Extended study of Gaussian process applications in smart grids

Abstract

Smart Grids become the next generation of environmentally beneficial and long-lasting electrical infrastructure. Forecasting grid demand is becoming more challenging as renewable energy and electric vehicles become more prevalent, and accuracy demands are rising. With the rapid advancement of artificial intelligence and big data technology in recent years, academic and industrial researchers have concentrated on how to apply these new technologies and theories to raise the level of intelligence in smart grid operation and administration. This research deduces and solves the classical power flow and offers a load forecasting framework based on a Gaussian Process, which is frequently utilized as a supervised learning tool in numerous deep learning applications. This work describes the Gaussian Process and its solution in two dimensions, as well as the content and use of GPML as a tool for the Gaussian process. Finally, the dissertation extends earlier work by investigating the influence of mean function and likelihood function on system performance in various combinations.