Advanced machine learning applications to modern power systems

Abstract

Abstract With the development of smart grid techniques, big data in power systems makes it possible to provide new solutions to power system operations by artificial intelligence (AI). AI technologies have contributed in various power system control and optimization problems. In this chapter, two state-of-the-art AI subfields, namely, ensemble learning and deep learning, are introduced. To demonstrate their applications in power systems, two groups of ensemble learning, that is, competitive and cooperative ensemble learning, are developed to provide short-term wind forecasting in both deterministic and probabilistic manner. In addition, a deep learning gated recurrent network is proposed to solve a network reconfiguration problem. At last, an advanced deep learning configuration with a convolutional neural network and a long short-term memory network is developed to automatically detect smart building occupancy condition by using advanced metering infrastructure data. Three sets of case studies showed that the developed ensemble learning and deep learning methodologies outperformed benchmarks. The success of the three AI applications promotes the grid integration of renewable energy and smart buildings.