Computationally efficient direct model predictive control with long prediction horizons for medium voltage drives.

对于中压(MV)交流驱动，逆变器的开关损耗应该保持在较低的水平。为了达到这一目的，可以降低开关频率，但这将导致电机电流的总谐波失真(THD)不可接受的增加。另一种选择是实施模型预测控制(MPC)算法来解决上述矛盾，因为MPC能够处理可能相互冲突的多个控制目标。此外，MPC通常作为直接控制器实现，即控制和调制问题是在一个阶段处理的，因此这种方案的灵活性很大。因此，预测控制被证明能够有效地控制电机的电流(或电磁转矩和定子磁链)，同时保持较低的开关频率和电流的THD。然而，在许多预测控制问题中(如上所述)，需要较长的预测周期才能获得足够的闭环性能。因此，即使使用功能强大的现代微处理器，也很难满足计算要求。因此，开发具有长预测周期和适度计算复杂性的预测控制方案的挑战仍然存在。本项目的目标是开发一种用于MV驱动(由多电平逆变器和感应电机-IM组成)的具有长预测周期的直接预测控制策略。控制目标是多方面的，并且同等重要：定子电流应紧密跟踪其基准电流，其THD应较低，开关频率(或相当于开关损耗)应保持较小。然而，为了实时实施所设计的策略，所需的计算量应该很小，为了实现所有的控制目标，目标是将问题描述为一个整数最小二乘优化问题(或类似的优化问题)。通过这种方式，可以采用离散搜索方法(例如球面译码算法)来高效地解决问题。该方法的一个重要优点是优化问题的复杂性不依赖于驱动电机的逆变器的电压水平。这是因为球面译码算法在一个超球面内寻找最优解，而超球面的结构与逆变器的电平无关。这意味着MPC算法可以很容易地适应不同的(多电平)逆变器拓扑结构，而不需要额外的计算开销。本项目的最后一个重要任务是在一个缩小的原型(低压驱动系统)中实现该算法，以验证其有效性。该驱动器将由一个三电平中点钳位逆变器和一个IM组成。

For medium-voltage (MV) ac drives the inverter's switching losses should be kept low. To achieve this, the switching frequency could be decreased, but this would lead to an unacceptable increase in the total harmonic distortion (THD) of the motor's current.An alternative is to implement model predictive control (MPC) algorithms to resolve the aforementioned contradiction, since MPC is capable of handling multiple control objectives that may be conflicting. Moreover, MPC is very often implemented as a direct controller, i.e. the control and modulation problems are tackled in one stage, thus the flexibility of such schemes is great. Therefore, MPC has been proven to be effective forcontrolling the currents (or the electromagnetic torque and the stator flux) of the motor, and, at the same time, for keeping the switching frequency and the currents' THD low.Nonetheless, in many MPC problems (as the one above), a long prediction horizon is required for adequate closed-loop performance. Hence, the computational requirements are difficult to meet, even with modern microprocessors that are powerful. Therefore, the challenge to develop MPC schemes with long prediction horizons and a modest computational complexity remains.The goal of this project is to develop a direct MPC strategy with long prediction horizons for MV drives (consisting of a multilevel inverter and an induction motor - IM). The control objectives are multiple and of equivalent importance: the stator current should closely track its reference, its THD should be low, and the switching frequency (or equivalently the switching losses) are to be kept small. However, in order to implement the devised strategy in real time the required computational effort should be small.To achieve all the control objectives, the goal is to formulate the problem as an integer least squares optimization problem (or a similar one). In this way a discrete search approach could be adapted (such as the sphere decoding algorithm) to solve the problem in an efficient manner. However, issues such as the initial radius selection of the hypersphere, or the factorization of the optimization problem matrices should be takeninto account and effectively tackled.A very important advantage of the proposed method is that the complexity of the formulated optimization problem does not depend on the voltage levels of the inverter that drives the motor. This is because the sphere decoding algorithm searches for the optimal solution in a hypersphere, the structure of which is independent of the levels of the inverter. This implies that the MPC algorithm can be easily adapted for different (multilevel)inverter topologies without additional computational overhead.A final, important task of this project is to implement the algorithm in a scaled-down prototype (low voltage drive system) in order to validate its effectiveness. The drive will consist of a three level neutral-point clamped inverter and an IM.