Short-Term Generation Forecasting for Solar fed DC Microgrids Using Machine Learning with OPTUNA

Desert regions experience significant exposure to solar radiation which can be harnessed effectively for sustainable grid operations. DC microgrids can be developed with an interconnected network to not only supply energy demands but also to store power. This research introduces a machine learning-based framework for forecasting PV power generation to feed operations in isolated/non-isolated DC microgrids. For microgrid operators, these predictive capabilities offer considerable operational advantages, such as improved energy dispatch scheduling, decreased reliance on backup generation, and enhanced management of battery storage. A comparison of different machine learning algorithms tuned by OPTUNA is used in this paper and the results are compared on the basis of R2, MAE and RMSE scores. By accurately predicting fluctuations in PV output, operators can reduce expensive reserve margins while upholding reliability standards. Generation forecasting allows for precise load balancing, enables smart demand response strategies, and supports optimal economic operations by guiding energy trading decisions with the main grid.