CN115313932B - A direct thrust control method for permanent magnet linear synchronous motor - Google Patents

Abstract

The invention relates to a control method of direct thrust of a permanent magnet linear synchronous motor, which belongs to the technical field of motors and comprises the steps of obtaining equivalent current and equivalent voltage under an alpha-beta coordinate system by three-phase current and three-phase voltage of the permanent magnet linear synchronous motor, obtaining thrust, flux linkage and flux linkage angle of the permanent magnet linear synchronous motor by utilizing the equivalent current and the equivalent voltage under the alpha-beta coordinate system, constructing ADHDP a speed controller to obtain a thrust difference value, obtaining a flux linkage difference value by making a given flux linkage amplitude value and the flux linkage difference value, determining the state of each switch of an inverter by signals output by a sector selection unit through the thrust difference value, the flux linkage difference value and the flux linkage angle, and generating three-phase alternating voltage to enable the motor to operate. Compared with the traditional PI controller with low convergence speed and low error precision, the direct thrust control method of the permanent magnet linear synchronous motor is a data driving control method, optimizes to a certain extent and improves the control precision of the permanent magnet linear synchronous motor.

Description

Direct thrust control method for permanent magnet linear synchronous motor

Technical Field

The invention relates to the technical field of motors, in particular to a control method for direct thrust of a permanent magnet linear synchronous motor.

Background

Along with the continuous enhancement of the mutual promotion and dependency relationship between cities and towns in China and the rapid expansion of urban core areas, the enhancement of the rapid, safe and comfortable traffic interconnection between the areas is more and more important, so that more requirements are put forward for the track traffic industry in China.

The traditional rail transit industry adopts a rotating motor as a power device, and drives a train by means of the adhesive force between wheels and a rail. This driving scheme has been developed and practiced for a long time, and the technology is relatively mature. However, since this method uses physical adhesion between wheel tracks, the speed, acceleration, climbing performance, etc. of the vehicle are limited. In addition, because the rotating motor is adopted for driving, the rotating moment is required to be converted into the traction force of the train through gear transmission, and the whole power system is huge due to the existence of a transmission system, so that the system structure is complex. The Permanent Magnet Linear Synchronous Motor (PMLSM) has the characteristics of non-adhesive force driving, simple structure, reliable performance and the like, and the linear motor is different from the rotary motor, so that a middle transmission link is eliminated, and electric energy is directly converted into required mechanical skills, so that the structure of the whole power system becomes simple, the size is reduced, and the PMLSM is increasingly applied to the rail transit industry.

At present, the control systems of the permanent magnet linear synchronous motor are basically closed-loop control systems, and the control modes are basically PI control, but the control method of the permanent magnet linear synchronous motor is susceptible to parameter time-varying and external unknown disturbance, most of the control modes of modern control theory are based on ideal mathematical models for control, and fluctuation can occur when the influence of parameter time-varying and the like occurs, so that the control system is unstable.

Disclosure of Invention

Aiming at the defects existing in the prior art, the invention provides a control method for the direct thrust of a permanent magnet linear synchronous motor, and aims to solve the technical problem of unstable control system caused by the fact that the control method of the permanent magnet linear synchronous motor is easily affected by parameter time variation and the like.

The technical scheme of the invention is as follows:

a direct thrust control method of a permanent magnet linear synchronous motor comprises the following steps,

S101, measuring three-phase current i_a、i_b、i_c and three-phase voltage u_a、u_b、u_c of a permanent magnet linear synchronous motor, and performing Clark transformation to obtain equivalent current i_α、i_β and equivalent voltage u_α、u_β under an alpha-beta coordinate system;

S102, calculating thrust fe, flux linkage psi_s and flux linkage angle theta_s of the permanent magnet linear synchronous motor by using equivalent current i_α、i_β and equivalent voltage u_α、u_β under an alpha-beta coordinate system;

S103, constructing ADHDP a speed controller, wherein a thrust reference value fe^* output by the ADHDP speed controller is differed from the calculated thrust fe to obtain a thrust difference delta fe, and a given flux linkage amplitude psi_s^* is differed from a flux linkage psi_s calculated in the step S102 to obtain delta psi_s;

S104, inputting the signals obtained by the thrust difference delta fe and the flux linkage difference delta phi_s in the step S103 through the thrust and flux linkage hysteresis regulator and the signals output by the flux linkage angle theta_s through the sector selection unit into the voltage vector switch selection unit, and further selecting the states of all the switches of the inverter, so that the inverter generates three-phase alternating voltage to enable the motor to operate.

Further, the construction method of ADHDP speed controller in step S103 is as follows:

Two BP neural networks, namely an execution network and an evaluation network, are constructed, and a ADHDP speed controller is constructed by using the execution network and the evaluation network.

Further, both the execution network and the evaluation network in ADHDP speed controllers contain two hidden layers.

Further, the hidden layers of the execution network and the evaluation network use bipolar sigmoidal functions, and the output layer uses linear functions purelin.

Further, the structure of the execution network is 3-12-12-1.

Further, the structure of the evaluation network was 4-10-10-1.

Further, the speed error e (k) and the variable e (k-1) at two times before the speed error e (k-2) are taken as the input of the execution network, and the thrust reference value fe^* is taken as the output of the execution network.

Further, the speed error e (k) and the variables e (k-1), e (k-2) at two times before the speed error and the thrust reference value fe^* for executing the network output are used as the input of the evaluation network, and the estimated value of the performance index function of the system is usedIs output.

Further, the speed error e (k) is obtained by obtaining the motor rotation speed v using a magnetic scale and making a difference from the desired speed v_ref to obtain the speed error e (k).

Further, the learning rate l_a,c (k) of the evaluation network and the execution network are each an adaptive learning rate, as shown in the following formula,

Where α, β are learning rate gains, the initial value of l_a,c (k) is 0.1, and E_a,c (k) is an objective function of the execution network and the evaluation network.

Compared with the prior art, the invention has the following advantages:

1. Compared with the traditional PI controller with low convergence speed and low error precision, the direct thrust control method of the permanent magnet linear synchronous motor is a data driving control method, optimizes to a certain extent and improves the control precision of the permanent magnet linear synchronous motor.

2. According to the direct thrust control method for the permanent magnet linear synchronous motor, the evaluation network and the execution network respectively adopt the BP neural network, two hidden layers are designed to improve the approximation precision, the adaptive learning rate is used, the convergence speed of the neural network is improved, and the BP neural network is prevented from being in local optimum.

Drawings

FIG. 1 is a flow chart of a method for controlling direct thrust of a permanent magnet linear synchronous motor according to the present invention;

FIG. 2 is a block diagram of a direct thrust control system of a permanent magnet linear synchronous motor according to the present invention;

FIG. 3 is a block diagram of internal logic structure of ADHDP;

FIG. 4 is a network architecture of an evaluation network;

FIG. 5 is a network architecture for an execution network;

Fig. 6 is a speed controller tracking curve based on adaptive dynamic programming.

Detailed Description

The invention is described in further detail below with reference to the accompanying drawings.

The invention provides a direct thrust control method of a permanent magnet linear synchronous motor, which is shown in fig. 1 to 6, and specifically comprises the following steps.

A direct thrust control method of a permanent magnet linear synchronous motor specifically comprises the following steps,

Step 1, measuring three-phase current i_a、i_b、i_c and three-phase voltage u_a、u_b、u_c of a permanent magnet linear synchronous motor, and performing Clark transformation to obtain equivalent current i_α、i_β and equivalent voltage u_α、u_β under an alpha-beta coordinate system;

Step 2, calculating the thrust fe and the flux linkage psi_s of the permanent magnet linear synchronous motor by using the equivalent current i_α、i_β and the equivalent voltage u_α、u_β under the alpha-beta coordinate system in the step 1, and then performing inverse trigonometric tangent calculation by using the flux linkage component psi_α、ψ_β under the alpha-beta coordinate system to obtain the flux linkage angle theta_s, wherein the specific expression is that,

Wherein p_n is the pole pair number of the motor, and τ is the pole pitch of the motor;

And 3, constructing two BP neural networks, namely an execution network and an evaluation network to form ADHDP speed controllers, wherein a control algorithm of the ADHDP speed controller consists of an evaluation network updating control strategy and an execution network adjusting speed error based on the idea of reinforcement learning, a thrust reference value fe^* output by the ADHDP speed controller is differed from the calculated thrust fe to obtain a thrust difference delta fe, and meanwhile, a given flux linkage amplitude phi_s^* is differed from the flux linkage phi_s calculated in the step 2 to obtain a flux linkage difference delta phi_s.

And 4, inputting a signal obtained by the thrust difference delta fe and the flux linkage difference delta phi_s in the step 3 through the thrust and flux linkage hysteresis regulator and a signal output by the flux linkage angle through the sector selection unit into the voltage vector switch selection unit together, and further selecting the states of all switches of the inverter, wherein the inverter generates three-phase alternating voltage to enable the motor to stably operate.

The specific construction process of constructing ADHDP the speed controller in the above step 3 is as follows:

The method comprises the steps of obtaining the rotating speed v of a motor by using a magnetic grating ruler, carrying out difference between the rotating speed v and the expected speed v_ref to obtain a speed error e (k), selecting values e (k-1) and e (k-2) of the two previous moments of the speed error, taking e (k), e (k-1) and e (k-2) as state variables of a ADHDP speed controller, and taking a thrust reference value fe^* output by the ADHDP speed controller as a control quantity.

Wherein the execution network and the evaluation network in the ADHDP speed controller both contain two hidden layers. A specific description of the evaluation network and the execution network is as follows.

The structure of the evaluation network is 4-10-10-1, a speed error, variables of two moments before the speed error and a thrust reference value fe^* output by a speed controller are selected as inputs of the evaluation network, and an estimated value of a performance index function of the system is usedFor output, the hidden layer of the evaluation network uses a bipolar sigmoidal function, and the output layer uses a linear function purelin.

The input vector defining the evaluation network is:

x_c(k)＝[e(k),e(k-1),e(k-2),fe^*(k)]

The process of evaluating the forward calculation of the network is as follows:

Where i, j represent the number of rows and columns of each matrix, j' represents the number of columns of each neural network input vector, and ch_1j(k),ch_2j(k),ch_3j(k),ch_4j (k) is an element in ch₁(k),ch₂(k),ch₃(k),ch₄ (k), respectively. ch_1j(k),ch_2j(k),ch_3j(k),ch_4j (k) is the input/output of the j-th node of the two hidden layers respectively. W_c1ij(k),W_c2ij(k),W_c3i (k) is an element in W_c1(k),W_c2(k),W_c3 (k), respectively. W_c1ij(k),W_c2ij(k),W_c3i (k) is the weight from the input layer to the hidden layer, the weight between the two hidden layers, and the weight from the hidden layer to the output layer.

Defining utility functions of the system as:

U(k)=e(k)Ae(k)^T+fe^*(k)Bfe^*(k)^T

Wherein A and B are positive diagonal matrices.

The objective function E_c (k) defining the evaluation network is:

Where e_c (k) is the error function of the evaluation network, where γ is the discount factor.

In the middle ofAnd (3) withAre respectively realized by two evaluation networks, the inputs of the two evaluation networks are respectively,

x_c(t)＝[e(k),e(k-1),e(k-2),fe^*(k)],x_c(t)＝[e(k-1),e(k-2),e(k-3),fe^*(k-1)], The two evaluation networks are identical in structure except for the input.

The weight updating process of the evaluation network is shown as follows

The weight change amount from the hidden layer to the output layer is defined as follows:

The updated weight matrix is:

W_c3(k+1)＝W_c3(k)+ΔW_c3(k)

defining the weight change from hidden layer to hidden layer as:

The updated weight matrix is:

W_c2(k+1)＝W_c2(k)+ΔW_c2(k)

The weight change amount from the input layer to the hidden layer is defined as follows:

The updated weight matrix is:

W_c1(k+1)＝W_c1(k)+ΔW_c1(k)

Where l_c (k) is the learning rate of the evaluation network and W_c3(k+1),W_c2(k+1),W_c1 (k+1) is the updated weight matrix.

The structure of the execution network is 3-12-12-1, the speed error and the variables at the two moments before the speed error are selected as the input of the execution network, the thrust reference value fe^* output by the speed controller is taken as the output, the hidden layer of the execution network adopts a bipolar sigmoidal function, and the output layer adopts a linear function purelin.

Defining the input vector of the execution network as:

x_a(t)=[e(k),e(k-1),e(k-2)]

the following is a calculation procedure to perform the forward direction of the network:

where ah_1j(k),ah_2j(k),ah_3j(k),ah_4j (k) are the elements within ah₁(k),ah₂(k),ah₃(k),ah₄ (k), respectively. ah_1j(k),ah_2j(k),ah_3j(k),ah_4j (k) are the input and output of the j-th node of the two hidden layers, respectively. W_a1ij(k),W_a2ij(k),W_a3i (k) is an element in W_a1(k),W_a2(k),W_a3 (k), respectively. W_a1ij(k),W_a2ij(k),W_a3i (k) is the weight from the input layer to the hidden layer, the weight from the hidden layer to the hidden layer and the weight from the hidden layer to the output layer, and the weight is updated by gradient descent method and minimizedPerforming weight updating of the network for the target implementation, i.e. performing an objective function of the network

The weight update procedure of the network is performed as follows,

The weight change amount from the hidden layer to the output layer is defined as follows:

The updated weight matrix is:

W_a3(k+1)＝W_a3(k)+ΔW_a3(k)

Defining the weight change quantity between two hidden layers as follows:

The updated weight matrix is:

W_a2(k+1)＝W_a2(k)+ΔW_a2(k)

The weight change amount from the input layer to the hidden layer is defined as follows:

The updated weight matrix is:

W_a1(k+1)＝W_a1(k)+ΔW_a1(k)

Where l_a (k) is the learning rate of the evaluation network and W_a3(k+1),W_a2(k+1),W_a1 (k+1) is the updated weight matrix.

Wherein in the updating of the weightThe method comprises the following steps:

In both the evaluation network and the execution network, the learning rate l_a,c (k) is an adaptive learning rate, as shown in the following formula:

Where α, β is the learning rate gain, the initial value of l_a,c (k) is 0.1, and E_a,c (k) is the objective function of the executive network and the evaluation network.

As shown in fig. 6, the direct thrust control method of the permanent magnet linear synchronous motor according to the present invention is a data driving control method, and only includes input and output of a control system in the control process, when there is a parameter change due to operation in the system or when a load disturbance is suddenly added, the control method ADHDP of the present invention can be used to feed back an accurate speed value and output an accurate thrust value, and the present invention has the characteristics of good robustness and no overshoot, so that the permanent magnet linear synchronous motor can stably operate.

Claims (7)

1. A direct thrust control method of a permanent magnet linear synchronous motor comprises the following steps,

S101, measuring three-phase current i_a、i_b、i_c and three-phase voltage u_a、u_b、u_c of a permanent magnet linear synchronous motor, and performing Clark transformation to obtain equivalent current i_α、i_β and equivalent voltage u_α、u_β under an alpha-beta coordinate system;

S102, calculating thrust fe, magnetic linkage psi_s and magnetic linkage angle theta_s of the permanent magnet linear synchronous motor by using equivalent current i_α、i_β and equivalent voltage u_α、u_β under an alpha-beta coordinate system;

S103, constructing ADHDP a speed controller, wherein a thrust reference value fe^* output by the ADHDP speed controller is differed from the calculated thrust fe to obtain a thrust difference value delta fe, and a given flux linkage amplitude value psi_s^* is differed from a flux linkage psi_s calculated in the step S102 to obtain a flux linkage difference value delta psi_s;

The construction method of the ADHDP speed controller comprises constructing two BP neural networks, namely an execution network and an evaluation network, forming a ADHDP speed controller by using the execution network and the evaluation network, wherein the execution network and the evaluation network in the ADHDP speed controller both contain two hidden layers, the learning rates l_a,c (k) of the evaluation network and the execution network are self-adaptive learning rates, as shown in the following formula,

Wherein alpha and beta are learning rate gains, the initial value of l_a,c (k) is 0.1, and E_a,c (k) is an objective function of an execution network and an evaluation network;

S104, inputting the signals obtained by the thrust difference delta fe and the flux linkage difference delta phi_s in the step S103 through the thrust and flux linkage hysteresis regulators and the signals output by the flux linkage angle theta_s through the sector selection unit into the voltage vector switch selection unit together, and further selecting the states of all switches of the inverter to generate three-phase alternating voltage to enable the motor to operate.

2. The method for controlling direct thrust of a permanent magnet linear synchronous motor according to claim 1, wherein the hidden layers of the execution network and the evaluation network are bipolar sigmoidal functions, and the output layer is linear function purelin.

3. The method for controlling direct thrust of a permanent magnet linear synchronous motor according to claim 2, wherein the execution network has a structure of 3-12-12-1.

4. The method for controlling direct thrust of permanent magnet linear synchronous motor according to claim 2, wherein the evaluation network has a structure of 4-10-10-1.

5. The method of claim 3, wherein the speed error e (k) and the variable e (k-1) at two times before the speed error e (k-2) are used as inputs of the execution network, and the thrust reference value fe^* is used as an output of the execution network.

6. The direct thrust control method of the permanent magnet linear synchronous motor as set forth in claim 4, wherein:

taking a speed error e (k) and a variable e (k-1) at two moments before the speed error, e (k-2) and a thrust reference value fe^* output by an execution network as inputs of an evaluation network, and taking an estimated value of a performance index function of a systemIs output.

7. The method for controlling direct thrust of permanent magnet linear synchronous motor according to claim 5 or 6, wherein the speed error e (k) is obtained by obtaining the motor rotation speed v by using a magnetic grating ruler and obtaining the speed error e (k) by making a difference with the expected speed v_ref.