This project addresses the challenge of maintaining grid stability in photovoltaic (PV) systems during voltage disturbances caused by grid faults. Low Voltage Ride Through (LVRT) requirements ensure that PV inverters remain connected to the grid and support voltage recovery, but conventional controllers with fixed parameters often struggle to perform effectively under severe and dynamic voltage sag conditions. To overcome these limitations, a Deep Reinforcement Learning (DRL) based control strategy is proposed for grid-connected PV inverters. The controller continuously monitors real-time grid conditions such as voltage variations and current limits, and dynamically regulates active and reactive power output. Through learning-based decision making, the system adapts to different fault scenarios without relying on detailed mathematical models, improving flexibility and performance. Simulation studies conducted under various voltage sag levels demonstrate that the DRL-based controller enhances LVRT capability, reduces inverter stress, and accelerates post-fault voltage recovery. Overall, the proposed approach improves grid support, robustness, and reliability, making it a promising solution for smart grids with high renewable energy penetration.