Rapid judgment method for offshore construction operation and maintenance operation window period based on deep learning
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- POWERCHINA HUADONG ENG CORP LTD
- Filing Date
- 2025-12-01
- Publication Date
- 2026-04-28
AI Technical Summary
Existing technologies make it difficult to quickly and accurately determine the window of opportunity for offshore construction and maintenance operations, leading to increased safety risks, decreased equipment utilization, and project delays.
Based on deep learning, a wind, wave and current database is constructed, a target parameter prediction model is trained, and the operation window is determined by combining the ship type and operation content.
It enables rapid and accurate assessment of offshore construction and maintenance operation windows, reducing safety risks and improving equipment utilization and operational efficiency.
Smart Images
Figure CN121936705A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to a method for rapidly determining the operational window for offshore construction and maintenance based on deep learning. It is applicable to the field of offshore construction and maintenance operations. Background Technology
[0002] With the rapid development of offshore wind power and the continuous growth of installed capacity, there is a significant demand for accurate judgment of the "operational window" for wind farm construction and maintenance. The spatiotemporal variations of wind, waves, and currents in the sea area are significant, and the region is frequently affected by extreme events such as typhoons. At the same time, due to the scarcity of measured data, insufficient site layout, and poor consistency of multi-source data, traditional judgment methods relying on experience or low-frequency, low-spatiotemporal resolution numerical model forecasts are unable to respond promptly to rapid environmental changes. This directly increases safety risks and standby costs, leading to decreased equipment utilization, project delays, and even contract disputes.
[0003] Furthermore, different types of work vessels exhibit varying sensitivities to marine environments (especially wind, wave, and current parameters), resulting in different upper limits for operational thresholds. Therefore, to ensure the smooth operation of construction and maintenance work, it is urgent to establish a work window forecasting method tailored to specific vessel types and work procedures, possessing higher spatiotemporal resolution, computational efficiency, and update frequency. Summary of the Invention
[0004] The technical problem to be solved by the present invention is to provide a method for rapid determination of the window period for offshore construction and maintenance operations based on deep learning, in order to address the above-mentioned problems.
[0005] The technical solution adopted in this invention is: a method for rapidly determining the operational window period for offshore construction and maintenance based on deep learning, comprising: Based on the requirements of each offshore construction and maintenance operation procedure, the target parameter combination corresponding to each procedure is determined from the wind field, wave field, and current field parameters; Based on the wind, wave and current database, relevant feature parameters that are correlated with the target parameter combination are selected. The wind, wave and current database includes latitude and longitude data of sample points in the engineering sea area, as well as time series data of the atmosphere, wave field and current field at the sample points. A target parameter prediction model is trained based on the time series data corresponding to the combination of target parameters and related feature parameters in the wind, wave and current database. Based on the current work process, obtain the time series data of the relevant feature parameters of the process within a preset time period starting from the current time. Input the time series data of relevant feature parameters into the target parameter prediction model corresponding to the current work procedure, and output the time series data of the target parameter combination within the forecast time period; Based on the type of vessel and the content of the current operation, determine the environmental threshold corresponding to the combination of operation duration and target parameters; Based on the time series data of target parameter combinations within the forecast period, combined with operation duration and environmental thresholds, the operation window period within the forecast period is determined.
[0006] The construction of the wind and wave current database includes: Based on the offshore construction and maintenance operations, the latitude and longitude range of the sea area of interest is determined, and historical atmospheric data, wave field data and current field data of the sea area of interest during the same period are obtained. A wind-wave-current coupled model was constructed using the SCHISM three-dimensional baroclinic module as the hydrodynamic model and WWMIII as the wave model. Based on historical wind field data, wave field data, and current field data of the sea area in question during the same period, the wind-wave-current coupling model was calibrated and validated. The latitude and longitude data of the calculation points included in the sea area of interest are extracted from the wind, wave and current coupled numerical model, and then the latitude and longitude data of the sample points are determined. Based on the latitude and longitude data of the sample points, time series data of wave field and flow field at the sample points within a specified time period are extracted from the longitude-calibrated and validated wind-wave-flow coupling model, and corresponding spatiotemporal atmospheric time series data are extracted from the atmospheric database.
[0007] The historical atmospheric data were obtained from an atmospheric database, which is an ERA5 reanalysis database corrected for typhoon season wind field data.
[0008] The ERA5 reanalysis database, corrected for typhoon season wind field data, includes: When a typhoon affects the sea area of concern, the ERA5 wind field is corrected using the typhoon path, typhoon central pressure, and maximum central wind speed during the typhoon's life, according to the Holland typhoon field correction algorithm.
[0009] The process of extracting latitude and longitude data of calculation points within the sea area of interest from the wind-wave-current coupled numerical model, and then determining the latitude and longitude data of sample points, includes: Based on the computational points contained in the sea area of interest, the computational points in the sea area of interest are resampled at equal intervals using the farthest point sampling on the spherical surface to obtain sample points.
[0010] The selection of relevant feature parameters based on the wind, wave, and current database that are correlated with the target parameter combination includes: Feature correlation analysis and screening based on the XGBoost algorithm.
[0011] The target parameter prediction model is built based on a machine learning algorithm applicable to time series prediction problems.
[0012] A device for rapidly determining the operational window period for offshore construction and maintenance based on deep learning, comprising: The target parameter determination module is used to determine the target parameter combination corresponding to each process based on the requirements of each offshore construction and maintenance operation process from wind field, wave field, and current field parameters; The relevant feature selection module is used to select relevant feature parameters that are correlated with the target parameter combination based on the wind, wave and current database. The wind, wave and current database includes the latitude and longitude data of sample points in the engineering sea area, as well as the time series data of the atmosphere, wave field and current field at the sample points. The prediction model training module is used to train a target parameter prediction model based on the time series data corresponding to the combination of target parameters and related feature parameters in the wind, wave and current database. The input data module is used to obtain time series data of the relevant feature parameters of the current operation within a preset time period starting from the current moment, based on the current operation procedure. The model prediction module is used to input the time series data of relevant feature parameters into the target parameter prediction model corresponding to the current work procedure, and output the time series data of the target parameter combination within the forecast period. The environmental threshold determination module is used to determine the environmental threshold corresponding to the combination of operation duration and target parameters based on the type of vessel and operation content required for the current operation procedure. The operation window period determination module is used to determine the operation window period within the forecast period based on the time series data of the target parameter combination within the forecast period, combined with the operation duration and environmental thresholds.
[0013] A storage medium storing a computer program executable by a processor, wherein the computer program, when executed, implements the steps of the deep learning-based method for rapidly determining the window period for offshore construction and maintenance operations.
[0014] A device for rapidly determining the operational window period for offshore construction and maintenance 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 deep learning-based method for rapidly determining the operational window period for offshore construction and maintenance.
[0015] The beneficial effects of this invention are as follows: Based on the refined numerical simulation results of wind, wave and current in the engineering sea area and reanalysis databases such as ERA5, this invention constructs a wind, wave and current database, and then establishes a target parameter prediction model. Thus, it can quickly predict the wind, wave and current parameters of the engineering sea area based on commonly used and publicly available weather forecast data (such as GFS forecast data). The predicted elements are quickly compared with the operational environment thresholds of different ship types to form an immediate determination of "operable / restricted".
[0016] On the one hand, this invention can identify the operational window in advance, reducing risky operations such as "not being able to work on site" and operating under critical conditions; on the other hand, it can reduce waiting and rescheduling costs, improve fleet scheduling and resource allocation efficiency, and substantially support the safety, economy and schedule controllability of offshore wind power projects. Attached Figure Description
[0017] Figure 1 The flowchart is for an example.
[0018] Figure 2 This is a technical roadmap for an example. Detailed Implementation
[0019] 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.
[0020] 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.
[0021] Example 1: As Figure 1 As shown in the figure, this embodiment is a method for rapidly determining the operational window period for offshore construction and maintenance based on deep learning, which specifically includes the following steps: S100. Based on the requirements of each offshore construction and maintenance operation procedure, determine the target parameter combination corresponding to each procedure from the wind field, wave field, and flow field parameters.
[0022] This example uses a certain process as an example. It is assumed that the target parameter combination corresponding to this process is four target parameters: wave height, wave direction, flow velocity, and flow direction.
[0023] S200. Based on the wind, wave and current database, select relevant feature parameters that are correlated with the target parameter combination.
[0024] In this embodiment, the wind, wave, and current database includes latitude and longitude data of sample points within the engineering sea area, as well as atmospheric time series data such as wind field at the sample points, and time series data of wave field and current field at the sample points.
[0025] Based on the wind, wave, and current database, characteristic parameters that significantly affect the combination of target parameters (such as wave height, wave direction, current velocity, and current direction for a specific process) are selected. These characteristic parameters include water depth, wind speed, wind direction, 2m atmospheric temperature, mean sea-level atmospheric pressure, 2m atmospheric temperature gradient, and mean sea-level atmospheric pressure gradient (where the gradient is obtained through linear interpolation of the corresponding parameters). Temporal and spatial characteristics (longitude and latitude) should also be considered. The XGBoost algorithm is used for feature correlation analysis and selection.
[0026] S300: Based on the time series data corresponding to the combination of target parameters and related feature parameters in the wind, wave and current database, train the target parameter prediction model.
[0027] The wind, wave, and flow database was divided into a validation set and a training set in a 2:8 ratio. The validation set was used to tune model hyperparameters and conduct preliminary performance evaluations, while the training set was used for model learning and training.
[0028] In this example, the target parameter prediction model is built based on machine learning algorithms suitable for time series prediction problems (such as DLinear), and the model is trained using a training set.
[0029] Unlike the recently popular Transformer-type models, the DLinear model simplifies time series analysis by decomposing it into a trend component and a residual component (long-term trend and short-term fluctuations). It then models each component using two single-layer linear networks, summing the results to output the predicted value. This approach improves the effectiveness and accuracy of time series forecasting while also offering higher efficiency and interpretability, making it arguably the most powerful model for time series prediction. Furthermore, compared to the complex Transformer model, DLinear offers faster training and inference speeds, lower computational resource consumption, and suitability for rapid forecasting, even in resource-constrained environments.
[0030] S310, Data Standardization and Sliding Window Construction
[0031] (1) Data standardization: For each input feature and each target parameter after screening, the Revin method is used to standardize the data in the time dimension, thereby obtaining the input feature tensor. (B is the batch size, L is the history lookback length, and F is the number of input features), target tensor ( (G is the target number, used to predict the step size). The Revin method has the advantage of high robustness.
[0032] (2) Sliding window construction: In order to make full use of the data, a sliding window construction was carried out hourly to obtain sample pairs ( ).
[0033] S320, DLinear model
[0034] (1) Time series decomposition: The moving average of each input channel in the time dimension is calculated. ( The size of the moving average kernel, such as (Trend) residual Then, by linearly mapping each value to the future, we obtain the target predicted value: ; in and The weights are linearly mapped. It is recommended that each channel have its own weight for robust parameter saving.
[0035] (2) Target output: First, use DLinear to independently predict the future for each input channel. Step one, then use the shared 1×1 linear header to connect the channel dimensions. By compressing to the target dimension G, cross-variable linear fusion is achieved with very few parameters.
[0036] S330, Model Training
[0037] (1) First, generate a label mask for each target and only participate in training at the position where there is a ground truth value to avoid interference from missing tests;
[0038] (2) Then RevIN standardization is used on both the input and the target to put different units on the same numerical scale to stabilize the gradient;
[0039] (3) The loss function adopts MAE or SmoothL1, and each task (or target parameter) is independently normalized by mask to ensure that the loss of each task is comparable; the overall loss is balanced by adaptive weights (such as uncertainty weighting or GradNorm) for different tasks.
[0040] (4) AdamW was used for optimization. The learning rate was reduced by cosine annealing. Appropriate batch size, gradient pruning and light regularization were used to improve convergence stability and generalization ability. Mixed precision was used during training to reduce memory usage and speed up computation. An early stopping mechanism was set to mainly verify the overall loss of the validation set.
[0041] (5) In the evaluation stage, the predicted values and true values are denormalized to physical dimensions. Finally, the key hyperparameters such as model weights, RevIN configuration, window and decomposition kernel, random seed and data time window are solidified. Then the model is saved to ensure that the results are reproducible, traceable and can be used in subsequent forecasting application stages.
[0042] S340, Model Validation and Optimization: The batch size B, historical lookback length L, and prediction step size in the above S330 process can be modified. Parameters such as weight decay, early stopping tolerance, and overall error are used to obtain the optimal model with the highest accuracy. Model performance is evaluated on a validation set (e.g., correlation coefficient R², root mean square error RMSE, mean absolute error MAE) to compare the performance of models under different parameters, select the optimal model, and prevent overfitting and underfitting.
[0043] Once the aforementioned target parameter prediction model has been trained and validated for a specific engineering sea area, it can be used to assess the feasibility of different construction and maintenance operations.
[0044] S400. Based on the current work process, obtain the time series data of the relevant characteristic parameters of the process within a preset time period starting from the current time.
[0045] In this example, time series data of relevant feature parameters corresponding to the process are obtained from commonly used and publicly available weather forecast data (such as GFS forecast data) and used as the spatiotemporal sequence of input features required in the target parameter prediction model.
[0046] S500: Input the time series data of relevant feature parameters into the target parameter prediction model corresponding to the current work process, and output the time series data of the target parameter combination within the forecast time period.
[0047] In this embodiment, the spatiotemporal sequence of input features is adjusted to the data input format required for model training, and the trained target parameter prediction model is run to obtain the spatiotemporal sequence of the predicted target parameter combination wave height, wave direction, flow velocity, and flow direction.
[0048] S600. Based on the type of vessel and the content of the current operation, determine the environmental threshold corresponding to the combination of operation duration and target parameters.
[0049] Based on the type of vessel required and the scope of operations (e.g., 3 different types of vessels, named Vessel 1, Vessel 2, Vessel...), and according to past operational experience and vessel safety requirements, the wind speed for each vessel's operation is determined. , wave height Flow rate Homework duration Restrictions.
[0050] Taking into account the environmental threshold conditions of different vessels, the final environmental threshold condition for this operation can be set as wind speed. <min( , , )* , wave height <min( , , )* Flow rate <min( , , )* Duration > min ( , , )* , This is for the safety factor.
[0051] S700: Based on the time series data of target parameter combinations within the forecast period, combined with the operation duration and environmental thresholds, determine the operation window within the forecast period.
[0052] Based on the comprehensive environmental threshold determined in step S600 and the spatiotemporal sequences of wave height, wave direction, flow velocity, and flow direction obtained in step S500, the sequences of each target parameter are divided according to whether the threshold conditions are met, and time periods that satisfy all target parameters and operation durations are selected. If a time period that meets the requirements exists, the operation can be carried out smoothly, and a defined construction window period is given. If no time period that meets the requirements exists, an early warning message is reported.
[0053] The construction of the wind, wave, and current database in this embodiment includes the following steps:
[0054] ① Based on the offshore construction and maintenance operations, determine the latitude and longitude range of the sea area of interest, and obtain historical wind field data, wave field data, and current field data of the sea area of interest during the same period.
[0055] In this embodiment, the sea area of interest is defined based on the engineering sea area of offshore construction and maintenance operations, its latitude and longitude range is clearly defined, and it is appropriately extended to the periphery, such as an open area along the coast of Zhejiang (123.2°E~123.8°E, 28.5°N~28.8°N).
[0056] In this example, the scope of the numerical model calculation area is determined based on the range of the sea area of interest. The latter should encompass the former, and the sea area of interest should be far enough away from the boundary of the numerical model to avoid the influence of the boundary conditions on the sea area of interest.
[0057] This embodiment acquires historical atmospheric data such as wind field, wave field, and current field data for the sea area of interest during the same period, including:
[0058] i. Multi-source data acquisition.
[0059] (1) Reanalysis data: Collect multi-year (e.g., 2020-2024) ERA5 reanalysis data of the numerical model calculation area, including meteorological elements such as wind speed, wind direction, temperature, and air pressure, for use in the atmospheric driving of the numerical model and the establishment of deep learning models.
[0060] (2) Obtain typhoon process data for the same period in the sea area of concern. Based on the China Coastal Typhoon Track Set Data (http: / / msdc.qdio.ac.cn), extract typhoon process data for the sea area of concern from 2020 to 2024, including the latitude and longitude of the actual typhoon track, typhoon intensity, air pressure, central wind speed, moving speed, and moving direction, etc., to use the wind field of the typhoon period revision reanalysis data to improve the accuracy of the numerical model.
[0061] (3) Ocean model data: Collect HYCOM global model data from the same period, extract ocean elements such as ocean currents (velocity, direction) and tide levels in the computational area, and use them for boundary driving and initial field establishment of the numerical model.
[0062] (4) Measured data:
[0063] Hydrometeorological observation: Collect hydrometeorological observation station data (such as buoy, ship survey, and fixed-point observation data) within the numerical model calculation area (especially the sea area of concern and its vicinity), including wind speed, wind direction, significant wave height, mean wave period, water temperature, current velocity, current direction, tide level, etc., for the calibration and verification of the numerical model.
[0064] Runoff data: Collect runoff data of major rivers that affect the area of interest, and use them as land-source inputs to drive the numerical model.
[0065] ii. Data preprocessing.
[0066] (1) Assess the quality of multi-source data and perform necessary preprocessing on possible outliers and missing values in the data, including using wavelet transform to decompose and denoise the original data, and using embedded data interpolation method to interpolate missing data.
[0067] (2) ERA5 wind field data correction during typhoon season. Based on the dataset of typhoon tracks in the coastal waters of China, when a typhoon exists, the initial ERA5 wind field can be corrected using the typhoon track, central pressure, and maximum central wind speed during the typhoon's lifespan, according to the Holland typhoon field correction algorithm. This includes: ; ; ; ; in, Wind speed; The pressure is represented by m, tc, mov, and ERA5, which refer to the wind speeds of the composite wind field, typhoon gradient wind field, typhoon moving wind speed, and initial ERA5 wind field, respectively; e, c, and n are parameters. ; ; in, , These are the central and peripheral air pressures of the typhoon. Distance from the center of the typhoon; The radius of maximum wind speed; For air density: Determines the intensity and peak size of a typhoon; ; ; ; in, To calculate the latitude of the point; ;
[0068] ② A wind-wave-current coupled model was constructed using the SCHISM three-dimensional baroclinic module as the hydrodynamic model and WWMIII as the wave model.
[0069] Based on the characteristics of the computational domain, horizontal grids (e.g., near-shore densification, densification of areas of interest) and vertical grids (e.g., using LSC) are implemented. 2 The division method is used to divide the data and set reasonable physical parameterization schemes (such as bottom friction, turbulent diffusion coefficient, etc.).
[0070] Driving field settings: The preprocessed ERA5 meteorological field and HYCOM ocean field are used as driving inputs for the model.
[0071] The SCHISM three-dimensional barometric compression module includes: ; ; ; ; in, It is a Cartesian coordinate system in the horizontal direction. The vertical coordinate is the one pointing upwards. For horizontal gradient operators; The substance derivative; For horizontal flow velocity, it includes , Components in two directions , ,unit ; Vertical flow velocity, unit ; For time, For gravitational acceleration, in units ; For free elevation surfaces, units ; The water depth measured from a fixed reference surface, in units ; , The vertical eddy viscosity coefficient and vertical turbulent diffusion coefficient are given in units of... ; For example, the concentration of tracers, such as salinity, temperature, and suspended sediment concentration; Represents horizontal diffusion term, unit ; For quality sources / sinks, units ; Density of water, including salinity ,temperature hydrostatic pressure The impact, unit ; Other forcing terms that affect fluid momentum include baroclinic gradient force, horizontal viscous force, Coriolis force, Earth's tidal potential, atmospheric pressure, and radiation stress.
[0072] The WWMⅢ model includes: ; in, Radiation stress is used to calculate other forcing terms that affect fluid momentum. , Used to calculate the momentum equation in the hydrodynamic module; It is a Cartesian coordinate system in the horizontal direction; , and The components of the radiation stress tensor defined for the spectrum.
[0073] ③ Based on historical wind field data, wave field data, and current field data of the sea area of interest during the same period, the wind-wave-current coupling model was calibrated and validated.
[0074] Historical measured data (especially wind field, waves, current velocity, and water level) from the area of interest and surrounding sea areas were selected to calibrate the simulation results. Parameters in the physical parameterization scheme were modified to make the model output closer to the observed values. The simulation period can be consistent with the collected HYCOM global model data (2020-2024), and the time step can be selected as 120s.
[0075] By selecting multiple non-overlapping time periods (such as different seasons and different weather backgrounds) and utilizing measured data, the accuracy of the model in simulating wind fields, wave fields (wave height, wave period, wave direction), flow fields (flow velocity, flow direction), and tides is comprehensively evaluated, and simulation errors (such as RMSE, MAE, correlation coefficient, etc.) are quantified. This ensures that the model can reasonably reproduce the wind, wave, and flow characteristics of the study area, especially the extreme conditions under typhoon processes.
[0076] ④ Extract the latitude and longitude data of the calculation points included in the sea area of interest from the wind-wave-current coupled numerical model, and then determine the latitude and longitude data of the sample points.
[0077] Due to the high resolution of the sea area of interest (approximately tens of meters), the large number of horizontal grids, and the long time scale, it is necessary to adjust the spatial resolution in the numerical model results in order to reduce the database size and improve the training efficiency of the deep learning model. Farthest Point Sampling (FPS) can be used to resample the simulated calculation points of the sea area of interest at equal intervals, resulting in n equidistant sample points.
[0078] (1) Calculation of the geographic centroid of the sea area of interest: Based on the latitude and longitude range of the sea area of interest, extract the latitude and longitude data (number of N) of the calculation points included in the sea area of interest from the wind-wave-current coupled numerical model, and use the arithmetic mean of the longitude and latitude of all calculation points as the approximate geographic centroid of the sea area of interest.
[0079] (2) Selection of the first sample point: In the first round, the distance between all calculated points and the geographic centroid is calculated according to the Haversine formula, and the point with the smallest distance is selected as the first sample point.
[0080] (3) Determining Remaining Sample Points: The general approach is to first calculate the minimum distance between the remaining calculation points and the sample point set in each round, and then increase the number of sample points in each round based on the maximum value of the minimum distance. Specifically, in the second round, calculate the distance from each unselected calculation point (N-1 points) to the first sample point. The point with the largest distance among these is defined as the second sample point. In the third round, the minimum distance between each of the remaining N-2 calculation points and the sample point set is calculated, and the minimum value among the results is taken as the minimum distance between the calculation point and the sample point set. .Will The point with the maximum value is defined as the 3rd sample point. This process is repeated for a total of n rounds to obtain n resampled equidistant sample points and their corresponding latitude and longitude.
[0081] ⑤ Based on the latitude and longitude data of the sample points, extract the time series data of wave field and flow field at the sample points within a specified time period from the longitude calibration and verification wind-wave-flow coupling model, and extract the corresponding spatiotemporal atmospheric time series data from the atmospheric database.
[0082] Based on the simulation period (2020-2024) set in step ①, to avoid errors before the model calculation stabilizes, the time window for the wind, wave, and current database can be set to 2021-2024. Based on the set time window, combined with the latitude and longitude data of the sample points in step ④, time series data of water depth, significant wave height, mean wave period, current velocity, and current direction at the sample points in the sea area of interest within these time periods are extracted from the calibrated and validated wind, wave, and current coupled model. Wind speed, wind direction, 2m atmospheric temperature, and mean sea level pressure are extracted from the atmospheric database (ERA5 reanalysis database corrected by typhoon season wind field data) to obtain the corresponding spatiotemporal wind speed, wind direction, 2m atmospheric temperature, and mean sea level pressure. The 2m atmospheric temperature gradient and mean sea level pressure gradient are calculated. By combining all the above parameters, a wind, wave, and current database for the sea area of interest is established.
[0083] Example 2: This example is a device for rapidly determining the operational window period for offshore construction and maintenance based on deep learning, specifically including: The target parameter determination module is used to determine the target parameter combination corresponding to each process based on the requirements of each offshore construction and maintenance operation process from wind field, wave field, and current field parameters; The relevant feature selection module is used to select relevant feature parameters that are correlated with the target parameter combination based on the wind, wave and current database. The wind, wave and current database includes the latitude and longitude data of sample points in the engineering sea area, as well as the time series data of the atmosphere, wave field and current field at the sample points. The prediction model training module is used to train a target parameter prediction model based on the time series data corresponding to the combination of target parameters and related feature parameters in the wind, wave and current database. The input data module is used to obtain time series data of the relevant feature parameters of the current operation within a preset time period starting from the current moment, based on the current operation procedure. The model prediction module is used to input the time series data of relevant feature parameters into the target parameter prediction model corresponding to the current work procedure, and output the time series data of the target parameter combination within the forecast period. The environmental threshold determination module is used to determine the environmental threshold corresponding to the combination of operation duration and target parameters based on the type of vessel and operation content required for the current operation procedure. The operation window period determination module is used to determine the operation window period within the forecast period based on the time series data of the target parameter combination within the forecast period, combined with the operation duration and environmental thresholds.
[0084] 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 rapid judgment method for offshore construction and maintenance operation window based on deep learning described in Example 1.
[0085] Example 4: This example is a device for quickly determining the offshore construction and maintenance operation window. It 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 deep learning-based method for quickly determining the offshore construction and maintenance operation window described in Example 1.
[0086] Furthermore, although the invention has been described in the context of functional modules, it should be understood that, unless otherwise stated, one or more of the aforementioned functions and / or features may be integrated into a single physical device and / or software module, or one or more functions and / or features may be implemented in a separate physical device or software module. It is also understood that a detailed discussion of the actual implementation of each module is unnecessary for understanding the invention. Rather, given the properties, functions, and internal relationships of the various functional modules in the apparatus disclosed herein, the actual implementation of the module will be understood within the scope of conventional skill of an engineer. Therefore, those skilled in the art can implement the invention as set forth in the claims using ordinary techniques without excessive experimentation. It is also understood that the specific concepts disclosed are merely illustrative and not intended to limit the scope of the invention, which is determined by the full scope of the appended claims and their equivalents.
[0087] 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.
[0088] 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.
[0089] 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.
[0090] 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.
[0091] 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.
[0092] 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.
[0093] 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 method for rapidly determining the operational window period for offshore construction and maintenance based on deep learning, characterized in that, include: Based on the requirements of each offshore construction and maintenance operation procedure, the target parameter combination corresponding to each procedure is determined from the wind field, wave field, and current field parameters; Based on the wind, wave and current database, relevant feature parameters that are correlated with the target parameter combination are selected. The wind, wave and current database includes latitude and longitude data of sample points in the engineering sea area, as well as time series data of the atmosphere, wave field and current field at the sample points. A target parameter prediction model is trained based on the time series data corresponding to the combination of target parameters and related feature parameters in the wind, wave and current database. Based on the current work process, obtain the time series data of the relevant feature parameters of the process within a preset time period starting from the current time. Input the time series data of relevant feature parameters into the target parameter prediction model corresponding to the current work procedure, and output the time series data of the target parameter combination within the forecast time period; Based on the type of vessel and the content of the current operation, determine the environmental threshold corresponding to the combination of operation duration and target parameters; Based on the time series data of target parameter combinations within the forecast period, combined with operation duration and environmental thresholds, the operation window period within the forecast period is determined.
2. The method for rapid determination of offshore construction and maintenance operation windows based on deep learning according to claim 1, characterized in that, The construction of the wind and wave current database includes: Based on the offshore construction and maintenance operations, the latitude and longitude range of the sea area of interest is determined, and historical atmospheric data, wave field data and current field data of the sea area of interest during the same period are obtained. A wind-wave-current coupled model was constructed using the SCHISM three-dimensional baroclinic module as the hydrodynamic model and WWMIII as the wave model. Based on historical wind field data, wave field data, and current field data of the sea area in question during the same period, the wind-wave-current coupling model was calibrated and validated. The latitude and longitude data of the calculation points included in the sea area of interest are extracted from the wind, wave and current coupled numerical model, and then the latitude and longitude data of the sample points are determined. Based on the latitude and longitude data of the sample points, time series data of wave field and flow field at the sample points within a specified time period are extracted from the longitude-calibrated and validated wind-wave-flow coupling model, and corresponding spatiotemporal atmospheric time series data are extracted from the atmospheric database.
3. The method for rapid determination of offshore construction and maintenance operation windows based on deep learning according to claim 2, characterized in that, The historical atmospheric data were obtained from an atmospheric database, which is an ERA5 reanalysis database corrected for typhoon season wind field data.
4. The method for rapid determination of offshore construction and maintenance operation windows based on deep learning according to claim 3, characterized in that, The ERA5 reanalysis database, corrected for typhoon season wind field data, includes: When a typhoon affects the sea area of concern, the ERA5 wind field is corrected using the typhoon path, typhoon central pressure, and maximum central wind speed during the typhoon's life, according to the Holland typhoon field correction algorithm.
5. The method for rapid determination of offshore construction and maintenance operation windows based on deep learning according to claim 2, characterized in that, The process of extracting latitude and longitude data of calculation points within the sea area of interest from the wind-wave-current coupled numerical model, and then determining the latitude and longitude data of sample points, includes: Based on the computational points contained in the sea area of interest, the computational points in the sea area of interest are resampled at equal intervals using the farthest point sampling on the spherical surface to obtain sample points.
6. The method for rapid determination of offshore construction and maintenance operation windows based on deep learning according to claim 1, characterized in that, The selection of relevant feature parameters based on the wind, wave, and current database that are correlated with the target parameter combination includes: Feature correlation analysis and screening based on the XGBoost algorithm.
7. The method for rapid determination of offshore construction and maintenance operation windows based on deep learning according to claim 1, characterized in that, The target parameter prediction model is built based on a machine learning algorithm applicable to time series prediction problems.
8. A device for rapidly determining the operational window period for offshore construction and maintenance based on deep learning, characterized in that, include: The target parameter determination module is used to determine the target parameter combination corresponding to each process based on the requirements of each offshore construction and maintenance operation process from wind field, wave field, and current field parameters; The relevant feature selection module is used to select relevant feature parameters that are correlated with the target parameter combination based on the wind, wave and current database. The wind, wave and current database includes the latitude and longitude data of sample points in the engineering sea area, as well as the time series data of the atmosphere, wave field and current field at the sample points. The prediction model training module is used to train a target parameter prediction model based on the time series data corresponding to the combination of target parameters and related feature parameters in the wind, wave and current database. The input data module is used to obtain time series data of the relevant feature parameters of the current operation within a preset time period starting from the current moment, based on the current operation procedure. The model prediction module is used to input the time series data of relevant feature parameters into the target parameter prediction model corresponding to the current work procedure, and output the time series data of the target parameter combination within the forecast period. The environmental threshold determination module is used to determine the environmental threshold corresponding to the combination of operation duration and target parameters based on the type of vessel and operation content required for the current operation procedure. The operation window period determination module is used to determine the operation window period within the forecast period based on the time series data of the target parameter combination within the forecast period, combined with the operation duration and environmental thresholds.
9. 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 method for rapidly determining the offshore construction and maintenance operation window period based on deep learning as described in any one of claims 1 to 7.
10. A device for rapidly determining the operational window period of offshore construction and maintenance, 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 method for rapidly determining the offshore construction and maintenance operation window period based on deep learning as described in any one of claims 1 to 7.