Deep deterministic policy gradient for adaptive power system stabilization and voltage regulation

Abstract

This paper focuses on the challenges posed by the time-variable and nonlinear nature of energy systems. In this systems set points fluctuate due to load variations, generator losses, and disturbances such as short circuit faults. Maintaining power system stability and effectively damping low-frequency oscillations are crucial objectives. To enhance damping and dynamic stability in modern systems power system stabilizers (PSSs) are typically employed. These facilities generate auxiliary control signals for the generator excitation system, diminishing oscillations. In this study, by involving a deep deterministic policy gradient (DDPG) based on a synergetic control approach to the excitation system of synchronous generators, a novel approach to achieving fast and adaptive stability improvement and voltage regulation is suggested. The presented technique incorporates an intelligent synergetic PSS that effectively stabilizes the voltage response of a single machine connected to an infinite bus power system. To leverage its adaptive capability in controller parameter adjustments, an auxiliary deep DDPG algorithm with an Actor-Critic architecture is designed. The suggested procedure offers a promising avenue for effectively handling electromechanical oscillations and achieving improved voltage regulation and transient stability in power systems.

Keywords

• Power system
• Deep learning
• Excitation control
• Deep deterministic policy gradient
• Voltage regulation
• Transient stability