Model Predictive Direct Power Control Of PWM Rectifier

Three-phase PWM rectifier has been widely used in many fields due to its merits of sinusoidal line current, bidirectional power flow and operation under unity power factor. As a new control strategy of PWM rectifier, model predictive control has superior performance. This paper studies deeply in model predictive control of PWM rectifier. which is compared with conventional control strategies, including voltage oriented control and direct power control. Two-vectors based direct power control has been proposed based on single-vector based direct power control. Two-vectors based model predictive control and three-vectors SVM based predictive control have been proposed based on single-vector based model predictive control. Furthermore, Fast model predictive control and fast two-vectors based model predictive control have been proposed based on conventional model predictive control and two-vectors based model predictive control.This paper analyzes the performance of voltage oriented control based PWM rectifier, and validates its effectiveness by simulation and experiment. The performance of direct power control strategy is also analyzed through the mechanism of vector table. According to the results of simulation and experimental results, it can be concluded that direct power control achieves better dynamic performance compared to voltage oriented control. Due to variable switching frequency and high steady state power ripples, two-vectors based direct power control has been proposed based on single vector based direct power control. The core of two-vectors based method is by allocating a fraction of control period for a non-zero voltage vector and the rest time for a zero vector, and its duration is obtained based on the principle of control target. Simulation and experimental results show that two-vectors based direct power control can reduce active power ripples and current total harmonic distortion, and achieve more sinusoidal line current and further steady state performance, while maintaining dynamic performance of the single-vector based direct power control.Model predictive control strategy of PWM rectifier is deeply researched, including theoretical foundation, simulation and experimental tests, which is also compared with conventional control methods. Compared to voltage oriented control, model predictive control has the same steady state performance but better dynamic response. Compared to direct power control, model predictive control replaces switch table with predictive model, which is more accurate and efficient in vector selection. It improves the steady state performance significantly, and achieves lower current THD. Based on the study of single-vector based model predictive control, this paper proposes two-vectors based model predictive control and three-vectors based predictive control. Two-vectors based model predictive control draws to two-vectors based direct power control, and can achieve further steady state performance. Three-vectors SVM based predictive control discretes the derivative of complex power, and deadbeat control is applied. The voltage vector is achieved in this control period, and SVPWM is used to synthesis the voltage vector. It can reduce the power ripples and current total harmonic distortion, while maintaining fast dynamic response, and fixed switching frequency is kept. Due to the complexity of model predictive control, fast model predictive control and fast two-vectors based model predictive control have been proposed based on conventional model predictive control and two-vectors based model predictive control, which can simplify the model. More argumentations show the equivalence between them.This paper designs a two-level three-phase voltage-type PWM rectifier experimental platform, including control circuit panel based on TI’s floating-point DSP TMS320F28335and main circuit based on Fuji IPM7MBP75RA120. All experiments are completed, and experimental results are analyzed, which verify the correctness and effectiveness of the proposed methods.