The game of Go is an ancient board game which is considered by-far the most complex board game for computer software or artificial intelligence (AI) to solve. AlphaGo, developed by the Google DeepMind team, is the first AI application to defeat a human professional world champion. The three components in AlphaGo (Policy Network, Value Network, and Monte-Carlo Tree Search) have similarities to some classic complex power system problems. In the proposed project, several potential AI applications in electric power systems are categorized into two groups: game-based problems and search-based problems. Detailed analysis of AlphaGo-like algorithms will be investigated for game-based and search-based power system AI applications. This is particularly important under the ongoing paradigm change in power systems evidenced by increasing variable renewable generations and demand-side participations, which lead to a larger amount of data, more uncertain scenarios, and more players. Thus, the success of the proposed project can solve the emerging challenges and potentially change the operation and planning of the power grid.
The intellectual merit of the proposed work includes: (1) similarity comparison of the game of Go and power system problems; (2) detailed algorithm investigations for AlphaGo-like algorithms for game-based and search-based power system problems considering multistage, multiplayer, and multi-scenario studies to address emerging challenges such as high-penetration renewables and demand-side participations; and (3) an open-source software package to implement the proposed work.
The broader impacts lie in the opening of a new door in AI for the area of energy science and engineering. For instance, it may provide new technologies and insights for integrated multi-energy systems involving many players from different energy sectors like gas, thermal, and electricity. From the educational perspective, the results of this project will be utilized to develop new teaching materials, to promote interdisciplinary collaboration in STEM areas, to recruit underrepresented minority and female students, to continue excellence in teaching, advising, and mentoring undergraduate and graduate students, and to develop a Web-based information center with social media to disseminate research results and the open-source software package for power system AI.
This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.
