Research Of Medium Frequency Power Supply Balance Controller Based On BP Neural Network

High power controllable rectifying circuit may cause large harmonic to the whole power supply system. This problem can be solved by changing the three phase rectifying circuit from single mode to double mode with Y, Δ parallel circuit so far. The balance of these two circuits output is the key point when it is performed, otherwise the rectify component will be damaged due to the large consumption between two circuits. So, it is necessary to produce a high precision balanced power controller to protect that. Based on lots of applications of high power controllable rectifying circuit in medium frequency heating system, the paper introduce the design procedure of one kinds of balanced power controller.Firstly, this paper analyses the structure and control methods of medium frequency inverter, point out the relationship between rectifying circuit and inverter circuit. Based on those theories, the author creates the mathematic model and circuit structure of the balanced power controller. The main idea is: with the precondition of the stabilization of medium frequency output, as the reference of loop-close regulator outputs be system, use the warp of those two power supply circuit as the disturb signal, to adjust to shift-phase controlled angle of the two rectifying trigger circuit, finally achieved the balance of those two Y, Δ rectifying circuit.According to lots of analyses of control algorithms, the paper discovers one kinds of control method based on BP neural network which is suitable to control nonlinear system and inertial system. That is the PID controller based on BP neural network. After the simulation of MATLAB, it is confirmed that this algorithm is better than traditional PID control algorithm. Meanwhile, this algorithm also overcomes the disadvantage and limitation of traditional PID control method.After the introduction of the detail design procedure of all hardware and software for the whole system, much waveforms and experiment results is provided. The results proved that the system works well and the whole design is valid.