两电平逆变器长时域有限控制集模型预测控制

Long Horizon Finite Control Set Model Predictive Control for Two-Level Inverter

传统的长时域有限控制集模型预测控制(LHFCS-MPC)采用穷举法来实现。由于其计算量随着预测时域的增加而呈指数增长,难以应用到逆变器控制系统中。针对这一问题,提出了一种新的LHFCS-MPC算法。利用球形译码思想,缩小了最优解的搜索范围,显著提高了算法搜索最优解的效率;结合回溯法的深度优先搜索策略,进一步降低了算法的计算量。同时,为了确保算法在每个采样周期内都能维持较低的计算量,给出了初始和动态时算法预测时域的切换策略。利用该LHFCS-MPC算法,实现了两电平逆变器的电流控制,并搭建仿真模型进行验证。仿真结果表明,其计算量比穷举法降低了2个数量级;与比例积分(PI)控制相比,该算法具有更优的静态和动态性能。该算法对于扩大LHFCS-MPC的工程适用范围(如高采样率或动态变化很快的场合)具有实际意义。

The traditional long horizon finite control set model predictive control(LHFCSMPC)is implemented by the exhaustive method.Its computational burden increases exponentially with the increase of prediction horizon,so it is difficult to be applied to the inverter control system.In order to solve this problem,a new algorithm of LHFCSMPC is proposed.The sphere decoding concept is adopted to reduce the search range of optimal solution,so the searching efficiency of the optimal solution is improved remarkably.By combining the depthfirst search strategy for backtracking,the computational burden of the algorithm is decreased further.Moreover,in order to ensure that the algorithm can maintain a low amount of computation in each sampling period,the prediction horizon switching strategy during initial and dynamic time is given.The LHFCSMPC algorithm is used to realize the current control for the twolevel inverter,and the simulation model is built for verification.Predictive simulation results show that comparing with the exhaustion method,the computational burden is reduced by two orders of magnitude,and the proposed algorithm has better static and dynamic performance than those of the proportional integral(PI)control.The algorithm has practical significance for expanding the scope of application of LHFCSMPC(for high sampling rate or rapid dynamic changing situation).