Enhancing Load Forecasting by Clustering in Distributed Microgrids Based Energy Internet Framework

The intermittent nature of distributed renewable en-ergy sources and varying patterns of end-user loads in microgrids necessitate the manufacturers to accommodate unforeseen and expected fluctuations in energy consumption and production. Lack of accurate load forecasting may result in ineffective harnessing and storage of renewable energy and complicates energy trading and dynamic pricing. Existing literature on load forecasting of microgrids is limited to single microgrids, and the possibility of inter-microgrid communication is not addressed sufficiently. This study explores the enhancement of load fore-casting in an Energy Internet (EI) framework among multiple interconnected microgrids. A novel approach is proposed which integrates k-means clustering with Support Vector Regression (SVR) to forecast the load in the EI. We also investigate the influence of the communication network of the EI in improving short-term load forecasting (STLF).