Application of Neural Network Inverse System in Motor Variable Frequency Speed ​​Regulating System

The traditional control system design is based on the known mathematical model of the system. Therefore, the quality of the control system has a lot to do with the accuracy of the mathematical model. However, the actual system is very diverse and complex, and it is difficult to find a suitable description model. For a nonlinear system, it is more difficult and sometimes impossible. Neural networks are used to control system design, but it is not necessary to control the mathematical model of the object, but to train the neural network online or offline, and then use the training results to design. Because the neural network has strong adaptability, parallel processing capability and essential non-linearity, the controllers designed with neural networks for nonlinear and uncertain systems will have stronger adaptability, better real-time performance and robustness. Sex. There are many methods for designing control systems based on neural networks, but no comprehensive theoretical systems and systematic design methods have been formed yet. The neural network controllers that have been proposed mainly include neural network PID control, neural network predictive control, and neural network internal control. Model control, neural network fuzzy control and so on.

Induction motors are typically multi-variable, nonlinear systems, coupled with variable frequency devices, the entire system is more complex. In this paper, the induction motor variable frequency speed control system is selected as the control object, and its mathematical model is analyzed for reversibility. The neural network is used to construct the inverse controller to control the induction motor variable frequency speed control system. The neural network inverse controller is a combination of inverse system method and neural network. Liu Guohai, male, born in 1964, Ph.D., professor and vice dean of the School of Electrical and Information Engineering, Jiangsu University. The research directions are motor control and complex system control.

The BP network is used to approximate the a-order inverse system of the object, and then it is connected in series with the object to form a complex pseudo-linear system. Then the existing linear system design method is used to design the control system. In this paper, the neural network inverse system method was used to test the actual induction motor variable frequency speed control system respectively for no-load/full-load start test, sudden/shock load test and follow-up test verification. The first-order pseudo-linear system is formed before the inverse system is serially connected to the original system, so that the entire compound system is transformed into a second-factory-type pseudo-linear single-input single-output system.

The simulation results of the simulation curve diagram of the unit under changing conditions indicate that the control system effectively solves the influence of boiler nonlinearity, large time delay, and load disturbance on the unit operation. Not only under normal load stability, the main steam pressure can remain fairly stable, and under the circumstances of large changes in unit peaking and combustion rate, the control unit can quickly track the load, effectively improving the control quality of the system and satisfying the actual conditions. The requirement of control has important practical significance for improving the peaking performance of the unit.

4 Conclusions The outstanding features of the cluster adaptive fuzzy control based on the neural network prediction model designed in this paper: () The use of neural networks to predict the system provides a guarantee for the precise control of nonlinear large time-delay systems.

(2) While using fuzzy control to implement fuzzy rules based on expert control strategy and experience, cluster adaptive control is used to compensate for the inadaptability and imperfection of fuzzy rules. The control method is concise, flexible and fast.

Simulation shows that the control has strong robustness, real-time performance and immunity, even if the peaking unit can maintain good control performance and operating results under variable conditions (variable load under a wide range).