PID Neural Network Sliding-mode Controller Of The Three-phase Voltage-source PWM Rectifier

Converter technology has been widely used in industrial and domestic in recent years, however, traditional uncontrolled diode rectifier and phased-controlled thyristor rectifier can’t reach the ideal control effect and create much pollution to power grid, it can’t meet the requirements of the development. There are the excellent characteristics of bidirectional flowing energy, high power factor, low harmonic, high power quality and less pollution to power grid for PWM rectifier, so that it conforms to the state of the requirements of "green electricity transformation". With respect to control algorithm, the double closed loop PI control is widely applicable and is easy to reach, but it spends much time to set the parameters and those parameter can’t be modified in real time with this control method, also, the parameters needs to be re-tuned under different application conditions, hence, double-loop PI control has poor environmental adaptability. In order to obtain better dynamic and static performance, a kind of PID neural network sliding mode algorithm is adopted in this paper.Firstly, the applications circumstances of rectifier and research trend is introduced in this paper, and the working principle is explained. As it has difficulty to control the mathematical model of the rectifier in ABC coordinate system which is highly nonlinear and strong coupling characteristic, therefore it is analysed under dq rotating coordinate system. Then the sliding mode variable structure algorithm is introduced, which can be applied well in PWM rectifier, because the system characteristic of structural discontinuity is consistent with the switching characteristics of power electronic devices. But similarly because of the feature, when system reaches a steady state, the output voltage will jitter back and forth on the steady-state value because of the discrete algorithm, which accordingly affects the system performance. To eliminate the shortcomings of sliding mode controller, a new PID neural network algorithm is added to the sliding mode controller, it didn’t calculate three parameters of PID through neural network, but PID control blend in neural network. In this way, the jitter is reduced and the system performance is improved. At last, the SIMULINK simulation platform is set up in this paper. What’s more, a comparative analysis is made between this two control methods. The result finds that the PIDNN sliding mode variable control has smaller jitter, better robustness, insensitive to external disturbance and change of rectifier parameters.Secondly, an experimental prototype is designed according to the actual demand, and the selection of prototype device as well as the designed hardware circuit is descripted. Then the framework of the overall program and the important module of TSM320F2812 chip is presented, the flow chart of the program are drawn as well. What’s more, the floating point speed is optimized through software programming. At last, experiments are conducted on the experimental prototype.