Offshore operation platform-oriented stabilization intelligent control method and system

Abstract

The invention particularly relates to an intelligent stabilization control method and system for an offshore operation platform, and belongs to the technical field of intelligent control. The method comprises the following steps: acquiring platform attitude change trend data; acquiring short-time attitude prediction data of the platform; calculating current attitude deviation data of the platform; calculating platform prediction attitude deviation data; constructing a platform attitude control fusion matrix; inputting the platform attitude control fusion matrix into an anti-rolling gyroscope control module to generate a platform attitude real-time control signal; according to the platform attitude real-time control signal, an anti-rolling gyroscope is driven to execute a platform attitude correction action; and after executing the platform attitude correction action, the stabilization gyroscope re-triggers the next round of attitude control of the platform so as to collect the updated current attitude data of the platform. The system implements the steps of the method. According to the invention, the real-time performance, the control precision and the robustness of stabilization control of the offshore operation platform are improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent control technology, and in particular to an intelligent control method and system for reducing roll on offshore operating platforms. Background Technology

[0002] Offshore platforms are widely used in energy extraction, marine engineering construction, and scientific research. However, the marine environment in which these platforms operate is complex and changeable, frequently affected by environmental factors such as wind, waves, and currents, causing continuous swaying motion. This swaying not only significantly reduces the efficiency and safety of platform operations, but long-term, severe swaying can also cause mechanical damage and structural fatigue to the platform and related equipment, and in severe cases, even threaten personnel safety.

[0003] To address the aforementioned issues, active roll stabilization using roll stabilization gyroscopes has gradually become the industry mainstream. Roll stabilization gyroscopes rely on a high-speed rotating flywheel; by controlling its precession angle and speed, they generate a counter-torque to counteract changes in the platform's attitude, achieving active stabilization of the platform's posture. However, existing roll stabilization gyroscope control systems generally suffer from several prominent problems. These include a significant time delay between sensing changes in platform attitude and outputting an effective roll stabilization control signal, severely impacting real-time control performance. Furthermore, the actual dynamic characteristics of the platform structure are complex, with significant rigid-flexible coupling effects. Existing technologies do not adequately consider the structural flexibility characteristics, leading to decreased control accuracy and robustness. Simultaneously, existing attitude prediction technologies fail to effectively capture the dynamic characteristics and trends of platform attitude data, reducing the accuracy of forward-looking roll stabilization control. Moreover, current methods for fusing real-time attitude data with short-term predicted attitude data are relatively simple, failing to effectively integrate time-frequency characteristic information in the data, further reducing the reliability of the simplified control signal. Finally, existing roll stabilization gyroscope actuators typically use a single fixed control parameter, neglecting load variation characteristics, making it difficult to adjust the control strategy in real time, thus affecting the accuracy of attitude correction actions. Summary of the Invention

[0004] The purpose of this invention is to address the problems of poor real-time performance, low control accuracy, and poor robustness in existing roll reduction control systems for offshore platforms, and to provide an intelligent roll reduction control method and system for offshore platforms, thereby improving the real-time performance, control accuracy, and robustness of roll reduction control for offshore platforms.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] A roll reduction intelligent control method for offshore operating platforms includes:

[0007] S101. Real-time acquisition of the platform's current attitude data; inputting the platform's current attitude data into a pre-built platform attitude feature extraction model to obtain platform attitude change trend data;

[0008] S102. Generate initial parameters for the platform attitude prediction model based on historical platform attitude data; obtain short-term platform attitude prediction data based on platform attitude change trend data and the initial parameters of the platform attitude prediction model.

[0009] S103. Calculate the current attitude deviation data of the platform based on the current attitude data and the platform structural stiffness parameters; calculate the predicted attitude deviation data of the platform based on the short-term attitude prediction data and the platform structural stiffness parameters.

[0010] S104. Construct a platform attitude control fusion matrix based on the platform's current attitude deviation data and the platform's predicted attitude deviation data; input the platform attitude control fusion matrix into the anti-roll gyroscope control module to generate a real-time platform attitude control signal;

[0011] S105. Drive the anti-roll gyroscope to perform platform attitude correction actions according to the real-time control signal of the platform attitude; after the anti-roll gyroscope performs the platform attitude correction actions, re-trigger S101 to collect the updated current attitude data of the platform.

[0012] As one feasible approach, the construction of a platform pose feature extraction model includes:

[0013] Empirical mode decomposition was performed on historical platform attitude data to obtain multiple modal components of different frequencies;

[0014] Perform Hilbert transform on each modal component to extract the instantaneous frequency and instantaneous amplitude of each modal component;

[0015] A feature matrix is ​​constructed using instantaneous frequency and instantaneous amplitude, which is then input into the autoencoder model to output platform attitude change trend data.

[0016] The training process of the autoencoder model includes:

[0017] Construct an autoencoder network consisting of an encoder and a decoder;

[0018] The autoencoder network is trained using the feature matrix until the reconstruction error converges, and the trained platform posture feature extraction model is output.

[0019] As one possible approach, the generation of initial parameters for the platform attitude prediction model includes:

[0020] A long short-term memory network is used to extract the temporal features of historical platform posture from historical platform posture data;

[0021] The parameters of the long short-term memory network are dynamically adjusted based on the historical platform attitude prediction error to determine the initial parameters of the platform attitude prediction model.

[0022] Specifically, the parameters of the Long Short-Term Memory network are dynamically adjusted based on historical platform attitude prediction errors to determine the initial parameters of the platform attitude prediction model, including:

[0023] A historical platform attitude prediction error trend function is established. Based on the gradient of the historical platform attitude prediction error trend function, the parameters of the long short-term memory network are dynamically adjusted using the backpropagation algorithm.

[0024] Repeat the adjustment until the historical platform attitude prediction error trend function meets the preset convergence condition. At this point, the long short-term memory network is the platform attitude prediction model, and the parameters of the long short-term memory network are the initial parameters of the platform attitude prediction model.

[0025] One possible approach is to calculate the platform's current attitude deviation data, including:

[0026] Extract the angular velocity differences of the platform's roll, pitch, and yaw based on the platform's current attitude data;

[0027] A rigid-flexible coupled dynamic model is established by combining the platform's structural stiffness parameters. The finite element numerical method is used to solve the rigid-flexible coupled dynamic model and output the platform's current attitude deviation data.

[0028] As one feasible method, obtaining the platform structure stiffness parameters includes:

[0029] Establish a finite element structural model of the platform and perform modal analysis of the platform structure;

[0030] The platform structure stiffness matrix is ​​obtained based on the platform structure modal analysis results, and the platform structure stiffness parameters are output.

[0031] As one feasible approach, the construction of the platform attitude control fusion matrix includes:

[0032] Fourier transforms are performed on the current attitude deviation data and the predicted attitude deviation data of the platform to obtain the spectrum of the platform attitude deviation data.

[0033] The platform attitude deviation data is subjected to joint time-frequency decomposition to obtain the platform attitude fusion weight matrix, and the platform attitude control fusion matrix is ​​constructed accordingly.

[0034] The acquisition of the platform posture fusion weight matrix includes:

[0035] The objective function of spectral entropy is defined based on the frequency band energy density matrix. The objective function is optimized using the differential evolution algorithm to obtain the fusion weight matrix of platform attitude.

[0036] One feasible approach is to drive the anti-roll gyroscope to perform platform attitude correction actions based on real-time platform attitude control signals, including:

[0037] The target rotational speed and target precession angular velocity of the anti-roll gyroscope flywheel are calculated based on the real-time control signal of the platform attitude.

[0038] A dual-closed-loop platform attitude controller for the rotational speed and precession angular velocity of the roll stabilizing gyroscope flywheel is constructed, which outputs the drive signal for the roll stabilizing gyroscope flywheel motor to control the roll stabilizing gyroscope to perform attitude correction actions.

[0039] One feasible approach involves constructing a dual-closed-loop platform attitude controller for the rotational speed and precession angular velocity of the roll stabilization gyroscope flywheel. This controller outputs drive signals to the roll stabilization gyroscope flywheel motor, controlling the roll stabilization gyroscope to perform attitude correction actions, including:

[0040] The inner-loop platform attitude control loop is constructed based on the difference between the target speed and the actual speed of the anti-roll gyroscope flywheel;

[0041] The outer-loop platform attitude control loop is constructed using the difference between the target precession angular velocity and the actual precession angular velocity of the anti-roll gyroscope flywheel.

[0042] The control parameters of the inner loop platform attitude control loop are dynamically adjusted using the output of the outer loop platform attitude control loop.

[0043] One feasible approach is to dynamically adjust the control parameters of the inner loop platform attitude control loop using the output of the outer loop platform attitude control loop, including:

[0044] Based on the output torque of the anti-roll gyroscope flywheel motor and the actual speed of the anti-roll gyroscope flywheel, the control parameters of the inner loop platform attitude control loop are dynamically adjusted using an incremental adaptive control algorithm.

[0045] The load model parameters of the anti-roll gyroscope flywheel are updated in real time using an online load model identification method, and the control parameters of the outer ring platform attitude control loop are dynamically adjusted.

[0046] This invention also provides an intelligent control system for roll reduction of offshore operating platforms, which implements the aforementioned intelligent control method for roll reduction of offshore operating platforms, including:

[0047] The data acquisition module is used to obtain real-time platform current attitude data; it inputs the platform's current attitude data into a pre-built platform attitude feature model to obtain platform attitude change trend data.

[0048] The prediction acquisition module is used to generate initial parameters for the platform attitude prediction model based on historical platform attitude data; and to obtain short-term platform attitude prediction data based on platform attitude change trend data and the initial parameters of the platform attitude prediction model.

[0049] The prediction calculation module is used to calculate the current attitude deviation data of the platform based on the platform's current attitude data and the platform's structural stiffness parameters; and to calculate the predicted attitude deviation data of the platform based on the platform's short-time attitude prediction data and the platform's structural stiffness parameters.

[0050] The signal generation module is used to construct a platform attitude control fusion matrix based on the platform's current attitude deviation data and the platform's predicted attitude deviation data; and inputs the platform attitude control fusion matrix into the anti-roll gyroscope control module to generate a real-time platform attitude control signal.

[0051] The correction module is used to drive the roll stabilizer gyroscope to perform platform attitude correction actions based on the real-time control signal of the platform attitude. After the roll stabilizer gyroscope performs the platform attitude correction actions, it re-triggers the data acquisition module to collect updated platform current attitude data.

[0052] Beneficial technical effects of the present invention:

[0053] The present invention provides an intelligent roll reduction control method and system for offshore operating platforms. By acquiring platform attitude data in real time and utilizing an attitude feature extraction method combining autoencoder networks and empirical mode decomposition, the signal processing delay from platform attitude perception to platform attitude control output is significantly reduced, effectively improving the real-time response performance of platform roll reduction control. By establishing a rigid-flexible coupling dynamic model and combining it with an attitude prediction method based on long short-term memory networks, high-precision short-time prediction of platform attitude is achieved, effectively solving the problems of insufficient consideration of structural flexible response and low attitude prediction accuracy in existing roll reduction control technologies, and improving the accuracy and reliability of platform attitude control signals. By constructing a platform attitude fusion weight matrix, using time-frequency joint decomposition and differential evolution algorithms for weight optimization, and further utilizing dual closed-loop dynamic adaptive control technology based on flywheel speed and precession angular velocity to adjust the parameters of the roll reduction gyroscope actuator in real time, the accuracy and robustness of platform attitude correction actions are ensured, effectively improving the overall stability and safety of the platform under complex sea conditions. Attached Figure Description

[0054] Figure 1 A flowchart of one embodiment of the intelligent roll reduction control method for offshore operation platforms of the present invention;

[0055] Figure 2 This is a schematic structural view of an embodiment of the intelligent control system for roll reduction of offshore operating platforms according to the present invention. Detailed Implementation

[0056] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains; the terminology used herein in the specification is for the purpose of describing particular embodiments only and is not intended to limit the application; the terms “comprising” and “equivalent to”, and any variations thereof, in the specification, claims, and foregoing description of the drawings are intended to cover non-exclusive inclusion.

[0057] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0058] The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings and specific embodiments.

[0059] See Figure 1 This embodiment provides a roll reduction intelligent control method for offshore operation platforms, including:

[0060] S101. Real-time acquisition of the platform's current attitude data; inputting the platform's current attitude data into a pre-built platform attitude feature extraction model to obtain platform attitude change trend data;

[0061] S102. Generate initial parameters for the platform attitude prediction model based on historical platform attitude data; obtain short-term platform attitude prediction data based on platform attitude change trend data and the initial parameters of the platform attitude prediction model.

[0062] S103. Calculate the current attitude deviation data of the platform based on the current attitude data and the platform structural stiffness parameters; calculate the predicted attitude deviation data of the platform based on the short-term attitude prediction data and the platform structural stiffness parameters.

[0063] S104. Construct a platform attitude control fusion matrix based on the platform's current attitude deviation data and the platform's predicted attitude deviation data; input the platform attitude control fusion matrix into the anti-roll gyroscope control module to generate a real-time platform attitude control signal;

[0064] S105. Drive the anti-roll gyroscope to perform platform attitude correction actions according to the real-time control signal of the platform attitude; after the anti-roll gyroscope performs the platform attitude correction actions, re-trigger S101 to collect the updated current attitude data of the platform.

[0065] In this embodiment, in the actual operating environment, the attitude data of the platform is acquired in real time by inertial attitude sensors deployed at key locations on the offshore operating platform. The inertial attitude sensors can be a combination of high-precision three-axis gyroscopes, three-axis accelerometers and three-axis tilt sensors to measure the platform's key attitude angles such as roll, pitch and yaw and their angular velocity information in real time, and continuously transmit the measured data to the data processing unit through real-time data communication protocols such as EtherCAT or CAN bus.

[0066] In practical implementation, a combined inertial navigation system, such as the MTi-G-710 inertial navigation system, is used to measure the platform's current roll, pitch, and yaw angles and angular velocities in real time, forming the platform's current attitude data. This data is continuously output at a frequency of 50Hz and formatted into a matrix.

[0067] X t =[θ x (t),θ y (t),θ z (t),ω x (t),ω y (t),ω z (t)];

[0068] Where: θ x (t),θ y (t),θ z (t) represents the current angles of the platform's roll, pitch, and yaw, respectively; ω x (t),ω y (t),ω z (t) represents the current angular velocities of the platform's roll, pitch, and yaw, respectively.

[0069] In this embodiment, the construction of the platform posture change feature extraction model includes:

[0070] Empirical mode decomposition was performed on historical platform attitude data to obtain multiple modal components of different frequencies;

[0071] Perform Hilbert transform on each modal component to extract the instantaneous frequency and instantaneous amplitude of each modal component;

[0072] A feature matrix is ​​constructed using instantaneous frequency and instantaneous amplitude, which is then input into the autoencoder model to output platform attitude change trend data.

[0073] In this embodiment, empirical mode decomposition is performed on historical platform attitude data to obtain multiple modal components of different frequencies, including:

[0074] First, historical platform attitude data is collected to form a historical platform attitude data sequence, which is denoted as:

[0075] X H =[X t-n ,X t-n+1 ,…,X t ];

[0076] Among them, X t The platform's attitude data at time t, where n is the length of the historical platform attitude data sequence, such as 10,000 sampling points;

[0077] Then, empirical mode decomposition algorithms, such as EMD, are used to decompose the historical platform attitude data sequence. The specific decomposition process is as follows:

[0078] The historical platform attitude data sequence is treated as the raw signal, and a filtering process is performed on it;

[0079] By identifying local extrema, the upper and lower envelopes of the signal are generated, and the average value of the upper and lower envelopes is calculated to obtain the first-order modal component IMF1.

[0080] Subtract IMF1 from the original signal to obtain the residual signal. Then, using the residual signal as the new signal, repeat the above steps to obtain multiple modal components (IMFs) step by step. K The remainder is a positive integer until the residual signal becomes monotonic;

[0081] After implementing the empirical mode decomposition algorithm, a series of modal components can be obtained:

[0082] X H →[IMF1,IMF2,…,IMF K ,R];

[0083] Among them, IMF K Let R represent the Kth modal component, and R be the final residual component.

[0084] In this embodiment, a Hilbert transform is performed on each modal component to extract the instantaneous frequency and instantaneous amplitude of each modal component, including:

[0085] For each modal component IMF k (t), perform Hilbert transform to obtain the analytic signal:

[0086] Z k (t)=IMF k (t)+jH[IMF k (t)];

[0087] Where H[i] represents the Hilbert transform operator, and j is the imaginary unit;

[0088] Then calculate the instantaneous amplitude a from the analytical signal. k (t) and instantaneous phase φ k (t), defined as follows:

[0089] Instantaneous amplitude:

[0090]

[0091] Instantaneous phase:

[0092]

[0093] The instantaneous frequency f is obtained from the instantaneous phase derivative. k (t):

[0094]

[0095] In this embodiment, a feature matrix is ​​constructed using instantaneous frequency and instantaneous amplitude, input to the autoencoder model, and outputs platform attitude change trend data, including:

[0096] Using the instantaneous amplitude and instantaneous frequency corresponding to all obtained modal components as basic features, an input feature matrix F is constructed:

[0097] F=[a1(t),f1(t),a2(t),f2(t),…,a K (t),f K (t)];

[0098] The feature matrix F has high dimensionality and high redundancy. In order to effectively capture the platform attitude change trend and reduce the dimensionality and redundancy of the platform attitude data, the feature matrix F is input into the autoencoder model. The autoencoder model performs dimensionality reduction and reconstruction on the feature matrix F and outputs the platform attitude change trend data.

[0099] In this embodiment, the training process of the autoencoder model includes:

[0100] Construct an autoencoder network, consisting of an encoder and a decoder;

[0101] The encoder takes a high-dimensional feature matrix F as input and outputs a low-dimensional hidden layer feature matrix H, i.e.:

[0102] H = g(W) e ·F+b e );

[0103] Among them, W e With b eThese are the encoder weight matrix and bias vector, respectively, and g is a non-linear activation function, such as the ReLU function;

[0104] The decoder takes the hidden layer feature matrix H as input and reconstructs it into low-dimensional pose trend feature data T:

[0105] T = g(W) d ·H+b d );

[0106] Among them, W d With b d Let's define the decoder weight matrix and bias vector respectively;

[0107] The autoencoder network is trained using a large amount of historical data. By continuously optimizing the weight parameters of the autoencoder network, it is ensured that the reconstructed data T can accurately describe the trend characteristics of the platform's posture change. When the autoencoder network converges and the reconstruction error is lower than the preset threshold, the autoencoder network parameters are solidified to obtain a trained platform posture feature extraction model. This model can be directly used for real-time posture data feature extraction of the platform. The current posture data of the platform is input in real time, and the posture change trend data of the platform is output.

[0108] In this embodiment, during practical application, the real-time acquired platform current attitude data is processed by empirical mode decomposition and Hilbert transform to obtain a real-time instantaneous feature matrix; the real-time instantaneous feature matrix is ​​input into the platform attitude feature extraction model; the platform attitude feature extraction model outputs the dimensionality-reduced platform attitude change trend data in real time.

[0109] The platform attitude change trend data is the attitude change trend data of the platform at the current moment. It is a set of low-dimensional feature data that can clearly reflect the change trend characteristics of the platform attitude in a short period of time, and is used for the next step of platform short-term attitude prediction and intelligent control signal generation.

[0110] In this embodiment, the platform attitude prediction model is constructed using a long short-term memory network. Its initial parameters include input weights, recurrent weights, and bias terms, which are obtained by extracting temporal features from historical platform attitude data and dynamically adjusting them based on historical platform attitude prediction errors. The specific implementation steps are as follows:

[0111] A long short-term memory network is used to extract the temporal features of historical platform posture from historical platform posture data;

[0112] The parameters of the long short-term memory network are dynamically adjusted based on the historical platform attitude prediction error to determine the initial parameters of the platform attitude prediction model.

[0113] In this embodiment, a Long Short-Term Memory (LSTM) network is used to extract the temporal features of historical platform attitude from historical platform attitude data, including:

[0114] Historical attitude data collected during continuous platform operation is used as a historical training dataset, represented as follows:

[0115] X train =[X1,X2,…,X n ];

[0116] Among them, the platform's attitude data X at each moment i Including the platform's attitude angle θ and angular velocity ω:

[0117] X i =[θ xi ,θ yi ,θ zi ,ω xi ,ω yi ,ω zi ];

[0118] Historical training data is normalized to eliminate differences in the dimensions of different postures and improve training stability. The method for normalizing historical training data is as follows:

[0119]

[0120] Among them, X min X max These represent the minimum and maximum values ​​of the historical data for each attitude variable of the platform, respectively.

[0121] A long short-term memory network was constructed using normalized historical training data to extract the temporal features of historical platform poses.

[0122] The Long Short-Term Memory (LSTM) network is designed as follows: Input layer: The dimension is consistent with the platform pose data dimension, with 6 nodes; Hidden layer: A 2-layer LSTM structure is used, with 128 hidden units in each layer; Output layer: 6 nodes, corresponding to the predicted output of the platform pose data at future time steps; Activation function: The tanh activation function is used in the hidden layer, and the output layer is a linear output.

[0123] The short-term memory network structure expression is:

[0124]

[0125] Among them, h t and c t The hidden layer consists of a hidden state and a cellular state. It uses a "forget gate", an "input gate", and an "output gate" to accurately capture the long-term and short-term dependencies of the platform's pose data.

[0126] In this embodiment, the parameters of the Long Short-Term Memory network are dynamically adjusted based on historical platform attitude prediction errors to determine the initial parameters of the platform attitude prediction model, including:

[0127] In the initial training process of the Long Short-Term Memory network, the predicted values ​​of historical platform attitude temporal features are obtained through forward computation and compared with the true values ​​of historical platform attitude temporal features to obtain the historical platform attitude prediction error sequence:

[0128]

[0129] A trend function for historical platform attitude prediction errors is constructed using the historical platform attitude prediction error sequence, defined as the mean square error form:

[0130]

[0131] Where θ represents the set of parameters of the Long Short-Term Memory network, including input weights, recurrent weights, and bias terms;

[0132] Based on the gradient of the historical platform attitude prediction error trend function, the parameters of the long short-term memory network are dynamically adjusted using the backpropagation algorithm;

[0133] Repeat the adjustment until the historical platform attitude prediction error trend function meets the preset convergence condition. At this point, the long short-term memory network is the platform attitude prediction model, and the parameters of the long short-term memory network are the initial parameters of the platform attitude prediction model.

[0134] In this embodiment, the gradient of the historical platform attitude prediction error trend function is used as a basis to dynamically adjust the parameters of the long short-term memory network using the backpropagation algorithm, including:

[0135] Calculate the gradient of the historical platform attitude prediction error trend function:

[0136]

[0137] The Adam optimizer is used to update the Long Short-Term Memory network parameters. The Adam optimization rules are as follows:

[0138]

[0139] Where, m t and v t β1 and β2 are the first and second moment estimates of the gradient, respectively; β1 and β2 are the correction parameters, for example, β1 = 0.6 and β2 = 0.4; η is the learning rate, generally set to 0.001; ∈ is a constant to prevent division by zero, with a value of 10. -8 .

[0140] In this embodiment, the real-time acquisition process of the platform's short-term attitude prediction data includes:

[0141] Using the initial parameters of the platform attitude prediction model and combining them with the real-time attitude change trend data of the platform obtained from S101, the platform's short-term attitude prediction data is output in real time. The specific implementation process is as follows:

[0142] The platform attitude change trend data is input into the platform attitude prediction model as the real-time input sequence; the platform attitude prediction model performs real-time forward propagation and outputs short-term predicted platform attitude data Y. t+1 This refers to the predicted data of the platform's attitude angle and angular velocity within a short future window, such as the next few seconds.

[0143] The platform's short-term attitude prediction data is expressed as follows:

[0144] Y t+1 =LSTM(T) t ,h t ,c t |θ * );

[0145] Among them, h t For the short-time hidden state memory data of the platform attitude prediction model at the current moment, c t For the platform's attitude prediction model, the long-term cell state memory data at the current moment, θ * The initial parameters for the platform attitude prediction model are used. The platform's short-term attitude prediction data is then used to generate control signals for the subsequent intelligent anti-roll gyroscope, providing real-time guidance for the prediction and dynamic planning of anti-roll actions.

[0146] In this embodiment, the platform attitude deviation data specifically refers to the displacement and angular deviation of the offshore operating platform in various directions such as roll, pitch, and yaw, relative to the ideal stable state of the platform being stable, stationary, without tilt or sway. The platform attitude deviation is specifically manifested as follows:

[0147] Position displacement deviation: The difference between the actual displacement of the platform structure under the action of external forces such as wind and waves and the zero displacement under the ideal steady state;

[0148] Attitude angle deviation: The difference between the actual attitude angles of the platform, such as roll, pitch, and yaw, and the zero angle under ideal stable conditions;

[0149] Attitude angular velocity deviation: The difference between the actual attitude angular velocity of the platform and the zero angular velocity under ideal steady-state conditions.

[0150] In this embodiment, the calculation of the platform's current attitude deviation data based on the platform's current attitude data and platform structural stiffness parameters includes:

[0151] Extract the angular velocity differences of the platform's roll, pitch, and yaw based on the platform's current attitude data;

[0152] A rigid-flexible coupled dynamic model is established by combining the platform's structural stiffness parameters. The finite element numerical method is used to solve the rigid-flexible coupled dynamic model and output the platform's current attitude deviation data.

[0153] In this embodiment, the platform structure stiffness parameters are obtained through the following steps:

[0154] Establish a finite element structural model of the platform and perform modal analysis of the platform structure;

[0155] The platform structure stiffness matrix is ​​obtained based on the platform structure modal analysis results, and the platform structure stiffness parameters are output.

[0156] In this embodiment, establishing the finite element structural model of the platform includes: based on the actual structural design drawings and parameters of the platform, constructing the finite element structural model of the platform using finite element analysis software such as ANSYS and ABAQUS. The specific process includes:

[0157] Based on the platform's design dimensions and the material parameters of the steel structure, such as the elastic modulus, Poisson's ratio, and density, a geometric model of the platform structure is established.

[0158] Select finite element types suitable for the platform's structural characteristics, such as beam elements, plate and shell elements, and combinations of three-dimensional solid elements, and rationally divide the mesh;

[0159] Define boundary conditions and constraints to accurately reflect the constraint characteristics of the platform's support connection to the seabed or the floating structure;

[0160] Appropriate wave loads, wind loads, and gravity loads are applied to simulate real working conditions;

[0161] After the above steps, the finite element structural analysis model of the platform is obtained.

[0162] In this embodiment, the platform structure stiffness matrix is ​​obtained based on the platform structure modal analysis results, and the platform structure stiffness parameters are output, including:

[0163] Using the finite element structural model of the platform, modal analysis of the platform structure is carried out, specifically including:

[0164] Eigenvalue modal analysis was performed to solve for the natural frequencies and corresponding mode shapes of the platform structural model. The Lanczos algorithm was used to solve the problem to ensure both efficiency and accuracy.

[0165] Extract the first few natural frequencies and mode shapes from the modal analysis results, for example, the first 10 natural frequencies, specifically represented as:

[0166] ω n ,φ n n = 1, 2, ..., 10;

[0167] Where, ω n Let φ be the nth natural frequency. n This corresponds to the nth mode shape;

[0168] Based on the above results, the platform structural stiffness matrix K is clearly obtained:

[0169]

[0170] Where [M] is the mass matrix of the platform and [K] is the structural stiffness matrix of the platform. The structural stiffness matrix of the platform is obtained through this characteristic equation and numerical solution.

[0171] Thus, the platform structural stiffness parameters, namely the platform structural stiffness matrix [K], are finally obtained and used for subsequent platform attitude deviation calculations.

[0172] In this embodiment, the platform's current attitude deviation data reflects the deviation of the platform's current actual attitude from the ideal stable state. The calculation steps are as follows:

[0173] Based on the platform's current attitude data collected by S101, the actual attitude angle and angular velocity information of the current platform are determined.

[0174] Using the platform's ideal stable state as a benchmark, calculate the angle difference and angular velocity difference between the platform's current attitude data and the ideal state;

[0175] Based on the established rigid-flexible coupled dynamic model, the finite element numerical calculation method, such as the Newmark-β method, is used to perform time-domain integration to obtain the current attitude deviation data of the platform.

[0176] In this embodiment, the actual attitude angle and angular velocity information of the current platform are defined as follows:

[0177] X t =[θ x (t),θ y (t),θ z (t),ω x (t),ω y (t),ω z (t)];

[0178] Where, θ x (t),θ y (t),θ z (t) represents the platform's roll, pitch, and yaw angles at the current moment; ω x (t),ω y (t),ω z (t) represents the angular velocities of the platform's roll, pitch, and yaw at the current moment, respectively;

[0179] The ideal state of the platform is zero angle and zero angular velocity, defined as:

[0180] X ideal = [0,0,0,0,0,0];

[0181] The angle and angular velocity differences between the platform's current attitude data and the ideal state are expressed as follows:

[0182] Platform current attitude angle difference: Δθ x (t)=θ x (t)-0,Δθ y (t)=θ y (t)-0,Δθ z (t)=θ z (t)-0;

[0183] The current angular velocity difference of the platform is: Δω x (t)=ω x (t)-0,Δω y (t)=ω y (t)-0,Δω z (t)=ω z (t)-0;

[0184] To accurately describe the platform's dynamic response characteristics, a rigid-flexible coupling dynamic modeling method is adopted, treating the platform as a structural system that simultaneously exhibits rigid motion and flexible deformation. The basic dynamic equations of the rigid-flexible coupling dynamic model are expressed as follows:

[0185]

[0186] In the formula, [M] is the mass matrix of the platform, obtained through the finite element model; [C] is the damping matrix, determined using the Rayleigh damping form; [K] is the structural stiffness matrix, determined by finite element modal analysis; {u} is the platform displacement vector, including rigid body motion displacement and flexible structure deformation displacement; F ext [K] represents the external load vector, determined through real-time measurement or preset sea state parameters, reflecting the wave load experienced by the platform at the current moment; [K] represents the previously obtained platform structural stiffness parameters.

[0187] In this embodiment, based on the constructed rigid-flexible coupling dynamic model, the Newmark-β method is used for time-domain integration. The specific implementation process is as follows:

[0188] Using the platform's current real-time attitude data and external environmental loads, determine the current external force F acting on the platform. ext 、;

[0189] According to the Newmark-β method, given the initial boundary conditions, select appropriate integration parameters, such as β = 0.25 and γ = 0.5, to ensure stability and accuracy, and perform stepwise integration calculations;

[0190] Solving the rigid-flexible coupling dynamic equations yields the dynamic displacement response vector of the platform at the current moment, which in turn provides the platform's current attitude deviation data.

[0191] X dex,t ={u t}

[0192] In this embodiment, the platform's predicted attitude deviation data is used to anticipate the trend of the platform's attitude deviating from the ideal state in the short term. Specific implementation steps are as follows:

[0193] Using the platform short-time attitude prediction data Y obtained from S102 t+1 :

[0194]

[0195] Following the same method used to calculate the platform's current attitude deviation data, the rigid-flexible coupling dynamic equations are used to replace the platform's current attitude data with the platform's predicted attitude data to obtain the platform's predicted attitude deviation data at the prediction time, i.e., time t+1:

[0196]

[0197] In this embodiment, a platform attitude control fusion matrix is ​​constructed based on the platform's current attitude deviation data and the platform's predicted attitude deviation data, including:

[0198] Fourier transforms are performed on the current attitude deviation data and the predicted attitude deviation data of the platform to obtain the spectrum of the platform attitude deviation data.

[0199] The platform attitude deviation data is subjected to joint time-frequency decomposition to obtain the platform attitude fusion weight matrix, and the platform attitude control fusion matrix is ​​constructed accordingly.

[0200] In this embodiment, Fourier transforms are performed on the current attitude deviation data and the predicted attitude deviation data of the platform to obtain the spectrum of the platform attitude deviation data, including:

[0201] Obtain the platform's current attitude deviation data X calculated by S103 respectively. dev,t and platform-predicted attitude deviation data They are respectively recorded as:

[0202] X dev,t =[x1(t),x2(t),…,x n (t)]

[0203]

[0204] Each component represents the attitude deviation of the platform in a certain direction;

[0205] Fast Fourier transform is performed on the current attitude deviation data and the predicted attitude deviation data of the platform to obtain the spectrum of the current attitude deviation and the spectrum of the predicted attitude deviation of the platform.

[0206] The spectrum of the platform's current attitude deviation includes the frequency domain characteristics of the platform's current attitude deviation, expressed as:

[0207] X f,t =FFT(X) dev,t );

[0208] The platform's predicted attitude deviation spectrum includes the frequency domain characteristics of the attitude deviation at the short-time prediction time, expressed as:

[0209]

[0210] In this embodiment, in order to effectively capture the time-frequency feature information in the platform attitude deviation spectrum data, a joint time-frequency decomposition method is used for further feature extraction, and a platform attitude fusion weight matrix is ​​constructed. The specific implementation method is as follows:

[0211] Build a time-frequency joint decomposition network, such as a CNN-LSTM hybrid network, with the following specific network architecture:

[0212] Input layer: Input the platform attitude deviation spectrum data X respectively f,t and

[0213] LSTM layer: Further captures the dynamic temporal correlation of platform attitude deviation spectrum data;

[0214] Output layer: Outputs multiple sub-spectrum data corresponding to different frequency bands;

[0215] After processing by this network, multiple sub-spectral components of the platform's current attitude deviation and the platform's predicted attitude deviation are obtained:

[0216] Current attitude deviation sub-spectrum of the platform:

[0217]

[0218] Platform-predicted attitude deviation sub-spectrum:

[0219]

[0220] Where m is the number of sub-spectrums decomposed, which can usually be set to 4-8 frequency bands;

[0221] Based on the obtained sub-spectrums, the energy density of each frequency band sub-spectrum is calculated, and the frequency band energy density matrix E is constructed:

[0222]

[0223] in, and These are the energy densities of the i-th frequency band of the current and predicted attitude deviation sub-spectrums of the platform, respectively. The specific calculation method is as follows:

[0224]

[0225] In this embodiment, obtaining the platform attitude fusion weight matrix includes:

[0226] The sub-spectrum of each frequency band is obtained using a time-frequency joint decomposition network;

[0227] The objective function of spectral entropy is defined based on the frequency band energy density matrix, and the objective function is optimized by the differential evolution algorithm to obtain the fusion weight matrix of platform attitude.

[0228] Define the objective function H for the spectral entropy value to characterize the concentration of energy density distribution in a frequency band:

[0229]

[0230] Where W = [w1, w2, ..., w m ] is the fusion weight vector to be optimized, representing the proportion of each frequency band contributing to the final fused attitude control signal.

[0231] In this embodiment, the weight optimization of the differential evolution algorithm includes:

[0232] Initial population generation: Randomly generate the initial population, with each individual representing a weight vector;

[0233] Mutation operation: Generate new individuals according to the rules of the DE algorithm:

[0234] W j ′(t)=W best (t)+F·[W r1 (t)-W r2 (t)];

[0235] Among them, W best (t) represents the current optimal weight; F is the variation factor, typically taken as 0.5-0.8; W r1 (t), W r2 (t) represents different individuals randomly selected from the population;

[0236] Crossover operation: Generate experimental individuals based on the crossover probability CR;

[0237] Selection operation: Select a better weight vector based on the entropy objective function H(W);

[0238] Repeat the above steps until the entropy objective function reaches the preset convergence threshold, and output the optimized platform attitude fusion weight matrix:

[0239]

[0240] In this embodiment, the optimized platform attitude fusion weight matrix is ​​used to construct the platform attitude control fusion matrix, and the current attitude deviation sub-spectrum and the predicted attitude deviation sub-spectrum are fused and calculated:

[0241]

[0242] The platform attitude control fusion matrix M fusion The signal is input to the roll stabilization gyroscope control module, where it undergoes inverse Fourier transform to obtain the platform's real-time attitude deviation signal. The roll stabilization gyroscope control module then generates the final real-time platform attitude control signal used to drive the roll stabilization gyroscope.

[0243] U control (t)=IFFT(M fusion );

[0244] The real-time attitude control signal of the platform can effectively drive the anti-roll gyroscope to correct the platform attitude in a timely manner, ensuring the stability of the platform.

[0245] The platform attitude correction action of the roll stabilizer gyroscope is jointly determined by the rotational speed and precession angular velocity of the roll stabilizer gyroscope flywheel. In this embodiment, the roll stabilizer gyroscope is driven to perform platform attitude correction actions according to the real-time platform attitude control signal, including:

[0246] The target rotational speed of the anti-roll gyroscope flywheel is calculated based on the platform's real-time attitude control signal. and target precession angular velocity

[0247] A dual-closed-loop platform attitude controller for the rotational speed and precession angular velocity of the roll stabilizing gyroscope flywheel is constructed, which outputs the drive signal for the roll stabilizing gyroscope flywheel motor to control the roll stabilizing gyroscope to perform attitude correction actions.

[0248] In this embodiment, the target rotational speed of the anti-roll gyroscope flywheel is calculated based on the platform attitude real-time control signal. and target precession angular velocity include:

[0249] Based on the amplitude and direction of the platform attitude real-time control signal, determine the magnitude and direction of the target rotational speed and target precession angular velocity of the anti-roll gyroscope flywheel;

[0250] The target rotational speed of the gyroscope flywheel is determined in real time using a preset control mapping function for the roll stabilization gyroscope. and target precession angular velocity

[0251] In this embodiment, the control mapping function of the roll-stabilizing gyroscope is:

[0252]

[0253] Where f(·) is the control mapping function of the roll stabilization gyroscope, used to convert the platform attitude real-time control signal U control (t) is converted to the target speed of the anti-roll gyroscope flywheel. and target precession angular velocity It was designed and determined in advance based on the platform's anti-roll principle;

[0254] For example, assume that the real-time attitude control signal of the platform is a three-dimensional vector, corresponding to roll, pitch, and yaw:

[0255] U control (t)=[u x (t),u y (t),u z (t)];

[0256] The empirical formula for platform roll reduction obtained from platform roll reduction experiments is used as the control mapping function for the roll reduction gyroscope:

[0257]

[0258] Where, k f To reduce the rotational speed gain of the gyroscope flywheel, for example, a value of 500 rpm per unit control input; k p The gain factor for reducing the precession angular velocity of the gyro flywheel, for example, is 0.1 rad / s / degree; k f and k p The process of determining is as follows:

[0259] Under actual platform operating conditions, the effects of platform attitude control signals of different amplitudes and directions on platform attitude correction were recorded. Using the least squares fitting method, and employing experimental data from multiple sets of platform attitude control signals and the actual flywheel speed and precession angular velocity of the roll-damping gyroscope, k was calculated through regression analysis. f and k p The optimal value; for example, the specific process is as follows:

[0260] Five sets of platform attitude control signals of different intensities were used to drive the roll stabilization gyroscopes. The rotational speed and precession angular velocity of the gyroscope flywheel were measured. The least squares method was used to solve for k in the above empirical formula. f and k p .

[0261] In this embodiment, a dual-closed-loop platform attitude controller is constructed, controlling the rotational speed and precession angular velocity of the roll stabilization gyroscope flywheel. This controller outputs drive signals to the roll stabilization gyroscope flywheel motor, controlling the roll stabilization gyroscope to perform attitude correction actions, including:

[0262] The inner-loop platform attitude control loop is constructed based on the difference between the target speed and the actual speed of the anti-roll gyroscope flywheel;

[0263] The outer-loop platform attitude control loop is constructed using the difference between the target precession angular velocity and the actual precession angular velocity of the anti-roll gyroscope flywheel.

[0264] The control parameters of the inner loop platform attitude control loop are dynamically adjusted using the output of the outer loop platform attitude control loop.

[0265] In this embodiment, the outer-loop platform attitude control loop is constructed based on the difference between the target precession angular velocity and the actual precession angular velocity of the anti-roll gyroscope flywheel, including:

[0266] The inner-loop platform attitude control loop adopts the PID control method, and the output of the inner-loop platform attitude control loop is the speed control quantity u of the anti-roll gyroscope flywheel motor. inner (t):

[0267]

[0268] Among them, K p,inner ,K i,inner ,K d,inner These are all PID parameters of the inner loop platform attitude control loop;

[0269] The difference between the target speed and the actual speed of the anti-roll gyroscope flywheel is used as the input signal for the inner-loop platform attitude control loop:

[0270]

[0271] In this embodiment, the outer-loop platform attitude control loop is constructed based on the difference between the target precession angular velocity and the actual precession angular velocity of the anti-roll gyroscope flywheel, including:

[0272] The outer ring platform attitude control loop also adopts the PID control method. The outer ring platform attitude control loop outputs the speed control quantity u of the anti-roll gyroscope flywheel motor. outer (t):

[0273]

[0274] The target precession angular velocity of the anti-roll gyroscope flywheel With actual precession angular velocity The difference serves as the input signal for the outer loop platform attitude control loop:

[0275]

[0276] In this embodiment, the control parameters of the inner loop platform attitude control loop are dynamically adjusted using the output of the outer loop platform attitude control loop, including:

[0277] Based on the output torque of the anti-roll gyroscope flywheel motor and the actual speed of the anti-roll gyroscope flywheel, the control parameters of the inner loop platform attitude control loop are dynamically adjusted using an incremental adaptive control algorithm.

[0278] The load model parameters of the anti-roll gyroscope flywheel are updated in real time using an online load model identification method, and the control parameters of the outer ring platform attitude control loop are dynamically adjusted.

[0279] In this embodiment, the control parameter K of the inner loop platform attitude control loop is adjusted in real time by the output of the outer loop platform attitude control loop. p,inner ,K i,inner ,K d,inner This enables the linkage between the outer and inner loop platform attitude control loops, as well as real-time adaptive control of the platform attitude.

[0280] In this embodiment, based on the output torque of the roll-stabilizing gyroscope flywheel motor and the actual rotational speed of the roll-stabilizing gyroscope flywheel, an incremental adaptive control algorithm is used to dynamically adjust the control parameters of the inner loop platform attitude control loop, including:

[0281] Real-time acquisition of the output torque of the anti-roll gyroscope flywheel motor and the actual speed of the anti-roll gyroscope flywheel;

[0282] The incremental adaptive control algorithm is used to control the parameter K of the inner loop platform attitude control loop. p,inner ,K i,inner ,K d,inner Dynamic adjustments will be made, specifically including:

[0283] The update rule for the control parameters of the inner-loop platform attitude control loop is defined as follows:

[0284] K p,inner (t+1)=K p,inner (t)+ΔK p

[0285] K i,inner (t+1)=K i,inner (t)+ΔK i

[0286] K d,inner (t+1)=K d,inner (t)+ΔK d

[0287] Increment ΔK p ,ΔK i ,ΔK dDetermined according to the following rules:

[0288]

[0289] Among them, the function g(·) is a pre-designed incremental adaptive adjustment mapping rule, which is used to adjust the control parameter K of the inner loop platform attitude control loop based on the real-time dynamic response data of the anti-roll gyroscope flywheel. p,inner ,K i,inner ,K d,inner ;

[0290] For example, the function input is defined as the real-time torque change rate of the anti-roll gyroscope flywheel motor. and the actual rotational speed change rate of the anti-roll gyroscope flywheel

[0291] By real-time acquisition of the output torque T of the anti-roll gyroscope flywheel motor motor (t) and the actual rotational speed of the anti-roll gyroscope flywheel

[0292] Numerical differentiation calculation:

[0293]

[0294] Based on actual operational experience, an incremental adaptive adjustment formula was designed:

[0295]

[0296] For example, the adaptive adjustment coefficients determined by experience are: α = 0.05, β = 0.02, γ = 0.01, δ = 0.03. The process of determining these coefficients is as follows: In actual operation, a large amount of torque and speed data of the anti-roll gyroscope flywheel are collected under different working conditions of the dynamic response of the anti-roll gyroscope flywheel, such as under different wave intensities, and the anti-roll effect of the platform is recorded at this time.

[0297] For example, in a practical operation: In a certain experiment, real-time data was measured as follows: The calculation is obtained based on the above formula:

[0298] ΔK p =0.05×2.5+0.02×50=0.125+1=1.125

[0299] ΔK i =0.01×2.5=0.025,ΔK d =0.03 × 50 = 1.5

[0300] Subsequently, the ΔK calculated above will be used... p ,ΔK i ,ΔK dThe updated PID parameters applied to the current inner-loop platform attitude control loop are as follows:

[0301] K p,inner (t+1)=K p,inner (t)+1.125

[0302] K i,inner (t+1)=K i,inner (t)+0.025

[0303] K d,inner (t+1)=K d,inner (t)+1.5

[0304] Machine learning regression or least squares fitting methods are used to determine the incremental adjustment coefficients for each item, so as to ensure that the PID control parameter adjustment can quickly converge and stably control the platform attitude correction action of the anti-roll gyroscope flywheel.

[0305] In this embodiment, the load model parameters of the anti-roll gyroscope flywheel are updated in real time using an online load model identification method, and the control parameters of the outer ring platform attitude loop are dynamically adjusted, specifically including:

[0306] Real-time acquisition of the actual precession angular velocity of the anti-roll gyroscope flywheel and the corresponding precession motor control input data;

[0307] Real-time identification of load characteristic parameters of the roll-damping gyroscope using a recursive least squares algorithm:

[0308]

[0309] In the formula, φ(t) is the load model parameter vector; P(t) is the error covariance matrix; φ(t) is the identification input data vector; e(t) is the identification error, representing the difference between the predicted precession response and the actual response.

[0310] Real-time updates of control parameters for the outer loop platform attitude control loop:

[0311]

[0312] Where h(·) is the parameter update rule function, ensuring that the control parameters of the outer ring platform attitude control loop match the actual working characteristics of the gyroscope flywheel load in real time.

[0313] In this embodiment, after the anti-roll gyroscope completes the platform attitude correction action in real time, the platform attitude is effectively corrected, and then S101 is triggered again. Specifically, the current attitude data of the platform is re-acquired through the real-time updated platform attitude sensor data; the next platform control cycle is entered, and the subsequent control methods are implemented again to form a complete closed-loop control process, thereby continuously ensuring the stability of the platform attitude.

[0314] refer to Figure 2 As an implementation of the above method, the present invention provides an embodiment of an intelligent control system for reducing roll on offshore operating platforms. This system embodiment corresponds to the embodiment of the above method, and the system can be specifically applied to various electronic devices.

[0315] The roll reduction intelligent control system for offshore operating platforms described in this embodiment implements the aforementioned roll reduction intelligent control method for offshore operating platforms, including:

[0316] The data acquisition module 201 is used to collect the platform's current attitude data in real time; input the platform's current attitude data into the pre-built platform attitude feature model to obtain the platform attitude change trend data;

[0317] The prediction acquisition module 202 is used to generate initial parameters for the platform attitude prediction model based on historical platform attitude data; and to obtain short-term platform attitude prediction data based on platform attitude change trend data and the initial parameters of the platform attitude prediction model.

[0318] The prediction calculation module 203 is used to calculate the current attitude deviation data of the platform based on the current attitude data of the platform and the platform structural stiffness parameters; and to calculate the predicted attitude deviation data of the platform based on the short-term attitude prediction data of the platform and the platform structural stiffness parameters.

[0319] The signal generation module 204 is used to construct a platform attitude control fusion matrix based on the platform's current attitude deviation data and the platform's predicted attitude deviation data; and input the platform attitude control fusion matrix into the anti-roll gyroscope control module to generate a real-time platform attitude control signal.

[0320] The correction module 205 is used to drive the roll stabilizer gyroscope to perform platform attitude correction actions according to the real-time control signal of the platform attitude; after the roll stabilizer gyroscope performs the platform attitude correction actions, it re-triggers the data acquisition module to collect updated platform current attitude data.

[0321] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via a wired or wireless network. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium, an optical medium, or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0322] In the embodiments provided by this invention, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative; for instance, the division of units is only one method, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0323] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0324] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0325] Some of the data in the above formula are calculated by removing dimensions and taking their numerical values. The formula is the closest to the real situation obtained by software simulation of a large amount of collected data. The preset parameters and preset thresholds in the formula are set by those skilled in the art according to the actual situation or obtained through simulation of a large amount of data.

[0326] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. A roll reduction intelligent control method for offshore operating platforms, characterized in that, include: S101. Real-time acquisition of the platform's current attitude data; inputting the platform's current attitude data into a pre-built platform attitude feature extraction model to obtain platform attitude change trend data; S102. Generate initial parameters for the platform attitude prediction model based on historical platform attitude data; obtain short-term platform attitude prediction data based on platform attitude change trend data and the initial parameters of the platform attitude prediction model. S103. Calculate the current attitude deviation data of the platform based on the current attitude data and the platform structural stiffness parameters; calculate the predicted attitude deviation data of the platform based on the short-term attitude prediction data and the platform structural stiffness parameters. S104. Construct a platform attitude control fusion matrix based on the platform's current attitude deviation data and the platform's predicted attitude deviation data; input the platform attitude control fusion matrix into the anti-roll gyroscope control module to generate a real-time platform attitude control signal; S105. Drive the anti-roll gyroscope to perform platform attitude correction actions according to the real-time control signal of the platform attitude; after the anti-roll gyroscope performs the platform attitude correction actions, re-trigger S101 to collect the updated current attitude data of the platform.

2. The intelligent control method for reducing sway according to claim 1, characterized in that, The construction of the platform pose feature extraction model includes: Empirical mode decomposition was performed on historical platform attitude data to obtain multiple modal components of different frequencies; Perform Hilbert transform on each modal component to extract the instantaneous frequency and instantaneous amplitude of each modal component; A feature matrix is ​​constructed using instantaneous frequency and instantaneous amplitude, which is then input into the autoencoder model to output platform attitude change trend data. The training process of the autoencoder model includes: Construct an autoencoder network consisting of an encoder and a decoder; The autoencoder network is trained using the feature matrix until the reconstruction error converges, and the trained platform posture feature extraction model is output.

3. The intelligent control method for reducing sway according to claim 1, characterized in that, The initial parameters for the platform attitude prediction model are generated, including: A long short-term memory network is used to extract the temporal features of historical platform posture from historical platform posture data; The parameters of the long short-term memory network are dynamically adjusted based on the historical platform attitude prediction error to determine the initial parameters of the platform attitude prediction model. Specifically, the parameters of the Long Short-Term Memory network are dynamically adjusted based on historical platform attitude prediction errors to determine the initial parameters of the platform attitude prediction model, including: A historical platform attitude prediction error trend function is established. Based on the gradient of the historical platform attitude prediction error trend function, the parameters of the long short-term memory network are dynamically adjusted using the backpropagation algorithm. Repeat the adjustment until the historical platform attitude prediction error trend function meets the preset convergence condition. At this point, the long short-term memory network is the platform attitude prediction model, and the parameters of the long short-term memory network are the initial parameters of the platform attitude prediction model.

4. The intelligent control method for reducing sway according to claim 1, characterized in that, The current attitude deviation data of the calculation platform includes: Extract the angular velocity differences of the platform's roll, pitch, and yaw based on the platform's current attitude data; A rigid-flexible coupled dynamic model is established by combining the platform's structural stiffness parameters. The finite element numerical method is used to solve the rigid-flexible coupled dynamic model and output the platform's current attitude deviation data.

5. The intelligent control method for reducing sway according to claim 1, characterized in that, Obtaining the platform structure stiffness parameters includes: Establish a finite element structural model of the platform and perform modal analysis of the platform structure; The platform structure stiffness matrix is ​​obtained based on the platform structure modal analysis results, and the platform structure stiffness parameters are output.

6. The intelligent control method for reducing sway according to claim 1, characterized in that, The construction of the platform attitude control fusion matrix includes: Fourier transforms are performed on the current attitude deviation data and the predicted attitude deviation data of the platform to obtain the spectrum of the platform attitude deviation data. The platform attitude deviation data is subjected to joint time-frequency decomposition to obtain the platform attitude fusion weight matrix, and the platform attitude control fusion matrix is ​​constructed accordingly. The acquisition of the platform posture fusion weight matrix includes: The objective function of spectral entropy is defined based on the frequency band energy density matrix. The objective function is optimized using the differential evolution algorithm to obtain the fusion weight matrix of platform attitude.

7. The intelligent control method for reducing sway according to claim 1, characterized in that, Based on the real-time control signal of the platform attitude, the anti-roll gyroscope is driven to perform platform attitude correction actions, including: The target rotational speed and target precession angular velocity of the anti-roll gyroscope flywheel are calculated based on the real-time control signal of the platform attitude. A dual-closed-loop platform attitude controller for the rotational speed and precession angular velocity of the roll stabilizing gyroscope flywheel is constructed, which outputs the drive signal for the roll stabilizing gyroscope flywheel motor to control the roll stabilizing gyroscope to perform attitude correction actions.

8. The intelligent control method for reducing sway according to claim 7, characterized in that, A dual-closed-loop platform attitude controller is constructed to measure the rotational speed and precession angular velocity of the roll stabilization gyroscope flywheel. This controller outputs drive signals to the roll stabilization gyroscope flywheel motor, controlling the roll stabilization gyroscope to perform attitude correction actions, including: The inner-loop platform attitude control loop is constructed based on the difference between the target speed and the actual speed of the anti-roll gyroscope flywheel; The outer-loop platform attitude control loop is constructed using the difference between the target precession angular velocity and the actual precession angular velocity of the anti-roll gyroscope flywheel. The control parameters of the inner loop platform attitude control loop are dynamically adjusted using the output of the outer loop platform attitude control loop.

9. The intelligent control method for reducing sway according to claim 8, characterized in that, The control parameters of the inner loop platform attitude control loop are dynamically adjusted using the output of the outer loop platform attitude control loop, including: Based on the output torque of the anti-roll gyroscope flywheel motor and the actual speed of the anti-roll gyroscope flywheel, the control parameters of the inner loop platform attitude control loop are dynamically adjusted using an incremental adaptive control algorithm. The load model parameters of the anti-roll gyroscope flywheel are updated in real time using an online load model identification method, and the control parameters of the outer ring platform attitude control loop are dynamically adjusted.

10. A roll reduction intelligent control system for offshore operating platforms, characterized in that, Implementing the anti-roll intelligent control method according to any one of claims 1-9, comprising: The data acquisition module (201) is used to obtain real-time platform current attitude data; it inputs the platform current attitude data into the pre-built platform attitude feature model to obtain platform attitude change trend data. The prediction acquisition module (202) is used to generate the initial parameters of the platform attitude prediction model based on historical platform attitude data; and to obtain short-term platform attitude prediction data based on platform attitude change trend data and the initial parameters of the platform attitude prediction model. The prediction calculation module (203) is used to calculate the current attitude deviation data of the platform based on the current attitude data of the platform and the platform structural stiffness parameters; and to calculate the predicted attitude deviation data of the platform based on the short-term attitude prediction data of the platform and the platform structural stiffness parameters. The signal generation module (204) is used to construct a platform attitude control fusion matrix based on the platform's current attitude deviation data and the platform's predicted attitude deviation data; and input the platform attitude control fusion matrix into the anti-roll gyroscope control module to generate a real-time platform attitude control signal. The correction module (205) is used to drive the anti-roll gyroscope to perform platform attitude correction actions according to the real-time control signal of the platform attitude; after the anti-roll gyroscope performs the platform attitude correction actions, it re-triggers the data acquisition module to collect updated platform current attitude data.

Images