Optimized reactive power compensation for enhanced power quality in grid-connected solar PV systems using Meta-Reinforced Graph-Based Multi-Task Learning (Meta-RG-MTL) approach

As the global transition toward sustainable energy holds, grid-connected solar photovoltaic (PV) systems have come to be of great use in strengthening the resilience and reliability of the power system. However, integrating those renewable energy sources into the grid creates enormous problems in reactive power compensation, crucial for voltage stabilization and power quality enhancement. The outline of this paper presents the Meta-Reinforced Graph-Based Multi-Task Learning (Meta-RG-MTL) framework, which is capable of resolving the issue of optimal reactive power compensation in grid-connected solar PV systems. The solution put forward integrates Graph Neural Networks (GNNs) for the purpose of simulating spatial–temporal relationships, Reinforcement Learning (RL) for the making of real-time adaptive control decisions and also for the provision of automatic feedback, and LSTM networks for the precise prediction of the conditions of the sun and the grid. This versatile learning model could continuously predict and consequently supply the needed reactive power, thus facilitating voltage stabilization and reduction of power losses. The test outcomes demonstrate the Meta-RG-MTL’s strength, adaptability, and superior performance in enhancing power quality and system reliability even under a wide range of varying environmental and operational conditions.