Study On Model Predictive Control For NPC Three Level Converter

At present,there is an increasing demand for high performance and high power converter,because of the rapid development of industry and transportation and the urgent need of energy saving and emission reduction.So it is very necessary to study the advanced control strategy of converter.And,in recent years,finite control set model predictive control(FCS-MPC)is widely used in the field of power electronics.Based on above background,the FCS-MPC strategy and its improved strategy for NPC three level converter are studied.A mathematical model of NPC three level converter is established.Then the basic principle of FCS-MPC is elaborated,including the establishment of cost function,algorithm implementation process and time delay compensation.The traditional FCS-MPC has some problems,such as large computational burden,high du/dt and static current error caused by model error.In order to solve those problems,An improved strategy of FCS-MPC is proposed.The computational burden is reduced to 1/3 of origin while du/dt drops by adding switch jumps and other constraints.The current prediction is replaced by voltage prediction,which further reduces the computational burden.By adding the current feedback compensation in the voltage prediction formula,the static current error is effectively reduced.The Multi-step FCS-MPC for NPC three level converter is also studied.The cost function is rearranged,equivalent to the switch sequence error,in order to simplify caculation.The multi-step FCS-MPC strategy is implemented by using the iterative backtracking algorithm in reference to the idea of sphere decoding algorithm.Taking the prediction horizons of 5 as an example,the computational burden is reduced by four orders of magnitude,compared to the standard exhaustive search approach.But,Multi-step FCS-MPC may encounter two problems during its operation.One problem is the smaller initial radius of the sphere decoding results in a lack of complete switch sequences,resulting in increased current harmonic.The other is in dynamic processes,for example,a step of reference current,the computational burden will increase dramatically.For problem 1,the present optimal switch sequence is obtained by shifting the last one forward by one time when no complete switch sequence is found.For Problem 2,the number of complete switch sequences searched is limited in the program.The two solutions above are validated by simulation.Finally,the performance of Multi-step FCS-MPC is simulated and analyzed,and the following conclusions are obtained:1.FCS-MP Multi-step shows good dynamic performance,and the increase of the predicted horizons would not slow down the dynamic response of the current.2.The static performance of FCS-MPC Multi-step is different in the two cases of large inductance and large resistance load,and,in general,the longer the prediction horizons is,the better the static performance is.Especially for the large inductance load,the increase of the prediction horizons significantly reduce the current harmonics at low switch frequency.The model predictive control strategy for PMSM fed by NPC three level converter is studied.The mathematical model of PMSM is established.On this basis,the speed deadbeat predictive controller and current FCS-MPC controller are designed respectively.Finally,the PMSM model predictive control system is simulated.The simulation results show that the control system can realize dq axis current decoupling control.And it has good dynamic and static performance,strong anti-interference ability,and good control performance at low speed.