A method for hoisting a four-pile jacket foundation for offshore wind power

Abstract

This invention relates to the field of offshore wind power construction and discloses a method for hoisting a four-pile jacket foundation for offshore wind power. The method includes: a parameter estimator calculating the compensation cable length to correct the Jacobian matrix of the jacket hoisting system; a feedforward solver extracting wave height sequence data to generate disturbance vectors and wave impact vectors; the parameter estimator combining tension data to perform variable-sensor observation and outputting a dynamic mass matrix; an optimization controller using the dynamic mass matrix, the jacket hoisting system Jacobian matrix, disturbance vectors, and wave impact vectors as boundary conditions to perform multivariate predictive control calculations and update the target rope velocity command data; and a main control computer issuing the target rope velocity command data to complete the pile insertion action. This invention reduces attitude deviation by introducing wave feedforward through the feedforward solver, resolves control divergence in the variable-parameter model by updating the dynamic mass matrix through the parameter estimator, and prevents traction cable loosening and breakage through the operation of tension constraints and tension protection by the optimization controller.

Description

Technical Field

[0001] This invention relates to the field of offshore wind power construction, specifically to a method for hoisting a four-pile jacket foundation for offshore wind power. Background Technology

[0002] Offshore wind turbine jacket foundations face complex sea conditions during the offshore construction and hoisting phase. Existing hoisting control methods typically rely on feedback adjustment mechanisms, only executing compensation actions after wave impacts cause positional deviations. The transmission of wave disturbances to the closed-loop control nodes introduces lag, leading to deviations in the spatial six-degree-of-freedom positional control of the jacket foundation, making it difficult to meet the positional requirements for pile driving operations.

[0003] During the installation and submersion phase of the four-pile jacket, the static gravity load, variable buoyancy load, and hydrodynamic added mass borne by the jacket continuously change due to variations in submersion depth. Conventional dynamic models often use fixed parameters for state evolution calculations, failing to incorporate real-time added mass and buoyancy values ​​during the submersion phase into the dynamic boundary conditions. Fixed-parameter models exhibit control divergence when faced with the dynamic changes in added mass and buoyancy during the submersion phase.

[0004] The four-pile jacket uses a parallel drive system composed of multiple traction cables for attitude control. Under wave impact loads, the multi-cable parallel drive system exhibits nonlinear time-varying characteristics. Existing tension distribution algorithms do not provide boundary constraints for the wave impact on the traction cable on the wave-facing side, nor do they actively limit the relaxation force component generated along the traction cable axis, making it prone to localized traction cable loosening under wave impact. When encountering abnormal environmental load impacts, the existing control system lacks a low-level intervention mechanism targeting the breaking tensile strength limit of the traction cable material. When the real-time tension of the traction cable exceeds the safety threshold, it cannot interrupt the attitude closed-loop adjustment and execute active cable release, resulting in tension overload and breakage of the traction cable. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a method for hoisting offshore wind turbine four-pile jacket foundations. This method solves the problems of large hoisting posture control deviations and localized loosening, breakage, and overload of traction cables caused by wave impact, dynamic changes in added mass and buoyancy, and nonlinear time-varying characteristics of multi-cable parallel systems during the hoisting and water entry phase of offshore wind turbine four-pile jacket foundations.

[0006] To achieve the above objectives, the present invention provides a method for hoisting a four-pile jacket foundation for offshore wind power, comprising the following steps: The parameter estimator extracts the measured cable length and the group of basic constants to calculate the equivalent cable length, constructs the cable net geometric mapping and compensation, analyzes the spatial positional relationship between the crane ship and the four-pile jacket, and uses the equivalent cable length to correct the Jacobian matrix of the jacket hoisting system. The feedforward solver reads wave height sequence data and six-degree-of-freedom motion attitude data, obtains horizontal offset distance values ​​based on spatial position relationships, performs wave feedforward and spatial mapping, and generates disturbance vectors, wave impact vectors and real-time wave characteristic data. The parameter estimator acquires real-time wave characteristic data, combines vertical tension data and release length data to carry out coordinated observation of water entry parameters, and outputs the state vector and dynamic mass matrix of the four-pile jacket; The optimized controller uses the dynamic mass matrix, the Jacobian matrix of the jacket hoisting system, the disturbance vector, and the wave impact vector as state evolution boundary conditions to perform multivariate predictive control calculations and update the target rope velocity command data. The main control computer sends target rope speed command data to the frequency conversion traction component to coordinate the completion of the stake insertion action.

[0007] Furthermore, the process of constructing the cable net geometric mapping and compensation includes: Obtain the starting reference point of the vertical line where the main hook is located in front of the bow of the crane vessel, extract the three-dimensional absolute coordinates of the first, second, third and fourth frequency conversion traction winches included in the frequency conversion traction assembly, and extract the three-dimensional relative coordinates of the first, second, third and fourth force nodes within the coordinate system of the jacket center of mass. The main sling and the first, second, third and fourth traction cables are combined and set as a multi-cable parallel drive topology network. Based on the spatial vector boundary constraints contained in the multi-cable parallel drive topology network, a kinematic differential transformation matrix between the control vector and the state vector is constructed in the global origin coordinate system. The kinematic differential transformation matrix is ​​defined as the Jacobian matrix of the jacket hoisting system.

[0008] Furthermore, the process of correcting the Jacobian matrix of the jacket hoisting system using equivalent cable lengths includes: Before substituting the real-time tension value at the end of the traction cable into the parameter estimator, a condition comparison and judgment is performed. If the real-time tension value at the end of the traction cable is lower than the minimum calculated tension constant, the minimum calculated tension constant is forcibly substituted into the denominator of the equivalent cable length calculation equation for calculation. The spatial geometric constraint side length values ​​inside the multi-cable parallel driving topology network are updated based on the equivalent cable length, and the set of constant coefficients of the Jacobian matrix of the jacket erection system is corrected using the equivalent cable length.

[0009] Furthermore, the process of performing wave feedforward and spatial mapping includes: The feedforward solver performs a convolution operation on the wave height sequence data and the crane ship amplitude response operator matrix in the time dimension to generate a disturbance vector. It then performs a deviation comparison operation between the disturbance vector and the six-degree-of-freedom motion attitude data, removes high-frequency measurement noise components, and outputs the filtered disturbance vector. The lag time of wave arrival at the wave front is calculated using the horizontal offset distance and wave propagation speed. Based on the lag time, a time axis translation transformation is performed on the wave height sequence data to generate real-time wave characteristic data. The real-time wave characteristic data and the effective pressure area are mapped and calculated to output the wave impact vector.

[0010] Furthermore, the process of conducting coordinated observation of water-borne parameters includes: The parameter estimator extracts three-dimensional absolute coordinate data and performs high- and low-frequency asynchronous filtering. It retrieves the historical state vector at the corresponding moment from the historical state memory and uses the three-dimensional acceleration and three-dimensional angular velocity values ​​to extrapolate to the current moment to generate the predicted state vector. Calculate the numerical residual between the three-dimensional absolute coordinate data and the predicted state vector, use the Kalman gain matrix to eliminate the cumulative prediction error caused by the communication delay of the positioning base station, correct the predicted state vector, and output the corrected state vector after eliminating the spatial position error to update the state vector.

[0011] The parameter estimator extracts the vertical axial acceleration component from the three-dimensional acceleration values, performs high-frequency time-domain differential processing on the vertical tension data to separate the static gravity load and the variable buoyancy load, evaluates the algebraic sum of the gravitational acceleration and instantaneous acceleration components, and enforces saturation limiting when the algebraic sum is lower than the minimum effective apparent weight acceleration constant. The real-time added mass value is calculated based on the hydrodynamic added mass separation equation applied to the pre-set safety limit boundary. The real-time buoyancy value is calculated by combining the release length data and wave height program sequence data. The real-time added mass value is written into the dynamic mass matrix to complete the real-time parameter update of the dynamic mass matrix.

[0012] Furthermore, the process of performing multivariate predictive control solutions includes: The controller internal setting prediction time domain constant is optimized and the target reference state vector value containing the target space six-degree-of-freedom pose parameters of the four-pile guide frame at the target pile position is pre-stored. At the low-level operation level, the continuous time domain integral model is equivalently transformed into a discrete algebraic summation model by using the fixed step size numerical integration rule, which helps to construct the target cost function of nonlinear model predictive control. By minimizing the numerical value of the objective cost function, we seek the optimal control sequence within the prediction time constant that makes the spatial six-degree-of-freedom pose information of the four-pile jacket approach the numerical value of the target reference state vector.

[0013] Furthermore, during the execution of multivariable predictive control solutions, the optimization controller uses the electromechanical characteristics of the first, second, third, and fourth variable frequency traction winches as hard constraint boundary conditions: The load torque is converted from the real-time tension value at the end of the traction cable. The maximum allowable acceleration value of the target rope speed command data is calculated using the electromechanical inertial acceleration constraint formula. The real-time change amplitude of the target rope speed command data is constrained within the closed interval formed by the negative maximum allowable acceleration value to the positive maximum allowable acceleration value. During the solution of the target cost function, if the first derivative of the target rope velocity command data exceeds the maximum allowable acceleration value, the rate of change is forced to converge to the maximum allowable acceleration value.

[0014] Furthermore, the algorithm for synchronously running the controller with asymmetric tension lower bound constraint is optimized: Obtain the instantaneous coordinate values ​​of the first, second, third, and fourth force-bearing nodes within the global origin coordinate system. Combine this with the installation point coordinates of the variable frequency traction component to calculate the unit direction vectors of the first, second, third, and fourth traction cables. The relaxation force components generated by wave impact load in the axial direction of each traction cable are identified by performing a dot product mapping between the wave impact vector and the unit direction vector. The real-time tension lower limit value of each traction cable is calculated by applying the asymmetric tension lower limit constraint formula. The real-time lower bound value of tension is written into the hard constraint condition set of the nonlinear model predictive control, which restricts the solution space of the control vector, so that the predicted value of the tension of the traction cable on the wave-facing side is always above the numerical range specified by the real-time lower bound value.

[0015] Furthermore, during the period when the main control computer issues target rope speed command data: The optimization controller uses a command smoothing filter to perform a second-order low-pass filter on the target rope speed command data and outputs smooth target rope speed data. Simultaneously, it cyclically extracts the real-time tension values ​​fed back by the first tension sensor, the second tension sensor, the third tension sensor, and the fourth tension sensor. The maximum safe tension threshold is pre-set based on the breaking tensile strength limit of the materials of the first, second, third, and fourth traction cables and the safety redundancy coefficient. If the real-time tension value of any traction cable exceeds the maximum safe tension threshold and the duration reaches the emergency degradation response duration constant, the optimized controller immediately interrupts the posture closed-loop adjustment and calculates the tension protection compensation component using the tension limit protection criterion equation. The tension protection compensation component is superimposed on the smooth target rope speed data, and the output torque target value of the variable frequency traction component is adjusted to execute the active cable release action.

[0016] This invention provides a method for hoisting a four-pile jacket foundation for offshore wind power. It has the following advantages: 1. This invention uses a feedforward solver to read wave height sequence data and six-degree-of-freedom motion attitude data to generate disturbance vectors and wave impact vectors. The optimized controller uses the disturbance vectors and wave impact vectors as state evolution boundary conditions to perform multivariate predictive control calculations. At the underlying computational level, it offsets the lag phenomenon caused by wave disturbances being transmitted to the closed-loop control node, thereby reducing the spatial six-degree-of-freedom pose control deviation caused by wave impacts.

[0017] 2. This invention uses a parameter estimator to combine vertical tension data and release length data to conduct joint observation of variable parameters during water entry. It separates static gravity load from variable buoyancy load and applies the hydrodynamic added mass separation equation to calculate the real-time added mass value, thereby completing the real-time variable parameter update of the dynamic mass matrix. This solves the control divergence problem caused by sudden changes in added mass and real-time buoyancy during the water entry stage of the four-pile jacket model.

[0018] 3. This invention identifies the relaxation force component generated by wave impact load in the traction cable axis direction by optimizing the controller's operation of the asymmetric tension lower bound constraint algorithm. It limits the control vector solution space to ensure that the predicted tension value of the traction cable on the wave-facing side is above the real-time tension lower bound value. Combined with the tension limit protection criterion equation, when the real-time tension value exceeds the maximum safe tension threshold, it superimposes the tension protection compensation component to perform active cable release action, thereby avoiding local traction cable loosening and tension overload breakage in the multi-cable parallel drive topology network. Attached Figure Description

[0019] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a diagram showing the overall architecture of the jacket erection system of the present invention; Figure 3 This is the multivariable predictive control logic diagram of the optimized controller of the present invention; Figure 4 This is a comparison diagram of the X-axis displacement convergence of the four-pile guide frame of the present invention; Figure 5 This is a comparison diagram of the stress state and boundary control of the traction cable in this invention. Detailed Implementation

[0020] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] See attached document Figure 1 and attached Figure 2 This invention provides a method for hoisting a four-pile jacket foundation for offshore wind power, which is implemented using a jacket hoisting system. The jacket hoisting system includes a hoisting execution layer mounted on a crane vessel.

[0022] The lifting operation layer consists of a main crane mounted on the bow of the crane vessel and a variable frequency traction assembly. The main crane is connected to the main sling at its bottom, and the main sling is connected to the main hook at its end. A load sensor is installed on the main hook to collect vertical tension data. A lifting encoder is installed inside the main crane to collect data on the extended length of the main sling.

[0023] The variable frequency traction assembly includes a first variable frequency traction winch, a second variable frequency traction winch, a third variable frequency traction winch, and a fourth variable frequency traction winch, which are respectively arranged in the first edge area, the second edge area, the third edge area, and the fourth edge area of ​​the crane ship deck, and respectively lead out the first traction cable, the second traction cable, the third traction cable, and the fourth traction cable.

[0024] The ends of the first to fourth traction cables are respectively connected to the first, second, third, and fourth force-bearing nodes of the four-pile guide frame. The first to fourth frequency-controlled traction winches are equipped with a first, second, third, and fourth absolute encoder, along with a first, second, third, and fourth tension sensor, respectively. The first to fourth absolute encoders independently collect the extension values ​​of their respective traction cables.

[0025] The environmental and attitude perception layer is divided into ship-side sensing components and jacket-side sensing components. The ship-side sensing components are mounted on the crane vessel body and include a wave-measuring radar fixedly arranged at the foremost bow of the crane vessel and a hull reference instrument rigidly installed at the center of gravity of the crane vessel.

[0026] The wave-measuring radar collects wave height sequence data in the sea area in front of the crane vessel; the hull reference instrument collects six-degree-of-freedom motion attitude data of the crane vessel. The jacket-side sensing components are mounted on the four-pile jacket, including a high-precision inertial navigation system (INS) and a positioning base station. The high-precision INS is securely connected to the geometric center node of the four-pile jacket to acquire its three-dimensional acceleration and three-dimensional angular velocity values; the positioning base station is mounted on the top platform structure of the four-pile jacket to receive external satellite signals and extract the three-dimensional absolute coordinate data of the four-pile jacket. The wave-measuring radar, hull reference instrument, high-precision INS, and positioning base station operate independently, each outputting sensing data to the outside world.

[0027] The main control and network communication layer establishes electrical connections with the hoisting execution layer and the environment and attitude perception layer, respectively, consisting of a main control computer and a fieldbus network. The main control computer is located inside the control cabin of the crane vessel and is equipped with a real-time operating system. The fieldbus network provides a data transmission channel, with one end connected to the main control computer's communication interface and the other end connected to load sensors, lifting encoders, tension sensors, absolute encoders, and in conjunction with wave measuring radar, hull reference instruments, high-precision inertial navigation systems, and positioning base stations. The fieldbus network synchronously transmits the mechanical operation data collected by the hoisting execution layer and the external environmental data collected by the environment and attitude perception layer to the main control computer.

[0028] The main control computer, based on a real-time operating system, is divided into a feedforward solver, a parameter estimator, and an optimization controller. The feedforward solver connects to the input ports of the wave measuring radar and the hull reference instrument to read wave height program data and six-degree-of-freedom motion attitude data. The parameter estimator connects to the data channels of the corresponding hardware to extract vertical tension data, release length data, three-dimensional acceleration values, three-dimensional angular velocity values, and three-dimensional absolute coordinate data. The optimization controller connects to the frequency conversion drive terminals of the first to fourth frequency conversion traction winches and sends down the corresponding command data.

[0029] The main control computer establishes a spatial reference model to provide a spatial calibration framework for data processing. The spatial reference model includes a global origin coordinate system and a jacket center-of-mass coordinate system. The global origin coordinate system is set as a static reference coordinate system, with its origin fixed at the planar geometric center of the four target points on the seabed, providing a stationary absolute reference in three-dimensional space. The jacket center-of-mass coordinate system is set as a dynamic follower coordinate system, with its origin fixed at the centroid nodes of the four jacket piles, changing synchronously with the spatial displacement and rotation of the four jacket piles.

[0030] The parameter estimator extracts the three-dimensional absolute coordinate data output from the positioning base station and maps it to the global origin coordinate system. Based on the global origin coordinate system, it calculates in real time the three-dimensional displacement parameters and three-dimensional Euler angle parameters of the jacket center of mass coordinate system relative to the global origin coordinate system. The global origin coordinate system and the jacket center of mass coordinate system together provide a unified data processing environment within the main control computer.

[0031] The main control computer internally constructs state vectors and control vectors. The parameter estimator writes real-time sensed values ​​into the state vectors; the optimization controller outputs drive data to the variable frequency traction components based on the control vectors. The state vector contains twelve dimensions of data, which sequentially represent the X-axis displacement, Y-axis displacement, Z-axis displacement, roll angle, pitch angle, yaw angle, and the time first derivative components corresponding to the six pose components in the jacket's center of mass coordinate system relative to the global origin coordinate system.

[0032] The state vector provides a digital representation of the six-degree-of-freedom spatial pose and velocity information for the four-pile jacket. The control vector contains four dimensions of data, representing the target rope speed commands issued by the optimization controller to the first to fourth frequency conversion traction winches. The main control computer calculates the four dimensions of data contained in the control vector based on the twelve dimensions of data contained in the state vector.

[0033] The main control computer pre-writes a set of fundamental constants. The parameter estimator synchronously calculates the dimensional boundary parameters based on the real-time received data stream, providing fundamental variable support for subsequent data calculations. The first to fourth absolute encoders each input independent extension values ​​to the parameter estimator. The parameter estimator extracts the extension values ​​as the measured cable length, subtracts the sag and elastic tensile deformation included within the measured cable length, and generates equivalent cable length data. The subscript variables are set to natural number constants, with values ​​including 1, 2, 3, and 4, corresponding to the first to fourth traction cables, respectively.

[0034] The fundamental constants include the elastic modulus, cable cross-sectional area, gravitational acceleration, traction cable linear density, intrinsic mass constant, and wave propagation speed. The elastic modulus characterizes the material's intrinsic constant in terms of the traction cable's resistance to tensile deformation; the cable cross-sectional area characterizes the geometric cross-sectional area of ​​the traction cable's end face; gravitational acceleration characterizes the gravitational field constant of the region where the four-pile jacket is located; the traction cable linear density characterizes the mass per unit length of a single traction cable; the intrinsic mass constant characterizes the static mass of the four-pile jacket in air; and the wave propagation speed characterizes the natural environmental constant of wave propagation in the operating sea area. The parameter estimator encapsulates the fundamental constants, measured cable length, and equivalent cable length as a set of computational input variables.

[0035] See attached document Figure 1 The present invention provides a method for hoisting a four-pile jacket foundation for offshore wind power, comprising the following steps.

[0036] S100: The parameter estimator extracts the measured cable length and the basic constant group, and calculates the equivalent cable length by combining the measured cable length and the basic constant group, and constructs the cable net geometric mapping and compensation.

[0037] S200: The feedforward solver reads the wave height program data collected by the wave measuring radar and the six-degree-of-freedom motion attitude data collected by the hull reference instrument, and performs wave feedforward and spatial mapping by combining the wave height program data and the six-degree-of-freedom motion attitude data.

[0038] S300: The parameter estimator reads the vertical tension data collected by the load sensor and the release length data collected by the lifting encoder to carry out coordinated observation of the water entry parameters for the four-pile jacket.

[0039] S400: The optimized controller receives the dataset output by the feedforward solver and parameter estimator, uses the dataset to perform multivariable predictive control calculations, and updates the target rope speed command data contained in the control vector.

[0040] S500: The main control computer sends the target rope speed command data contained in the control vector to the variable frequency traction component through the fieldbus network. The first to fourth variable frequency traction winches receive the target rope speed command data and work together to complete the stake insertion action.

[0041] Combined with S100, the parameter estimator inside the main control computer extracts the spatial calibration data of the main crane and the static position data of the variable frequency traction components, obtains the three-dimensional absolute coordinates of the first to fourth variable frequency traction winches arranged on the deck of the crane ship, and extracts the three-dimensional relative coordinates of the first, second, third, and fourth force nodes within the coordinate system of the jacket center of mass. It obtains the vertical starting reference point of the main hook located in front of the bow of the crane ship, sets the main sling and the combination of the first to fourth traction cables as a multi-cable parallel drive topology network, and constructs the kinematic differential transformation matrix between the control vector and the state vector within the global origin coordinate system based on the spatial vector boundary constraints contained in the multi-cable parallel drive topology network.

[0042] The control vector includes the launch and recovery speed variables of the first to fourth variable frequency traction winches; the state vector includes the spatial six-degree-of-freedom pose and velocity information of the four-pile jacket. The parameter estimator defines the kinematic differential transformation matrix as the Jacobian matrix of the jacket hoisting system, providing a set of constant coefficients for numerical conversion between control vector data and state vector data.

[0043] The Jacobian matrix operation of the jacket erection system relies on high-precision geometric span variables. The parameter estimator extracts the independent tensile values ​​collected in real-time by the first to fourth tension sensors and defines them as the real-time tension value at the end of the traction cable. Combined with the traction cable linear density constant pre-written into the main control computer, it synchronously reads the measured cable length, elastic modulus, cable cross-sectional area, and gravitational acceleration, encapsulates and substitutes these values ​​into the equivalent cable length calculation equation, and outputs the equivalent cable length. The mathematical expression of the equivalent cable length calculation equation is set as follows: ; In the formula, Characterizing the first The equivalent cable length of the root traction cable; Characterizing the first The measured length of the root traction cable; Characterizing the first Real-time tension value at the end of the root traction cable; The elastic modulus shared by the first, second, third, and fourth traction cables; Characterizes the cross-sectional area of ​​the cable shared by the first, second, third, and fourth traction cables; The linear density constant of the traction cable, which is shared by the first, second, third, and fourth traction cables; Characterizes the gravitational acceleration of the work area.

[0044] In this embodiment, the derivation of the equivalent cable length calculation equation is based on the principle of cable stress deformation. The equivalent cable length calculation equation includes an elastic tension reduction term and a catenary sag compensation term. To avoid data overflow caused by the denominator approaching zero in the equivalent cable length calculation equation due to a momentary complete relaxation of the traction cable, a non-zero minimum calculation tension constant is pre-written into the main control computer.

[0045] Before substituting the real-time tension value at the end of the traction cable, the parameter estimator performs a conditional comparison. If the real-time tension value at the end of the traction cable is lower than the minimum calculated tension constant, the minimum calculated tension constant is forcibly substituted into the denominator of the equivalent cable length calculation equation for calculation. The parameter estimator updates the spatial geometric constraint side length values ​​within the multi-cable parallel drive topology network based on the equivalent cable length, and uses the equivalent cable length to correct the set of constant coefficients of the Jacobian matrix of the jacket erection system, so that the Jacobian matrix of the jacket erection system obtains dynamic mapping data that eliminates the deformation error of the flexible cable.

[0046] Combined with S200, the feedforward solver extracts the wave height program data continuously input by the wave measuring radar, retrieves the crane ship amplitude response operator matrix pre-written inside the main control computer, performs convolution operation on the wave height program data and the crane ship amplitude response operator matrix in the time dimension, and generates a disturbance vector based on the value output by the convolution operation.

[0047] The disturbance vector includes the three-dimensional linear displacement disturbance and the three-dimensional angular displacement disturbance caused by wave heave and swaying of the crane vessel's bow. The feedforward solver reads the six-degree-of-freedom motion attitude data collected by the hull reference instrument, performs a deviation comparison calculation between the disturbance vector and the six-degree-of-freedom motion attitude data, isolates the high-frequency measurement noise components inside the disturbance vector, removes the high-frequency measurement noise components, and outputs the filtered disturbance vector as the feedforward compensation data for optimizing the controller output reference.

[0048] The feedforward solver acquires the horizontal offset distance of the geometric center node of the four-pile jacket relative to the bow of the crane vessel. Using this horizontal offset distance and wave propagation speed, it calculates the lag time of wave arrival at the wave-facing surface of the four-pile jacket. Based on this lag time, it performs a time-axis translation transformation on the wave height sequence data, generating real-time wave characteristic data at the spatial location of the four-pile jacket. Simultaneously, the feedforward solver retrieves the pre-stored force-bearing section parameters of the four-pile jacket from the main control computer and, combined with the released length data, extracts the effective pressure area below the waterline. It then maps the real-time wave characteristic data and the effective pressure area, outputting the wave impact vector acting on the four-pile jacket. This wave impact vector contains the three-dimensional wave impact force value experienced by the four-pile jacket in the global origin coordinate system.

[0049] The feedforward solver sends the wave impact vector to the optimization controller. The optimization controller receives the wave impact vector and predicts the displacement trend of the four-pile jacket at future moments. The wave impact vector, together with the disturbance vector, constitutes a spatial mapping data source, assisting the optimization controller in adjusting the target rope speed command data from the first to the fourth variable frequency traction winch.

[0050] In conjunction with the S300, the parameter estimator performs asynchronous high- and low-frequency filtering, receiving high-frequency motion data from the high-precision inertial navigation system (INS) and simultaneously receiving low-frequency position data from the positioning base station. The sampling frequency of the high-precision INS is higher than that of the positioning base station. Due to signal transmission delays, the three-dimensional absolute coordinate data output by the positioning base station exhibits a time lag relative to the real-time time.

[0051] In this embodiment, the multi-source data fusion operation is carried out based on the optimal state estimation theoretical framework. The parameter estimator internally sets up a historical state memory to cache state vectors in time sequence. When the three-dimensional absolute coordinate data of the positioning base station is received, the historical state vector at the corresponding time is retrieved from the historical state memory. The retrieval time is set as the starting point, and the predicted state vector is generated by using the three-dimensional acceleration and three-dimensional angular velocity values ​​collected by high-precision inertial navigation to extrapolate to the current time.

[0052] The main control computer internally sets a fixed basic sampling time period, and based on this period, transforms the continuous-time system kinematics into a discretized matrix form. The parameter estimator synchronously calculates the numerical residual between the three-dimensional absolute coordinate data and the predicted state vector, and the mathematical expression followed by the state update is: ; In the formula, Characterized at discrete sampling time Estimated condition values ​​of the lower four-pile jacket; Characterizes the state transition matrix of a four-pile jacket that includes the basic sampling time period attribute; This represents the current discrete sampling time of the main control computer; Characterizes the previous discrete sampling moment of the main control computer; Characterized at discrete sampling time Estimated condition values ​​of the lower four-pile jacket; Characterizes the input control matrix of a four-pile jacket containing discretization coefficients; Characterized at discrete sampling time The control input value is below; The Kalman gain matrix represents the current sampling time. Characterizes the three-dimensional absolute coordinate data with communication delay compensation output by the positioning base station; Characterize the observation matrix of the four-pile jacket.

[0053] The parameter estimator uses the Kalman gain matrix to correct the predicted state vector, outputs the corrected state vector that eliminates the spatial position deviation caused by the communication delay of the positioning base station, and synchronizes it to the optimization controller in real time.

[0054] As a preferred approach, the parameter estimator performs hydrodynamic parameter separation and matrix update, receives vertical tension data collected by load sensors and three-dimensional acceleration values ​​collected by high-precision inertial navigation, extracts the vertical axial acceleration component from the three-dimensional acceleration values, combines it with the pre-stored inherent mass constants of the four-pile jacket, performs high-frequency time-domain differential processing on the vertical tension data to separate the static gravity load and the variable buoyancy load, and synchronously calculates the hydrodynamic resistance value in conjunction with real-time wave characteristic data.

[0055] To avoid data divergence where the vertical axial acceleration component cancels out gravitational acceleration due to severe wave pull on the four-pile jacket, leading to a denominator approaching zero, the main control computer presets a non-zero minimum effective apparent weight acceleration constant. The parameter estimator evaluates the algebraic sum of gravitational acceleration and instantaneous acceleration components, and enforces saturation limiting when the algebraic sum is lower than the minimum effective apparent weight acceleration constant. Based on the pre-set safety limit boundary, the parameter estimator applies the hydrodynamic added mass separation equation to calculate the real-time added mass value. The mathematical expression of the hydrodynamic added mass separation equation is defined as follows: ; In the formula, Characterization in continuous time The real-time added mass values ​​exhibited by the lower four pile jackets; Characterizing the load sensor in continuous time The vertical tension data of the main hook was collected from below; The real-time buoyancy values ​​of the four jacket piles are represented. The numerical values ​​representing the hydrodynamic resistance experienced by the four-pile jacket; Characterizes the gravitational acceleration of the work area; Characterizes the instantaneous acceleration components of the four-pile jacket in the vertical axis; Characterizes the intrinsic mass constant of the four-pile jacket in air.

[0056] The parameter estimator combines the deployed length data collected by the lifting encoder with the wave height sequence data input by the wave measuring radar to calculate the volume of the four-pile jacket entering the seawater and obtain the real-time buoyancy value. The main control computer has a pre-built dynamic mass matrix containing the inherent mass parameters of the jacket. The parameter estimator writes the calculated real-time additional mass value into the pre-built dynamic mass matrix, completing the real-time parameter update of the dynamic mass matrix. The updated dynamic mass matrix is ​​then transmitted to the optimization controller. The optimization controller uses the values ​​of the dynamic mass matrix to correct the prediction model parameters of the control vector, achieving real-time compensation for the dynamic characteristics of the four-pile jacket entering the water.

[0057] See attached document Figure 3 Combined with S400, the controller internal settings predict the time domain constant and pre-store the target reference state vector value; the target reference state vector value includes the target space six-degree-of-freedom pose parameters of the four-pile guide frame at the target pile insertion position.

[0058] At the current sampling moment, the optimized controller reads the state vector and dynamic mass matrix synchronously output by the parameter estimator, and simultaneously reads the disturbance vector and wave impact vector synchronously output by the feedforward solver. It then performs rolling optimization calculations within the predictive time domain constant formed from the current sampling moment to future moments. Considering the discrete computational characteristics of the digital controller, the optimized controller uses a fixed-step numerical integration rule at the low-level computational level to equivalently transform the continuous-time domain integral model into a discrete algebraic summation model, assisting in constructing the target cost function for nonlinear model predictive control. The mathematical expression of the target cost function is set as follows: ; In the formula, Characterization in continuous time The numerical value of the objective cost function of the optimized controller; Represents the current continuous time; Characterize the prediction time-domain constant; Characterizing the time-integral independent variable; Characterization in integration time The predicted numerical value of the state vector under the given conditions; Characterization in integration time The target reference state vector value is shown below; The weighted matrix representing the state deviation; The weighted matrix representing the control input; Characterization in integration time The predicted values ​​of the control vectors are as follows; The matrix transpose operator represents the matrix.

[0059] As a preferred implementation method, in order to ensure that the nonlinear model predictive control has a unique optimal analytical value in the solution process, and at the same time to avoid the singularity risk inside the matrix inversion operation, both the state deviation weighting matrix and the control input weighting matrix are constrained to be positive definite diagonal matrices.

[0060] The controller adjusts the diagonal element values ​​inside the state deviation weighting matrix. Based on the spatial motion tolerance of the four-pile guide frame required by the pile insertion guide structure, the weight of the corresponding dimension is assigned. The smaller the tolerance of the spatial degree of freedom direction, the larger the weight of the diagonal element is assigned. In this way, the control priority values ​​of the four-pile guide frame in different degree of freedom directions in three-dimensional space are set. By adjusting the diagonal element values ​​inside the control input weighting matrix, the change amplitude of the target rope speed command data from the first frequency conversion traction winch to the fourth frequency conversion traction winch is constrained.

[0061] The optimized controller uses the mass change constraints contained in the dynamic mass matrix, the geometric constraints contained in the Jacobian matrix of the jacket hoisting system, the hull disturbance contained in the disturbance vector, and the external thrust contained in the wave impact vector as the evolution boundary conditions for the predicted state vector values. By minimizing the target cost function value, it seeks the optimal control sequence within the predicted time domain constant that makes the spatial six-degree-of-freedom pose information of the four-pile jacket approximate the target reference state vector value.

[0062] During the multivariable predictive control (MRC) calculation, the optimization controller uses the electromechanical characteristics of the first to fourth variable frequency traction winches as hard constraint boundary conditions. It extracts a set of pre-stored electromechanical parameters from the main control computer, including the winch drum's moment of inertia, the winch reduction ratio, the drive motor's peak torque, the drum's winding radius, and the mechanical transmission efficiency constant. Simultaneously, it calculates the load-equivalent torque value based on the real-time tension value at the end of the traction cable. It sets an extreme value for the time rate of change of the target rope speed command data within the control vector and calculates the maximum allowable acceleration value of the target rope speed command data using the electromechanical inertial acceleration constraint formula. The mathematical expression of the electromechanical inertial acceleration constraint formula is set as follows: ; In the formula, The maximum permissible acceleration value representing the target rope velocity command data; Characterizes the peak torque of the drive motor; Characterizes the reduction ratio of the traction winch; Characterizing the efficiency constant of mechanical transmission; Characterizes the equivalent torque value of the load; Characterizes the moment of inertia of the traction winch drum; Characterizes the inherent mass constant of the four-pile jacket; Characterizing the four-pile jacket in continuous time Real-time additional mass value; Characterizes the winding radius of the traction winch drum.

[0063] The optimized controller constrains the real-time variation amplitude of the target rope speed command data within a closed interval defined by the negative maximum allowable acceleration value to the positive maximum allowable acceleration value. During the solution of the target cost function, if the first derivative of the target rope speed command data exceeds the maximum allowable acceleration value, the rate of change is forced to converge to the maximum allowable acceleration value. The electromechanical inertial acceleration constraint formula ensures that the control sequence output by the control vector conforms to the torque response limit of the drive motor. The electromechanical inertial hard constraint setting prevents overload shutdown of the frequency conversion traction component.

[0064] The optimized controller synchronously runs the asymmetric tension lower bound constraint algorithm, reads the wave impact vector output by the feedforward solver, extracts the basic pre-tension lower bound constant and tension lifting gain coefficient pre-stored in the main control computer, and obtains the instantaneous coordinate values ​​of the first force node to the fourth force node in the global origin coordinate system.

[0065] Based on the installation point coordinates of the first to fourth frequency conversion traction winches, the controller optimizes the calculation of the unit direction vector of the first to fourth traction cables. It then uses a dot product mapping between the wave impact vector and the unit direction vector to identify the relaxation force components generated by the wave impact load along the axial direction of each traction cable. Finally, it applies the asymmetric tension lower bound constraint formula to calculate the real-time tension lower bound value of each traction cable. The mathematical expression of the asymmetric tension lower bound constraint formula is set as follows: ; In the formula, Characterizing the first root traction cable in continuous time The lower limit value of the real-time tension; Characterizing the lower bound constant of the foundation pretension; Characterizing the tension lift gain coefficient; Characterizing wave impact vector; Characterizing the first The unit direction vector of the root traction cable.

[0066] The optimized controller incorporates the real-time tension lower bound value into the hard constraint set of the nonlinear model predictive control. During rolling optimization within the prediction time constant, it restricts the control vector solution space, ensuring that the predicted tension value of the traction cable on the wave-facing side always remains above the range specified by the real-time tension lower bound value. The asymmetric tension lower bound constraint algorithm counteracts the cable force drop caused by wave loads by raising the tension constraint benchmark of the traction cable on the wave-impacted side. It utilizes the asymmetric tension lower bound value to maintain the overall structural stiffness of the multi-cable parallel drive topology network, eliminating cable slack and displacement overshoot generated during the wave-impacted phase of the four-pile jacket structure.

[0067] In conjunction with the S500, the optimized controller completes multivariate predictive control calculations and outputs the optimal control sequence at the current sampling time, extracting the target rope speed command data corresponding to the next time step within the optimal control sequence. The optimized controller's internal preset command smoothing filter utilizes preset damping ratio coefficients, cutoff angular frequency constants, and the data transmission cycle of the fieldbus network to perform a second-order low-pass filter on the target rope speed command data. This eliminates numerical jumps in the target cost function at the rolling optimization time domain boundary, outputting smoothed target rope speed data. The mathematical expression followed by the command smoothing filter is set as follows: ; In the formula, Characterized at discrete sampling time The smoothed target rope velocity data below; Characterized at discrete sampling time The target rope speed command data is calculated by the optimization controller. Characterizing the damping ratio coefficient; Characterized by the cutoff angular frequency constant; Characterizes the data transmission cycle of a fieldbus network; Characterized at discrete sampling time The smoothed target rope velocity data below; Characterized at discrete sampling time The smoothed target rope velocity data.

[0068] The main control computer uses a fieldbus network to send the smoothing target rope speed data to the frequency conversion traction component. The first to fourth frequency conversion traction winches synchronously receive the smoothing target rope speed data and use the frequency converter to drive the motor to adjust the speed, so that the first to fourth traction cables work together to pull the four-pile guide frame closer to the target reference state vector value according to the measured deviation of the state vector.

[0069] When the bottom of the four-pile jacket enters the interior of the preset pile insertion guide structure, the optimization controller lowers the velocity component value inside the target reference state vector value, and the frequency conversion traction component works with the smooth target rope speed data to maintain the verticality and horizontal displacement of the four-pile jacket within the deviation threshold limited by the target reference state vector value based on the pose information contained in the state vector.

[0070] In this embodiment, to match the electromechanical transmission characteristics of the variable frequency traction assembly, the damping ratio coefficient is preset to an overdamped value range of 0.7 to 0.9, and the cutoff angular frequency constant is set according to the inherent mechanical resonant frequencies of the first to fourth variable frequency traction winches. The command smoothing filter reduces the high-frequency components of the target rope speed command data and eliminates step jumps at the data change edges, preventing the target rope speed command data from exciting resonance phenomena within the multi-cable parallel drive topology network, effectively eliminating mechanical oscillations, and ensuring that the four-pile jacket can complete high-precision pile insertion.

[0071] During the lowering of the four-pile jacket into the splash zone, the jacket is subjected to wave impact. The parameter estimator calculates the real-time added mass value, and the feedforward solver outputs the real-time changing wave impact vector based on the wave height sequence data. The optimization controller inputs the dynamic mass matrix and wave impact vector into the target cost function of the nonlinear model predictive control, dynamically adjusts the control input weighting matrix and the state deviation weighting matrix, and generates updated smoothed target rope velocity data through the command smoothing filter and sends it out.

[0072] The first to fourth frequency conversion traction winches execute speed adjustment commands, and combined with the lower limit value of asymmetric tension to counteract the overturning moment generated by the wave impact vector, constrain the horizontal displacement and verticality deviation of the four-pile jacket within the splash zone based on the state vector representation to be within the set value range, thus completing the splash zone crossing.

[0073] As the four-pile jacket approaches the pre-set subsea pile guide structure, the positioning base station and high-precision inertial navigation system output high-frequency motion data and low-frequency position data. The parameter estimator outputs a corrected state vector through high- and low-frequency asynchronous filtering. The optimized controller compares the pose information contained in the state vector with the target reference state vector value, and reduces the velocity component value contained in the target reference state vector value within the prediction time domain constant, limiting the amplitude of the smoothed target rope velocity data. The first to fourth frequency-converter traction winches perform low-speed adjustment actions according to the smoothed target rope velocity data, making the axis of the bottom pile legs of the four-pile jacket coincide with the central axis of the pre-set pile guide structure, completing the pile leg docking action.

[0074] The parameter estimator synchronously monitors the signal feedback status of the positioning base station, high-precision inertial navigation system, and the first to fourth tension sensors in real time. Based on the data update cycle threshold set internally by the main control computer, if the three-dimensional absolute coordinate data output by the positioning base station does not change within the data update cycle threshold, the parameter estimator determines that the positioning base station has lost sensor data packets. During the specified duration, it stops executing the measurement correction term calculation and starts the short-time pure prediction algorithm. It extracts the state estimation value of the previous sampling moment, the state transition matrix of the four-pile jacket, the input control matrix of the four-pile jacket, and the control input value, and performs recursive calculations to generate short-time pure prediction state values. These values ​​are encapsulated into a state vector and sent to the optimization controller to maintain the rolling optimization calculation of the target cost function, ensuring that the first to fourth frequency conversion traction winches maintain the continuous output of the target rope speed command data during the signal loss period.

[0075] When the positioning base station resumes signal output and receives new three-dimensional absolute coordinate data, the parameter estimator recalculates the numerical residual between the three-dimensional absolute coordinate data and the short-time pure prediction state value, uses the Kalman gain matrix to eliminate the accumulated prediction deviation value, completes the numerical reset of the state vector, and exits the short-time pure prediction algorithm, thus eliminating the risk of control vector mutation caused by signal interruption.

[0076] During the distribution of smooth target rope speed data, the optimization controller synchronously and cyclically extracts real-time tension values ​​from the first to the fourth tension sensors. In this embodiment, the main control computer pre-sets a maximum safe tension threshold based on the material breaking tensile strength limit of the first to fourth traction cables and a safety redundancy coefficient, and simultaneously sets the emergency degradation response duration constant and tension control gain coefficient for the corresponding time judgment window. If the real-time tension value of any traction cable exceeds the maximum safe tension threshold and the duration reaches the emergency degradation response duration constant, the optimization controller immediately interrupts the posture closed-loop adjustment and calculates the tension protection compensation component using the tension limit protection criterion equation. The mathematical expression of the tension limit protection criterion equation is set as follows: ; In the formula, Characterizing the first Tension protection compensation component of the root traction cable; The tension control gain coefficient is a proportional amplification constant used to convert the tension over-limit difference into a speed compensation command. Characterizing the first Real-time tension value of the root traction cable; Characterizes the maximum safe tension threshold.

[0077] The optimization controller superimposes the tension protection compensation component onto the smoothed target rope speed data, adjusts the output torque target value of the variable frequency traction component to execute the active cable release action of the traction cable. The frequency converters inside the first to fourth variable frequency traction winches switch to torque limiting operation mode.

[0078] When extreme surges cause a sudden increase in tension in the traction cable on the wave-facing side, the variable frequency traction assembly releases the accumulated load stress within the first to fourth traction cables by increasing the cable release speed, forcibly limiting the real-time tension value within the maximum safe tension threshold range. After the real-time tension value falls back below the maximum safe tension threshold, the optimized controller withdraws the tension protection compensation component and reconnects to the multivariable predictive control solution. By relying on torque constraints combined with speed compensation, the risk of traction cable breakage and damage to the variable frequency traction assembly under extreme surge conditions is eliminated.

[0079] Specific application examples: Based on the specific on-site hoisting and pile driving operations of the four-pile jacket foundation for offshore wind power, the constants and initial parameters set inside the main control computer are as follows: The minimum calculated tension constant in the equivalent cable length calculation equation is set to a minimum constant of 0.01N; the damping ratio coefficient in the target cost function is set to 0.8; the maximum safe tension threshold of the variable frequency traction component is set to 80000.0N; the lower limit constant of the foundation pretension is set to 30000.0N; the tension lifting gain coefficient is set to 0.5; and the tension control gain coefficient is set to 0.005.

[0080] During the process of cable net geometry mapping and compensation performed by the parameter estimator, the main control computer eliminates deformation coupling factors from the measured cable length. At a certain discrete sampling moment, the parameter estimator extracts the measured cable length value of the first traction cable fed back by the first absolute encoder as 50.0 m. Combining this with the current mechanical operating state, the parameter estimator extracts the real-time tension value at the end of the traction cable collected by the first tension sensor as 50000.0 N, and extracts the traction cable linear density constant as 20.0 kg / m from the basic constant group, and the gravitational acceleration of the working area as 9.8 m / s². 2 The elastic modulus is 1.0 × 10¹¹ Pa, and the cable cross-sectional area is 0.005 m². The parameter estimator is substituted into the equivalent cable length calculation equation to calculate the true geometric span after deformation elimination. First, the sum of the elastic tensile reduction term and the catenary sag compensation term is calculated, and this algebraic sum is set as the traction cable deformation compensation amount. : ; Then, substituting the measured cable length into the actual cable length, the equivalent cable length of the first traction cable is calculated by continuous subtraction: ; The parameter estimator uses the calculated 49.915m as a purely rigid spatial span and adds it to the Jacobian matrix of the jacket erection system, thereby eliminating the false extension caused by the elastic tension of the traction cable and the sag of its own weight, and preventing the mapping distortion of the topology network driven by multiple cables in parallel.

[0081] When the main control computer deals with wave impacts encountered by the four-pile jacket as it traverses the splash zone, the optimized controller dynamically quantifies the anti-slackening lower limit. The feedforward solver extracts the real-time changing wave impact vector based on the wave height sequence data, determining the peak value of the wave impact vector at that sampling moment to be 100000.0N. Simultaneously, the dot product projection coefficient between the unit direction vector of the first traction cable and the wave impact vector at that moment is calculated. 0.8. The real-time lower limit value of the tension of the traction cable on the wave-facing side is calculated by substituting the asymmetric tension lower limit constraint formula into the optimized controller: ; The calculated real-time tension lower limit value of 70000.0N is written into the hard constraint condition set of the nonlinear model predictive control to constrain the optimal control sequence search space of the optimization controller, ensuring that the traction cable is completely relaxed instantaneously without causing damage during the multivariable predictive control solution process.

[0082] As the smoothing target rope speed data was continuously transmitted and executed, the first to fourth frequency conversion traction winches completed multiple rounds of dynamic adjustments. In the latest round of mechanical response encountering extreme swells, the main control computer extracted a real-time tension value of 82000.0N from the first traction cable. Since the real-time tension value had exceeded the preset maximum safe tension threshold of 80000.0N, and the duration had reached the emergency degrade response duration constant, the optimization controller immediately substituted the tension limit protection criterion equation to calculate the tension protection compensation component: ; Once the judgment condition is triggered, the optimization controller immediately interrupts the posture closed-loop adjustment, adds the calculated 10.0 mm / s as the tension protection compensation component to the smooth target rope speed data, and forces the frequency conversion traction component to perform the active cable release action of the traction cable to eliminate the risk of traction cable breakage under extreme swell conditions.

[0083] Actual operation tests were conducted and data were compared. Specific verification results are detailed in the appendix. Figure 4 and attached Figure 5 The experiment compared field operation data from the traditional manual experience-based constant tension hoisting method with the method of this invention under the same sea state and initial distribution conditions of the same crane vessel. The experimental data are recorded in Table 1.

[0084] Table 1. Comparison of operational data between traditional manual experience-based constant tension hoisting method and the method of this invention. See attached document Figure 4 And Table 1, Appendix Figure 4 This diagram compares the X-axis displacement convergence of four guide frames during the hoisting and insertion of piles through the splash zone. The horizontal axis represents the continuous operation time in seconds (s), ranging from 0 to 100. The vertical axis represents the X-axis displacement in meters (m), ranging from -1.0 to 1.0. The diagram includes displacement curves from the conventional method and the displacement convergence curves from the method of this invention.

[0085] During the initial water entry phase of continuous operation from 0 to 20 seconds, traditional methods fail to decouple the deformation of the traction cable from spatial wave disturbances, leading to a reference offset in the determination of the target reference state vector. This results in severe fluctuations in the X-axis displacement, with swaying occurring later, and the maximum displacement overshoot reaching 0.85m, which cannot be quickly eliminated. The method of this invention extracts the equivalent cable length to eliminate the elastic expansion and contraction of the traction cable and performs feedforward spatial mapping based on wave height sequence data. The displacement convergence curve of this method exhibits a stable and small-amplitude oscillation decay trend, quickly penetrating the 0.05m allowable tolerance threshold at 40 seconds. Ultimately, the maximum X-axis displacement overshoot is strictly controlled within 0.12m, completely eliminating the spatial displacement deviation accumulation phenomenon caused by blind constant tension following.

[0086] See attached document Figure 5 And Table 1, Appendix Figure 5 This diagram illustrates the force state and boundary control of the traction cable during the adjustment process of the variable frequency traction assembly. The horizontal axis represents the continuous operating time in seconds (s), ranging from 0 to 100. The vertical axis represents the real-time tension value of the first traction cable in kilometres (N), ranging from 10,000 to 90,000. The diagram includes tension fluctuation curves using conventional methods, the real-time tension control curve of this invention, and the lower bound constant line of the basic pretension and the maximum safe tension threshold line configured in this invention.

[0087] When encountering concentrated wave impacts during continuous operation for approximately 60 seconds, the hydrodynamic resistance of the four-pile jacket increases sharply. Due to the lack of a multivariate constraint prediction model, the traditional method causes the main hook cable release action to be delayed, directly exceeding the mechanical bearing limit and leading to sudden over-limit or slack. The tension fluctuation curve of the traditional method shows a maximum tension range of 68500.0N, and slack phenomena below the safety benchmark occur at the wave trough.

[0088] The method of this invention dynamically updates the real-time tension lower bound value through a feedforward solver and constructs a set of hard constraint conditions in the optimization controller. As can be seen from the real-time tension control curve of this invention, when the real-time tension approaches the maximum safe tension threshold of 80000.0N, the system automatically adds a tension protection compensation component to actively release the cable, strictly limiting its peak value to near the threshold (the highest extreme value only momentarily reaches 82000.0N before being forcibly pulled back); simultaneously, during the wave pullback phase, an asymmetric tension lower bound algorithm is used to maintain the tension above the foundation pretension lower bound constant of 30000.0N. This ensures that the force on the traction cable remains within the safe tension resistance envelope throughout the entire hoisting adjustment cycle, with zero overload and relaxation occurrences, guaranteeing the safety of on-site pile driving operations.

Claims

1. A method for hoisting a four-pile jacket foundation for offshore wind power, characterized in that, Includes the following steps: The parameter estimator extracts the measured cable length and the group of basic constants to calculate the equivalent cable length, constructs the cable net geometric mapping and compensation, analyzes the spatial positional relationship between the crane ship and the four-pile jacket, and uses the equivalent cable length to correct the Jacobian matrix of the jacket hoisting system. The feedforward solver reads wave height sequence data and six-degree-of-freedom motion attitude data, and obtains the horizontal offset distance value based on the spatial position relationship, so as to perform wave feedforward and spatial mapping, and generate disturbance vector, wave impact vector and real-time wave characteristic data. The parameter estimator acquires the real-time wave characteristic data, combines the vertical tension data and the release length data to carry out water entry parameter collaborative observation, and outputs the state vector and dynamic mass matrix of the four-pile jacket; The optimized controller uses the dynamic mass matrix, the Jacobian matrix of the jacket hoisting system, the disturbance vector, and the wave impact vector as state evolution boundary conditions to perform multivariate predictive control calculations and update the target rope velocity command data. The main control computer sends the target rope speed command data to the frequency conversion traction component to coordinate the completion of the stake insertion action.

2. The method for hoisting a four-pile jacket foundation for offshore wind power according to claim 1, characterized in that, The process of constructing cable net geometry mapping and compensation includes: Obtain the starting reference point of the vertical line where the main hook is located in front of the bow of the crane vessel, and extract the three-dimensional absolute coordinates of the first, second, third and fourth variable frequency traction winches included in the variable frequency traction assembly, as well as the three-dimensional relative coordinates of the first, second, third and fourth force-bearing nodes within the coordinate system of the jacket center of mass. The main sling and the first, second, third and fourth traction cables are combined and set as a multi-cable parallel drive topology network. Based on the spatial vector boundary constraints contained in the multi-cable parallel drive topology network, a kinematic differential transformation matrix between the control vector and the state vector containing the target rope speed command data is constructed in the global origin coordinate system. The kinematic differential transformation matrix is ​​defined as the Jacobian matrix of the jacket hoisting system.

3. The method for hoisting a four-pile jacket foundation for offshore wind power according to claim 2, characterized in that, The process of correcting the Jacobian matrix of the jacket erection system using the equivalent cable length includes: Before substituting the real-time tension value at the end of the traction cable into the parameter estimator, a condition comparison and judgment is performed. If the real-time tension value at the end of the traction cable is lower than the minimum calculated tension constant, the minimum calculated tension constant is forcibly substituted into the denominator of the equivalent cable length calculation equation for calculation. The spatial geometric constraint side length values ​​inside the multi-cable parallel driving topology network are updated based on the equivalent cable length, and the set of constant coefficients of the Jacobian matrix of the jacket erection system is corrected using the equivalent cable length.

4. The method for hoisting a four-pile jacket foundation for offshore wind power according to claim 1, characterized in that, The process of performing wave feedforward and spatial mapping includes: The feedforward solver performs a convolution operation on the wave height program sequence data and the crane ship amplitude response operator matrix in the time dimension to generate the disturbance vector. It then performs a deviation comparison operation between the disturbance vector and the six-degree-of-freedom motion attitude data, removes high-frequency measurement noise components, and outputs the filtered disturbance vector. The lag time of the wave reaching the wave front is calculated using the horizontal offset distance and wave propagation speed. Based on the lag time, a time axis translation transformation is performed on the wave height sequence data to generate the real-time wave feature data. The real-time wave feature data and the effective pressure area are mapped and calculated to output the wave impact vector.

5. The method for hoisting a four-pile jacket foundation for offshore wind power according to claim 1, characterized in that, The process of conducting coordinated observation of water-entry parameters includes: The parameter estimator extracts the three-dimensional absolute coordinate data output by the positioning base station and performs high- and low-frequency asynchronous filtering. It retrieves the historical state vector at the corresponding moment from the historical state memory and uses the three-dimensional acceleration value and the three-dimensional angular velocity value to calculate and generate the predicted state vector for the current moment. Calculate the numerical residual between the three-dimensional absolute coordinate data and the predicted state vector, use the Kalman gain matrix to eliminate the cumulative prediction deviation value of the spatial position deviation caused by the communication delay of the positioning base station, correct the predicted state vector, and output the corrected state vector after eliminating the spatial position deviation to update the state vector.

6. The method for hoisting a four-pile jacket foundation for offshore wind power according to claim 5, characterized in that, The process of conducting coordinated observation of water-entry parameters also includes: The parameter estimator extracts the vertical axial acceleration component from the three-dimensional acceleration values, performs high-frequency time-domain differential processing on the vertical tension data to separate the static gravity load and the variable buoyancy load, evaluates the algebraic sum of the gravitational acceleration and instantaneous acceleration components, and forcibly executes saturation limiting when the algebraic sum is lower than the minimum effective apparent weight acceleration constant. The real-time added mass value is calculated based on the hydrodynamic added mass separation equation applied to the pre-safe amplitude limit boundary. The real-time buoyancy value is calculated by combining the released length data and the wave height program sequence data. The real-time added mass value is written into the dynamic mass matrix to complete the real-time parameter update of the dynamic mass matrix.

7. The method for hoisting a four-pile jacket foundation for offshore wind power according to claim 1, characterized in that, The process of performing multivariate predictive control solution includes: The optimized controller internally sets a prediction time domain constant and pre-stores the target reference state vector value containing the target space six-degree-of-freedom pose parameters of the four-pile guide frame at the target pile insertion position. At the underlying computation level, it uses the fixed step size numerical integration rule to transform the continuous time domain integral model into a discrete algebraic summation model, which helps to construct the target cost function of nonlinear model predictive control. By minimizing the value of the target cost function, the optimal control sequence is sought within the predicted time domain constant, such that the spatial six-degree-of-freedom pose information of the four-pile jacket approximates the value of the target reference state vector.

8. The method for hoisting a four-pile jacket foundation for offshore wind power according to claim 7, characterized in that, During the execution of multivariable predictive control solution, the optimization controller uses the electromechanical and dynamic characteristics of the first, second, third, and fourth variable frequency traction winches included in the variable frequency traction assembly as hard constraint boundary conditions: The load-converted torque value is calculated based on the real-time tension value at the end of the traction cable. The maximum allowable acceleration value of the target rope speed command data is calculated using the electromechanical inertial acceleration constraint formula. The real-time change amplitude of the target rope speed command data is constrained within a closed interval consisting of the negative maximum allowable acceleration value to the positive maximum allowable acceleration value. During the solution of the target cost function, if the first derivative of the target rope speed command data exceeds the maximum allowable acceleration value, the rate of change is forced to converge to the maximum allowable acceleration value.

9. A method for hoisting a four-pile jacket foundation for offshore wind power according to claim 7, characterized in that, The optimization controller synchronously runs the asymmetric tension lower bound constraint algorithm: Obtain the instantaneous coordinate values ​​of the first, second, third, and fourth force-bearing nodes included in the four-pile jacket in the global origin coordinate system. Combined with the installation point coordinates of the variable frequency traction component, calculate the unit direction vectors of the first, second, third, and fourth traction cables connected between the variable frequency traction component and the four-pile jacket. The relaxation force components generated by the wave impact vector and the unit direction vector are identified by performing a dot product mapping. The real-time tension lower limit value of each traction cable is calculated by applying the asymmetric tension lower limit constraint formula. The real-time lower bound value of tension is written into the hard constraint condition set of the nonlinear model predictive control, which restricts the solution space of the control vector, so that the predicted value of the tension of the traction cable on the wave-facing side is always above the numerical range specified by the real-time lower bound value.

10. The method for hoisting a four-pile jacket foundation for offshore wind power according to claim 1, characterized in that, During the period when the main control computer issues the target rope speed command data: The optimization controller uses a command smoothing filter to perform a second-order low-pass filter on the target rope speed command data and outputs smooth target rope speed data, while simultaneously and cyclically extracting the real-time tension values ​​fed back by the first tension sensor, second tension sensor, third tension sensor and fourth tension sensor configured in the frequency conversion traction component; Based on the material breaking tensile strength limit of the first, second, third, and fourth traction cables connected between the variable frequency traction assembly and the four-pile guide frame, and in conjunction with the safety redundancy coefficient, a maximum safe tension threshold is pre-set. If the real-time tension value of any traction cable exceeds the maximum safe tension threshold and the duration reaches the emergency degradation response duration constant, the optimization controller immediately interrupts the posture closed-loop adjustment and calculates the tension protection compensation component using the tension limit protection criterion equation. The tension protection compensation component is superimposed on the smooth target rope speed data, and the output torque target value of the variable frequency traction component is adjusted to perform an active cable release action.

Images