Intelligent control method and system for dynamic balance of three floating bodies for crude oil transfer in deep and far sea

Abstract

The invention discloses an intelligent control method and system for dynamic balance of three floating bodies for crude oil transfer in deep and far sea. The method comprises the following steps: (1) data acquisition and state vector construction; (2) constructing a dynamic response prediction model: constructing a floating body historical motion database based on the initial motion data set, fusing hydrodynamic interference coefficient matrix calculation and multi-body motion mixed constraint analysis, and designing a space-time prediction algorithm based on a graph neural network and a recurrent nerve process to obtain a dynamic response prediction model; establishing a dynamic response prediction model under close-range coupling interference of the three floating bodies; (3) evaluating and correcting data confidence; (4) self-adaptive thrust control and early warning are carried out; and (5) visual monitoring and log management. The problems that in the prior art, positioning precision is low, response lags behind, the abnormal rate of sensor data is high, hydrodynamic coupling interference is not fully considered, and visual monitoring and intelligent early warning are lacked are solved, and the safety, stability and operation efficiency of the deep and far sea crude oil transfer process are effectively improved.

Description

Technical Field

[0001] This invention relates to the field of marine energy equipment control, and more specifically to a dynamic balance intelligent control method and system for three floating bodies used in deep-sea crude oil transfer. Background Technology

[0002] With the depletion of global onshore oil resources, deep-sea oil exploration has become a strategic priority in energy development. According to statistics from the Organization of the Petroleum Exporting Countries (OPEC), the total global seabed oil reserves amount to 140-200 billion tons, of which 44% are located in deep water. my country's South China Sea basin alone holds approximately 23-30 billion tons of oil resources, with 70% located in deep-sea areas. Against this backdrop, a three-hull transshipment system consisting of a floating production storage and offloading vessel (FPSO), a crude oil transshipment vessel (CTV), and a conventional oil tanker (VLCC) has become the core mode for deep-sea crude oil transportation. However, its dynamic balance control under complex sea conditions faces severe challenges. Existing three-hull dynamic balance control technologies mainly suffer from the following defects: insufficient positioning accuracy and response capability: Traditional dynamic positioning (DP) systems employ a fixed-parameter PID control strategy, resulting in positioning deviations exceeding 30m and response times greater than 5s under the coupling effects of wind, waves, and currents, failing to meet the control requirements for a safe distance (≥100m) between the three hulls. The core reasons are: insufficient consideration of hydrodynamic interference between multiple floating bodies and poor adaptability to sudden changes in sea state, resulting in insufficient stability at low speeds. Inadequate multi-sensor data fusion technology: existing systems mostly rely on single-sensor positioning and lack sensor confidence assessment models, leading to high data anomaly rates in complex sea states and difficulty in accurately reflecting the real-time motion status of the three floating bodies. For example, traditional DP systems can have sensor data errors of up to 20% in sea state 6, directly affecting the accuracy of control strategies. Lack of visualization and intelligent decision-making: the absence of a 3D visualization monitoring interface for the relative motion of multiple floating bodies prevents operators from intuitively assessing collision risks, resulting in low emergency response efficiency. Furthermore, existing control algorithms do not incorporate machine learning models, making it impossible to optimize control parameters based on historical data and resulting in weak anti-interference capabilities. Inadequate safety warning mechanisms: existing systems mostly use fixed threshold warnings, failing to combine multi-dimensional parameters such as the relative speed and attitude angles of the three floating bodies to assess collision risks, leading to high false alarm rates and delayed warnings, making it difficult to ensure operational safety in extreme sea states. Furthermore, existing technologies fail to establish dynamic response models that consider positional constraints when dealing with the hydrodynamic coupling of three floating bodies. This leads to unreasonable thrust distribution of the propellers under wind, wave, and current loads, further exacerbating the collision risk. Statistics show that traditional transshipment systems have a collision accident rate as high as 15% in deep-sea operations, posing a significant risk of oil spills and severely restricting the safety and efficiency of deep-sea energy development. Summary of the Invention

[0003] The purpose of this invention is to provide a dynamic balance intelligent control method and system for three-buoy systems in deep-sea crude oil transshipment that can improve the safety, stability and operational efficiency of the three-buoy system during deep-sea crude oil transshipment.

[0004] A dynamic balance intelligent control method for three-buoy system in deep-sea crude oil transshipment. Includes the following steps: Step S1. Data Acquisition and State Vector Construction: Real-time acquisition of the relative position coordinates, attitude angles and environmental load data of the three buoys of the FPSO, CTV shuttle and conventional tanker; construction of a relative coordinate system with the FPSO mooring center as the dynamic reference origin; and temporal alignment to form an initial state vector dataset. Step S2. Construction of dynamic response prediction model: Based on the initial state vector dataset, a floating body historical motion database is constructed. The calculation of hydrodynamic interference coefficient matrix and multi-body motion hybrid constraint analysis are integrated. By designing a spatiotemporal prediction algorithm based on graph neural network and recurrent neural process, a dynamic response prediction model with confidence estimation under close-range coupling interference of three floating bodies is established. Step S3. Data confidence assessment and correction: A robust adaptive filtering algorithm is used to assess the confidence of the sensor data and correct for anomalies, outputting high-precision corrected real-time motion data; Step S4. Adaptive thrust control and early warning: Based on the predicted data and corrected real-time motion data of the dynamic response prediction model, the thrust of each float's main thruster and side thruster is dynamically adjusted through an adaptive control algorithm to achieve active collision avoidance control. A three-level safety early warning and automatic switching of operation mode are achieved by combining relative distance and angular acceleration trends for comprehensive judgment. Step S5. Visual Monitoring and Log Management: The user interface module presents a three-dimensional visual scene of the relative motion of the three floating bodies, supporting emergency control parameter intervention and traceable control log management.

[0005] Further, in step S1, the attitude angles include the roll angle, pitch angle, and bow angle of the three-hulled hull; the environmental load data includes wind speed and direction vectors, wave energy spectral density, and ocean current drag force, and the calculation formula for the wind speed and direction vectors is as follows: ,in For real-time wind speed, This is the wind direction angle.

[0006] The wave energy spectral density is based on the JONSWAP model: ,in For the effective wave height, For wave cycles, .

[0007] Furthermore, the formula for calculating the ocean current drag force is as follows: ; in, The density of seawater is (unit: kg / m³, usually taken as 1025). Floating body The drag coefficient (dimensionless, determined by the shape of the floating body). This is the wake interference correction factor (dimensionless, with a value of 0.6-0.8). The surface area of ​​the floating body facing the current. The relative ocean current speed, Floating body The formula for motion speed addresses the problem that traditional single-buoy drag force models cannot be applied to near-range interference.

[0008] Furthermore, in step S2, the hydrodynamic interference coefficient matrix is ​​as follows: ; in Floating body The drag force of a single floating body, Floating body buoyant body Drag force after interference The matrix represents the dimensionless interference coefficients and provides dynamic edge weights for the spatiotemporal prediction model.

[0009] Further, in step S2, the multibody motion hybrid constraint analysis refers to simplifying the FPSO to planar motion locking, retaining only the vertical and attitude degrees of freedom; the CTV and the tanker are constrained by a cable tension model. When the relative distance exceeds a threshold, the cable tension is substituted as a nonlinear spring force into the dynamic equation and calculated using the following formula: ; in This refers to the relative displacement between the CTV and the tanker. This is the maximum permissible length of the cable. The cable stiffness coefficient is denoted as . When the relative distance exceeds a threshold, the cable tension is substituted into the dynamic equation as a nonlinear spring force.

[0010] Further, in step S2, the specific method of the spatiotemporal prediction algorithm is as follows: A graph neural network-recurrent neural process (GNN-RNP) architecture is used to process the timestamped dataset. Using FPSO, CTV, and conventional tanker as the three buoys in the graph, the node characteristics include relative positional deviations. , This is the wake interference coefficient. The side weights are estimated values ​​for cable tension, derived from the hydrodynamic interference coefficient matrix. Dynamically updated; this model extracts spatial coupling features through a graph convolutional network and processes temporal evolution patterns through a recurrent neural process, ultimately outputting the predicted minimum relative distance between the three floating bodies in the next 10 seconds. Relative angular velocity prediction and the corresponding prediction confidence interval , The mapping relationship is expressed by the following formula: ; In the formula: The minimum relative distance between the three floaters is predicted for the next 10 seconds. This is the predicted value of relative angular velocity. The corresponding prediction confidence interval is used to achieve collision risk prediction and uncertainty quantification.

[0011] Furthermore, the recurrent neural process maps the context subsequences of each node to latent representations through a bidirectional LSTM encoder, aggregates them into a global context vector through graph convolution, calculates the distribution parameters of the temporal random process variables using an MLP and samples them, and then generates the predicted distribution parameters through an autoregressive LSTM decoding network. The predicted distribution parameters are obtained by synchronous decoding of independent output layers, and the mapping constraints are implemented through end-to-end differentiability training using reparameterization techniques.

[0012] Furthermore, the spatiotemporal prediction algorithm is optimized using an evidence lower bound loss function: ; In the formula: This represents the expectation operation. For variational posterior distribution, To predict the likelihood, for divergence, Given the context dataset, this loss function ensures that the prediction confidence accurately reflects the uncertainty of the multi-buoy coupling dynamics.

[0013] Furthermore, in step S4, the three-level safety warning system that combines relative distance and angular acceleration trend for comprehensive judgment is as follows: the safe state corresponds to a minimum relative distance greater than 120 meters and a stable trend; the warning state corresponds to a distance less than or equal to 120 meters or a sudden increase in angular acceleration, triggering CTV thrust pre-adjustment; the dangerous state corresponds to a distance less than 100 meters or an angular acceleration exceeding the limit, activating emergency thrust saturation output and audible and visual alarms, supporting operators to customize thresholds.

[0014] A dynamic balance intelligent control system for three floating bodies used in deep-sea crude oil transshipment is established for a triangular collaborative operation mode consisting of a floating production, storage and offloading (FPSO), a crude oil transfer vessel (CTV), and a conventional tanker. The system establishes a complete link from environmental perception to collision avoidance control. The system includes a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor executes the computer program within the following system modules: The data acquisition module is used to acquire and store the relative motion state vector, position coordinates, attitude angles and environmental load data between the three floating bodies in real time through the Beidou differential positioning system, gyroscopes and a unified sea state sensor array deployed on the FPSO. The data processing module includes a three-body dynamic response prediction model and a sensor confidence assessment and anomaly correction unit. The three-body dynamic response prediction model comprises a GNN-RNP module, a CFD calculation unit, and a multibody dynamics simulation unit. The sensor confidence assessment and anomaly correction unit performs confidence assessment and anomaly correction on sensor data, outputting high-precision corrected state information. The CFD calculation unit calculates the hydrodynamic interference coefficient between the floats. The multibody dynamics simulation unit performs multibody motion mixed constraint analysis. The GNN-RNP module performs dynamic response prediction with confidence estimation under close-range coupling interference of the three floats. The control execution module is used to dynamically adjust the thrust of each float's main thruster and side thruster through an adaptive control algorithm based on the predicted data and real-time correction status output by the data processing module, thereby achieving active collision avoidance control. The user interface module is used to receive status warning information from the control execution module and present a three-dimensional visualization scene of the relative motion of the three floating bodies.

[0015] Furthermore, the control execution module outputs a set of thrust commands containing safety indicators. The main thruster controls the longitudinal relative position, and the side thrusters suppress lateral deviation. When the warning level is dangerous, the mode switching trigger value forces a switch to the traction control mode. Each command is bound to the relative position prediction value at the corresponding moment, forming a traceable control log that supports accident playback analysis.

[0016] The beneficial effects of this invention are: This application introduces a graph neural network-recurrent neural process (GNN-RNP) spatiotemporal prediction architecture to achieve early prediction and uncertainty quantification of collision risks under the dynamic coupling of three floating bodies. This improves the safety, stability and operational efficiency of the three floating body system during deep-sea crude oil transfer. It solves the problems of low positioning accuracy, slow response, high sensor data anomaly rate, insufficient consideration of hydrodynamic coupling interference, lack of visualization monitoring and intelligent early warning in the existing technology. More importantly, it overcomes the limitations of traditional Kalman filtering in that it cannot handle nonlinear interference of multiple floating bodies and does not have the ability to estimate prediction confidence. Attached Figure Description

[0017] Figure 1 This is a flowchart of the three-buoy dynamic balance intelligent control method of the present invention; Figure 2 This is a system framework diagram of the present invention; Figure 3 This is a schematic diagram of the three-buoy cooperative operation of the present invention. Detailed Implementation

[0018] like Figure 3 As shown in the diagram, the three-hull cooperative operation schematic of this embodiment clearly illustrates the triangular layout and key control elements of the FPSO, CTV, and conventional tanker. In the diagram, the FPSO is located at the origin of the coordinate system, and its mooring center serves as the dynamic reference point for the entire system. The CTV and conventional tanker are located to the side and rear of the FPSO, forming a typical triangular cooperative operation mode. The overall system workflow is as follows: The data acquisition module uses the BeiDou differential positioning system, gyroscopes, and a unified sea state sensor array deployed on the FPSO to acquire the relative motion state vectors of the three hulls in real time at a sampling frequency of 1Hz. After time-series alignment, an initial motion dataset is constructed. Historical data is summarized every 10 seconds to construct a historical motion database of the hulls, providing training samples for the data processing module.

[0019] like Figure 1 As shown, a dynamic balance intelligent control method for three-buoy deep-sea crude oil transshipment includes the following steps: Step S1. Data Acquisition and State Vector Construction: Real-time acquisition of the relative position coordinates, attitude angles and environmental load data of the three buoys of the FPSO, CTV shuttle and conventional tanker; construction of a relative coordinate system with the FPSO mooring center as the dynamic reference origin; and temporal alignment to form an initial state vector dataset. In this embodiment, the attitude angles include the roll angle, pitch angle, and bow angle of the three floats. Specifically, the roll angle is directly output by the gyroscope built into the three floats. Pitch angle Bow roll angle No acceleration integral calculation is required, thus avoiding accumulated errors.

[0020] The environmental load data includes wind speed and direction vectors, wave energy spectral density, and ocean current drag. The formula for calculating the wind speed and direction vector is as follows: ,in For real-time wind speed, The wind direction angle is given. In this embodiment, the wind vane is installed 10m above the sea surface to avoid the splash zone, with a sampling frequency of 1Hz. The data is then averaged over 10 points for calculation. Wave parameters are obtained from wave height sensors. Wave cycle and wave direction angle .

[0021] The wave energy spectral density is based on the JONSWAP model: ,in For the effective wave height, For wave cycles, .in This spectral function is used to calculate the wave-induced force of the floating body in six degrees of freedom. Ocean current parameters are obtained by measuring the current velocity using a current meter. and flow angle For the special scenario of CTV operating in the wake region of FPSO, a wake interference correction factor is introduced into the drag force calculation. The formula for calculating the drag force of the ocean current is as follows: ; in, This refers to the density of seawater, in kg / m³, with a value of 1025. Floating body The drag coefficient is dimensionless and determined based on the shape of the floating body. This is the wake interference correction factor, dimensionless, with a value of 0.6-0.8; The surface area of ​​the floating body facing the current. The relative ocean current speed, Floating body The formula for motion speed addresses the problem that traditional single-buoy drag force models cannot be applied to near-range interference.

[0022] Step S2. Construction of dynamic response prediction model: Based on the initial state vector dataset, a historical motion database of the floating body is constructed. The calculation of hydrodynamic interference coefficient matrix and multi-body motion hybrid constraint analysis are integrated. By designing a spatiotemporal prediction algorithm based on graph neural network and recurrent neural process, a dynamic response prediction model with confidence estimation under close-range coupling interference of three floating bodies is established.

[0023] To address the close-range coupling characteristics of the three floating bodies, the hydrodynamic interference coefficient matrix between the floating bodies is calculated. ,in D is obtained by looking up a table in the CFD offline database. Floating body The drag force of a single floating body used in the calculation The coefficients are dynamically adjusted based on the Reynolds number of the floating body; then the fast multipole method is used to calculate the floating body. Dragging force after being disturbed by the floating body j Finally, the dimensionless interference coefficient is obtained. .

[0024] The multibody motion hybrid constraint analysis refers to simplifying the FPSO to planar motion locking, retaining only the vertical and attitude degrees of freedom; the CTV and the tanker are constrained by a cable tension model. When the relative distance exceeds a threshold, the cable tension is substituted as a nonlinear spring force into the dynamic equation and calculated using the following formula: ; in This refers to the relative displacement between the CTV and the tanker. This is the maximum permissible length of the cable. The cable stiffness coefficient is denoted as . When the relative distance exceeds a threshold, the cable tension is substituted into the dynamic equation as a nonlinear spring force.

[0025] The specific method of the spatiotemporal prediction algorithm is as follows: A graph neural network-recurrent neural process (GNN-RNP) architecture is used to process timestamped datasets. ,in This is the relative positional deviation. This is the wake interference coefficient. This is an estimated value for cable tension.

[0026] To meet the contextual modeling requirements of circulatory neural processes, the relative velocity sequence of the past 30 seconds was used. Divided by sliding window There are subsequences, each subsequence having a length of _ ... (Corresponding to three floating bodies), forming a contextual dataset. Constructing an undirected graph ,node It is a three-floating body, with sides The weights are determined by the hydrodynamic disturbance coefficient matrix. Dynamically assigned values ​​characterize the real-time hydrodynamic coupling strength. Edge weights are determined by the hydrodynamic interference coefficient matrix. Dynamically updated; this model extracts spatial coupling features through a graph convolutional network and processes temporal evolution patterns through a recurrent neural process, ultimately outputting the predicted minimum relative distance between the three floating bodies in the next 10 seconds. Relative angular velocity prediction and the corresponding prediction confidence interval , The mapping relationship is expressed by the following formula: ; In the formula: The minimum relative distance between the three floaters is predicted for the next 10 seconds. This is the predicted value of relative angular velocity. The corresponding prediction confidence interval is used to achieve collision risk prediction and uncertainty quantification.

[0027] For each node The context subsequence is encoded using a bidirectional LSTM encoder. Generate hidden state sequence Take the hidden state at the last time step. As a subsequence summary. Passed through a multilayer perceptron. Mapping to node-level latent representation Among them, the latent space dimension This process uses bidirectional LSTM to fully capture the forward and backward temporal dependencies of subsequences, avoiding information loss.

[0028] The neighbor node information is aggregated through two graph convolutional layers. The layer update formula is: ,in For nodes The neighborhood group, , For node degree, For learnable weight matrix, The ReLU activation function ultimately outputs a spatially coupled feature vector. The aggregation process dynamically incorporates the hydrodynamic interference coefficients between the three floating bodies, enabling the model to capture nonlinear coupling characteristics.

[0029] Will Input aggregate functions (Using average pooling) to obtain the global context vector Through a two-layer MLP Calculate time-stochastic process variables Distribution parameters: ,in Output mean vector and standard deviation vector Sampling using reparameterization techniques: This ensures that the gradient can be backpropagated. The latent variable... The timing dynamic mode of the entire three-body system was encoded.

[0030] The decoder is an LSTM decoding network. ,by As the initial hidden state, combined with future timestamp features Perform 10-step autoregressive prediction: via independent output layer Simultaneous generation of predicted distribution parameters ,in Quantitative characterization of prediction uncertainty.

[0031] The final mapping relationship is: This design enables the model to predict not only the minimum relative distance over the next 10 seconds. and relative angular velocity It can also output confidence intervals. This provides a quantitative basis for risk assessment in the three-level early warning system.

[0032] The spatiotemporal prediction model is trained end-to-end using the Evidence Lower Bound (ELBO) loss function: The first term is the predicted likelihood expectation, and the second term is the variational posterior. with prior The KL divergence between the two sets of objects is calculated. During training, the context set and the target set are randomly divided to simulate a meta-learning scenario and improve the generalization ability for unsteady conditions in the deep sea.

[0033] Step S3. Data confidence assessment and correction: A robust adaptive filtering algorithm is used to assess the confidence of the sensor data and correct for anomalies, outputting high-precision corrected real-time motion data.

[0034] To address the issues of large absolute positioning errors and intermittent satellite signals in deep-sea areas, a joint confidence model of satellite count and sea state level is established for positioning data, automatically reducing the weight of low-confidence sensors. For abnormally fluctuating data, a weighted correction is performed using correlation verification of nearby floating body data, outputting high-precision status information with a positioning accuracy better than 10 meters and a packet loss rate of less than 5%.

[0035] The control execution module receives the predicted values ​​output by GNN-RNP. The state information after robust filtering correction is used to dynamically generate thrust distribution commands for the main thruster and side thrusters through an adaptive PID control algorithm. The main thruster controls the longitudinal relative position, the side thrusters suppress lateral deviation, and the thrust command calculation comprehensively considers the prediction confidence interval. When the thrust is large, the thrust adjustment range is automatically reduced to avoid over-response.

[0036] By controlling the execution module to force the mode switching unit to traction control mode under dangerous conditions, each thrust command is bound to the relative position prediction value at the corresponding moment. and confidence level This creates a traceable control log, supporting accident playback analysis. The log records timestamps, float status, predicted values, confidence levels, thrust commands, and warning levels, providing a complete data chain for accident investigation.

[0037] Step S4. Adaptive Thrust Control and Early Warning: Based on the predicted data and corrected real-time motion data from the dynamic response prediction model, the thrust of each float's main thruster and side thruster is dynamically adjusted through an adaptive control algorithm to achieve active collision avoidance control. This is combined with a three-level safety early warning system and automatic switching of operating modes, based on a comprehensive judgment of relative distance and angular acceleration trends. The early warning logic uses the predicted value and angular acceleration trend output by the GNN-RNP as input. The judgment process first determines the distance: a safe state corresponds to a minimum relative distance greater than 120 meters and a stable angular acceleration trend; a warning state corresponds to a distance less than or equal to 120 meters or a sudden increase in angular acceleration, triggering CTV thrust pre-adjustment; a dangerous state corresponds to a distance less than 100 meters or angular acceleration exceeding the limit, activating emergency thrust saturation output and audible and visual alarms, supporting operator-defined thresholds. Predictive confidence is introduced as a weighting factor in the early warning judgment to improve the reliability of the early warning.

[0038] Step S5. Visual Monitoring and Log Management: The user interface module presents a three-dimensional visual scene of the relative motion of the three floating bodies, supporting emergency control parameter intervention and traceable control log management.

[0039] The user interface module, based on Three.js and WebGL technology, implements 3D visualization, rendering the relative positions, motion trajectories, and safety zones of the three floating bodies in real time, dynamically updating the scene with the FPSO mooring center as the origin. The parameter adjustment unit supports manual or automatic adjustment of PID control parameters, allowing operators to adjust parameters according to confidence intervals. The system assesses and adjusts control strategies in real time based on predictive reliability. The early warning unit outputs real-time safety / warning / hazard level alerts based on a three-level judgment logic, and supports historical data playback and chart analysis.

[0040] The significant advantages of this embodiment are: by directly learning the nonlinear coupling characteristics of three floating bodies from historical data through the GNN-RNP architecture, collision risks can be predicted 10 seconds in advance and a confidence interval can be output; compared with the traditional Kalman filter, which only performs linear state estimation, this system can not only capture the hydrodynamic interference of multiple floating bodies, but also provide a reliable risk assessment basis for the three-level early warning through uncertainty quantification, solving the limitations of traditional methods that cannot handle multi-body interference and lack confidence estimation, and significantly improving the safety, stability and operational efficiency of deep-sea crude oil transfer.

[0041] like Figure 2As shown, a dynamic balance intelligent control system for three floating bodies used in deep-sea crude oil transshipment is established for a triangular collaborative operation mode consisting of a floating production, storage and offloading (FPSO), a crude oil transfer vessel (CTV), and a conventional tanker. The system establishes a complete link from environmental perception to collision avoidance control. The system includes a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor executes the computer program within the following system modules: The data acquisition module is used to acquire and store in real time the relative motion state vector, position coordinates, attitude angles and environmental load data between the three floating bodies through the Beidou differential positioning system, gyroscopes and a unified sea state sensor array deployed on the FPSO.

[0042] The data processing module includes a three-body dynamic response prediction model and a sensor confidence assessment and anomaly correction unit. The three-body dynamic response prediction model includes a GNN-RNP module, a CFD calculation unit, and a multibody dynamics simulation unit. The sensor confidence assessment and anomaly correction unit performs confidence assessment and anomaly correction on sensor data and outputs high-precision corrected state information. The CFD calculation unit calculates the hydrodynamic interference coefficient between the floats. The multibody dynamics simulation unit performs multibody motion mixed constraint analysis. The GNN-RNP module performs dynamic response prediction with confidence estimation under close-range coupling interference of the three floats.

[0043] The control execution module is used to dynamically adjust the thrust of the main and side thrusters of each float using an adaptive control algorithm based on the predicted data and real-time correction status output by the data processing module, thereby achieving active collision avoidance control. The control execution module outputs a set of thrust commands containing safety indicators. The main thrusters control the longitudinal relative position, and the side thrusters suppress lateral deviation. When the warning level is dangerous, the mode switching trigger value forces a switch to the traction control mode. Each command is bound to the relative position prediction value at the corresponding time, forming a traceable control log that supports accident playback analysis.

[0044] The user interface module is used to receive status warning information from the control execution module and present a three-dimensional visualization scene of the relative motion of the three floating bodies.

[0045] This invention is based on a dynamic balance intelligent control system for three floating bodies used in deep-sea crude oil transshipment, which can run on computing devices such as desktop computers, laptops, handheld computers, and cloud servers.

[0046] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor. This processor is the control center of the deep-sea crude oil transshipment three-buoy dynamic balancing intelligent control system, connecting various parts of the operational system through various interfaces and lines.

[0047] The memory can be used to store the computer programs and / or modules. The processor, by running or executing the computer programs and / or modules stored in the memory, and by calling the data stored in the memory, realizes various functions of the deep-sea crude oil transfer three-buoy dynamic balance intelligent control system. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system and at least one application program required for a function, such as sound playback function, image playback function, etc.; the data storage area may store data created based on the use of the mobile phone, such as audio data, phonebook, etc.

[0048] The memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

Claims

1. A deep sea crude oil transfer three-floater dynamic balance intelligent control method, characterized in that, Comprising the following steps: Step S1. Data acquisition and state vector construction: Real-time acquisition of relative position coordinates, attitude angles and environmental load data of FPSO, CTV shuttle ship and conventional tanker three floating bodies, construction of a relative coordinate system with FPSO mooring center as the dynamic reference origin, and time sequence alignment to form an initial state vector data set; Step S2. Dynamic response prediction model construction: Based on the initial state vector data set, a floating body historical motion database is constructed, a water dynamic interference coefficient matrix is calculated, and a multi-body motion mixed constraint analysis is performed, and through the design of a spatio-temporal prediction algorithm based on a graph neural network and a recurrent neural process, a dynamic response prediction model with confidence estimation under the near-distance coupling interference of three floating bodies is established; Step S3. Data confidence evaluation and correction: A robust adaptive filtering algorithm is used to evaluate and correct the confidence of sensor data, and high-precision corrected real-time motion data is output; Step S4. Adaptive thrust control and early warning: Based on the prediction data of the dynamic response prediction model and the corrected real-time motion data, the thrust of each floating body main propeller and side propeller is dynamically adjusted through an adaptive control algorithm to realize active control for collision prevention, and a three-level safety early warning and operation mode automatic switching are realized through comprehensive judgment of relative distance and angular acceleration trend; Step S5. Visual monitoring and log management: Through a user interface module, a three-dimensional visual scene of the relative motion of the three floating bodies is presented, supporting emergency control parameter intervention and traceable control log management.

2. The deep sea crude oil transfer three floating body dynamic balance intelligent control method according to claim 1, characterized in that in step S1, the attitude angles include the roll angle, pitch angle and yaw angle of the three floating bodies; the environmental load data includes wind speed and direction vector, wave energy spectrum density and sea current drag force, and the calculation formula of the wind speed and direction vector is as follows: The wave energy spectrum density adopts the JONSWAP model: Further, the calculation formula of the sea current drag force is as follows: wherein is the real-time wind speed, is the wind direction angle; 3. The deep sea crude oil transfer three floating body dynamic balance intelligent control method according to claim 2, characterized in that the calculation formula of the sea current drag force is as follows: wherein is the significant wave height, is the wave period, ; 4. The deep sea crude oil transfer three floating body dynamic balance intelligent control method according to claim 3, characterized in that in step S2, the water dynamic interference coefficient matrix is as follows: ； wherein is the seawater density (unit: kg / m³, usually taken as 1025), is the floating body is the drag coefficient (dimensionless, determined according to the shape of the floating body), is the wake interference correction coefficient (dimensionless, taken as 0.6-0.8), is the frontal area of the floating body, is the relative seawater velocity, is the floating body moving speed, and the formula solves the problem that the traditional single floating body drag force model cannot be applied to close-range interference.

5. The deep sea crude oil transfer three floating body dynamic balance intelligent control method according to claim 4, characterized in that in step S2, the multi-body motion mixed constraint analysis refers to simplifying the FPSO as a plane motion lock, only keeping the vertical and attitude degrees of freedom; the CTV and the tanker are constrained through a cable tension model, and when the relative distance exceeds a threshold value, the cable tension is taken as a nonlinear spring force and substituted into the dynamics equation, which is calculated by the following formula:

6. The deep sea crude oil transfer three floating body dynamic balance intelligent control method according to claim 5, ； wherein is the seawater density (unit: kg / m³, usually taken as 1025), is the floating body is the drag coefficient (dimensionless, determined according to the shape of the floating body), is the wake interference correction coefficient (dimensionless, taken as 0.6-0.8), is the frontal area of the floating body, is the relative seawater velocity, is the floating body is the moving velocity, and this formula solves the problem that the traditional single floating body drag force model cannot be applied to close-range interference.

7. The deep sea crude oil transfer three floating body dynamic balance intelligent control system according to claim 6, ​ ； wherein is the single body drag force of the floating body is the single body drag force of the floating body is the single body drag force of the floating body is the single body drag force of the floating body is the single body drag force of the floating body is the dimensionless interference coefficient, which provides dynamic edge weights for the space-time prediction model. ​ ​ ； wherein is the CTV relative displacement to the tanker, is the maximum allowed length of the cable, is the cable stiffness coefficient, when the relative distance exceeds the threshold, the cable tension is substituted as a non-linear spring force into the dynamics equations. ​ The specific method of the space-time prediction algorithm in step S2 is as follows: a graph neural network-recurrent neural process (GNN-RNP) architecture is used to process the timestamped data set , the node characteristics include relative position deviation , , the wake interference coefficient is , the cable tension estimation value is , the edge weight is updated dynamically by the hydrodynamic interference coefficient matrix , the model extracts spatial coupling features through a graph convolution network, processes the time evolution law through a recurrent neural process, and finally outputs the prediction value of the minimum relative distance between the three floating bodies in the future 10 seconds , the relative angular velocity prediction value , , and the corresponding prediction confidence interval , the mapping relationship is represented by the following formula: ； wherein: is the predicted minimum relative distance between the three floating bodies in the next 10 seconds, is the relative angular velocity prediction, is the corresponding prediction confidence interval, realizing the collision risk prediction and uncertainty quantification. ​ The characteristic is that the recurrent neural process maps the node context subsequence into a latent representation through a bidirectional LSTM encoder, aggregates into a global context vector through a graph convolution, and then uses an MLP to calculate the distribution parameters of the time stochastic process variable and sampling, and then generates the predicted distribution parameters through an LSTM decoding network autoregression, wherein the mapping constraint is realized by the end-to-end differentiable training through the reparameterization trick.

8. The deep sea crude oil transfer three-floater dynamic balance intelligent control method of claim 7, wherein The characteristic is that in step S4, the three-level safety early warning of the combination of relative distance and angular acceleration trend is as follows: the safe state corresponds to the minimum relative distance greater than 120 meters and the trend is stable; the warning state corresponds to the distance less than or equal to 120 meters or the sudden increase of angular acceleration, triggering the CTV thrust pre-adjustment; the dangerous state corresponds to the distance less than 100 meters or the angular acceleration exceeding the limit value, starting the emergency thrust saturation output and the sound and light alarm, and supporting the operation personnel to define the threshold value.

9. A deep sea crude oil transfer three-floater dynamic balance intelligent control system, characterized in that, For the triangular collaborative operation mode of the floating production, storage and offloading (FPSO), the crude transfer vessel (CTV) and the conventional oil tanker, a complete link from environmental perception to anti-collision control is established, and the system comprises a memory, a processor and a computer program stored in the memory and executable on the processor. The processor executes the computer program to run in the following system modules: A data acquisition module is configured to acquire and store the relative motion state vector, position coordinates, attitude angle and environmental load data among the three floaters in real time through the Beidou differential positioning system, the gyroscope and the unified sea state sensor array deployed on the FPSO. A data processing module comprises a three-floater dynamic response prediction model, a sensor confidence assessment and abnormal correction unit, the three-floater dynamic response prediction model comprises a GNN-RNP module, a CFD calculation unit and a multi-body dynamics simulation unit, the sensor confidence assessment and abnormal correction unit is configured to perform confidence assessment and abnormal correction on the sensor data and output corrected high-precision state information; the CFD calculation unit is configured to calculate the hydrodynamic interference coefficient between the floaters; the multi-body dynamics simulation unit is configured to analyze the mixed constraint of multi-body motion; and the GNN-RNP module is configured to predict the dynamic response with confidence estimation under the near-distance coupling interference of the three floaters. A control execution module is configured to dynamically adjust the thrust of the main propeller and the side propeller of each floater based on the predicted data and the real-time corrected state output by the data processing module, so as to realize active anti-collision control. A user interface module is configured to receive the state early warning information of the control execution module and present a three-dimensional visual scene of the relative motion of the three floaters.

10. The system of claim 9, wherein the deep sea crude oil transfer three-floater dynamic balance intelligent control is characterized in that ​ The application is characterized in that the control execution module outputs a thrust instruction set containing a safety mark, the main propeller controls the longitudinal relative position, and the side propeller suppresses the lateral deviation; when the early warning level is dangerous, the mode switching trigger value is forced to switch to the traction control mode; each instruction is bound to a relative position prediction value at a corresponding time to form a traceable control log, thereby supporting accident playback analysis.

Images