This technology develops an evolutionary multi-agent deep reinforcement learning voltage regulation technique, combining population evolution and curriculum learning strategies to reframe voltage regulation as a Markov game. By designing an extensible agent architecture, it achieves adaptive collaborative regulation of distributed photovoltaic nodes and multi-resource systems, effectively resolving three-phase imbalance and voltage fluctuation issues. This method does not rely on precise grid models, significantly enhancing the operational stability and power quality of distribution networks with high photovoltaic penetration.

Technology provider:Nanjing University of Posts and Telecommunications