.png){kind=icon}
.png){kind=icon}
Return
Robust Deep Reinforcement Learning Based Coordinated Wide-Area Damping Control for Multimode Inter-Area Oscillations With Randomly Delayed PMU Measurements
.png){kind=icon}
PRE
AI
Q
S
C
H
A
DOI：10.1109/tpwrs.2026.3670916.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Although the integration of phasor measurement units (PMUs) and wide-area transmission systems improves the system stability by transmitting the remote control signals to dampen inter-area low-frequency oscillations (LFOs), the reliance on communication networks for signal transmission induced time delays, significantly degrading wide-area damping control (WADC) performance. This paper proposes a coordinated WADC scheme to damp multi-mode LFOs. For LFO modes with small disturbances, this work designs a deep reinforcement learning (DRL) based WADC with a long short-term memory (LSTM) delay compensator. DRL agent adapts to the unknown power systems by designing an optimal stabilization voltage action set. For LFOs with large disturbances, this paper proposed robust DRL-based WADC to stabilize power systems with severe LFOs with online training for a continuous optimal action set. The minimax DRL algorithm optimizes the worst-case performance of the proposed policy and therefore provides robustness to uncertainties. The cyber layer metric (time delay) and the physical system metrics (phase angle and speed) are uniquely combined into the state set and provide sufficient cyber-physical information for the DRL agent. Extensive comparisons on two large IEEE systems under multimode LFOs demonstrate that the proposed coordinated WADCs can improve 20<inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$\%$</tex-math></inline-formula> to 35<inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$\%$</tex-math></inline-formula> of the stabilization effectiveness and save 50<inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$\%$</tex-math></inline-formula> of the training costs.
Keywords:
Wide-area damping control
uncertain time delays
deep reinforcement learning
minimax optimization algorithm
LSTM delay compensator
Journal
IF:
7.2
Papers:
1.1W
Citations:
5.0W
Organization
Cited Papers