Development of an Adaptive Low-Voltage Control Algorithm for Rural Power Grids Based on Deep Reinforcement Learning

Against the background where the high penetration of distributed photovoltaic (PV) integration poses severe challenges to the voltage stability of rural power grids, this study aims to develop an intelligent voltage control strategy capable of autonomously adapting to the dynamic changes of the power grid. To this end, this paper proposes an adaptive voltage control algorithm based on Deep Reinforcement Learning (DRL) and constructs a multi-device coordinated control framework involving On-Load Tap Changers (OLTC), Static Var Generators (SVG), and distributed PV inverters. By designing a multi-dimensional state space representation that integrates real-time grid operation information and a multi-objective composite reward function, the mapping relationship between the grid operating environment and the control strategy is established. Furthermore, an online experience replay mechanism is innovatively introduced to achieve the dynamic updating of the control strategy.