Large inertia flywheel active anti-roll control method and system based on model predictive control

Abstract

The application provides a large inertia flywheel active roll reduction control method and system based on model predictive control, real-time acquisition of motion state data of a water surface platform and a roll reduction gyro; calculation of historical real-time disturbance torque data based on the motion state data, and then secondary interpolation to obtain a predicted disturbance torque sequence in a prediction time domain; input of the predicted disturbance torque sequence into a model predictive control system MPC to generate an optimal control sequence; the model predictive control system is composed of a water surface platform rolling motion equation and a roll reduction gyro motion nonlinear equation; mapping of the optimal control sequence into an output torque of a motor, control of a precession axis of the roll reduction gyro, and formation of a closed loop feedback. The roll reduction performance is optimized under the premise of ensuring safe and reliable operation, the roll reduction efficiency of the roll reduction gyro is improved, and the stability of the water surface platform under wind and wave is effectively enhanced.

Description

Technical Field

[0001] This invention relates to the field of water surface platform roll reduction control technology, specifically to a large inertia flywheel active roll reduction control method and system based on model predictive control. Background Technology

[0002] Surface platforms, such as ships, marine nuclear power platforms, and offshore wind power platforms, inevitably experience six degrees of freedom of motion during offshore operations due to environmental disturbances such as wind, waves, and currents: roll, pitch, yaw, sway, heave, and heave. Among these, the platform's rolling motion—roll, pitch, and yaw—can affect the normal operation of equipment on the platform and, in severe cases, may even damage machinery. Furthermore, excessive rolling motion can directly impact the work efficiency and personal safety of personnel on the platform. Therefore, reducing the rolling motion of surface platforms is of great significance for their safe and stable operation.

[0003] High-speed rotating, high-inertia flywheels utilize their precession effect to effectively suppress the swaying of surface platforms in wind and waves. Technology based on this principle is called a roll-damping gyroscope. Compared to the anti-roll fins or anti-roll rudders commonly used in ships, roll-damping gyroscopes have the advantage of reducing sway even at zero speed, thus they can also be used in offshore floating wind power platforms for stabilization. Currently, roll-damping gyroscopes are mainly divided into passive and active types. Passive roll-damping gyroscopes primarily rely on the natural precession of a high-speed rotating rotor under the action of external disturbance torque to generate gyroscopic torque and achieve a roll-damping effect. However, because the precession is uncontrolled, it may become excessive in high sea states, damaging the equipment or causing loss of control. In such cases, it not only fails to reduce sway but may even have the opposite effect. Therefore, to maximize roll-damping capability, active control of the gyroscope's precession is usually required.

[0004] Existing precession control strategies for active roll stabilization gyroscopes mainly include constant gain control, PD control, variable gain control, nonlinear robust control, and optimization algorithms for roll stabilization gyroscope parameters. Constant gain control proportionally controls the precession angular velocity of the roll stabilization gyroscope to the platform's yaw angular velocity. Under the premise of a physical model based on linearization assumptions, the joint system of the platform and gyroscope can be transformed into a transfer function, simplifying the overall control. It has a good roll stabilization effect when the precession angle is small. However, in order to prevent the system from exceeding the drive limit or to keep the precession within ±π / 2 rad, an integral control term or a penalty function control term is usually required. PD control is an optimization and improvement on the traditional constant gain control. PD control directly controls the output of the control motor on the precession axis of the roll stabilization gyroscope, making the output of the control motor proportional to the precession angle and the precession angular velocity. Through formula derivation, it can be found that PD control can be transformed into constant gain control within a certain parameter range.

[0005] The precession control algorithm of active roll stabilization gyroscopes is a key technology that determines whether floating platforms, ships, etc. can operate stably. However, most existing active control strategies for roll stabilization gyroscopes are based on the principle that the precession angular velocity is proportional to the yaw angular velocity, and most of the proposed control methods are offline control based on the assumption of a linearized physical model. At the same time, the parameter tuning of the controller is complicated, cannot be adjusted in real time, and is difficult to adapt to complex and ever-changing sea conditions. In addition, it cannot flexibly handle precession angle constraints. Summary of the Invention

[0006] This invention proposes a large-inertia flywheel active roll reduction control method and system based on model predictive control to solve the technical problems that existing control methods are based on strategies where precession angular velocity and roll angular velocity are proportional, and most of the proposed control methods are offline control based on linearized assumptions of physical models, which leads to complex parameter tuning of the controller, inability to adjust in real time, and difficulty in adapting to complex and ever-changing sea conditions.

[0007] To address the aforementioned technical problems, this invention provides a method for active roll reduction control of a large inertia flywheel based on model predictive control, comprising the following steps:

[0008] Step S1: Acquire real-time motion state data of the water surface platform and the anti-roll gyroscope;

[0009] Step S2: After calculating the historical real-time disturbance torque data based on the motion state data, the predicted disturbance torque sequence in the prediction time domain is obtained by secondary interpolation;

[0010] Step S3: Input the predicted disturbance torque sequence into the model predictive control system (MPC) to generate the optimal control sequence; the model predictive control system consists of the water surface platform swaying motion equation and the anti-sway gyroscope motion nonlinear equation;

[0011] Step S4: Map the optimal control sequence to the output torque of the motor, control the precession axis of the roll reduction gyroscope, and form a closed-loop feedback.

[0012] Preferably, the expression for calculating the historical real-time disturbance torque data based on the motion state data in step S2 is as follows:

[0013]

[0014] In the formula, W k-1 I represents the actual external disturbance torque experienced during times k-2 to k-1; b C b and K b These represent the ship's lateral moment of inertia, roll damping coefficient, and roll restoring moment, respectively; for a single gyroscope anti-roll device, m = 1, and for multiple gyroscope anti-roll devices, m > 1; H represents the angular momentum of the anti-roll gyroscope in the direction of its rotation axis; Δt represents the controller sampling time. and φ represents the angular velocity of the water platform at time k and time k-1, respectively; k-1 θ represents the rocking angle of the water platform at time k-1; k-1 and These represent the precession angle and precession angular velocity of the gyroscope at time k-1, respectively.

[0015] Preferably, the expression for obtaining the predicted disturbance torque sequence in the prediction time domain through quadratic interpolation in step S2 is as follows:

[0016]

[0017]

[0018] In the formula, i = {0, 1, 2, ..., N} P-1};N p This indicates the prediction time domain.

[0019] Preferably, the cost function of the model predictive control system (MPC) is expressed as follows:

[0020]

[0021] J i =J(φ)+J(θ),(i=k,k+1,…,k+N p );

[0022]

[0023] a = bs 2 ;

[0024] In the formula, J(φ) represents the cost function for the yaw angle φ; J(θ) represents the cost function for the precession angle θ; a, b, c, and d represent the weighting coefficients of the cost function; s and n represent the thresholds; N p This indicates the prediction time domain.

[0025] Preferably, the expression for the model predictive control system in step S3 is:

[0026]

[0027] In the formula, M represents the output torque of the precession control motor; W represents the external disturbance torque; H and I g C g and K g and represent the angular momentum, precession axis moment of inertia, precession axis damping coefficient, and precession axis restoring torque coefficient of the roll-damping gyroscope, respectively; I b C b and K bThese represent the ship's lateral moment of inertia, roll damping coefficient, and roll restoring moment, respectively; φ, and Represent the platform's yaw angle, yaw rate, and roll acceleration, respectively; θ, and These represent the precession angle, precession angular velocity, and precession angular acceleration of the gyroscope, respectively; for a single gyroscope anti-roll device, m = 1, and for multiple gyroscope anti-roll devices, m > 1.

[0028] Preferably, the model predictive control system is improved by using a variable gain control law.

[0029] Preferably, the expression for the model predictive control system improved by the variable gain control law is:

[0030]

[0031] In the formula, α k and β k The control parameters are the variable gain control parameters at time k.

[0032] Preferably, a control loop time domain N is introduced into the model predictive control system MPC. c Solve the problem.

[0033] Preferably, a control loop time domain N is added. c The subsequent solution methods include: in the control loop time domain N c Keep the control parameters constant, every N c The control parameters are updated once using the nine-square grid method, and the corresponding cost is obtained at each step. The costs are accumulated and N is used to predict the time domain. p The total cost within the range is used to obtain the optimal control sequence.

[0034] The present invention also provides a large inertia flywheel active roll reduction control system based on model predictive control, comprising: one or more processors and a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, and the one or more programs include methods for performing the above.

[0035] The beneficial effects of this invention include at least the following: By combining the swaying motion equations of the surface platform and the nonlinear motion equations of the anti-roll gyroscope, a model predictive control system is constructed. This retains the nonlinear terms, accurately describes the dynamic coupling between the surface platform and the anti-roll gyroscope, avoids linearization errors, predicts the swaying state in the future time domain in real time, and dynamically adjusts control parameters to adapt to the time-varying nature of wave disturbances. The control structure is simple, the control performance is excellent, it can adapt to more complex and variable deep-sea conditions, and it is highly practical. Most importantly, it can optimize anti-roll performance while ensuring safe and reliable operation, improving the anti-roll efficiency of the anti-roll gyroscope and effectively enhancing the stability of the surface platform under wind and waves. Attached Figure Description

[0036] Figure 1 This is a schematic diagram of the method flow according to an embodiment of the present invention;

[0037] Figure 2 This is a schematic diagram of the overall control flow of the MPC variable gain control for the anti-roll gyroscope according to an embodiment of the present invention;

[0038] Figure 3 This is a schematic diagram of the simulation curve of the external disturbance torque in an embodiment of the present invention;

[0039] Figure 4 This is a comparative schematic diagram of external disturbance torque prediction in an embodiment of the present invention;

[0040] Figure 5 This is a schematic diagram of the simulation curve of the roll angle of a ship without roll stabilization according to an embodiment of the present invention;

[0041] Figure 6 A schematic diagram of the simulated roll angle curve of a ship with an additional anti-roll gyroscope is provided for an embodiment of the present invention.

[0042] Figure 7 This is a schematic diagram of the logic control in the time domain of the MPC control loop according to an embodiment of the present invention;

[0043] Figure 8 This is a schematic diagram of prediction and control at time k in an embodiment of the present invention;

[0044] Figure 9 This is a schematic diagram of the second round of control parameter selection for variable gain control according to an embodiment of the present invention;

[0045] Figure 10 This is a schematic diagram of the MPC variable gain control nine-square grid traversal optimization logic in an embodiment of the present invention;

[0046] Figure 11 This is a schematic diagram of the overall optimization logic of MPC variable gain control in an embodiment of the present invention;

[0047] Figure 12 The diagram shows the trajectory of the control parameters for the MPC variable gain control in an embodiment of the present invention. Detailed Implementation

[0048] The technical solutions of 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 protection scope of the present invention.

[0049] Example 1

[0050] This invention provides a method for active roll reduction control of a large inertia flywheel based on model predictive control, comprising the following steps:

[0051] Step S1: Acquire the motion status data of the water surface platform and the anti-roll gyroscope in real time.

[0052] Specifically, real-time motion data of the surface platform and the anti-roll gyroscope are acquired. The real-time motion data includes the yaw angle and yaw rate of the surface platform, and the precession angle and precession rate of the anti-roll gyroscope. The surface platform includes ships, marine nuclear power platforms, and offshore wind power platforms, etc.

[0053] Step S2: After calculating the historical real-time disturbance torque data based on the motion state data, the predicted disturbance torque sequence in the prediction time domain is obtained by secondary interpolation.

[0054] Specifically, the process of obtaining the actual external disturbance torque based on real-time motion state data includes: at time k-1, acquiring the real-time motion state data X of the platform and the anti-roll gyroscope at time k-1 through sensors. k-1 , i.e. φ k-1 , θ k-1 and Where φ and Let θ be the yaw angle and yaw rate of the water platform; and θ and To reduce the precession angle and angular velocity of the gyroscope, the subscript k reflects the sampling time; during the time interval from k-1 to k, the combined system is subjected to an external disturbance torque W. k The effect is that the motion state data changes, that is, the real-time motion state data becomes X. k , i.e. φ k , θ k and The actual external disturbance torque W experienced during the time intervals from k-2 to k-1 can be derived from the joint dynamic equations. k-1 The expression is:

[0055]

[0056] Among them, I b C b and K b These represent the ship's lateral moment of inertia, roll damping coefficient, and roll restoring moment, respectively; H represents the angular momentum of the roll-damping gyroscope in the direction of its rotation axis; Δt is the controller sampling time. For a single gyroscope roll-damping device, m = 1; for multiple gyroscope roll-damping devices, m > 1.

[0057] Then, based on historical data, the external disturbance sequence within the prediction time domain is determined as follows:

[0058] At time k, a quadratic interpolation method is used, employing the actual external disturbance torque W stored from the previous three time points. k-1 W k-2 and W k-3 The external disturbance torque at time k is predicted. The quadratic interpolation expression is:

[0059]

[0060] Define an accurate historical sequence of disturbance torques. With the estimated prediction time domain N p The unknown external interference sequence They are respectively:

[0061]

[0062] The entire prediction time domain N is derived using the same method. p Internal and external interference sequences for:

[0063]

[0064] In the formula, i = {0, 1, 2, ..., N} P-1}

[0065] Step S3: Input the predicted disturbance torque sequence into the model predictive control system (MPC) to generate the optimal control sequence; the model predictive control system consists of the water surface platform swaying motion equation and the anti-sway gyroscope motion nonlinear equation.

[0066] Specifically, the expression for the Model Predictive Control System (MPC) is:

[0067]

[0068] In the formula, M represents the output torque of the precession control motor; W represents the random wave disturbance torque; H and I g C g and K gand represent the angular momentum, precession axis moment of inertia, precession axis damping coefficient, and precession axis restoring torque coefficient of the roll-damping gyroscope, respectively; I b C b and K b These represent the ship's lateral moment of inertia, roll damping coefficient, and roll restoring moment, respectively; φ, and Represent the platform's yaw angle, yaw rate, and roll acceleration, respectively; θ, and These represent the precession angle, precession angular velocity, and precession angular acceleration of the gyroscope, respectively. For a single gyroscope anti-roll device, m = 1; for multiple gyroscope anti-roll devices, m > 1.

[0069] In this embodiment, the following cost function is proposed for solving the problem: the highest priority performance indicator for the roll reduction gyroscope is the roll angle of the water platform, and the smaller this angle, the better. Furthermore, considering that when the roll angle is less than a critical value, the impact of the platform's roll on the human body and mechanical equipment is minimal, and that obtaining a smaller roll angle at this point requires a greater cost in precession angle, it is desirable that the cost corresponding to the roll angle is equal when the roll angle is less than ±s (rad), thereby obtaining a better precession angle. The cost function for the roll angle is:

[0070]

[0071] Precession angle, while not a performance indicator, reflects the stability and lifespan of the roll stabilization gyroscope. It needs to be transformed from a hard constraint to a soft constraint through a cost function design. When the precession angle exceeds ±π / 2 rad, the roll stabilization gyroscope will exhibit runaway behavior. Furthermore, the greater the precession angle, the greater the wear on its precession bearing. Therefore, the precession angle also needs to be expressed as a piecewise function. When the precession angle exceeds the critical value ±n (rad), a penalty function is used to prevent the precession angle from increasing further. The cost function for the roll angle is:

[0072]

[0073] Taking into account both the platform's yaw angle and the precession angle of the anti-yaw gyroscope, the cost function for MPC variable gain control is as follows:

[0074] J i =J(φ)+J(θ),(i=k,k+1,…,k+N p );

[0075] Where a, b, c, and d are the weighting coefficients of the cost function, a and b reflect the degree of emphasis on the roll reduction performance of the roll reduction gyroscope, and c and d reflect the degree of emphasis on the operational stability of the roll reduction gyroscope, and a = bs 2 .

[0076] MPC prediction time domain N p The total cost within is:

[0077]

[0078] Generally, anti-roll gyroscopes prioritize anti-roll performance, so b >> d >> c.

[0079] Step S4: Map the optimal control sequence to the output torque of the motor, control the precession axis of the anti-roll gyroscope, and form a closed-loop feedback.

[0080] Example 2

[0081] The output torque M of the motor is the manipulated variable of the model predictive control. Within the motor's processing range, M varies considerably, which is detrimental to subsequent optimization. To improve the optimization speed, this embodiment, based on Embodiment 1, combines a variable gain control law to obtain a new joint model as shown below. The optimization process is as follows: Figure 2 As shown.

[0082]

[0083] Here, α and β are parameters of the variable gain control law, which are the manipulated variables of model predictive control. A set of control parameters (α, β) corresponds to a control motor output torque M. Since the precession angular velocity of the gyroscope and the yaw angle and yaw velocity of the platform have a general range of variation, the parameters α and β can be determined to have a general range of values, which greatly narrows the optimization range compared with the range of values ​​of the motor output M.

[0084] The discretized prediction model then adopts a nonlinear model, taking into account nonlinear terms. The impact on the joint dynamics model, specifically the discretized nonlinear model, is as follows:

[0085]

[0086] As can be seen from the above formula, once the control parameters (α, β) are determined, a control torque M of the motor can be obtained. Therefore, it is only necessary to determine the optimal control parameters to obtain the optimal control torque.

[0087] As can be seen from the above expression, the larger the values ​​of the control parameters (α, β), the greater the precession angular velocity of the roll stabilization gyroscope. Therefore, the initial values ​​of α and β should be located as close as possible to the lower left corner of the constrained quadrilateral, and should be avoided at the upper right corner. This method of value selection is more conservative. Although a larger value of (α, β) results in a larger precession angular velocity and a larger roll stabilization torque fed back to the platform by the roll stabilization gyroscope, a larger precession angular velocity is difficult to control, and the controller may be unable to constrain the precession angle within ±π / 2 rad. Conversely, smaller initial values ​​of α and β, although the roll stabilization effect is poor when the roll stabilization gyroscope is first started, result in better stability, and the roll stabilization effect will gradually improve over time.

[0088] Taking a certain ship as an example, this paper compares the implementation effect of model predictive control combined with variable gain control law with the traditional roll reduction gyroscope precession control algorithm.

[0089] The rolling moment of inertia I of a certain ship b = 1,800,000 kg.m 2 Roll damping coefficient C b = 411040 Nms, roll restoring torque coefficient K b =1113200Nm, angular momentum H of the gyro anti-roll device =175930Nms, moment of inertia I along the precession axis g =1200kg.m 2 Precession shaft damping coefficient C g = 4500 Nms, restoring torque coefficient K in the precession axis direction g =2000Nm.

[0090] The disturbance torque of ocean waves is proportional to the wave dip angle, which is defined by the wave dip angle spectral density function.

[0091]

[0092] In the formula, ω is the angular frequency, ε is the random phase angle within (0, 2π), and the encounter frequency ω e =ω-ω 2 / g·Vcosμ, where V is the ship's speed and μ is the encounter angle; wave tilt angle spectrum function S σ (ω)=ω 4 / g 2 ·S ξ (ω), Ocean wave energy spectrum S ξ (ω) uses the ITTC two-parameter spectrum, and its expression is:

[0093]

[0094] In the formula, h 1 / 3 Let h be the significant wave height, and T1 be the average period of the wave. 1 / 3The roll reduction effect of different control laws was tested at sea state 4m. The external disturbance torque is shown in the attached figure. Figure 3 As shown.

[0095] Figure 4 A comparison chart of predicted external disturbance torques at a specific moment was extracted. As can be seen from the chart, the prediction error increases with the increase of the prediction time domain. However, considering the control characteristics of the MPC and the application scenarios of the anti-roll gyroscope, the prediction time domain is generally not taken to be too large. The maximum error between the predicted external disturbance torque and the actual disturbance torque is 0.0031%, which does not exceed 0.01%, indicating that the fitting formula has a good prediction effect.

[0096] With meaningful wave height h 1 / 3 The sea state of 4m represents an external disturbance. Simulation calculations were performed, and the simulated roll angle curve is shown in the attached figure. Figure 5 and Figure 6 As shown.

[0097] Simulation curves show that the anti-roll gyroscope device using MPC variable gain control law for active control exhibits the best anti-roll effect. To quantify the anti-roll effect of the gyroscope device, the anti-roll rate is used as the evaluation criterion to reflect the actual anti-roll performance of the gyroscope. The formula for calculating the anti-roll rate R is:

[0098]

[0099] Where X represents the mean square error of the roll angle of the ship without the gyro-based roll reduction device, and Y represents the mean square error of the roll angle of the ship after the addition of the gyro-based roll reduction device.

[0100] The statistical table of the roll reduction effect under different control laws is shown in Appendix Table 1. It can be found that under the same meaningful wave height, the roll reduction effect of the active control gyroscope precession using the MPC variable gain control law is the best, with a roll reduction rate of 99.29%, a maximum roll angle of only 0.43°, and a root mean square error of only 0.02°.

[0101] Table 1

[0102]

[0103]

[0104] Example 3

[0105] Based on Example 2, this embodiment adds a new parameter to the MPC variable gain control, namely the control cycle time domain N. c It also provides a nine-square grid optimization method.

[0106] Control Cyclic Time Domain N cThis is because the frequency of external disturbance torque changes from ocean waves in the actual ocean is much smaller than the sampling frequency of the control, and combined with the traditional MPC control time domain concept, a quantity representing the time domain length of the control parameters is set.

[0107] The reason for not using the traditional MPC control time domain is that external disturbances in the joint dynamics model change constantly. Using the traditional control time domain approach, where the control variable remains consistent with the previous value after the control time domain expires, will result in significant deviations in later prediction time periods due to larger changes in external disturbances. By using a control cyclic time domain, updating the control parameters at regular intervals, we achieve the computational efficiency of the traditional MPC control time domain while avoiding excessive errors caused by an overly long control time domain.

[0108] In this embodiment, the control loop time domain N c The specific value needs to be set with reference to the actual sea conditions. Generally, due to the complexity and variability of sea conditions, the prediction time domain N... p Excessive size can lead to large deviations between the data in the later part of the external interference sequence and the actual data. Therefore, the prediction time domain and the control loop time domain should not be too large. This embodiment is only for illustration and does not limit the prediction time domain and the control loop time domain.

[0109] Figure 7 This explains when N p =5, N c When = 2, a simplified diagram of the prediction and control logic at time k. As can be seen from the diagram, the control loop time domain N... c =2 means that the control torque is updated every two steps within the prediction time domain. That is, in the control loop time domain N c The internal control torque remains unchanged.

[0110] The specific steps of the algorithm for optimizing MPC variable gain control in the prediction time domain using the nine-square grid algorithm include:

[0111] Appendix Figure 8 Reflects the prediction time domain N p =5, controlling the time domain N c When W = 2, the prediction and control of the joint system at time k is considered. It should be noted that, due to the backdifference of the second derivative in the equations, the influence of the external disturbance torque W and the output torque M of the precession control motor on the real-time motion state X of the water platform and the anti-roll gyroscope will be delayed by one step, such as W... k and M k Influence X k+1 .

[0112] Predicting time domain N p=5, it is necessary to obtain the motion state of the joint system from time k+1 to k+5 based on the prediction model. The yaw angle and precession angle of the joint system at time k+5 can be directly derived from the state at time k+4. The state variables at time k+4 and and M k+3 Since they are directly related, the control parameters of the model only need to be derived up to time k+3.

[0113] The real-time motion data X of the platform and the anti-roll gyroscope from time k-3 to k is used. k-3 X k-2 X k-1 X k The actual external disturbance torque W at times k-3 to k-1 was calculated. k-1 W k-2 W k-3 The external disturbance torque in the prediction time domain was predicted using the quadratic interpolation method. and

[0114] At time k, based on the prediction model and the external disturbance torque... and control motor control parameters (α) k ,β k ), that is, controlling the output power M of the motor. k The motion state data of the water surface platform and the anti-roll gyroscope at time k+1 are predicted.

[0115]

[0116] Since this step does not involve selecting new control parameters, the cost function is not calculated.

[0117] At time k+1, with Δα and Δβ as the moving step size, take (α... k ,β k ) 8 surrounding points, obtaining (α) k ,β k A total of 9 sets of control parameters (α) are to be determined, including α. k+1 ,β k+1 );

[0118]

[0119] Based on the prediction model and external disturbance torque The predicted motion state data of the water surface platform and the anti-roll gyroscope at time k+2 were obtained under the action of 9 sets of control parameters (α, β). And control the motor output torque M k+1 ;

[0120]

[0121] Although this step involves the selection of new control parameters, the quantities directly related to the cost function are... and It is not associated with it, and the cost function is not calculated.

[0122] At time k+2, at position N c In the time domain, to save algorithm resources and keep the control torque M constant, the control motor output M obtained at time k+1 is directly used at time k+2. k+1 =M k+2 , i.e. α k+1 =α k+2 and β k+1 =β k+2 From this, the motion state data of the water surface platform and the anti-roll gyroscope at time k+3 can be estimated.

[0123]

[0124] The predicted motion state data at time k+3 Substituting the cost function, we obtain the cost J corresponding to the 9 sets of control parameters. k+3 ;

[0125] At time k+3, using Δα and Δβ as the moving step sizes, as follows: Figure 9 As shown, 9 groups (α) were selected respectively. k+1 ,β k+1 Taking into account the 8 surrounding points and itself, a total of 9x9 sets of undetermined control parameters (α) are obtained. k+2 ,β k+2 Since the initial values ​​of the state variables are different, (α) k -Δα,β k -Δβ) through +Δα and +Δβ to get (α k ,β k ) and (α k +Δα,β k +Δβ) through -Δα and -Δβ to get (α k ,β k The final values ​​of the state variables are not the same.

[0126] Based on the prediction model and the external disturbance torque The predicted motion state data of the water surface platform and the anti-roll gyroscope at time k+4 were obtained under the action of the 81 sets of control parameters (α, β). And control the output torque of the motor;

[0127]

[0128] The predicted motion state data at time k+4 Substituting the cost function, we obtain the cost J corresponding to the 9 sets of control parameters. k+4 ;

[0129] Considering that the cost function only includes the yaw angle and the precession angle, and the yaw angle and precession angle at time k+5 can be obtained by simply using the real-time motion state data of the water surface platform and the anti-roll gyroscope at time k+4, it is not necessary to obtain the control torque at time k+4 to obtain the yaw angle and precession angle at time k+5.

[0130]

[0131] Unlike the cost function described above, different initial values ​​for state variables... With control parameter (α) k+2 ,β k+2 The combination of these factors yields 81 sets of cost function values ​​J. k+5 .

[0132] The costs of the three steps are summed to obtain the total cost ∑J = J for the 81 sets of control parameters. k+3 +J k+4 +J k+5 The group with the lowest cost is selected, and its corresponding control parameters are chosen as the control parameters (α) at ​​time k. k ,β k ).

[0133] The optimal control parameters (α) at ​​time k k+1 ,β k+1 Substituting the nonlinear system joint model, we obtain the output torque command M of the control motor at that moment. k+1 .

[0134] In summary, when choosing other prediction time domains, such as prediction time domain N... p When the value is >5, repeat the above steps: in the control loop time domain N c Keep the control parameters constant, every N c The control parameters are updated once using the nine-square grid method, and the corresponding cost is obtained at each step. The costs are accumulated and N is used to predict the time domain. p The total cost within the time frame is compared to find the minimum cost. Based on this, the optimal control sequence is selected, and the control torque M at time k is applied to the system only. k+1 The remaining control sequences are discarded.

[0135] Appendix Figure 10 This diagram illustrates the determination of MPC control parameters using the nine-square grid method. The nine blue squares represent the nine control parameters in the first round, and the 81 black squares represent the 81 control parameters in the second round. Each square reflects the change path of a control parameter and corresponds to the total cost in the prediction time domain under that path.

[0136] Appendix Figure 11 This reflects the overall optimization logic of MPC variable gain control, which is to gradually approach the optimal control parameters through a process of traversal and selection.

[0137] Appendix Figure 12 In a specific case, the control parameters change path. Since the frequency of external disturbances is much lower than the sampling frequency of the controller, after a period of optimization, the control parameters are optimized from the lower left corner to the upper right corner, and the optimal control sequence is found on the upper and right sides of the constraint quadrilateral as the external disturbances change.

[0138] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described; only preferred embodiments of the present invention are illustrated. The descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the present invention. As long as the combination of these technical features does not contradict each other, it should be considered within the scope of this specification.

[0139] It should be noted that those skilled in the art can make various modifications and improvements without departing from the inventive concept, and these all fall within the scope of protection of this invention. Therefore, the scope of protection of this invention should be determined by the appended claims.

Claims

1. A method for active roll reduction control of a large inertia flywheel based on model predictive control, characterized in that: Includes the following steps: Step S1: Acquire real-time motion state data of the water surface platform and the anti-roll gyroscope; Step S2: After calculating the historical real-time disturbance torque data based on the motion state data, the predicted disturbance torque sequence in the prediction time domain is obtained by secondary interpolation; Step S3: Input the predicted disturbance torque sequence into the model predictive control system (MPC) to generate the optimal control sequence; the model predictive control system consists of the water surface platform swaying motion equation and the anti-sway gyroscope motion nonlinear equation; Step S4: Map the optimal control sequence to the output torque of the motor, control the precession axis of the roll reduction gyroscope, and form a closed-loop feedback.

2. The active roll reduction control method for a large inertia flywheel based on model predictive control according to claim 1, characterized in that: The expression for calculating the historical real-time disturbance torque data based on the motion state data in step S2 is as follows: In the formula, W k-1 I represents the actual external disturbance torque experienced during times k-2 to k-1; b C b and K b These represent the ship's lateral moment of inertia, roll damping coefficient, and roll restoring moment, respectively; for a single gyroscope anti-roll device, m = 1, and for multiple gyroscope anti-roll devices, m > 1; H represents the angular momentum of the anti-roll gyroscope in the direction of its rotation axis; Δt represents the controller sampling time. and φ represents the angular velocity of the water platform at time k and time k-1, respectively; k-1 This represents the sway angle of the water platform at time k-1; θ k-1 and These represent the precession angle and precession angular velocity of the gyroscope at time k-1, respectively.

3. The active roll reduction control method for a large inertia flywheel based on model predictive control according to claim 2, characterized in that: The expression for the predicted disturbance torque sequence in the prediction time domain obtained by quadratic interpolation in step S2 is as follows: In the formula, i = {0, 1, 2, ..., N} P-1 };N p This indicates the prediction time domain.

4. The active roll reduction control method for a large inertia flywheel based on model predictive control according to claim 1, characterized in that: The expression for the cost function of the Model Predictive Control System (MPC) is as follows: Ji=J(φ)+J(θ), (i=k,k+1,…,k+Np); a=bs 2 In the formula, J(φ) represents the cost function for the yaw angle φ; J(θ) represents the cost function for the precession angle θ; a, b, c, and d represent the weighting coefficients of the cost function; s and n represent the thresholds; N p This indicates the prediction time domain.

5. The active roll reduction control method for a large inertia flywheel based on model predictive control according to claim 1, characterized in that: The expression for the model predictive control system described in step S3 is: In the formula, M represents the output torque of the precession control motor; W represents the external disturbance torque; H and I g C g and K g and represent the angular momentum, precession axis moment of inertia, precession axis damping coefficient, and precession axis restoring torque coefficient of the roll-damping gyroscope, respectively; I b C b and K b These represent the ship's lateral moment of inertia, roll damping coefficient, and roll restoring moment, respectively; φ, and Represent the platform's yaw angle, yaw rate, and roll acceleration, respectively; θ, and These represent the precession angle, precession angular velocity, and precession angular acceleration of the gyroscope, respectively; for a single gyroscope anti-roll device, m = 1, and for multiple gyroscope anti-roll devices, m >

1.

6. The active roll reduction control method for a large inertia flywheel based on model predictive control according to claim 5, characterized in that: The model predictive control system is improved by using a variable gain control law.

7. The active roll reduction control method for a large inertia flywheel based on model predictive control according to claim 6, characterized in that: The expression for the model predictive control system improved by the variable gain control law is: In the formula, α k and β k The control parameters are the variable gain control parameters at time k.

8. The active roll reduction control method for a large inertia flywheel based on model predictive control according to claim 1, characterized in that: In the model predictive control system MPC, a control loop time domain N is introduced. c Solve the problem.

9. The active roll reduction control method for a large inertia flywheel based on model predictive control according to claim 8, characterized in that: Add control loop time domain N c The subsequent solution methods include: in the control loop time domain N c Keep the control parameters constant, every N c The control parameters are updated once using the nine-square grid method, and the corresponding cost is obtained at each step. The costs are accumulated and N is used to predict the time domain. p The total cost within the range is used to obtain the optimal control sequence.

10. A large-inertia flywheel active roll reduction control system based on model predictive control, characterized in that: include: One or more processors and a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs comprising methods for performing any one of claims 1 to 9.

Images