Research On Efficiency Optimization And Robust Control Of Three-Level T-Type Grid-Connected Converter Based On Model Predictive Control

In recent years,new energy generation is more and more applied in microgrid system to alleviate resource shortage and environmental pollution.The DC generated by the new energy needs the energy conversion with converter before connecting to the grid,so the selection and control of the converter is very important.In recent years,the Three-level T-type converter(3LT~2C)is increasingly used in low and medium voltage applications because of its higher power quality,larger capacity and less electromagnetic interference compared with the two-level converter.After passing 3LT~2C,there are switching harmonics in the current which cannot be directly used.LCL-filter is often used in new energy grid-connected systems because of its good filtering effect and small size.In this thesis,3LT~2C grid-connected system with LCL filter(LCL-3LT~2C)is studied.The operation of LCL-3LT~2C requires the balance of neutral-point voltage and the output power quality to meet the standard.Therefore,the control of LCL-3LT~2C belongs to the multi-objective optimization problem(MOOP).In this case,the design of the controller will be complicated if the traditional control method is used.In recent years,Model Predictive Control(MPC)has been widely used in power electronics because of its good dynamic response and simple multi-objective optimization process.MPC is oriented to multi-input and multi-output system,which accords with the characteristics of power electronic system.Therefore,MPC was used as the research control method in this thesis.With MPC control,LCL-3LT~2C mainly has two problems:(1)applying high switching frequency,the power loss of 3LT~2C power devices is large,and the system output efficiency is affected;(2)Although LCL filter has good filtering effect,it has resonance peak,which affects the stability of the system.In addition,the MPC control performance depends on the precise model,so the output power quality is coupled with the system model.If the established model is not accurate,the current ripple will increase,and the system robustness will seriously decrease when internal and external disturbances occur.The following will explain the two issues respectively:LCL-3LT~2C integrates 12 power electronic devices.Under high switching frequency,the total loss of the device cannot be ignored,the power loss of devices will be converted into heat,threatening the life of the device.In addition,part of the input energy becomes the loss of the device,and the overall output efficiency of the system is reduced,resulting in a waste of resources.Thus,it is necessary to analyze and control the loss of LCL-3LT~2C.The second problem is the weak robustness of LCL-3LT~2C with MPC control.There are three main reasons for the low robustness of the system:(1)The LCL filter belongs to the third-order system and there is a peak at the resonant point,leading to system instability;(2)The control performance of MPC is highly dependent on the system model.When the system model is not accurate or parameter mismatch occurs,sub-optimization problems will occur in the MPC optimization process.(3)The stability of LCL-3LT~2C with MPC control also faces challenges when the grid inductance changes and the grid voltage distorts.(1)and(2)belong to the internal disturbance of the system,while(3)belongs to the external environment disturbance.In order to reduce loss and maintain high quality current output,a low loss tolerance sequential model predictive control(LL-TSMPC)is proposed in Chapter 3.Firstly,based on the analysis of the energy loss under switching transitions per commutation,the prediction models of switching loss and conduction loss are established.Secondly,the total loss is regarded as a control objective of LL-TSMPC.Finally,the multi-objective model predictive control needs to select appropriate weight factors,which complicates the controller design.In order to solve this problem,a lexicographic optimization method based on TSMPC is introduced to eliminate the tradeoff factors and simplify the MPC controller design.The advantages of the proposed method for loss reduction are verified by the example analysis.Aiming at the robustness of LCL-3LT~2C with MPC control,the thesis mainly studies high robust model predictive control and model-free predictive control.Firstly,the high robust model predictive control is studied.By introducing passivity-based control,dynamic damping injection rules are designed by using the prediction error of converter output current,and the passive converter output voltage calculated by passivity-based control is used as the reference value of MPC for tracking.In this thesis,a method combining passivity-based control with dynamic damping injection and partial MPC(DDPB-PSMPC)is proposed to improve the anti-disturbance ability of the system from the perspective of energy.In the Chapter 4,the advantages of the proposed method are verified by showing the performance of controllers under different working conditions.Secondly,a high robust Model-free Predictive Control(MFPC)is studied.Because the traditional neutral-point voltage predictive control uses converter parameters,this method cannot be used for MFPC of LCL-3LT~2C.In this thesis,the relationship between neutral-point voltage balance and converter side current is analyzed first,then anti-disturbance factor is introduced to dynamically change the current reference value.Secondly,the novel ultra-local theory based on extended sliding mode observer is used to realize the highly robust model-free control(ADMFPC)of LCL-3LT~2C.By tracking the enhanced reference value of converter side current,the optimization of the three control objectives:converter side current,grid current and neutral-point voltage is realized,and the weight factor is eliminated.The high robustness of the proposed ADMFPC method is verified by comparing the performance with two traditional controllers under different working conditions.