A Machine Learning-Driven Framework for Voltage Disturbance Mitigation in Dual PV–Grid Integration

Modern power grids have undergone significant transformation due to the increasing integration of photovoltaic systems. As a result, significant challenges have arisen because of the intermittency of solar photovoltaic systems. Changes in solar irradiance and environmental factors can introduce voltage disturbances, power quality issues, and a reduction in the reliability of power grids. This is more noticeable in networks having multiple grid-connected photovoltaic systems. In this research, a machine learning-driven method is proposed for the mitigation of voltage disturbances in a dual PV–grid system. The leading objective is to increase the voltage stability under realistic operation scenarios of the dual PV–grid system. The test system is modeled in MATLAB/Simulink 2017b. The analysis of the test system is conducted comprehensively to assess performance under fault, varying temperature, and changing irradiance scenarios. Total harmonic distortion and reliability indices are evaluated to demonstrate the working conditions of the test system under intermittency and interruptions of solar photovoltaic systems. Moreover, the treatment strategy based on a battery energy storage system is proposed and applied to the test system. Not only is transient peak shaving up to 77.8202% achieved, but also power grid energy import is reduced by 95% under disturbances and interruptions of grid-connected solar photovoltaic systems. A machine learning algorithm is applied to classify the realistic conditions, thereby effectively employing the battery energy storage system to cope with intermittent renewable energy conditions. A 30% reduction in peak grid power under transient disturbances is obtained by accurately predicting these conditions.