Design Of Soft Switching Power Supply Based On Artificial Neural Network

Power electronics technology is paid great attention in current society.Switching power supply plays an important role in power electronic technology,and is now used in a wide variety of fields,such as communications,computers science and power systems.The research of soft switching technology is the key way for switching power supply to achieve its goals of high frequency,digital,modular and green energy saving.Power supply is a nonlinear system,and its structure is very complex.The resonant device is used in the circuit,so that the old control method is difficult to meet the demand index.Because of the rapid development of control technology,artificial neural networks have attracted a wide range of attention.Its approximation to nonlinear systems makes it widely used in various systems.The phase shifted full bridge ZVZCS DC/DC converter is used as the main circuit of switch power supply.Combining the artificial neural network with conventional PID controller law,a new type of BP neural network PID controller is designed.And on this basis,a soft switching power supply is developed with the output of 960W(48V/20A)and the switching frequency of 100 k Hz.The overall design block diagram of switching power supply is presented in this thesis.Several basic topologies of power converter are discussed in detail.The realization of soft switching of phase shifted full bridge converter is studied.The structure of the series diode with lagging arm is mainly analyzed,and the principle of soft switching is expounded.The design of the main circuit structure,and the design and selection of its parameters and components are completed.The simulation experiment of the main circuit is completed by using the software of PSPICE.Combining the artificial neural network with the PID control,a neural network controller based on chip of TMS320F2407 is developed.The structure of the controller is designed,and its implementation algorithm is expounded.At the same time,the simulation experiment of the control circuit of the system is done by using MATLAB,and the experiment is compared with the conventional PID control law.The simulation results of the main circuit show that the circuit has been successfully implemented ZVZCS.The efficiency of the power supply is effectively improved,and the reliability of the switch tube is greatly enhanced.The design parameters are correct and reasonable,and the design requirements of the switching power supply can be satisfied.The control circuit simulation experiment show that compared with the conventional PID control,the neural BP neural network PID controller designed in this thesis has some advantages.It almost has no overshoot,its steady-state error is very small,and it is relatively stable and has good robustness.The experimental results show that the design of the controller is reasonable.