Research On Five-level Active Neutral Point Clamped Inverter Based On Model Predictive Control

With the promotion and application of multilevel converters in many fields such as new energy power generation,harmonic control,and electric drive.More and more experts and scholars pay attention to the performance of multilevel converters.In order to meet the applications in various occasions such as high performance,high reliability,high-voltage and high-power,high-step multilevel converters are also used.However,with the increase of the number of levels,the switching states that need to be controlled also increase.There is no doubt that it greatly increases the control difficulty of the multilevel converter.In this paper,the active neutral-point-clamped five-level converter is taken as the research object,and focus on optimizing control strategy.Finally,the theoretical analysis and experimental verification of the control strategy are completed.1.Firstly,this paper describes the development history of some multilevel converters and their background and significance.The topology and working principle of some common multilevel converters are analyzed,and the advantages and disadvantages of these converters are pointed out.Then it briefly introduces the research status of MPC algorithms in multilevel converters in recent years.2.Taking the five-level ANPC inverter as an example.In this paper,the working principle of the five-level inverter topology is introduced,and the effects of various switching states on the neutral potential and flying capacitor voltages of the inverter are analyzed.Based on the analysis of inverter topology,a traditional predictive control algorithm for five-level ANPC inverter is given.However,there are many problems and shortcomings in the control algorithm,such as the calculation burden caused by more switching states,and the multiple weighting factors make the coupling between control targets strong,etc.3.In view of the shortcomings of traditional control algorithm,an improved model predictive control algorithm is proposed.The algorithm uses the method of dividing space vectors into several regions to reduce the original 512 switching states to 31,which greatly reduces the calculation time of the model predictive control algorithm.In traditional control algorithm,two weighting factors are used to balance the neutral-point potential and the flying capacitor voltages.The improved algorithm proposed in this paper adopts the idea of step-bystep balance to deal with the balance of neutral-point potential and voltage of flying capacitors.Firstly,the voltage vectors that satisfies the condition is determined by judging the magnitude relationship between the flying capacitor voltages and its reference voltage.Then these voltage vectors which meet the conditions are brought into the value function for rolling optimization,and the optimal solution is selected.Therefore,in the value function,only one weighting factor is required to constrain the neutral-point potential,which also avoids the interference between the two voltage balances.The efficiency of calculation and the stability of the system are greatly improved.