Predictive Control Of NPC Three-level High-power PWM Rectifier

With the development of large-scale new energy power stations,the speed-up of electrified transportation railways,and the advancement of the strategy of deep-sea mining of mineral resources,high-power pulse width modulation rectifiers have become an important interface between power generation and power grids.Apart from the operating efficiency,the quality of current and longevity of devices need to be considered.When the rectifier performs electrical energy conversion,part of the energy is consumed in the rectifier system,resulting in power loss.If the auxiliary heat dissipation conditions are the same.Reducing the switching frequency is an effective method to reduce the switching loss and improve the operating life of switching devices,which is the main part of power loss.However,in the traditional vector control scheme,the accuracy of the system model is low and the error of the volt-second balance equivalent modulation principle becomes larger at low switching frequency.Therefore,this dissertation takes a more accurate description of the dynamic behavior of current as an entry point,and conducts research on improving the performance of the NPC three-level PWM rectifier under the low switching frequency.A predictive control method based on convex optimization for single-phase three-level PWM rectifier is proposed.Through switch optimization,a three-segment switching sequence that can control the balance of the midpoint potential is obtained.Moreover,the dynamic behavior of the current in the carrier cycle is described,and the traditional current error minimization at the end of the carrier cycle is changed to the integrated evaluation of error of the maxmium value and the erros of start and end of the carrier cycle.A convex problem that can evaluate the global performance of the controlled variables in a switching period is established and solved.A dual-vector current harmonic minimization predictive control method is proposed,which achieves the lowest current harmonics distortion in the carrier cycle by minimizing the error area.The correlation between the current harmonic and the control quantity is established.The current quality is better than the traditional dualvector method that minimize the current error at the end of carrier cycle.A satisfactory optimized FCS-MPDPC method for single-phase three-level PWM rectifier is studied.By introducing the concept of priority in the cost function,a global optimization combining all the control objectives and constraints of the PWM rectifier is realized.In order to improve the accuracy of the FCS-MPDPC method for describing the future behavior of PWM rectifiers,the prediction time domain is expanded from the traditional single step to multiple steps.Aiming at solving the problem of large computational burden of the satisfactory optimized control scheme,the FCS-MPC method with optimized switch is studied.Through switch control,the controlled quantity is converged to a bounded invariant set to ensure the stability of the system.And various performance indicators of the system are adjusted by setting boundaries.Compared with the traditional weighting factor method,when the performance of any two of three indexes,which are midpoint potential balance,current harmonic content,and average switching frequency,are similar,the other one is better.Moreover,the computational burden is reduced.The optimized switch FCS-MPC scheme is applied to a three-phase system.By optimizing the switch,the controlled quantity of the converter converges to the maximum bounded invariant set in the shortest time,which ensures the stability of the converter.The maximum bounded invariant set is dynamically and adaptively adjusted by the variable step search algorithm to realize the control of the switching frequency and the midpoint potential.The proposed predictive control scheme is experimentally verified in a high-power medium-voltage three-level PWM rectifier.In the high-power experiment,the proposed method can be realized by using TMS320F2812,which makes it easier to realize localization.