Deep Learning-based Load Forecasting and Monitoring in the Smart Grid

Abstract

With the rapid advances in sensing and acquisition, transmission, storage, computing, and analytics, the era of big data has come. Many advanced data analytic techniques, especially machine learning and deep learning techniques, have been proposed and found wide applications in our society.

In the power industry, data analytics play an essential role in daily power system operation and planning. One major challenge for energy management in the emerging smart grid is the uncertainty in both power supply (e.g., renewable energy generation) and demand (e.g., load demand from the service area). There is a compelling need to accurately predict generation and load for efficient power management. Such predictions will help to make intelligent decisions for improving power quality, saving energy, better utilizing renewable energy sources, and reducing cost.

This thesis develops data-driven solutions by using the latest deep learning and machine learning technology, including ensemble learning, meta-learning, and transfer learning, for energy management system issues, such as short-term load forecasting and non-intrusive load monitoring problems. Real-world datasets are tested on proposed models compared with state-of-the-art schemes, which demonstrates the superior performance of the proposed model.