Marine three-phase asynchronous motor neural network inversion sliding mode control method

Abstract

The invention provides a marine three-phase asynchronous motor neural network inversion sliding mode control method, which combines the advantages of nonlinear approximation capability and good real-time performance of an RBF neural network and good dynamic performance and strong anti-interference performance of sliding mode control, makes full use of the back-off decoupling characteristic of a back-stepping method, and improves the reliability of the control system. The neural network inversion sliding mode control method for the three-phase asynchronous motor for the ship is designed, and the purpose of improving the performance of a motor control system is achieved. For the problem that parameters cannot be determined in the modeling process of the marine three-phase asynchronous motor, a forward stability augmentation channel is designed. A backstepping method is adopted to perform decoupling control on the system, sliding mode control solves the disturbance problem of the control system, and the effect of an RBF neural network performs real-time estimation and compensates disturbance between system coupling of contact surfaces between control system frames and nonlinear friction force.

Description

Technical Field

[0001] The invention belongs to the technical field of ship electrical equipment, and in particular relates to a neural network inversion sliding mode control method for a ship-used three-phase asynchronous motor. Background Art

[0002] The three-phase asynchronous motor has the advantages of light weight, simple structure, good dynamic performance and high efficiency, and is increasingly widely used in the shipbuilding industry. The three-phase asynchronous motor is a variable parameter, multi-variable, strongly coupled, nonlinear system, and better control of the motor speed is a very critical technology. Conventional speed loop PI control has high speed overshoot and poor anti-external interference performance, which cannot meet the control system requirements at all. Especially when the system parameters change and the external load changes suddenly, the control effect will be very poor. In response to the above problems, many scholars have proposed advanced control methods, such as PID control, adaptive control, fuzzy control and sliding mode control.

[0003] Among the methods for controlling marine three-phase asynchronous motors, sliding mode control is highly robust to parameter changes and model uncertainty and unknown disturbances, and is widely used in the field of motor control. When the control system state is on the sliding mode surface, the sliding mode control method will experience chattering, which will have an adverse effect on the motor control system, not only affecting the motor's service life, but also affecting the speed tracking accuracy.

[0004] Usually, the PID controller used in the speed loop is just a very common linear PID regulator. Traditional PID control cannot meet the requirements and needs to rely on experience to continuously adjust the PID parameters. The parameters of marine three-phase asynchronous motors are time-varying and easily affected by external disturbances. The conventional PID speed controller parameter adjustment and speed tracking effects are not ideal. Summary of the invention

[0005] In order to solve the technical problems in the prior art, the present invention provides a marine three-phase asynchronous motor neural network inversion sliding mode control method, comprising the following steps:

[0006] Establish mathematical models of three-phase asynchronous motors, including mathematical models in a stationary three-phase coordinate system, mathematical models in a two-phase stationary coordinate system, and mathematical models in a two-phase synchronous rotating coordinate system;

[0007] Neural network sliding mode adaptive control, introducing RBF neural network algorithm to estimate the uncertain part of the marine motor dynamics model;

[0008] Control system analysis and design, used to make the motor control system have good static and dynamic performance, including backstepping decoupling control, sliding mode-based anti-disturbance control, nonlinear friction and system coupling estimator.

[0009] Furthermore, the mathematical model in the stationary three-phase coordinate system includes the following features:

[0010] The mathematical model of a typical motor includes the following equations:

[0011] Stator voltage equation

[0012]

[0013] In the formula, u a 、u b 、u c is the three-phase winding voltage, i a 、i b 、i c is the three-phase winding current, R s is the phase winding resistance, ψ a , b , c It is the full flux linkage of three-phase stator;

[0014] Stator flux equation

[0015]

[0016] In the formula, ψ f is the permanent magnet flux, L a , L b , L c is the self-inductance of the stator three-phase winding, L ab , L ba , L ac , L ca , L bc , L cb is the mutual inductance of the three-phase stator winding, and L ab =L ba , L ac =L ca , L cb =L cb .

[0017] Furthermore, the mathematical model in the two-phase stationary coordinate system includes the following features:

[0018] According to the principle of equal phase amplitude transformation, the Clarke transformation expression is as follows:

[0019]

[0020] i a +i b +i b =0, and we can get:

[0021]

[0022] The corresponding inverse Clarke transform is:

[0023]

[0024] The voltage equation is:

[0025]

[0026] The magnetic flux equation is:

[0027]

[0028] Where U α , U β is the αβ axis voltage, i α 、i β is the α and β axis current, Ψ α ,Ψ β is the voltage of α and β axes, L is the voltage of α and β axes, ω e is the electrical angular velocity of the rotor.

[0029] Furthermore, the mathematical model in the two-phase synchronous rotating coordinate system includes:

[0030] The Park transformation expression is as follows:

[0031]

[0032] The Park inverse transform expression is as follows:

[0033]

[0034] Where P 2s / 2r is the Park transformation matrix, P 2r / 2s is the Park inverse transformation matrix;

[0035] Stator voltage equation:

[0036]

[0037] Stator flux equation:

[0038] ψ d =L d i d +ψ f

[0039] ψ d =L q i q

[0040] where u d 、u q are the dq axis components of the stator respectively; i d 、iq are the dq axis components of the stator current respectively; Rs is the stator winding resistance; ψ d , q is the axial component of the stator flux; ω r is the rotor electrical angular velocity; L d , L q They are the shaft inductance components; ψ f represents the permanent magnet flux;

[0041] The electromagnetic torque equation is:

[0042] T e =nP n [ψ d i q -ψ q i d ]P n i q [ψ f +(L d -L q )i d ]

[0043] The mechanical motion equation is:

[0044]

[0045] In the formula, J is the moment of inertia, B is the friction coefficient, T e is the electromagnetic torque, T L Indicates the load torque of the motor.

[0046] Furthermore, the sliding mode adaptive control of the neural network includes the radial basis neural network design:

[0047] The actual output of the radial basis neural network estimated marine motor model is fNN∈n, and the ideal output is f*NN∈n, as shown in the following formula:

[0048]

[0049] f NN =w T δ(x)

[0050] Where: f is the actual output of the neural network; w is the optimal weight of the neural network * ;δ is the basis function of the neural network;L is the number of nodes in the hidden layer of the neural network;

[0051] The Gaussian basis function expression is:

[0052]

[0053] i=1,2,…,n;j=1,2,…,L

[0054]

[0055] c i =[c i1 ,c i2 ,…,c iL ] T

[0056] b i =[b i1 ,b i2 ,…,b iL ]

[0057] Where: x i is the input vector of the i-th neural network; c i is the central matrix of the ith neural network; b i is the basis width vector of the i-th neural network;

[0058] Unknown nonlinear functions of motor control systems:

[0059]

[0060] Where: f1 is the unknown nonlinear function; M(q) is the positive definite inertia matrix; is the motor angular velocity; is the angular acceleration of the motor; are centrifugal force and Coriolis force; G(q) is the gravity term vector; τ d To interfere; is the expected trajectory angular acceleration.

[0061] Furthermore, sliding mode control requires a high degree of mathematical model accuracy. The unknown nonlinear function of the motor control system contains modeling errors and interference from unknown factors. RBF approximation of nonlinear functions requires a low degree of mathematical model accuracy and can control the motor without an accurate model.

[0062] Neural networks have approximation errors. When the number of nodes in the middle layer is small, it will affect the performance of the control system. In order to make fNN infinitely close to f*NN, the adaptive compensation update law and the design weight update law are as follows:

[0063]

[0064] Where: δ is the basis function of the neural network.

[0065] Furthermore, the backstepping decoupling control includes:

[0066] Assume that the expected attitude signal is x d =[x dx x dy xdz ] T , define the tracking angle error:

[0067]

[0068] Where: z1 is the tracking angle error; x1 is the inertial angle of the system; x2 is the inertial angular velocity of the system; x d is the expected posture signal;

[0069] The Lyapunov function is designed using the backstepping control method:

[0070]

[0071] Where: L1 is the designed Lyapunov function; z1 is the defined tracking angle error;

[0072] Differentiating L1 yields:

[0073]

[0074] Where: L1 is the designed Lyapunov function; z1 is the defined tracking angle error; x2 is the inertial angular velocity of the system; x d is the expected posture signal;

[0075] Pick z2 is the control amount of the virtual control item, that is:

[0076]

[0077] Differentiating L2 gives:

[0078]

[0079] Design system control law so that

[0080] Where: A and B are system parameters respectively; C is a 3×3 positive definite diagonal matrix; T M is the system control rate; δ is the basis function of the neural network; z1 is the tracking angle error; z2 is the control amount of the virtual control item; d t is the unknown external disturbance; x2 is the inertial angular velocity of the system; x d is the expected posture signal;

[0081]

[0082] Where: A and B are system parameters respectively; C is a 3×3 positive definite diagonal matrix; T M is the system control rate; δ is the basis function of the neural network; z1 is the tracking angle error; z2 is the control amount of the virtual control item; dt is the unknown external disturbance; x2 is the inertial angular velocity of the system; x d is the expected attitude signal; C ′ is a 3×3 positive definite diagonal matrix;

[0083] The control rate T M Substitution We can get:

[0084]

[0085] Where: C is a 3×3 positive definite diagonal matrix; C ′ is a 3×3 positive definite diagonal matrix; z1 is the definition of the tracking angle error; z2 is the control amount of the virtual control item.

[0086] When t→∞, z1→0 and z2→0, according to the tracking angle error, we have:

[0087] x1→x d and

[0088] Furthermore, the anti-interference control based on sliding mode includes:

[0089] The sliding surface is defined as:

[0090] σ=Kz1+z2

[0091] Where: σ is the sliding surface; K is a 3×3 positive definite diagonal matrix; z1 is the tracking angle error; z2 is the control amount of the virtual control item;

[0092] Substituting the Lyapunov function into the sliding surface formula, we can get:

[0093]

[0094] Where: σ is the sliding surface; K and C are 3×3 positive definite diagonal matrices; z1 is the tracking angle error;

[0095] if but Redesign the Lyapunov function of backstepping control:

[0096]

[0097] Where: L1, L2 are Lyapunov functions; σ is the sliding surface defined;

[0098] Differentiating L2 gives:

[0099]

[0100] Differentiate σ and substitute the stator flux equation into it to obtain:

[0101]

[0102] To make Based on the tracking angle error, the control rate is updated as:

[0103]

[0104] Where: ф, ψ are 3×3 diagonal matrices. A and B are system parameters respectively; C is a 3×3 positive definite diagonal matrix; T M is the system control rate; δ is the basis function of the neural network; z1 is the tracking angle error; z2 is the control amount of the virtual control item; d t is the unknown external disturbance; x2 is the inertial angular velocity of the system; x d is the expected attitude signal.

[0105] Furthermore, the nonlinear friction and system coupling estimator specifically includes: using the RBF network structure to establish an adaptive law to approximate the nonlinear friction force δ, and the RBF network algorithm is:

[0106]

[0107] Where: δ is the nonlinear friction force; j = 1, 2, ... ... 7, x is the network input; is the ideal weight of the network; h(x) 7×1 is the network Gaussian basis function output; ε 3×1 is the network approximation error; c j is the coordinate of the center point of the jth neuron basis function in the hidden layer of the network, b j is the propagation range of the jth neuron basis function in the network hidden layer;

[0108] The network input is the motor angular velocity x = [ωp x ,ωp y ,ωp z ] T , then the network output is:

[0109]

[0110] Where: Output for the network; is the ideal weight of the network; h(x) 7×1 is the network Gaussian basis function output;

[0111] Pick Then the error of the estimator is:

[0112]

[0113] Where: is the ideal weight of the network; h(x) is the output of the network Gaussian basis function; ε is the network approximation error;

[0114] The RBFNN update law is:

[0115]

[0116] Where: B 3×3 , G 3×3 are positive definite diagonal matrices; h(x) 7×1 is the network Gaussian basis function output;

[0117] After applying the disturbance estimator, the control tracking angle error is updated as:

[0118]

[0119] Where: ф, ψ are 3×3 diagonal matrices; A and B are system parameters; C is a 3×3 positive definite diagonal matrix; T M is the system control rate; σ is the sliding surface; z1 is the tracking angle error; z2 is the control amount of the virtual control item; d t is the unknown external disturbance; x2 is the inertial angular velocity of the system; x d is the expected attitude signal.

[0120] Compared with the prior art, the present invention has the following beneficial effects:

[0121] The present invention aims at the complex nonlinear system of marine three-phase asynchronous motors, reduces the influence of external unknown interference, facilitates real-time estimation and compensation for the disturbance between the nonlinear friction force between control system frames and system coupling, and ensures system stability. BRIEF DESCRIPTION OF THE DRAWINGS

[0122] Figure 1 It is the NNBSMC control flow chart of the present invention. DETAILED DESCRIPTION

[0123] In order to enable those skilled in the art to better understand the technical solution of the present invention, the present invention is further described below in conjunction with the accompanying drawings and specific implementation methods.

[0124] The present application proposes a neural network back-stepping sliding mode control method for a marine three-phase asynchronous motor, which specifically includes:

[0125] 1 Mathematical model of three-phase asynchronous motor

[0126] 1.1 Mathematical model in the stationary three-phase coordinate system

[0127] Usually the mathematical model of the motor includes the following equations, stator voltage equation:

[0128]

[0129] In the formula, u a 、u b 、u c is the three-phase winding voltage, i a 、i b 、i c is the three-phase winding current, R s is the phase winding resistance, ψ a , b , c It is the full flux linkage of three-phase stator.

[0130] Stator flux equation:

[0131]

[0132] In the formula, ψ f is the permanent magnet flux, L a , L b , L c is the self-inductance of the stator three-phase winding, L ab , L ba , L ac , L ca , L bc , L cb is the mutual inductance of the three-phase stator winding, and L ab =L ba , L ac =L ca , L cb =L cb .

[0133] 1.2 Mathematical model in two-phase stationary coordinate system

[0134] According to the principle of equal phase amplitude transformation, the Clarke transformation expression is as follows:

[0135]

[0136] i a +i b +i b =0, and we can get:

[0137]

[0138] The corresponding inverse Clarke transform is:

[0139]

[0140] The voltage equation is:

[0141]

[0142] The magnetic flux equation is:

[0143]

[0144] Where U α , U β is the αβ axis voltage, i α 、i β is the α and β axis current, Ψ α , β is the voltage of α and β axes, L is the voltage of α and β axes, ω e is the electrical angular velocity of the rotor.

[0145] 1.3 Mathematical model in two-phase synchronous rotating coordinate system

[0146] The Park transformation expression is as follows:

[0147]

[0148] The Park inverse transform expression is as follows:

[0149]

[0150] Where P 2s / 2r is the Park transformation matrix, P 2r / 2s Inverse Park transform matrix.

[0151] Stator voltage equation:

[0152]

[0153] Stator flux equation:

[0154]

[0155] where u d 、u q are the dq axis components of the stator respectively; i d 、i q are the dq axis components of the stator current respectively; Rs is the stator winding resistance; ψ d , q is the axial component of the stator flux; ω r is the rotor electrical angular velocity; L d , L q They are the shaft inductance components; ψ f Represents the permanent magnet flux.

[0156] The electromagnetic torque equation is:

[0157] Te =nP n [ψ d i q -ψ q i d ]P n i q [ψ f +(L d -L q )i d ] (12)

[0158] The mechanical motion equation is:

[0159]

[0160] In the formula, J is the moment of inertia, B is the friction coefficient, T e is the electromagnetic torque, T L Indicates the load torque of the motor.

[0161] 2 Neural Network Sliding Mode Adaptive Control

[0162] 2.1 Radial Basis Function (RBF) Neural Network Design

[0163] The purpose of introducing the RBF neural network algorithm into the control system in the design of the present invention is to estimate the uncertain part of the marine motor dynamics model.

[0164] The actual output of the RBF neural network estimated marine motor model is fNN∈n, and the ideal output is f*NN∈n, as shown in the following formula:

[0165]

[0166] f NN =w T δ(x) (15)

[0167] Where: f is the actual output of the neural network; w is the optimal weight of the neural network * is an estimate of ; δ is the basis function of the neural network; L is the number of nodes in the hidden layer of the neural network.

[0168] The Gaussian basis function expression is:

[0169]

[0170] i=1,2,…,n;j=1,2,…,L

[0171]

[0172] c i =[c i1 ,c i2 ,…,ciL ] T (18)

[0173] b i =[b i1 ,b i2 ,…,b iL ] (19)

[0174] Where: x i is the input vector of the i-th neural network; c i is the central matrix of the ith neural network; b i is the basis width vector of the ith neural network.

[0175] Unknown nonlinear functions of motor control systems:

[0176]

[0177] Where: f1 is the unknown nonlinear function; M(q) is the positive definite inertia matrix; is the motor angular velocity; is the angular acceleration of the motor; are centrifugal force and Coriolis force; G(q) is the gravity term vector; τ d To interfere; is the expected trajectory angular acceleration.

[0178] Sliding mode control requires high accuracy of mathematical models. Formula (20) contains modeling errors and interference from unknown factors. RBF approximation of nonlinear functions requires low accuracy of mathematical models and can control motors without accurate models.

[0179] Neural networks have approximation errors. When the number of nodes in the middle layer is small, it will affect the performance of the control system. In order to make fNN infinitely close to f*NN, the adaptive compensation update law and the design weight update law are shown as follows:

[0180]

[0181] Where: δ is the basis function of the neural network.

[0182] 3. Control System Analysis and Design

[0183] RBF neural network meets the real-time requirements of the control system. It is difficult to establish an accurate mathematical model, and the neural network has an approximation error. The robustness of the system is improved by designing a forward stabilization channel. The present invention designs a neural network backstepping sliding mode controller (NNBSMC) based on RBF neural network, which enables the motor control system to have good static and dynamic performance. The controller principle is as follows Figure 1 shown.

[0184] 3.1 Inversion decoupling control

[0185] Assume that the expected attitude signal is x d =[x dx x dy x dz ] T , define the tracking angle error

[0186]

[0187] Where: z1 is the tracking angle error; x1 is the inertial angle of the system; x2 is the inertial angular velocity of the system; x f is the expected attitude signal.

[0188] The Lyapunov function is designed using the backstepping control method:

[0189]

[0190] Where: L1 is the designed Lyapunov function; z1 is the defined tracking angle error.

[0191] Differentiating L1 yields:

[0192]

[0193] Where: L1 is the designed Lyapunov function; z1 is the defined tracking angle error; x2 is the inertial angular velocity of the system; x d is the expected attitude signal.

[0194] Pick z2 is the control amount of the virtual control item, that is

[0195]

[0196] Where: z2 is the control quantity of the virtual control item; x2 is the inertial angular velocity; x d is the expected attitude signal; z1 is the defined tracking angle error; C is a 3×3 positive definite diagonal matrix.

[0197] If z2=2, then Therefore, the system needs to be further designed. Design the Lyapunov function:

[0198]

[0199] Differentiating L2 gives

[0200]

[0201] Design system control law so that

[0202] Where: A and B are system parameters respectively; C is a 3×3 positive definite diagonal matrix; T M is the system control rate; δ is the basis function of the neural network; z1 is the tracking angle error; z2 is the control amount of the virtual control item; d t is the unknown external disturbance; x2 is the inertial angular velocity of the system; x d is the expected attitude signal.

[0203]

[0204] Where: A and B are system parameters respectively; C is a 3×3 positive definite diagonal matrix; T M is the system control rate; δ is the basis function of the neural network; z1 is the tracking angle error; z2 is the control amount of the virtual control item; d t is the unknown external disturbance; x2 is the inertial angular velocity of the system; x d is the expected attitude signal; C ′ is a 3×3 positive definite diagonal matrix.

[0205] Substituting the control rate (30) into equation (29), we can get

[0206]

[0207] Where: C is a 3×3 positive definite diagonal matrix; C ′ is a 3×3 positive definite diagonal matrix; z1 is the definition of the tracking angle error; z2 is the control amount of the virtual control item.

[0208] When t→∞, z1→0 and z2→0. According to equation (23), x1→x d and

[0209] 3.2 Anti-interference control based on sliding mode

[0210] Single backstepping control is easily affected by external interference and has poor robustness. In order to improve the robustness, the sliding surface is introduced. The sliding surface is defined as:

[0211] σ=Kz1+z2 (32)

[0212] Where: σ is the sliding surface; K is a 3×3 positive definite diagonal matrix; z1 is the tracking angle error; z2 is the control amount of the virtual control item.

[0213] Substituting equation (24) into equation (32), we can obtain

[0214]

[0215] Where: σ is the sliding surface; K and C are 3×3 positive definite diagonal matrices respectively; z1 is the tracking angle error.

[0216] if but Redesign the Lyapunov function of the second step of inversion control:

[0217]

[0218] Where: L1, L2 are Lyapunov functions; σ is the defined sliding surface.

[0219] Differentiating L2 gives

[0220]

[0221] Differentiate σ and substitute equation (11) into the equation:

[0222]

[0223] To make Based on formula (23), the control rate is updated as

[0224]

[0225] Where: ф, ψ are 3×3 diagonal matrices. A and B are system parameters respectively; C is a 3×3 positive definite diagonal matrix; T M is the system control rate; δ is the basis function of the neural network; z1 is the tracking angle error; z2 is the control amount of the virtual control item; d t is the unknown external disturbance; x2 is the inertial angular velocity of the system; x d is the expected attitude signal.

[0226] 3.3 Nonlinear friction and system coupling estimator

[0227] The anti-interference control based on sliding mode has both advantages and disadvantages. It has good dynamic performance, but the system output has jitter. The friction force in the control law (24) is unknown. Since the subsystems interfere with each other, the control system has coupling, and it is difficult to build an accurate model. Because the angular velocity and friction force are not linearly related, there will be jitter in the system output. Eliminating jitter can improve system performance. Neural networks have strong environmental adaptability and good nonlinear approximation performance. The disturbance estimator uses a neural network, which is combined with a sliding mode anti-interference controller to achieve the purpose of reducing system jitter.

[0228] RBF network is one of the local approximation networks. It uses Gaussian function as the action function. It has fast learning speed, good real-time performance, good convergence effect, and is not easy to fall into local optimality.

[0229] This design uses the RBF network structure to establish an adaptive law to approximate the nonlinear friction force δ. The RBF network algorithm is:

[0230]

[0231] Where: δ is the nonlinear friction force; j = 1, 2, ... ... 7, x is the network input; is the ideal weight of the network; h(x) 7×1 is the network Gaussian basis function output; ε 3×1 is the network approximation error; c j is the coordinate of the center point of the jth neuron basis function in the hidden layer of the network, b j is the propagation range of the jth neuron basis function in the hidden layer of the network.

[0232] The network input is the motor angular velocity x = [ωp x ,ωp y ,ωp z ] T , then the network output is

[0233]

[0234] Where: Output for the network; is the ideal weight of the network; h(x) 7×1 is the network Gaussian basis function output.

[0235] Pick The error of the estimator is

[0236]

[0237] Where: is the ideal weight of the network; h(x) is the output of the network Gaussian basis function; ε is the network approximation error.

[0238] The RBFNN update law is

[0239]

[0240] Where: B 3×3 , G 3×3 are positive definite diagonal matrices; h(x) 7×1 is the network Gaussian basis function output.

[0241] After applying the disturbance estimator, the control law (36) is updated to

[0242]

[0243] Where: ф, ψ are 3×3 diagonal matrices; A and B are system parameters; C is a 3×3 positive definite diagonal matrix; T M is the system control rate; σ is the sliding surface; z1 is the tracking angle error; z2 is the control amount of the virtual control item; d t is the unknown external disturbance; x2 is the inertial angular velocity of the system; x d is the expected attitude signal.

[0244] This design combines the nonlinear approximation ability and good real-time performance of the RBF neural network with the good dynamic performance and strong anti-interference ability of the sliding mode control, and makes full use of the backstepping decoupling characteristics of the backstepping method to design a neural network backstepping sliding mode control method for marine three-phase asynchronous motors to achieve the purpose of improving the performance of the motor control system.

[0245] In order to solve the problem of parameter determination in the process of modeling marine three-phase asynchronous motors, a forward stabilization channel is designed. The backstepping method is used to decouple the system, and the sliding mode control solves the disturbance problem of the control system. The RBF neural network is used to estimate and compensate the disturbance between the system coupling and nonlinear friction of the contact surface between the control system frames in real time.

[0246] In summary, the above are only preferred embodiments of the present invention, and the protection scope of the present invention is not limited thereto. All equivalent changes and modifications made according to the patent scope of the present invention and the contents of the specification are within the scope covered by the patent of the present invention.

Claims

1. A neural network backstepping sliding mode control method for a marine three-phase asynchronous motor, characterized in that: The following steps are involved: Establish mathematical models of three-phase asynchronous motors, including mathematical models in a stationary three-phase coordinate system, mathematical models in a two-phase stationary coordinate system, and mathematical models in a two-phase synchronous rotating coordinate system; Neural network sliding mode adaptive control, introducing RBF neural network algorithm to estimate the uncertain part of the marine motor dynamics model; Control system analysis and design, used to make the motor control system have good static and dynamic performance, including backstepping decoupling control, sliding mode-based anti-disturbance control, nonlinear friction and system coupling estimator.

2. The method according to claim 1, characterized in that The mathematical model in the stationary three-phase coordinate system includes the following features: The mathematical model of a typical motor includes the following equations: Stator voltage equation In the formula, u a 、u b 、u c is the three-phase winding voltage, i a 、i b 、i c is the three-phase winding current, R s is the phase winding resistance, ψ a , b , c It is the full flux linkage of three-phase stator; Stator flux equation In the formula, ψ f is the permanent magnet flux, L a , L b , L c is the self-inductance of the stator three-phase winding, L ab , L ba , L ac , L ca , L bc , L cb is the mutual inductance of the three-phase stator winding, and L ab =L ba , L ac =L ca , L cb =L cb .

3. The method according to claim 2, characterized in that The mathematical model in the two-phase stationary coordinate system includes the following features: According to the principle of equal phase amplitude transformation, the Clarke transformation expression is as follows: i a +i b +i b =0, and we can get: The corresponding inverse Clarke transform is: The voltage equation is: The magnetic flux equation is: Where U α , U β is the αβ axis voltage, i α 、i β is the α and β axis current, Ψ α , β is the voltage of α and β axes, L is the voltage of α and β axes, ω e is the electrical angular velocity of the rotor.

4. The method according to claim 3, characterized in that The mathematical model in the two-phase synchronous rotating coordinate system includes: The Park transformation expression is as follows: The Park inverse transform expression is as follows: Where P 2s / 2r is the Park transformation matrix, P 2r / 2s is the Park inverse transformation matrix; Stator voltage equation: Stator flux equation: ψ d =L d I d +ψ f ψ d =L q I q where u d 、u q are the dq axis components of the stator respectively; i d 、i q are the dq axis components of the stator current respectively; Rs is the stator winding resistance; ψ d , q is the axial component of the stator flux; ω r is the rotor electrical angular velocity; L d , L q They are the shaft inductance components; ψ f represents the permanent magnet flux; The electromagnetic torque equation is: T e =nP n [ψ d i q -ψ q i d ]P n i q [ψ f +(L d -L q )i d ] The mechanical motion equation is: In the formula, J is the moment of inertia, B is the friction coefficient, T e is the electromagnetic torque, T L Indicates the load torque of the motor.

5. The method according to claim 4, characterized in that The sliding mode adaptive control of neural network includes radial basis neural network design: The actual output of the radial basis neural network estimated marine motor model is fNN∈n, and the ideal output is f*NN∈n, as shown in the following formula: f NN =w T δ(x) Where: f is the actual output of the neural network; w is the optimal weight of the neural network * ;δ is the basis function of the neural network;L is the number of nodes in the hidden layer of the neural network; The Gaussian basis function expression is: i=1,2,…,n;j=1,2,…,L c i =[c i1, c i2 ,…,c iL ] T b i =[b i1 ,b i2 ,…,b iL ] Where: x i is the input vector of the i-th neural network; c i is the central matrix of the ith neural network; b i is the basis width vector of the i-th neural network; Unknown nonlinear functions of motor control systems: Where: f1 is the unknown nonlinear function; M(q) is the positive definite inertia matrix; is the motor angular velocity; is the angular acceleration of the motor; are centrifugal force and Coriolis force; G(q) is the gravity term vector; τ d To interfere; is the expected trajectory angular acceleration.

6. The method according to claim 5, characterized in that: Sliding mode control requires high accuracy of mathematical models. The unknown nonlinear functions of the motor control system contain modeling errors and interference from unknown factors. RBF approximation of nonlinear functions requires low accuracy of mathematical models and can control motors without accurate models. Neural networks have approximation errors. When the number of nodes in the middle layer is small, it will affect the performance of the control system. In order to make fNN infinitely close to f*NN, the adaptive compensation update law and the design weight update law are as follows: Where: δ is the basis function of the neural network.

7. The method according to claim 6, characterized in that Backstepping decoupling control includes: Assume that the expected attitude signal is x d =[x dx x dy x dz ] T , define the tracking angle error: Where: z1 is the tracking angle error; x1 is the inertial angle of the system; x2 is the inertial angular velocity of the system; x d is the expected posture signal; The Lyapunov function is designed using the backstepping control method: Where: L1 is the designed Lyapunov function; z1 is the defined tracking angle error; Differentiating L1 yields: Where: L1 is the designed Lyapunov function; z1 is the defined tracking angle error; x2 is the inertial angular velocity of the system; x d is the expected posture signal; Pick z2 is the control amount of the virtual control item, that is: Differentiating L2 gives: Design system control law so that Where: A and B are system parameters respectively; C is a 3×3 positive definite diagonal matrix; T M is the system control rate; δ is the basis function of the neural network; z1 is the tracking angle error; z2 is the control amount of the virtual control item; d t is the unknown external disturbance; x2 is the inertial angular velocity of the system; x d is the expected posture signal; Where: A and B are system parameters respectively; C is a 3×3 positive definite diagonal matrix; T M is the system control rate; δ is the basis function of the neural network; z1 is the tracking angle error; z2 is the control amount of the virtual control item; d t is the unknown external disturbance; x2 is the inertial angular velocity of the system; x d is the expected attitude signal; C ′ is a 3×3 positive definite diagonal matrix; The control rate T M Substitution We can get: Where: C is a 3×3 positive definite diagonal matrix; C ′ is a 3×3 positive definite diagonal matrix; z1 is the definition of the tracking angle error; z2 is the control amount of the virtual control item. When t→∞, z1→0 and z2→0, according to the tracking angle error, we have: x1→x d and 8. The method according to claim 7, characterized in that The anti-disturbance control based on sliding mode includes: The sliding surface is defined as: σ=Kz1+z2 Where: σ is the sliding surface; K is a 3×3 positive definite diagonal matrix; z1 is the tracking angle error; z2 is the control amount of the virtual control item; Substituting the Lyapunov function into the sliding surface formula, we can get: Where: σ is the sliding surface; K and C are 3×3 positive definite diagonal matrices; z1 is the tracking angle error; if but Redesign the Lyapunov function of backstepping control: Where: L1, L2 are Lyapunov functions; σ is the sliding surface defined; Differentiating L2 gives: Differentiate σ and substitute the stator flux equation into it to obtain: To make Based on the tracking angle error, the control rate is updated as: Where: ф, ψ are 3×3 diagonal matrices. A and B are system parameters respectively; C is a 3×3 positive definite diagonal matrix; T M is the system control rate; δ is the basis function of the neural network; z1 is the tracking angle error; z2 is the control amount of the virtual control item; d t is the unknown external disturbance; x2 is the inertial angular velocity of the system; x d is the expected attitude signal.

9. The method according to claim 8, characterized in that The nonlinear friction and system coupling estimator specifically includes: using the RBF network structure to establish an adaptive law to approximate the nonlinear friction force δ, and the RBF network algorithm is: Where: δ is the nonlinear friction force; j = 1, 2, ·······7, x is the network input; is the ideal weight of the network; h(x) 7×1 is the network Gaussian basis function output; ε 3×1 is the network approximation error; c j is the coordinate of the center point of the jth neuron basis function in the hidden layer of the network, b j is the propagation range of the jth neuron basis function in the network hidden layer; The network input is the motor angular velocity x = [ωp x ,ωp y ,ωp z ] T , then the network output is: Where: Output for the network; is the ideal weight of the network; h(x) 7×1 is the network Gaussian basis function output; Pick Then the error of the estimator is: Where: is the ideal weight of the network; h(x) is the output of the network Gaussian basis function; ε is the network approximation error; The RBFNN update law is: Where: B 3×3 , G 3×3 are positive definite diagonal matrices; h(x) 7×1 is the network Gaussian basis function output; After applying the disturbance estimator, the control tracking angle error is updated as: Where: ф, ψ are 3×3 diagonal matrices; A and B are system parameters; C is a 3×3 positive definite diagonal matrix; T M is the system control rate; σ is the sliding surface; z1 is the tracking angle error; z2 is the control amount of the virtual control item; d t is the unknown external disturbance; x2 is the inertial angular velocity of the system; x d is the expected attitude signal.

Images