Permanent magnet motor robust predictive position control method with low observation noise characteristic

Abstract

The application belongs to the technical field of permanent magnet motor, and relates to a robust prediction position control method for a permanent magnet motor with low observation noise characteristics. The method comprises the following steps: S1, constructing a surface-mounted permanent magnet motor position control system total disturbance mathematical model considering parameter and load disturbance, and on this basis, constructing a disturbance considered position control system prediction model and a linear constraint condition; S2, constructing a low noise characteristic observer to observe disturbance and disturbance differential information; S3, disturbance observation based prediction model and position control system constraint compensation; S4, constructing a robust position control system convex optimization model containing constraints; and S5, solving the constraint optimization model based on the Hildreth method. The method is simple to implement, can enhance the robustness of the prediction position control strategy under the action of parameter and load disturbance, and can improve the steady positioning accuracy of the prediction position control strategy under the action of parameter disturbance and load disturbance compared with an algorithm model using motor nominal parameters.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of permanent magnet motors, and particularly relates to a robust predictive position control method for a permanent magnet motor with low observation noise characteristics. BACKGROUND

[0002] Position servo control is a typical application form of permanent magnet motors in high-end equipment manufacturing scenes such as numerical control machine tools, mechanical arms and simulation turntables, aiming to realize fast position following or high-precision positioning. The traditional control strategy is a cascade three-closed-loop structure control scheme based on field-oriented control, which can realize independent control of current, speed and position decoupling. However, the cascade multi-loop control requires the inner loop bandwidth to be much faster than the outer loop, thus limiting the dynamic response of the position servo control; the linear control scheme relying on frequency domain analysis is difficult to consider the influence of dead zone, nonlinearity and other factors on the control system, and the mismatch of electrical parameters and mechanical parameters will lead to deterioration of control performance. The position servo system is a typical high-order multivariable system and also a multi-constrained control system, and model predictive control has good adaptability to multivariable control and system constraints, so it has good application prospect in position servo control and can be used to realize integrated control of position, speed and current, eliminate control loops and improve position dynamic response, that is, predictive position control strategy. The current research focuses on the application and implementation of model predictive control strategy in position servo control. However, model predictive control relies on an accurate mathematical model, so it has some shortcomings in parameter robustness. At present, there is no research on the robustness improvement scheme for predictive position control.

[0003] Currently, there are two technical routes to improve the robustness of the model predictive control strategy: one is a data-based scheme, and typical ones include data-driven and neural network-based schemes. The control process can be independent of the model, but this kind of scheme relies on historical data or training data, and the implementation complexity is high and the universality is poor; the other kind of scheme is a model-based scheme, which improves the control robustness by improving the accuracy of the model. The scheme based on various observers is a typical method to improve the accuracy of the model, which is simple and easy to implement, but this scheme faces the contradiction between observation bandwidth and observation noise. In addition, the current robustness research on predictive control generally does not consider the robustness of the constraint condition, so it is not easy to improve the influence of parameter disturbance on the position dynamic process. SUMMARY

[0004] The purpose of the present application is to solve the parameter robustness problem of model predictive control in position servo application, and fully consider the parameter disturbance of the position control system constraint condition, and at the same time improve the noise problem of the observer applied to the model predictive control, and then propose a robust predictive position control method for a permanent magnet motor with low observation noise characteristics.

[0005] The low observation noise refers to a better noise suppression characteristic under the same observer bandwidth.

[0006] To achieve the above object, the technical scheme adopted by the present application is:

[0007] The method comprises the following steps:

[0008] S1: constructing a surface-mounted permanent magnet motor position control system total disturbance mathematical model considering parameter and load disturbance, and on this basis, constructing a disturbance considering position control system prediction model and position system linear constraint condition;

[0009] S2: constructing a low noise characteristic observer to observe disturbance and disturbance differential information;

[0010] S3: disturbance observation based prediction model and position control system constraint compensation;

[0011] S4: constructing a robust position system constraint convex optimization model;

[0012] S5: solving the constraint optimization model based on the Hildreth method.

[0013] Further, in step S1, the surface-mounted permanent magnet motor position system total disturbance mathematical model considering parameter and load disturbance is constructed, and on this basis, the disturbance considering position control system prediction model and position control system linear constraint condition are constructed; comprising the following steps:

[0014] S11: constructing a surface-mounted permanent magnet motor position control system total disturbance mathematical model considering parameter and load disturbance; as shown in the following formula (2):

[0015]

[0016] Where, i d ,i q ,u d ,u q are d-axis current, q-axis current, d-axis voltage and q-axis voltage, L s is motor inductance, λ f is flux, θ is motor electrical angle, ω e is electrical angular velocity, J and p are respectively rotational inertia and pole pair number, disturbances f d ,f q ,f ω are respectively d-axis current total disturbance, q-axis current total disturbance and speed total disturbance caused by parameter change, and the disturbance at time k is represented as f d,k ,f q,k ,f ω,k, whose expression is:

[0017]

[0018] wherein: R s , B are motor resistance and friction coefficient respectively, ΔR s , ΔL s , Δλ f , ΔJ and ΔB are parameter variation of corresponding variable; i d,k , i q,k , u d,k , u q,k are d-axis current, q-axis current, d-axis voltage and q-axis voltage at k moment respectively; ω e,k , θ k , T l are electrical angular velocity at k moment, electrical angle at k moment and load torque respectively;

[0019] S12: constructing a position control system prediction model considering disturbance;

[0020] Discretize formula (2) by using forward Euler discrete method shown in formula (4) as follows:

[0021]

[0022] wherein: x represents state variable, T s is discrete time step, and k represents k moment;

[0023] The position control system prediction model is expressed as:

[0024]

[0025] wherein: x k , y k are state variable matrix and output variable matrix, A m , B m , C m are position control system matrix, ε k is disturbance matrix, u k is control voltage at k moment, and state variable with subscript k represents state value at k moment, and the definition of related variable is as follows:

[0026]

[0027] S13: constructing a position control system linear constraint condition considering parameter disturbance;

[0028] The position control system needs to consider dq-axis current constraint I dmax , I qmax , and speed constraint ω maxand dq-axis voltage constraint u dmax , qmax , is expressed as:

[0029]

[0030] The truncated Taylor expansion method shown in equation (7) is used to discretize the current and speed system in equation (2) with order N j = 1 and N j = 2, respectively:

[0031]

[0032] In combination with equation (6), the current and speed constraints are derived as follows:

[0033]

[0034] Speed constraint

[0035] In the formula: is the first-order differential of the speed disturbance;

[0036] The voltage constraint uses a rectangular approximation scheme:

[0037]

[0038] In the formula: ε is the rectangular coefficient, 0≤ε≤1, V dc is the bus voltage.

[0039] Further, in step S2, the low-noise characteristic observer is constructed to observe the disturbance and the differential information of the disturbance; the specific construction process is:

[0040] The current and speed dynamic model in equation (2) is regarded as a disturbance model as shown below:

[0041]

[0042] In the formula: x1 is a state variable, representing i d ,i q ,ω e in equation (2); x2 is a state variable, representing the total disturbance f d ,f q ,f ω in equation (2); u is a control input, representing u d ,u q ,i q in equation (2); b0 is the gain before input, f (1) is the first-order differential of the disturbance;

[0043] Based on the disturbance model (11), the low-noise observer is designed as follows:

[0044]

[0045] In the formula, l1, l2 are observer gains, m3 is a gain coefficient for adjusting the disturbance and the disturbance observation information observation effect, τ is a first-order low-pass filter filtering coefficient for adjusting the noise suppression performance; x e is the filtered value of the input error .

[0046] Further, in step S3, the disturbance observation-based prediction model and position control system constraint compensation are performed; specifically:

[0047] The designed observer (12) is used to apply to the d-axis current equation, q-axis current equation and speed dynamic equation in formula (2) respectively.

[0048] The corresponding disturbance observer is designed as follows:

[0049]

[0050] In the formula, the superscript ^ represents the estimated value of the corresponding variable, x ed , x eq , x eω are the filtered values of the errors between the actual values and the observed values of the d-axis current, q-axis current and speed respectively, τ d , τ q , τ ω are the filtering coefficients of the corresponding variables, are the first-order derivatives of the total disturbance of the d-axis current, q-axis current and speed respectively, m d3 , m q3 , m ω3 are disturbance observation value adjustment gains, l d1 , l d2 , l q1 , l q2 , l ω1 , l ω2 are observer gains, whose values are:

[0051]

[0052] In the formula, ω d , ω q and ω s are the d-axis current observer bandwidth, q-axis current observer bandwidth and speed observer bandwidth respectively;

[0053] The obtained disturbance information f d , f q , fω and perturbation differential information Observed values It is used to compensate the prediction model (5) and the position control system constraints (8) and (9).

[0054] Furthermore, in step S4, the robust position control system includes a constrained convex optimization model; the specific construction process is:

[0055] Consider the prediction time domain N p When , the value function of position control is designed as follows:

[0056]

[0057] stx j+1 =A m x j +B m u j +ε j

[0058] Where: y r,j+1 where j=k,k+1,…,k+N p -1,y r,j+1 is the target reference value, that is: Q and R are the corresponding weight coefficient matrices, respectively: k id ,k θ ,k u are the d-axis current weight coefficient, position weight coefficient and control weight coefficient respectively; in order to smooth the control input, a smoothing factor a∈(0,1) is introduced, then the target reference value at different times Expressed as:

[0059]

[0060] Where: i dref,j+1 ,θ ref,j+1 where j=k,k+1,…,k+N p -1,i dref,j+1 ,θ ref,j+1 is the given value of d-axis current and position at different times. For the sake of simplicity, in the whole prediction time domain N p Take the same value within;

[0061] definition: For the sake of simplicity, it is assumed that: Take time k as the initial state and output y at different times j+1 =C m x j+1 , j=k,k+1,…,k+N p-1, determined using the following formula:

[0062] Y = Yx k + Γu + Ωε k (15)

[0063] wherein:

[0064]

[0065] wherein I is the identity matrix.

[0066] Based on the formula (8), (9), (10) and (13), a robust position control system constraint convex optimization model is constructed, the solution of the model is regarded as a linear quadratic programming problem, in order to simplify the solution, only the constraint at time k is considered, and a constraint convex optimization model is obtained, as shown in the following formula:

[0067]

[0068] In the formula: and is the disturbance estimation value at time k.

[0069] Further, in step S5, the constraint optimization model based on the Hildreth method is solved; specifically:

[0070] Based on the Hildreth method, the optimization model shown in the formula (16) is solved, first, without considering the constraints of the position control system, the unconstrained solution u of the position control system is n Solved according to the following formula:

[0071]

[0072] The unconstrained solution u n is expressed as:

[0073]

[0074] In the formula: The symbol represents the Kronecker product, Y r is the reference value matrix, expressed as: Definition and F = -Γ T Θ (Y r - Yx k - Ωε k ), the constraint optimal solution u c (k) is expressed as:

[0075] u c = -E -1 (F + M T λ un) (19)

[0076] Where: un is the Lagrange multiplier;

[0077] Based on formula (19), the optimal solution with constraints at time k is used c,k Applied to position control systems to achieve robust predictive position control,

[0078]

[0079] Where: c,k is the control voltage when considering the constraints of the position control system, u cd,k for u c,k The d-axis component, u cq,k for u c,k The q-axis component of .

[0080] The present invention offers advantages over existing technologies in that it provides a low-observation-noise disturbance compensation scheme for a predictive position control strategy, offering improved high-frequency noise suppression at the same bandwidth. A unique feature of the present invention is the implementation of a robust predictive position control scheme. The compensation model simultaneously considers the predictive model and constraints of the position control system, resulting in a moderately effective compensation effect on both dynamic and steady-state performance. The integrated solution eliminates the series loop, allowing the motor position, speed, and current to be controlled by a single controller, resulting in improved dynamic performance.

[0081] The method of the present invention is simple to implement and can enhance the robustness of the predicted position control strategy under parameter and load disturbances. Compared with an algorithm model that uses the nominal motor parameters (which does not consider parameters when parameters change), it can improve the steady-state positioning accuracy of the predicted position control strategy under parameter and load disturbances. The method of the present invention is of great significance for the practical application of the predicted position control strategy in the field of high-end equipment manufacturing. BRIEF DESCRIPTION OF THE DRAWINGS

[0082] Figure 1 This is an overall control block diagram of the robust predictive position control method for a permanent magnet motor with low observation noise characteristics of the present invention;

[0083] Figure 2 This is a schematic diagram of the low-noise observer structure. DETAILED DESCRIPTION

[0084] Specific embodiment 1: This embodiment discloses a method for robust predictive position control of a permanent magnet motor with low observation noise characteristics, the method comprising the following steps:

[0085] S1: Construct a surface-mounted permanent magnet motor position control system total disturbance mathematical model considering parameter and load disturbance, and on this basis, construct a position control system prediction model considering disturbance and position control system linear constraint condition;

[0086] S2: Construct a low-noise characteristic observer (improve the noise suppression performance of the observer), observe the disturbance and disturbance differential information (observe the speed and current dynamic equation disturbance and disturbance differential term);

[0087] S3: Disturbance observation-based prediction model and position control system constraint compensation (based on real-time observed disturbance, compensate the position control system prediction model, and modify the position control system constraint condition using disturbance and disturbance differential information);

[0088] S4: Construct a robust position control system constraint convex optimization model (use the modified prediction model and constraint condition to construct a position control system constraint convex optimization model, including the value function and constraint condition of integrated control);

[0089] S5: Hildreth method-based constraint optimization model solution (use the Hildreth method to solve the modified constraint convex optimization model online, and output the voltage for the position control system).

[0090] Further, in step S1, the surface-mounted permanent magnet motor position control system total disturbance mathematical model considering parameter and load disturbance is constructed, and on this basis, the position control system prediction model considering disturbance and the position control system linear constraint condition are constructed; including the following steps:

[0091] S11: Construct a surface-mounted permanent magnet motor position control system total disturbance mathematical model considering parameter and load disturbance; as shown in the following formula (2):

[0092]

[0093] Where, i d ,i q ,u d ,u q are the d-axis current, q-axis current, d-axis voltage and q-axis voltage, L s is the motor inductance, λ f is the flux, θ is the motor electrical angle, ω e is the electrical angular velocity, J and p are the moment of inertia and the number of pole pairs respectively, and the disturbances f d ,f q ,f ω are the total disturbances of the d-axis current, q-axis current and speed caused by parameter changes, respectively, and the disturbance at time k is represented as f d,k ,fq,k ,f ω,k , whose expression is:

[0094]

[0095] wherein: R s , B are motor resistance and friction coefficient respectively, ΔR s , ΔL s , Δλ f , ΔJ and ΔB are parameter variation amounts of corresponding variables; i d,k , i q,k , u d,k , u q,k are d-axis current, q-axis current, d-axis voltage and q-axis voltage at k moment respectively; ω e,k , θ k , T l are electrical angular velocity at k moment, electrical angle at k moment and load torque respectively;

[0096] S12: constructing a position control system prediction model considering disturbance;

[0097] Discretize formula (2) by using forward Euler discretization method shown in formula (4) as follows:

[0098]

[0099] wherein: x represents state variable, T s is discrete time step, and k represents k moment;

[0100] The position control system prediction model is expressed as:

[0101]

[0102] wherein: x k , y k are state variable matrix and output variable matrix, A m , B m , C m are position control system matrix, ε k is disturbance matrix, u k is control voltage at k moment, and state variable with subscript k represents state value at k moment, and the definition of related variables is as follows:

[0103]

[0104] S13: constructing a position control system linear constraint condition considering parameter disturbance;

[0105] The position control system needs to consider dq-axis current constraint I dmax , I qmax, speed constraint ω max and dq-axis voltage constraint u dmax ,u qmax , is expressed as:

[0106]

[0107] The truncated Taylor expansion method shown in equation (7) is used to discretize the current and speed system in equation (2) with order N j = 1 and N j = 2, respectively:

[0108]

[0109] In combination with equation (6), the current and speed constraints are derived as follows:

[0110]

[0111] where: is the first-order differential of the speed disturbance;

[0112] The voltage constraint uses a rectangular approximation scheme:

[0113]

[0114] where ε is the rectangular coefficient, 0 ≤ ε ≤ 1, and V dc is the bus voltage.

[0115] Further, in step S2, the low-noise characteristic observer is constructed to observe the disturbance and the differential information of the disturbance; the specific construction process is:

[0116] The current and speed dynamic model in equation (2) is regarded as a disturbance model as shown below:

[0117]

[0118] where x1 is a state variable representing i d ,i q ,ω e in equation (2); x2 is a state variable representing the total disturbance f d ,f q ,f ω in equation (2); u is a control input representing u d ,u q ,i q in equation (2); b0 is the gain before input, and f (1) is the first-order differential of the disturbance;

[0119] Based on the disturbance model equation (11), the low-noise observer is designed as follows:

[0120]

[0121] Where: l1, l2 are observer gains, m3 is the gain coefficient, which is used to adjust the observation effect of disturbance and disturbance observation information, τ is the first-order low-pass filter coefficient, which is used to adjust the noise suppression performance; x e is the input error The filter value of .

[0122] Furthermore, in step S3, the prediction model based on disturbance observation and position control system constraint compensation are specifically:

[0123] The designed observer formula (12) is used to apply the d-axis current equation, q-axis current equation, and speed dynamic equation in formula (2) respectively;

[0124] The corresponding disturbance observers are designed as follows:

[0125]

[0126] Where: superscript ^ represents the estimated value of the corresponding variable, x ed 、x eq 、x eω are the error filtering values ​​between the actual and observed values ​​of d-axis current, q-axis current and speed, τ d , τ q , τ ω is the filter coefficient of the corresponding variable, are the first-order differentials of the total disturbance of d-axis current, q-axis current and speed, respectively, m d3 、m q3 、m ω3 Adjust the gain for the disturbed observations, l d1 、l d2 、l q1 、l q2 、l ω1 、l ω2 are the observer gains respectively, and their values ​​are:

[0127]

[0128] Where: ω d 、ω q and ω s They are the d-axis current observer bandwidth, q-axis current observer bandwidth and speed observer bandwidth respectively;

[0129] Using the obtained disturbance information f d ,f q ,f ω and perturbation differential information Observed values to compensate the prediction model (5), and the position control system constraint conditions (8) and (9).

[0130] Further, in step S4, the robust position control system is constructed with a constraint convex optimization model, and the specific construction process is as follows:

[0131] Considering the prediction horizon N p The value function of the position control is designed as follows:

[0132]

[0133] wherein y r,j+1 , j = k, k + 1, …, k + N p -1, y r,j+1 is the target reference value, i.e.: Q and R are the corresponding weight coefficient matrices, respectively: k id , k θ , k u are the d-axis current weight coefficient, position weight coefficient and control weight coefficient respectively; in order to smooth the control input, a smooth factor a ∈ (0, 1) is introduced, then the target reference value at different time is expressed as:

[0134]

[0135] wherein i dref,j+1 , θ ref,j+1 , j = k, k + 1, …, k + N p -1, i dref,j+1 , θ ref,j+1 are the d-axis current and position given values at different time, in order to calculate conveniently, the same values are taken in the whole prediction horizon N p ;

[0136] Definition: In order to calculate simply, it is assumed that: Taking k time as the initial state, the output y j+1 = C m x j+1 , j = k, k + 1, …, k + N p -1, is determined by the following formula:

[0137] Y = Yx k + Γu + Ωε k (15)

[0138] Wherein:

[0139] ​

[0140] where I is an identity matrix.

[0141] Based on the equations (8), (9), (10) and (13), a robust position control system with constraint convex optimization model is constructed, and the solution of the model is regarded as a linear quadratic programming problem. In order to simplify the solution, only the constraint at time k is considered, and a convex optimization model with constraint is obtained as follows:

[0142]

[0143]

[0144] In the equation, Y and are the disturbance estimation values at time k.

[0145] Further, in step S5, the constraint optimization model based on the Hildreth method is solved; specifically:

[0146] Based on the Hildreth method, the optimization model shown in equation (16) is solved. First, without considering the constraints of the position control system, the unconstrained solution υ n is solved according to the following equation:

[0147]

[0148] The unconstrained solution υ n is expressed as:

[0149]

[0150] In the equation, Y is a symbol represents the Kronecker product, Y r is a reference value matrix, and is expressed as: is defined as and F = -Γ T Θ(Y r -Υx k -Ωε k ), then the constrained optimal solution υ c (k) is expressed as:

[0151] υ c = -E -1 (F+M T λ un )(19)

[0152] In the equation, λ un is a Lagrange multiplier;

[0153] Based on equation (19), the constrained optimal solution υ c,kThe application is applied to a position control system to realize robust predictive position control,

[0154]

[0155] wherein: u c,k is the control voltage considering the constraints of the position control system, cd,k is the d-axis component of u c,k , and u cq,k is the q-axis component of u c,k .

[0156] Figure 1 Description: Figure 1 The overall control block diagram proposed is an integrated current, speed and position control structure, which cancels the multi-loop control structure in the traditional control scheme. The current and speed measurement information is used as the input of the predictive model, the position control system constraints and the disturbance observer. The low-noise disturbance observer is used to observe the model disturbance information caused by parameter mismatch and load changes The observed disturbance is used to compensate and correct the predictive model and the constraint conditions of the position control system. The corrected predictive model and the constraint conditions constitute a convex optimization model with constraints. The optimal solution u c,k is obtained by solving the model via the Hildreth method, and then the space vector pulse width modulation algorithm (SVPWM) is used to modulate the output to the inverter, which ensures that the positioning performance is still good when the parameter mismatch or load disturbance acts.

[0157] Figure 2 Description: Figure 2 The low-noise observer structure designed for the predictive position control strategy is suitable for first-order dynamic models such as current and speed, i.e.,

[0158]

[0159] The structure of the observer can be considered as embedding a first-order low-pass filter with a time constant of τ in a three-order generalized proportional integral observer. The filter can be used to adjust the noise suppression performance of the observer, and m3 is used to adjust the disturbance and the observed value of the disturbance differential information. By embedding the filter and adjusting the time constant of the filter, the noise suppression characteristics of the observer at high frequencies can be improved. At the same time, the three-order generalized proportional integral observer can be used to observe the disturbance term x2 and the disturbance differential term f (1) in the first-order position control system. Figure 2 ).

[0160] ​The above merely provides the preferred embodiments of the present application, and the protection scope of the present application is not limited thereto, and any person skilled in the art can make equivalent replacements or changes according to the technical scheme and the inventive concept of the present application within the technical scope disclosed by the present application, which should be covered in the protection scope of the present application.

Claims

1. A robust predictive position control method for a permanent magnet motor with low observation noise characteristics, characterized by: The method comprises the following steps: S1: Construct a mathematical model of the total disturbance of the surface-mounted permanent magnet motor position control system under the influence of parameter and load disturbances. On this basis, construct a position control system prediction model and linear constraints of the position control system under the influence of disturbances. Step S1 includes the following steps: S11: Construct a mathematical model of the total disturbance of the surface-mounted permanent magnet motor position control system considering the parameters and load disturbances; as shown in the following formula (2): （2） in, are d-axis current, q-axis current, d-axis voltage and q-axis voltage respectively, is the motor inductance, is the magnetic link, is the motor electrical angle, is the electrical angular velocity, 、 are the moment of inertia and pole pair number, respectively, the disturbance are the total disturbances of d-axis current, q-axis current and speed caused by parameter changes, respectively. The disturbance at time k is expressed as , whose expression is: （3） in: , B are the motor resistance and friction coefficient respectively, 、 、 、 and are the parameter changes of the corresponding variables respectively; are the d-axis current, q-axis current, d-axis voltage and q-axis voltage at time k respectively; , , are the electrical angular velocity at time k, the electrical angle at time k, and the load torque; S12: Construct a prediction model for the position control system considering disturbances; The forward Euler discretization method shown in the following equation (4) is used to discretize equation (2): （4） Among them: x represents the state quantity, is the discrete time step, k represents the kth moment; The position control system prediction model is expressed as: （5） in: 、 are the state variable matrix and the output variable matrix, 、 、 is the position control system matrix, is the perturbation matrix, is the control voltage at time k, and the state quantity with subscript k represents the state value at time k. The definitions of related quantities are as follows: ； S13: Construct linear constraints for position control systems considering parameter disturbances; The position control system needs to consider the dq axis current constraints , speed constraint and dq axis voltage constraints , expressed as: （6） The truncated Taylor expansion method shown in formula (7) is used to take the order and Discretize the current and speed system in equation (2): （7） Combined with formula (6), the current and speed constraints are derived as follows: (8) (9) Where: is the first-order differential of the speed disturbance; The voltage constraint uses a rectangular approximation scheme: （10） Where: is the rectangular coefficient, 0≤ε≤1, is the bus voltage; S2: Construct a low-noise characteristic observer to observe disturbances and disturbance differential information; the specific construction process is as follows: The current and speed dynamic models in equation (2) are considered as disturbance models as follows: （11） Where: is the state variable, representing the ; is the state variable, representing the total disturbance in equation (2) ; is the control input, representing the ; is the gain before input, is the first-order differential of the disturbance; Based on the perturbation model (11), the low-noise observer is designed as follows: (12) Where: is the observer gain, is the gain coefficient, which is used to adjust the observation effect of disturbance and disturbance observation information. is the first-order low-pass filter coefficient, used to adjust the noise suppression performance; is the input error The filtered value of S3: Prediction model and position control system constraint compensation based on disturbance observation; using the obtained disturbance information and perturbation differential information Observed values , used to compensate the prediction model (5), and the position control system constraints (8) and (9); S4: Construct a constrained convex optimization model for a robust position control system. The specific construction process is as follows: Consider the prediction time domain When , the value function of position control is designed as follows: （13） Where: in , is the target reference value, that is: , Q and R are the corresponding weight coefficient matrices, respectively: , They are the d-axis current weight coefficient, position weight coefficient and control weight coefficient respectively; in order to smooth the control input, the smoothing factor is introduced , then the target reference value at different times Expressed as: （14） Where: 、 in , 、 The d-axis current and position given values ​​at different times are given in the entire prediction time domain for the sake of simplicity of calculation. Take the same value within; definition: , for the sake of simplicity of calculation, assume that: , with time k as the initial state, output at different times , , use the following formula to determine: （15） in: Where I is the identity matrix; Based on equations (8), (9), (10) and (13), a constrained convex optimization model for the robust position control system is constructed. The solution of this model is regarded as a linear quadratic programming problem. To simplify the solution, only the constraints at time k are considered, and the constrained convex optimization model is obtained as follows: （16） Where: 、 、 and is the estimated value of the disturbance at time k; S5: Solving constrained optimization models based on the Hildreth method.

2. The method for robust predictive position control of a permanent magnet motor with low observation noise characteristics according to claim 1, characterized in that: In step S3, the prediction model based on disturbance observation and position control system constraint compensation is specifically: The designed observer formula (12) is used to apply the d-axis current equation, q-axis current equation, and speed dynamic equation in formula (2) respectively; The corresponding disturbance observers are designed as follows: , , Where: the superscript ^ represents the estimated value of the corresponding variable, 、 、 are the error filtering values ​​between the actual and observed values ​​of d-axis current, q-axis current and speed, respectively. 、 、 is the filter coefficient of the corresponding variable, 、 、 are the first-order differentials of the total disturbance of d-axis current, q-axis current and speed, respectively. 、 、 Adjust the gain for the perturbed observations, 、 、 、 、 、 are the observer gains respectively, and their values ​​are: ， ， Where: 、 and They are the d-axis current observer bandwidth, q-axis current observer bandwidth and speed observer bandwidth respectively.

3. The method for robust predictive position control of a permanent magnet motor with low observation noise characteristics according to claim 1, characterized in that: In step S5, the constrained optimization model based on the Hildreth method is solved; specifically: The optimization model shown in Equation (16) is solved based on the Hildreth method. First, when the position control system constraints are not considered, the unconstrained solution of the position control system is Solve according to the following formula: （17） Unconstrained solution Expressed as: （18） Where: ,symbol represents the Kronecker product, is the reference value matrix, expressed as: ,definition ,as well as , Then the optimal solution with constraints Expressed as: （19） Where: is the Lagrange multiplier; Based on formula (19), the optimal solution with constraints at time k is used. Applied to position control systems to achieve robust predictive position control, （20） in: is the control voltage considering the constraints of the position control system, for The d-axis component of for The q-axis component of .

Images