Study Of Embedded PID Neural Network Regulator

In this paper, I have researched and designed an embedded intelligent controller, what used MSP430F149 singlechip as the core, and embedded PID neural network algorithm.Regulator is widely used in industrial sectors such as chemical, petroleum, light industry, electric power, metallurgy, building materials, textile, and is also used in scientific fields, to control parameters like temperature, pressure, flow, liquid level, composition. PID(Proportional-Integral-Derivative) controller is widely used because of its simple structure, easy implementation and strong robustness. Although there are many new control methods emerging in the control field, but 90% of the regulator is PID control or its modified type in currently running control circuit. The core issue of the application of the PID regulator is parameter setting, and the difficulty of setting parameter restrict its application. In addition, the structural characteristics determine its limitations. It can get good control effect only in the simple linear univariate system, but ineffective in the complex systems. In order to overcome the weaknesses of traditional PID control, the control industry has made a number of improvements to the PID control. There are mainly self-tuning PID control, general predictive PID control, fuzzy PID control, expert PID control and intelligent PID control. These programs are aimed at how to select and tuning the PID parameters, and determine PID parameters online or offline by new methods on the traditional PID control structure. These methods improve the performance of PID controller to some extent, but they are generally aiming at certain specific issues, lack of general character, restricting their applications.In recent years, with the research and application of the neural network, combining the neural network and PID control to improve the performance of traditional PID control becomes a new research direction, and has achieved some results. PID neural network is proposed by Professor Shu Huailin in 1997, which is made of proportional neural, integral neural and differetial neural. it possesses the advantages of PID law and neural network, and shows good control performance in the complex systems.Based on PIDNN studied and discussed on its algorithm, some improvements have been made, and simulation have been done for its realization. Then I produced a embedded regulator with MSP430F149 singlechip as the control core. The regulator was embedded PID neural network algorithm, and was added acquisition circuit and input/output channels, man-machine interface, communication interface and the keyboard circuit. And operated it in the actual liquid level controlling system. Theoretical and experimental data show that the embedded PIDNN control system can achieve the control characteristics as no overshooting, no static, high response speed, and short transition time.