There are difficulties in integrating solar energy into home grids, such as frequency deviations, harmonics, and voltage swings. A 2.5 MW solar-integrated grid system that uses a machine learning method to handle these problems is presented in this research. In order to maximize solar inverter performance and stabilize the grid, the system uses previous data to analyze grid parameters in real time. Important metrics like frequency, voltage regulation, and total harmonic distortion (THD) are tracked, and remedial measures are taken to guarantee that power quality criteria are met. Results from simulations show that the machine learning method improves grid power quality by lowering harmonic distortion, increasing voltage stability, and guaranteeing frequency regulation in a range of scenarios. This work shows how machine learning may be used to develop robust and effective solar energy systems, which will help home grids adopt renewable energy sources more widely.