Machine learning and big data analytics for the smart grid

Abstract

As numerous sensors, such as smart meters and PMUs, continue to be added to the grid, the emerging information collected is becoming a valuable source to researchers and grid operators who seek to conduct advanced analytics on the smart grid. This research combines the latest machine learning and big data analytics techniques with the domain knowledge of the smart grid to explore the added value of the emerging power system data. By exploiting the emerging smart grid database, we can develop data-driven solutions for the most pressing issues, such as load modeling, demand side management, and distributed energy resource hosting capacity analysis. This research first develops a methodology to apply data science technologies to smart grid applications. Then, it provides a set of examples to illustrate how the smart grid may benefit from the emerging data. These examples cover a broad range of smart grid analyses and applications, including residential photovoltaic system detection, electrical vehicle charging demand modeling, time-variant load modeling, and hosting capacity analysis. Different data analytics techniques are implemented in these examples, including clustering, statistical inference, change-point detection, parameter estimation, stochastic modeling, and statistical learning methods.