Design Of DC-DC Controller Based On BP Neural Network PID

With the development of science as well technology and the improvement of people's living standard,portable electronic products such as mobile phones are basically inseparable to modern people.The challenge is the power design of portable electronic devices.A good electronic device cannot be separated from a high performance power supply which has promoted the development of high performance digital switching power supply.In this thesis,we propose to introduce the error Back Propagation(BP)neural network into the traditional PID control,and obtain an improved control strategy which is called the PID control strategy based on BP neural network.A Buck dc-dc converter based on BP neural network is designed.When the input voltage or load changes cause the converter output voltage changes,will be sampling to the reference voltage and output voltage of calculated error signal as the input into the BP neural network,the choice in design of three layer BP neural network through self-learning training process real-time setting three adjustable parameters of PID controller K_P,K_I,K_D.BP neural network in the adjustment process according to the gradient descent method to adjust each layer has a weight function makes the error of the mean square value decreased rapidly,so we can improve the regulating ability of PID control algorithm,so as to make the converter for faster response speed and small overshoot.Both dynamic performance and static performance are improvements to conventional PID.After the system simulation in the Matlab/Simulink,the control algorithm realized by the hardware language,then downloaded to the FPGA and design of the hardware circuit structures,the closed-loop test platform,through the test,it showed that the dynamic performance,when the load current in 1 A and 0.5 A jump between the two current value,the conventional PID algorithm of step-down converter system the highest recovery time for the 680?s,the max overswing is 277 mV.Under the same test condition,the maximum recovery time of the step-down converter system based on BP neural network PID algorithm is 219?s,which is the max overswing is 224mV.In terms of dynamic performance,BP neural network is used to optimize PID algorithm.In steady-state performance,when the input voltage gradually increased from 2V to 5V,the output voltage is stable at 1.8v,and the maximum steady-state error is about 0.67%.When the load current gradually increased from 0A to 1.2A,the output voltage is stable at 1.8v,and the maximum steady-state error is about0.77%.Meet design requirements.