Offset-Free Direct Power Control of DFIG Under Continuous-Time Model Predictive Control

Rachid Errouissi, Ahmed Al-Durra, S. M. Muyeen, Siyu Leng, Frede Blaabjerg

Research output: Contribution to journalArticlepeer-review

83 Citations (Scopus)


This paper presents a robust continuous-time model predictive direct power control for doubly fed induction generator (DFIG). The proposed approach uses Taylor series expansion to predict the stator current in the synchronous reference frame over a finite time horizon. The predicted stator current is directly used to compute the required rotor voltage in order to minimize the difference between the actual stator currents and their references over the predictive time. However, as the proposed strategy is sensitive to parameter variations and external disturbances, a disturbance observer is embedded into the control loop to remove the steady-state error of the stator current. It turns out that the steady-state and the transient performances can be identified by simple design parameters. In this paper, the reference of the stator current is directly calculated from the desired stator active and reactive powers without encompassing the parameters of the machine itself. Hence, no extra power control loop is required in the control structure to ensure smooth operation of the DFIG. The feasibility of the proposed strategy is verified by the experimental results of the grid-connected DFIG and satisfactory performances are obtained.

Original languageEnglish
Article number7458905
Pages (from-to)2265-2277
Number of pages13
JournalIEEE Transactions on Power Electronics
Issue number3
Publication statusPublished - Mar 2017
Externally publishedYes


  • Continuous-time model predictive control (CTMPC)
  • direct power control (DPC)
  • disturbance observer
  • doubly fed induction generator (DFIG)
  • renewable energy

ASJC Scopus subject areas

  • Electrical and Electronic Engineering


Dive into the research topics of 'Offset-Free Direct Power Control of DFIG Under Continuous-Time Model Predictive Control'. Together they form a unique fingerprint.

Cite this