The Control Strategy Of Three-level Buck Converter Based On PID Neural Network

Multi- level converter has been widely used and rapid development for many advantages such as low switching voltage stress, low output voltage harmonics and large capacity. Too many levels causes the system much complex, lower reliability higher cost and can also make control becomes complicated. Three- level DC-DC converter has been widely used, it reduced voltage stress on the switch by least switch to the maximum extent, and improving the output performanc e, reduces the output filter volume, improve the dynamic characteristics of the system,so it very suitable for applications such as requiring high power but not very complex system. However, in practical application, the control circuit, drive circuit and the various parameters switch can not completely consistent, causing the switch turned or off each time is not equal, system parameter variations and change of load also resulting the flying capacitor voltage unstable, causing the switch voltage stress is not equal, and damage the capacitor switching device, three- level waveform not symmetrical, makes three-level converter lost its fundamental strengths. Therefore, in order to ensure the three-level converter work, it must take some certain measures to control the flying capacitor voltage stable.The output voltage and the flying capacitor voltage are strong coupling in three- level inverter, switching makes it a non- linear system, how to control the flying capacitor voltage and the output voltage efficiently is the key to success. PID controller is simple, stable and has high precision, reliability effect, now it has a wide range of applications in many ways, but it must rely on accurate mathematical model and can elaborate its advantages only when it has the proper PID controller parameters. Artificial neural network makes of large numbers of simple neurons, it forms intelligent network in accordance connections, it formed with non- linear nature of the multilayer neural network, so it has the ability to approximate any function, and it's control of the system does not rely on the mathematical model of the object, very suitable for the control of the flying capacitor voltage and output voltage of three- level switching converter. Therefore, we can obtain PID ne ural network controller by integrate PID control as the basic unit into the neural network. It has the advantages of neural network and PID controller, for it can automatically compensate for nonlinear system model order times change, parameters and other aspects of the input signal, has a very good adaptability for the controlled object, so it very suitable for three- level inverter control, to enhance its robustness and improve the control performance.This paper analyzes the basic principle and the perfor mance advantages of three- level DC-DC converter, gives the principle of PID neural network controller. For the three-level step-down DC-DC converter, designed the PID neural network to controller output voltage and the flying capacitor voltage. O n the basis of theoretical research, make a simulation of the controller. Analyzed the results and the results verify the theoretical analysis that neural network PID controller has advantage of the three-level DC-DC converter control.