Decentralized data-driven estimation of generator rotor speed and inertia constant based on adaptive unscented Kalman filter

Bendong Tan, Junbo Zhao, Vladimir Terzija, Yingchen Zhang

Research output: Contribution to journalArticlepeer-review

1 Citation (Scopus)


This paper proposes an online decentralized and data-driven method for synchronous generator rotor speed and inertia estimation. First, an analytical relationship between the generator terminal voltage/current phasors and its internal rotor speed is derived using the Thevenin equivalent. The parameters of the latter are obtained via a recursive estimator, which allows us to estimate the generator's internal rotor speed from voltage/current phasors without the need for any generator information. Second, we reformulate the equivalent swing equation into the Kalman filter form, which enables us to use a priori-estimated internal rotor speed along with measured real and reactive power injections to obtain an estimate of the inertia for each online synchronous generator. This is achieved through an adaptive unscented Kalman filter. We demonstrate that the use of terminal bus frequency to approximate the rotor speed for inertia estimation by existing methods leads to large estimation biases/errors. Numerical simulations carried out on the IEEE 39-bus and Texas 2000-bus systems reveal that our proposed method is consistently more accurate than existing methods. Moreover, our proposed method is insensitive to both the type and location of system disturbances, and it is robust to different levels of measurement noise.

Original languageEnglish
Article number107853
JournalInternational Journal of Electrical Power and Energy Systems
Publication statusPublished - May 2022


  • Bus frequency
  • Dynamic state estimation
  • Inertia estimation
  • Phasor measurement unit
  • Rotor speed
  • Thevenin equivalent
  • Unscented Kalman filter


Dive into the research topics of 'Decentralized data-driven estimation of generator rotor speed and inertia constant based on adaptive unscented Kalman filter'. Together they form a unique fingerprint.

Cite this