핵심 개념
The Physics-informed Actor-Critic (PI-AC) algorithm can efficiently coordinate the provision of Virtual Inertia from renewable Inverter-based Resources in power distribution systems without requiring an accurate model of the grid.
초록
The paper presents the Physics-informed Actor-Critic (PI-AC) algorithm for the model-free coordination of Virtual Inertia (VI) provision from power distribution systems. The key highlights are:
The PI-AC integrates the physical behavior of the power system, represented by the swing equation, into the Actor-Critic (AC) reinforcement learning approach to achieve faster learning compared to the purely data-driven AC.
The PI-AC is able to coordinate the VI from multiple renewable Inverter-based Resources (IBRs) in the distribution system to provide inertial support to the transmission system, while considering economic and operational constraints.
The performance of the PI-AC is evaluated through case studies on the CIGRE 14-bus and IEEE 37-bus distribution grids under various grid conditions, such as different IBR penetration levels and load variations.
The results show that the PI-AC achieves better rewards and faster learning compared to the AC and the metaheuristic Genetic Algorithm (GA) approaches, especially in scenarios with high IBR penetration.
The physics-based regularization in the PI-AC loss function is more influential in high IBR scenarios, where the system behavior deviates more from the swing equation assumptions.
The PI-AC is considered a model-free approach, as the swing equation formulation is generic and does not require specific knowledge of the individual grid structure or parameters.
통계
The maximum frequency deviation is limited to ∆ωmax.
The maximum voltage deviation is limited to ∆Vmax.
인용구
"The vanishing inertia of synchronous generators in transmission systems requires the utilization of renewables for inertial support."
"To this end, this paper presents the Physics-informed Actor-Critic (PI-AC) algorithm for coordination of Virtual Inertia (VI) from renewable Inverter-based Resources (IBRs) in power distribution systems."
"The PI-AC is able to achieve better rewards and faster learning than the exclusively data-driven AC algorithm and the metaheuristic Genetic Algorithm (GA)."