Artificial intelligence for resilience enhancement of power distribution systems

Abstract

The threat of high impact low probability (HILP) events on power distribution system is substantial but quite unpredictable. Enhancing the resilience of power distribution grids against such events requires solving combinatorial planning and operational problems in stochastic spaces, as well as classifying system conditions based on high-dimensional input data. Since traditional mathematical solutions struggle with both uncertainty and the curse of dimensionality, data-driven techniques based on artificial intelligence (AI) are gaining momentum for solving those problems. This paper reviews AI capabilities for decision making in uncertain and high-dimensional spaces in general, and their particular application in resilient enhancement problems such as damage detection and estimation, cyber-physical anomaly detection, stochastic operation, and cyber security enhancement. Efficient data structures and AI approaches are suggested for each problem, which depend on the type of input signals, search-based or game-based structure of the problem, as well as the uncertainty sources involved. In particular, potential applications of supervised and unsupervised deep learning combined with Monte Carlo Tree Search and ε-greedy search is explored to find near optimal operational decisions that help enhance the resilience of power distribution systems.