Permanent Magnet Synchronous Motor Vector Control Based on Neural Network

The permanent magnet synchronous motor adopts high-energy permanent magnets, which eliminates the influence of the mechanical transmission chain of rotary servo motor from rotary motion to linear motion. It has high thrust strength, low loss, small electrical time constant and fast response, and becomes a high-precision, micro- One of the best actuators in the feed servo system. At present, due to the continuous improvement of the performance of modern permanent magnet materials, it is relatively simple to control the relative asynchronous motor, and it is easy to achieve high-performance control. Therefore, in the low-power applications such as numerical control machine tools and industrial robots, permanent magnet synchronous motors have been more widely used. application. In recent years, AC speed regulation technology has been changing with the development of power electronics, microelectronics and automatic control theory: 1 2 Vector Control Principles First proposed in the early 1990s, it was first applied to induction motors and has been widely applied. It fundamentally solves the high-performance control problem of AC motor torque.

For a separately excited DC motor, the space angle between the excitation magnetic field and the armature magnetomotive force is fixed by brushes and mechanical commutators. Normally, the two are orthogonal. Therefore, the direction between armature current and electromagnetic torque is mechatronic engineering.

In a linear relationship. By adjusting the armature current, the torque can be directly controlled. Different from the DC motor, in the synchronous motor, the spatial angle between the excitation magnetic field and the armature magnetomotive force changes with the load, and thus it is not possible to directly control the electromagnetic torque simply by adjusting the armature current. If the physical model of the AC motor can be equivalently changed to a DC motor model and then controlled, the problem can be greatly simplified. Comparing the mathematical models of DC motors and AC motors, we can find that the complex mathematical model of AC motors is because there is a complex inductance matrix. Therefore, to simplify the mathematical model, it is necessary to simplify the magnetic flux relationship. The coordinate change can realize the simplification of the AC motor model. The principle of changing the coordinates is to make the magnetomotive force generated in different coordinate systems the same. AC motor three-phase symmetrical stationary winding A, ", C through the three-phase balanced sinusoidal current generated when the rotating magnetomotive force. Its Calarke change, and the three-phase sinusoidal current is equivalent to two phases, so that it also The formula for the change is: In the new structure, the expression of the torque still depends on the rotor flux and it needs to undergo Park change. After the above two changes, the physical model of the model and the DC motor has no essence. The difference is that the phase and amplitude of the stator current space phasor can be conveniently controlled, ie, vector control is achieved.

Block diagram for vector control.

3 Single neuron adaptive PID control %12 Although neural network control technology has many potential advantages, the study of purely neural network control methods still needs further development. It is usually necessary to combine artificial neural network technology with traditional control theory or Integrated use of intelligent control technology. However, due to its mature technology, traditional PID regulators have been widely used in process control. However, for some complex and time-varying systems, PID parameters are difficult to adjust online in real time, so the application will affect the control quality of the system. In this paper, a single neuron adaptive PID control is used to achieve the speed control of the permanent magnet synchronous motor.

A block diagram of a single neuron implementing adaptive PID control is shown.

Since the synchronous motor can be equivalently converted to a DC motor through coordinate transformation, the control amount of the DC motor can be obtained by imitating the control method of the DC motor, and then the synchronous motor can be controlled after the coordinate inverse transformation. When designing the vector control system, it can be considered that the inverse transformation added after the controller is offset by the transformation of the feedback link, so the speed regulation thereof is very similar to that of the DC motor.

Here, a single neuron adaptive PID control is used to realize motor speed regulation, and a simulation block diagram implemented with MATLAB/SIMULINK is shown. Among them: nn is a function of +, which is used to adjust the weight coefficient of the neuron.

The operation effect of the neuron adaptive PID learning algorithm has a great relationship with the adjustable parameter k,"i,"p,"d. k is the most sensitive parameter of the system, and the change of k value is equivalent to P, I, D 3 items. Change at the same time, so adjust k in the first step, see the specific adjustment rules for parameters.

"!=0.08,"=37. The unit step response curve is shown.

5 hardware platform design and experimental results The servo driver designed in this paper is suitable for medium and low power motor drive. The hardware design part adopts the DSP+IPM mode. In the design, the one produced by Texas Instruments (turned to page 61) is used, then the secondary winding voltage value is 15.8V, and because the primary flyback operation is visible, when the input voltage is low, though) the voltage is 150V, according to the primary and The relation of the secondary number of turns: 150V/hour, but because the input voltage is low, its total voltage is still smaller than 600V, p=15.8V/! S, U! p = 212 匝 into the available secondary winding turns 23 匝.