.png){kind=icon}
.png){kind=icon}
Return
Combining predictor neural network and extended state observer for robust FCS-MPC in power converters
.png){kind=icon}
PRE
AI
Z
L
Z
F
DOI：10.1016/j.epsr.2026.113011.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Finite control-set model predictive control is widely used in power converter applications. However, its performance deteriorates significantly under model uncertainties caused by parameter drift and unmodeled dynamics. To overcome this limitation, a novel robust predictive control framework is proposed, which is endowed with the merits of a predictor neural network (PNN) and an extended state observer (ESO). Specifically, a predictor neural network, which has a good potential to learn the dominant system dynamics online from real-time input-output data, is deployed to handle system uncertainties. Meanwhile, the extended state observer treats neural network approximation errors and external disturbances as a lumped disturbance, thereby estimating and compensating for it. The distinct advantage of this framework is that it ensures superior transient stability during the initial learning phase of the PNN in the absence of offline pre-training, and also enhances robustness against disturbances and steady-state accuracy through error compensation of ESO. Finally, the proposed solution is validated by simulation and experimentation on a three-level neutral-point-clamped inverter, demonstrating improved performance compared to conventional methods.
Keywords:
Model-free predictive control (MFPC)
Predictor neural network (PNN)
Extended state observer (ESO)
Online learning
Journal
IF:
4.2
Papers:
1.1W
Citations:
2.2W
Organization
Cited Papers