Title: Compensation for periodic disturbances beyond control bandwidth: A phase-compensated machine learning approach

Abstract: Control of periodic disturbances is a classic, long-lasting issue in control engineering. Up to date, the fundamental principle for this issue is the celebrated internal model principle. However, the phase-lag inherent in internal model hinders the enhancement of control bandwidth, making it extremely involved and difficult to strike a balance in control design. This issue is even more severe in the face of model uncertainty. There is a consensus that it is impossible to cancel a periodic disturbance when its frequency is higher than the system bandwidth. But in reality, such as power electronics and high-speed trains, there is a quite high need to suppress disturbances beyond bandwidth.

 This talk will outline the recent advances in this direction made by the authors. The key is to introduce a phase-lead compensation mechanism into the machine learning algorithm. This algorithm is extremely robust to model uncertainty and its design is almost independent of the feedback loop. Applications to high-speed trains and inverters will also be touched on.

Kang-Zhi Liu graduated from Northwestern Polytechnical University in 1984 and obtained a Ph.D. from Chiba University in 1991. Since then, he joined Chiba University and is now a full professor at the Department of Electrical and Electronic Engineering. His research interests include robust control, machine learning and their applications to industrial systems. Dr. Liu was awarded four academic awards by SICE and is a Fellow of SICE.

标题：控制带宽外周期性扰动的补偿方法：一种相位补偿机器学习方案

摘要

作者简介