Advanced learning-based energy policy and management of dispatchable units in smart grids considering uncertainty effects

Abstract In this paper, a new machine learning based framework is developed for the energy policy and operation management of the smart grids, utilizing advanced support vector networks in the renewable smart grids (RSGs), considering storage unit, wind and tidal systems and dispatchable units. The proposed system first develops a support vector regression (SVR) for prediction of the tidal and wind units output power with high accuracy. In the second step, an energy policy system is devised which forces the system operator to support renewable sources by guaranteeing the full purchase of their generation. In the third step, the optimal energy management framework is launched which optimizes the operation costs when considering the practical constraints. In the proposed novel framework, a new optimization method based on fuzzy dragonfly algorithm (FDA) is developed to enhance the search performance by creating adjusting fuzzy version this algorithm. In order to handle the uncertainty effects, a reduced scenario based approach is developed which shows high accuracy of 95% confidence level but with trivial computational time. The system quality is assessed on a test RSG system. The results prove the contributing claims of the research, clearly.