模块化多电平换流器优化模型预测控制策略研究

发布时间：2018-09-09 13:18

【摘要】：模块化多电平换流器(modular multilevel converter,MMC)具有效率高、谐波小、模块化设计、易级联等优点,在高压大容量电能变换领域得到了日益广泛的应用。作为一种先进的控制策略,模型预测控制(model predictive control,MPC)通过目标函数可同时控制多个系统变量,具有建模直观、动态响应快等优点。传统MMC模型预测控制通过计算所有开关状态组合以实现最优控制目标,但随着桥臂模块数量的增多,计算量将呈几何级数增长,严重制约MPC的工程推广应用。针对N+1电平MMC,提出一种优化的模型预测控制算法,在对子模块电压、交流电流、相间环流、器件开关频率有效控制的同时,将开关状态组合计算量从C N2N降至N+1。针对子模块数高达数百的MMC,进一步提出分组排序优化模型预测控制(grouping-sorting algorithm combined OMPC,GSOMPC)策略,在降低桥臂子模块电压整体排序对硬件资源苛刻需求的同时,将开关状态组合计算量从N+1降至2X+M+3(N=M×X)。基于2.7 k V/60 k W 23电平MMC背靠背动模实验平台的实验结果证明了所提优化模型预测控制(optimized model predictive control,OMPC)及GSOMPC策略的正确性与有效性。
[Abstract]:Modular multilevel converter (modular multilevel converter,MMC) has many advantages, such as high efficiency, low harmonic, modular design and easy cascade. It has been widely used in the field of high-voltage and high-capacity power conversion. As an advanced control strategy, Model Predictive Control (model predictive control,MPC) can control multiple system variables simultaneously by objective function, which has the advantages of intuitive modeling and fast dynamic response. The traditional MMC model predictive control achieves the optimal control objective by calculating all the switch state combinations. However, with the increase of the number of bridge arm modules, the computational complexity will increase in geometric series, which seriously restricts the engineering application of MPC. An optimized model predictive control algorithm is proposed for N 1 level MMC,. When the voltage, AC current, interphase circulation and switching frequency of the device are effectively controlled, the switching state combination computation is reduced from C N 2N to N 1. For MMC, with hundreds of sub-modules, a grouping scheduling Optimization Model Predictive Control (grouping-sorting algorithm combined OMPC,GSOMPC) strategy is further proposed, which reduces the harsh requirement of hardware resources for the whole voltage ordering of the submodules. The switching state combination computation is reduced from N 1 to 2X M 3 (NM 脳 X). The experimental results based on 2.7 kV / 60kW 23 level MMC back-to-back dynamic simulation platform prove the correctness and effectiveness of the proposed optimal model predictive control (optimized model predictive control,OMPC) and the GSOMPC strategy.
【作者单位】：西安交通大学电气工程学院;
【基金】：国家863高技术基金项目(2012AA050206)~~
【分类号】：TM46

【参考文献】