Online stability analysis of power system networks with renewable source of energy

  • Chloe Jones

Student thesis: Phd


The penetration of renewable energy sources is ever higher in large power grids. Weather- driven renewable generators are often far from the load and fail to provide physical inertia. Due to the imbalance of supply and demand and limited transfer of energy across a weakly interconnected network, inter-area oscillations are common. Poorly damped inter-area oscillations can breech protection thresholds and could violate the critical voltage in modern systems, which are operated close to maximum power transfer. Traditional models for stability analysis of synchronous generators can use very high computational power and fail to include effects of non-synchronous renewable generation. In recent years, some transmission system operators have introduced data-driven techniques for identification of low frequencies and damping. There is still a need, however, for data-driven techniques which are robust to noise, to identify low frequencies, damping and mode shape. Data-driven techniques must identify unfamiliar modal behaviour in systems with reduced inertia and variable renewable generation. It is proposed that variants of a data-driven technique, the dynamic mode decomposition, can meet these new requirements for stability analysis. Variants of higher order dynamic mode decomposition (HODMD), are applied in this project for short-term prediction and stability analysis of power systems with renewable source of energy. Real historical recordings of power systems during disturbance events are used for modal analysis. Noise-robust variants of HODMD are used to extract and to identify physically relevant dynamics, suitable for prediction. Due to the noise, the optimal sampling window is unknown and so multiple sampling time frames are used for comparison and robustness. Noisy and insignificant modes can over-fit reconstruction of the sampling window but fail to be suitable for short-term prediction. Spurious, physically irrelevant frequencies and components of measured data are therefore discarded. An energy criterion, for selection of the HODMD modes, is applied in this work for prediction of power system measurements, such as frequency and current. Higher-order and noise-robust variants of dynamic mode decomposition are calibrated and are compared for stability analysis and short-term prediction of power system recordings. This thesis proposes that variants of HODMD can identify poorly-damped oscillations, using noisy measurements of power systems. The variants of HODMD applied in this project are therefore suitable for power systems with renewable sources of energy. It is proposed that noise-robust HODMD can be used for short-term prediction of power systems, close to real-time, especially to monitor low frequency oscillations which could lead to blackout.
Date of Award31 Dec 2023
Original languageEnglish
Awarding Institution
  • The University of Manchester
SupervisorSergey Utyuzhnikov (Supervisor)


  • Kalman filtering dynamic mode decomposition
  • total-least-squares higher-order dynamic mode decomposition (THDMD)
  • Extended Kalman filtering dynamic mode decomposition
  • noise-robust
  • Higher order dynamic mode decomposition (HODMD)
  • modal analysis
  • oscillations
  • power system stability analysis
  • data-driven

Cite this