Machine Learning Based Seamless Solar-Grid Transition and Voltage Regulation Using Power Electronics in Industrial DC Grids

The use of solar photovoltaic (PV) systems in industrial DC microgrids is expanding and is facilitating the sustainable use of energy, but poses serious operation problems because solar irradiance is stochastic. The instability of the DC bus voltage due to sudden changes in the PV output may result in the disruption of the sensitive industrial loads and lead to the instability of the system. In this paper, the authors introduce a predictive control system using machine learning to provide a smooth transition of solar grids to grids and control voltage in DC grids in industrial settings. The KNearest Neighbors (KNN) and Random Forest algorithms are used to predict short-term PV power fluctuations using realtime irradiance, temperature, and voltage measurements to train the algorithms. The utility grid is synchronized and linked together via a rectified DC interface in advance of the predicted power deficit to prevent a voltage dip before it is realized. With the revival of solar generation, a gradual decrease in grid contribution is achieved so that the maximum use of renewable energy is obtained without discontinuing the load. MATLAB/Simulink is used for system modeling and validation. The simulation findings demonstrate that the suggested predictive control technique significantly reduces switching transients, suppresses voltage fluctuations, enhances DC bus stability, and boosts overall power supply dependability. The suggested method provides a viable solution to smart energy control and synchronous coordination of the source in the future DC microgrid industrial applications.