights of the last fully connected layer. Reduce the learning rate to a low level (typically 0.0005) and train for several epochs (typically 20 epochs, until the loss function converges). Calculate the parameter update amount through backpropagation. With new model parameters The original parameters are replaced to achieve dynamic correction of prediction bias, whereby... These are the parameters for the old model.

[0066] As a further explanation of this step, firstly, based on the characteristic of offshore wind power data that "transient changes (such as gusts, gearbox impacts) coexist with trend evolution (such as tidal cycles, equipment temperature drift)," this embodiment uses a dual sliding window + wavelet decomposition strategy to extract multi-scale temporal features: a 10-minute short window (5-minute step size, 50% overlap) is set to capture second-level gusts, instantaneous equipment vibrations, and other sudden changes; a 60-minute long window (30-minute step size, 50% overlap) is used to fit the daily variation of wind speed and the gradual trend of temperature change, and the overlap rate ensures the continuity of time series; the long window data is decomposed into three levels using the Daubechies-4 wavelet basis, with low-frequency components characterizing trend patterns (such as hourly wind speed decay and equipment temperature drift), and high-frequency components analyzing transient fluctuations (such as typhoon pulsating winds and abnormal bearing vibrations), naturally adapting to the multi-timescale characteristics of the environment and equipment data.

[0067] Then, considering the spatial characteristics of wind farms, namely the heterogeneity of nearshore and offshore environments and the differences in equipment distribution density, a spatial modeling scheme of double-layer nested grid + Kriging interpolation is constructed: the 2km×2km outer grid covers the entire wind farm area, depicting the spatial distribution of regional wind fields and large-scale ocean currents; the 500m×500m inner grid focuses on key equipment areas such as wind turbine clusters and cable joints, analyzing the fine changes in local wave height and vibration; the monitoring data of different grids (such as wind speed in the outer area and wave height near the wind turbines in the inner area) are optimized for spatial continuity through Kriging interpolation, and then multi-scale spatial information is integrated through feature stitching.

[0068] Next, this algorithm introduces a dual-stream attention architecture to achieve dynamic weighted fusion of spatiotemporal features: the temporal stream learns the temporal dependencies such as hourly wind speed evolution and equipment vibration attenuation through an LSTM network, while the spatial stream extracts spatial correlations such as wave height region propagation and vibration coupling between wind turbines through 2D convolution; a spatiotemporal attention module is designed to automatically assign high weights to "times of sudden change (such as typhoon landfall, precursors to equipment failure)" and "critical areas (such as densely populated wind turbine areas, cable bends)" to enhance the model's ability to characterize complex working conditions; the fused features are mapped to a prediction sequence of "wind speed, wave height (environmental dimension) + vibration, temperature (equipment dimension)" through a fully connected layer to ensure the spatiotemporal consistency of multi-parameter predictions and directly support the environmental dynamic simulation and equipment health extrapolation of the model building module 220.

[0069] Finally, this embodiment constructs an error feedback-lightweight optimization closed loop: when the average relative error of wind speed exceeds 15% and the wave height RMSE exceeds the reasonable range of wave height in engineering design, the correction process is automatically initiated; the training set is constructed by extracting the historical data of the most recent 72 hours (covering normal and extreme conditions, fully supporting the analysis of a typical weather process), freezing the core parameters of temporal flow LSTM and spatial flow convolution, and only updating the spatiotemporal attention weights and fusion layer parameters—this strategy balances computational efficiency and model adaptability, meeting the real-time calculation requirements of offshore wind power; the corrected model is verified by 12 hours of real-time data, and after passing the verification, the online running model is replaced, forming a continuous iteration of "prediction-verification-optimization", thereby ensuring the dynamic adaptation of the algorithm to changing sea conditions.

[0070] S3. Based on the predicted results of environmental change trends, the predicted results of equipment operation status, and the pre-stored operation and maintenance rules, the equipment fault warning and environmental risks are quantitatively classified and graded. The real-time operation and maintenance resource status is matched and scheduled. A collaborative operation and maintenance plan is generated through a spatiotemporal optimization algorithm. The collaborative operation and maintenance plan is executed and the execution status is fed back in real time to realize the iterative optimization of the operation and maintenance plan.

[0071] Based on the prediction results of environmental change trends and equipment operating status, this embodiment extracts two core tasks: equipment operation and maintenance tasks and environmental response tasks, in combination with pre-stored preset operation and maintenance rules.

[0072] Furthermore, equipment operation and maintenance tasks include: for equipment fault warnings such as abnormal vibration and excessive temperature, marking them as urgent, important, or general levels according to the degree of impact of the fault;

[0073] Environmental response tasks: For environmental risks such as strong winds and high waves, they are marked as short-term and medium-term response tasks according to the duration of the risk;

[0074] As a further explanation of this step, this embodiment achieves quantitative classification of faults and risks through multi-dimensional feature fusion: equipment fault early warning integrates parameters such as fault location (core equipment / auxiliary equipment) and development rate (daily vibration increase, temperature rise slope), and constructs a hierarchical judgment model using the industry-standard fault tree analysis (FTA) method; environmental risk combines prediction duration and risk intensity (such as whether the wind speed exceeds the operational safety threshold), and matches the response level through a risk matrix.

[0075] This embodiment relies on a structured resource database to achieve dynamic matching: the personnel skill database stores qualification types (electrical / mechanical maintenance, high-altitude operations, etc.), real-time location (obtained via GPS positioning), and task status (idle / busy), prioritizing the allocation of fully skilled personnel within 10 kilometers of the target wind turbine; the equipment resource database plans operation paths according to the ship's wave resistance level (distinguishing between nearshore / offshore operations) and the drone's wind resistance level, avoiding areas with excessive water depth and prohibited navigation areas; the material inventory management sets safety thresholds based on historical consumption patterns, and automatically initiates a replenishment process to the port logistics center when the inventory of spare parts such as gearbox bearings falls below the threshold.

[0076] This embodiment uses the real-time operation and maintenance resource status data (personnel location, equipment availability, and material inventory) obtained in S1 to match and schedule personnel, equipment, and materials for equipment operation and maintenance tasks and environmental response tasks; based on the time window prediction results determined by the environmental change trend prediction results in S2, a collaborative operation and maintenance plan including inspection, maintenance, and resource allocation is generated according to task priority, environmental risk period, and wind turbine spatial distribution.

[0077] As a further explanation of this step, this embodiment uses a spatiotemporal collaborative genetic algorithm to arrange the operation and maintenance sequence, with the "task-resource-time" triple as the chromosome code. The fitness function weights and considers task priority (emergency task weight is not less than 0.6), environmental risk period avoidance (such as operation wind speed ≤15m / s), and wind turbine spatial clustering (batch processing of tasks in the same area to reduce the cost of ship round trip).

[0078] The spatiotemporal collaborative genetic algorithm iteratively optimizes the operation through roulette wheel selection, two-point crossover, and random mutation. When a wave height exceeding 2m (based on wind turbine operation design parameters) or a resource allocation conflict is detected during the operation period, the crossover rate is automatically adjusted to 0.8 and the mutation rate to 0.1, driving the plan to be rearranged. The validated plan generates a structured file containing time nodes, resource lists, and path coordinates, which is synchronized to the control execution unit 400.

[0079] In this step, generating a collaborative operation and maintenance plan through an optimization algorithm includes the following steps: S31. Based on environmental parameter prediction information and equipment status prediction data, combined with pre-stored operation and maintenance rules, a multi-dimensional feature fusion algorithm is used to mark equipment fault warnings as emergency level, important level, and general level according to the fault impact degree quantitative model, and to mark environmental risks as short-term and medium-term response tasks according to the risk duration statistical algorithm. As a further explanation of this step, in this embodiment, based on pre-stored operation and maintenance rules (such as the GB / T36547 industry standard), task classification is achieved through a multi-dimensional feature fusion algorithm: First, extract the feature vector of the degree of impact of the fault. ;in, is the weight of the fault location; is the parameter deviation degree; is the fault development rate; where the parameter deviation degree is determined based on the difference between the predicted result of the equipment operation state and the equipment operation state threshold corresponding to the predicted result of the environmental change trend, and the fault development rate is the development rate after the predicted result of the equipment operation state deviates from the threshold; The fault level value is calculated using a weighted summation formula ; where, , , are the weight coefficients, which can be set based on the experience of "core equipment priority" and "fault development trend" in offshore wind power operation and maintenance. When 0.8 is marked as the emergency level, 0.5 R 0.8 is the important level, R [[ID=**********]] 0.5 is the general level; Then, based on the duration of deviation from the safety threshold in the predicted result of the environmental change trend, the environmental risk level is determined. The output risk duration T and intensity S (such as the duration of wind speed exceeding 15 m / s, wave height value), are marked according to the following logic: If T ≤ 6 hours and S is lower than the operation safety threshold (such as wave height ≤ 2 m), it is marked as a short-term response task; if 6 hours < T < 24 hours, or S is close to the safety threshold, it is marked as a medium-term response task.

[0080] S32. Call the real-time operation and maintenance resource status data for resource matching; The resource matching specifically includes: Screen suitable personnel according to the task skill requirements and the personnel skill library (qualifications, locations, task status, etc.); Plan the operation path (considering channel restrictions such as water depth) according to the task geographical coordinates and available equipment resources such as ships and drones; Check the inventory of spare parts and consumables, and trigger the replenishment process for shortage materials.

[0081] S33. Integrate the time window prediction result and the fan geographical coordinate clustering information, and use the spatio-temporal collaborative genetic algorithm to arrange the operation and maintenance plan according to the task priority, environmental risk period, and regional task load; As a further explanation of this step, integrate the time window prediction result of the dynamic simulation unit (such as the environmental safety window for each hour in the next 24 hours) and the fan geographical coordinate clustering information (divide the fans into clusters such as Cluster1 and Cluster2 with a radius of 5 km), and arrange the plan through the genetic algorithm: First, map the tasks, resources, and time into gene sequences; Then, define the fitness function ;in, Score the task priority; For time window matching degree; For spatial clustering efficiency; , , These are the weighting coefficients; Finally, the optimal plan sequence is output after 100 iterations using roulette wheel selection, two-point crossover (crossover rate 0.7), and mutation (mutation rate 0.1).

[0082] S34. Adaptive feedback verification algorithm is used to verify the environmental risks and execution efficiency of the plan. If the environmental risks or execution efficiency do not meet the requirements, return to S33 for re-arrangement. If the verification is successful, a structured operation and maintenance plan for compatible control and execution units is generated and output.

[0083] Furthermore, this embodiment relies on a multi-dimensional feature fusion algorithm to classify equipment failures and environmental risks into tasks. Then, real-time resource data is invoked to match tasks with resources based on personnel skill adaptation, equipment path planning, and material inventory verification. Next, time window prediction results and wind turbine geographic clustering information are integrated, and an operation and maintenance plan is arranged using a spatiotemporal collaborative genetic algorithm. Then, an adaptive feedback verification algorithm is used to perform closed-loop verification of the plan's environmental risk compatibility and execution efficiency. If the verification fails, the process returns to the plan arrangement stage for iterative optimization. Finally, the verified plan is generated as structured data for a compatible control execution unit and output. In the above process, the weights, thresholds, and other parameters involved in the algorithm are all set based on conventional engineering standards in the field of offshore wind power operation and maintenance.

[0084] In this embodiment, the collaborative operation and maintenance plan generated by the collaborative operation and maintenance method for offshore wind farms is executed by the control and execution unit. The control and execution unit automatically adjusts the operating parameters of the wind farm equipment according to the collaborative operation and maintenance plan, and coordinates the operation and maintenance vessels, drones and robots to perform operation and maintenance tasks through the wireless communication network.

[0085] In this step, the control execution unit includes an instruction processing module, an equipment control module, an operation and maintenance resource coordination module, and a status feedback module, wherein: The instruction processing module is used to receive collaborative operation and maintenance plans and parse equipment parameter adjustment instructions and operation and maintenance resource operation instructions. After receiving the collaborative operation and maintenance plan, the instruction processing module decomposes the equipment parameter adjustment instructions (such as wind turbine speed, converter power factor, and Modbus control protocol adapted to the wind farm SCADA system) and operation and maintenance resource operation instructions (including task type, target coordinates, and time window constraints), and synchronously distributes them to the equipment control module and the operation and maintenance resource coordination module.

[0086] The equipment control module automatically adjusts the operating parameters of the wind farm equipment and executes equipment fault early warning related tasks based on the parsed equipment parameter adjustment instructions and operation and maintenance resource operation instructions. As a further explanation of this step, the equipment control module in this embodiment relies on the existing SCADA system interface of the wind farm to realize parameter adjustment and fault response: Automatic parameter adjustment: For wind turbines, the speed is adjusted through the pitch angle control system (e.g., reducing the speed to a safe range as instructed); for converters, the reactive power compensation value output by the power module is adjusted; for transformers, the tap changer motor is driven to switch gears—all operations are completed through the “control-feedback” channel of the SCADA system, and new parameters are transmitted back in real time after adjustment (e.g., the current wind turbine speed is 12.1 rpm, and the converter power factor is 0.98).

[0087] Fault warning response: When an "emergency-level fault repair" instruction is received (such as gearbox vibration exceeding limits), a local audible and visual alarm is triggered and a work order is pushed to the maintenance terminal. At the same time, the remote control permissions of the faulty equipment are automatically locked (only emergency repair operations are allowed) to prevent misoperation from escalating the fault.

[0088] The operation and maintenance resource coordination module uses a wireless communication network to schedule operation and maintenance vessels, drones and robots, and execute operation and maintenance tasks according to instructions. The status feedback module is used to collect the results of equipment parameter adjustment and the execution status of operation and maintenance tasks, and feed the data back to step S3.

[0089] In this step, the operation and maintenance resource coordination module includes a resource matching submodule, a path planning submodule, an instruction issuance submodule, and a status monitoring submodule, wherein: The resource matching submodule receives the operation and maintenance task instructions output by the instruction processing module, filters suitable operation and maintenance vessels, drones and robots based on real-time resource status information, and completes the initial allocation of tasks and resources. The path planning submodule, in conjunction with the sea state data provided in step S2, plans the operation path for the scheduled operation and maintenance resources; when path conflicts occur, the path is adjusted according to the task priority. The instruction sending submodule sends operation instructions to maintenance vessels, drones, and robots via a wireless communication network and confirms the instruction reception status. The status monitoring submodule collects the operating status of maintenance resources in real time, and when an abnormal situation is detected, it feeds back information to step S3.

[0090] As a further explanation of this step, the collaborative process of the operation and maintenance resource coordination module in this embodiment is as follows: First, it receives maintenance task instructions (including task type, target coordinates, and time window) distributed by the instruction processing module; then, it calls the real-time resource status pool (which stores the location, battery life, and task status of maintenance vessels, drones, and robots) and filters suitable resources according to the following rules: Maintenance vessels: Prioritize vessels with a wave resistance rating ≥ 200 predicted wave height from the dynamic simulation unit (e.g., if the wave height is 1.5m, select vessels with a wave resistance rating of 2m) and idle vessels ≤ 15km from the target wind turbine. Drones: Select models with a battery life redundancy of ≥20% (to offset the impact of wind speed on flight time) and carry the necessary sensors for the mission (such as infrared thermal imagers for fault inspection). Maintenance robots: Matching functionally compatible models (such as climbing robots for tower maintenance, and underwater robots for inspecting cable joints).

[0091] After the initial allocation is completed, the resource status is marked as "allocated", and the resource status pool is updated synchronously. Next, the resource matching results and task parameters are received, and combined with the real-time sea state data (wave height, water depth, current direction) and electronic nautical charts from the dynamic simulation unit 200, the operation path is planned using the A* algorithm: Basic obstacle avoidance: Forcefully avoid shallow waters with a depth of less than 5m, restricted waterways (such as military control zones), and areas with dense wind turbine foundations; Conflict handling: If multiple task paths conflict, they will be adjusted according to task priority (urgent > important > general), and the path of the lower priority task will be postponed or detoured. Dynamic updates: Receives sea state updates (such as sudden changes in wave height and current changes) from the dynamic simulation unit every 10 minutes, triggering path replanning to ensure operational safety.

[0092] Then, based on the planned path and task steps, standardized work instructions (including coordinates, actions, and time constraints) are generated and issued via multi-protocol communication: Maintenance vessels: Commands are sent via AIS+4G network and require an ACK confirmation within 10 seconds of receipt; if no response is received within 1 minute, satellite communication is automatically switched to resend the command. Drones: Send flight commands (take-off and landing points, inspection routes) via a 5.8G private network wireless link and simultaneously receive real-time image transmission data; Maintenance robot: Sends action commands (such as climbing speed, tool start / stop) via wired / Bluetooth channels and transmits torque, position and other status data back in real time; Finally, the operational status of maintenance resources is collected in real time and categorized into "normal status + abnormal warning": Normal state: Collect the ship's GPS position (error ≤10m, in line with the normal accuracy of maritime positioning) every 30 seconds, the drone's battery level (low battery warning triggered when 20% remaining), and the robot's operation progress (such as climbing height and tool usage time), and upload them in a package; Anomaly warning: If a communication interruption is detected (no response for 1 minute) or equipment failure (such as ship engine alarm or drone attitude loss), remote reset should be attempted first (such as restarting drone flight control or switching robot working mode); if the repair fails, the task is marked as interrupted and real-time feedback is sent to step S3 to trigger the plan iteration.

[0093] Example 2: This example is a collaborative operation and maintenance device for offshore wind farms based on dynamic simulation of the atmospheric and marine environment, including: The data acquisition module is used to collect atmospheric and marine environmental data, wind farm equipment operation status data, and operation and maintenance resource status data of offshore wind farm areas through multiple types of monitoring points. It performs clock synchronization, coordinate transformation, anomaly removal, missing value repair, and dimensional standardization on the collected multi-source heterogeneous data to generate a standardized dataset in a unified format. The dynamic simulation module is used to construct a dynamic model of the atmospheric and marine environmental field and a model of equipment status evolution based on the standardized dataset. The two models are coupled and trained through a multi-scale spatiotemporal fusion algorithm to generate prediction results of environmental change trends and equipment operating status. The model parameters are dynamically corrected according to the prediction error. The operation and maintenance decision module is used to quantify and classify equipment fault warnings and environmental risks based on the predicted results of environmental change trends, equipment operating status, and pre-stored operation and maintenance rules. It also performs matching and scheduling in combination with real-time operation and maintenance resource status, generates collaborative operation and maintenance plans through spatiotemporal optimization algorithms, executes the collaborative operation and maintenance plans, and provides real-time feedback on the execution status to achieve iterative optimization of the operation and maintenance plans.

[0094] Example 3: This example is a storage medium that stores a computer program that can be executed by a processor. When the computer program is executed, it implements the steps of the collaborative operation and maintenance method for offshore wind farms described in Example 1.

[0095] Example 4: This example is a collaborative operation and maintenance device for offshore wind farms, which has a memory and a processor. The memory stores a computer program that can be executed by the processor. When the computer program is executed, it implements the steps of the collaborative operation and maintenance method for offshore wind farms described in Example 1.

[0096] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0097] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-including system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0098] More specific examples (a non-exhaustive list) of computer-readable media include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the aforementioned program can be printed, because the aforementioned program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0099] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0100] In the foregoing description of this specification, references to terms such as "one embodiment," "another embodiment," or "some embodiments" indicate that a specific feature, structure, material, or characteristic described in connection with an embodiment or example is included in at least one embodiment or example of the present invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0101] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

[0102] The above is a detailed description of the preferred embodiments of the present invention. However, the present invention is not limited to the above embodiments. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of the present invention. All such equivalent modifications or substitutions are included within the scope defined by the claims of this application.

Claims

1. A collaborative operation and maintenance method for offshore wind farms based on dynamic simulation of the atmospheric and marine environment, characterized in that, include: S1. Collect atmospheric and marine environmental data, wind farm equipment operation status data, and operation and maintenance resource status data of offshore wind farm areas through multiple types of monitoring points. Perform clock synchronization, coordinate transformation, anomaly removal, missing value repair, and dimensional standardization on the collected multi-source heterogeneous data to generate a standardized dataset in a unified format. S2. Based on the standardized dataset, construct a dynamic model of the atmospheric and marine environmental field and a model of equipment status evolution. Couple the two models through a multi-scale spatiotemporal fusion algorithm to generate prediction results of environmental change trends and equipment operating status. Dynamically correct the model parameters according to the prediction error. S3. Based on the predicted results of environmental change trends, the predicted results of equipment operation status, and the pre-stored operation and maintenance rules, the equipment fault warning and environmental risks are quantitatively classified and graded. The real-time operation and maintenance resource status is matched and scheduled. A collaborative operation and maintenance plan is generated through a spatiotemporal optimization algorithm. The collaborative operation and maintenance plan is executed and the execution status is fed back in real time to realize the iterative optimization of the operation and maintenance plan.

2. The method for collaborative operation and maintenance of offshore wind farms based on dynamic simulation of the atmospheric and marine environment according to claim 1, characterized in that: The atmospheric and marine environmental data includes wind field parameters, wave field parameters, and flow field parameters, which are collected collaboratively by monitoring equipment deployed on the sea surface, underwater observation platforms on the seabed, and onshore wind measurement towers. The equipment operating status data includes the wind turbine's rotational speed, vibration amplitude, temperature, and the insulation resistance of the submarine cable, which are collected by sensors deployed in key parts of the equipment. The operation and maintenance resource status data includes the location of operation and maintenance personnel, equipment availability, and material inventory, which are collected through GPS positioning devices and inventory management systems.

3. The method for collaborative operation and maintenance of offshore wind farms based on dynamic simulation of the atmospheric and marine environment according to claim 1, characterized in that, In step S1, clock synchronization is achieved through a satellite time synchronization system to control timestamp errors; coordinate transformation converts the latitude and longitude of the monitoring point into planar coordinates through the UTM projected coordinate system.

4. The method for collaborative operation and maintenance of offshore wind farms based on dynamic simulation of the atmospheric and marine environment according to claim 1, characterized in that, In step S1, the anomaly removal uses the 3σ principle to identify abnormal data points and removes impulse noise through sliding median filtering; the missing value repair uses linear interpolation, and data segments that are continuously missing for more than 2 hours are marked as invalid; the dimensional standardization uses Min-Max standardization to transform the data to the [0,1] interval.

5. The method for collaborative operation and maintenance of offshore wind farms based on dynamic simulation of the atmospheric and marine environment according to claim 1, characterized in that, In step S2, the atmospheric and oceanic environmental field dynamic model includes a wind speed field model, a wave field model, and an ocean current field model. The wind speed field model is constructed by integrating the WRF meteorological model and the CFD method, the wave field model is constructed based on the SWAN model, and the ocean current field model is constructed based on the FVCOM model. The equipment state evolution model extracts the equipment fault characteristic frequency and combines it with the exponential degradation model to quantify the evolution law of equipment health status.

6. The method for collaborative operation and maintenance of offshore wind farms based on dynamic simulation of the atmospheric and marine environment according to claim 1, characterized in that, In step S2, the multi-scale spatiotemporal fusion algorithm includes the following sub-steps: S21. Synchronize the standardized dataset by timestamp and resample using a sliding time window to generate a unified time series; S22. Perform a 3-level Daubechies-4 wavelet decomposition on the unified time series to obtain the low-frequency approximate components. With high-frequency detail components Calculate the energy percentage of each high-frequency component. ; S23. Generate a grid covering the offshore wind farm area, and construct a Gaussian semivariogram based on the Kriging interpolation method. Combine the location of each monitoring point and the collected atmospheric and marine environmental data to generate the environmental field matrix of each grid cell. S24. The splicing of time decomposition components and the environmental field matrix forms a spatiotemporal feature tensor, which is then combined with the energy proportion. Input spatiotemporal convolutional neural network, using The model is trained using a weighted mixed loss function, and the prediction results are output.

7. The method for collaborative operation and maintenance of offshore wind farms based on dynamic simulation of the atmospheric and marine environment according to claim 6, characterized in that, In step S2, the prediction error is obtained through continuous... The average relative error of each time window Quantification; when When the value exceeds the preset value, extract historical data within the most recent preset time period to construct an updated training set, freeze the parameters of the front 3 layers of the spatiotemporal convolutional neural network, fine-tune the weights of the last fully connected layer, and dynamically correct the model parameters.

8. The method for collaborative operation and maintenance of offshore wind farms based on dynamic simulation of the atmospheric and marine environment according to claim 1, characterized in that, In step S3, the quantitative classification of equipment fault warnings and environmental risks based on the environmental change trend prediction results, equipment operating status prediction results, and pre-stored operation and maintenance rules includes: Construct a device fault feature vector, including fault location weight, parameter deviation, and fault development rate. The parameter deviation is determined based on the difference between the device operating status prediction result and the device operating status threshold corresponding to the environmental change trend prediction result. The fault development rate is the development rate of the device operating status prediction result after it deviates from the threshold. Based on the equipment fault feature vector, the fault level value is calculated, and the equipment fault warning level is determined based on the fault level value. The environmental risk level is determined based on the duration of deviation from the safe threshold in the environmental change trend prediction results.

9. The method for collaborative operation and maintenance of offshore wind farms based on dynamic simulation of the atmospheric and marine environment according to claim 1, characterized in that, In step S3, the resource matching includes: selecting suitable maintenance personnel based on task skill requirements, planning the operation paths of maintenance vessels and drones based on environmental prediction results, checking material inventory, and triggering the shortage material replenishment process.

10. The method for collaborative operation and maintenance of offshore wind farms based on dynamic simulation of the atmospheric and marine environment according to claim 1, characterized in that, In step S3, the spatiotemporal optimization algorithm is a spatiotemporal cooperative genetic algorithm, which uses "task-resource-time" as chromosome encoding, and the fitness function weightedly considers task priority, environmental risk period and spatial clustering efficiency.

11. A collaborative operation and maintenance device for offshore wind farms based on dynamic simulation of the atmospheric and marine environment, characterized in that, include: The data acquisition module is used to collect atmospheric and marine environmental data, wind farm equipment operation status data, and operation and maintenance resource status data of offshore wind farm areas through multiple types of monitoring points. It performs clock synchronization, coordinate transformation, anomaly removal, missing value repair, and dimensional standardization on the collected multi-source heterogeneous data to generate a standardized dataset in a unified format. The dynamic simulation module is used to construct a dynamic model of the atmospheric and marine environmental field and a model of equipment status evolution based on the standardized dataset. The two models are coupled and trained through a multi-scale spatiotemporal fusion algorithm to generate prediction results of environmental change trends and equipment operating status. The model parameters are dynamically corrected according to the prediction error. The operation and maintenance decision module is used to quantify and classify equipment fault warnings and environmental risks based on the predicted results of environmental change trends, equipment operating status, and pre-stored operation and maintenance rules. It also performs matching and scheduling in combination with real-time operation and maintenance resource status, generates collaborative operation and maintenance plans through spatiotemporal optimization algorithms, executes the collaborative operation and maintenance plans, and provides real-time feedback on the execution status to achieve iterative optimization of the operation and maintenance plans.

12. A storage medium having a computer program stored thereon that can be executed by a processor, characterized in that, When the computer program is executed, it implements the steps of the collaborative operation and maintenance method for offshore wind farms as described in any one of claims 1 to 10.

13. A collaborative operation and maintenance device for offshore wind farms, comprising a memory and a processor, wherein the memory stores a computer program executable by the processor, characterized in that, When the computer program is executed, it implements the steps of the collaborative operation and maintenance method for offshore wind farms as described in any one of claims 1 to 10.

Images