A comparison between Machine learning algorithms for the application of micro-grids Energy management

Advanced micro-grids are envisioned to be a critical part of future smart grids due to their local intelligence and capability on hosting distributed energy resources. To control a community of micro-grids so that the energy is distributed in a proper manner and reliability in supplying critical loads is increased, a smart decision making should be applied to the system. By using machine learning algorithms, the system can make intelligent, precise, and fast decisions leading to achieving the above mentioned goals. In this paper, five machine learning algorithms are applied to a set of data related to a community of three micro-grids. The data set including test data and training data is obtained based on grid voltage, weather, state of charge (SOC) of the storage system in both micro-grid and it's neighboring unit, and time of the day. The modes of operation are achieved from rational decisions that brings about a wide range of energy distribution from available sources into the most preferable loads. Random Forests, Decision Tree, Logistic Regression, SVM, and Gradient Boosting are the five various algorithms used to manage the distribution of energy and will be compared.