Development a partially observable Markov decision processes-based intelligent assistant for power grids using Monte Carlo tree search

Autonomous control systems will make much ”smarter” used automatic controls of modern power grids, as well as partially or completely replace the system operator, which may not be able sometimes to adequately respond to critical conditions due to psychological stress. Development of such systems can be solved by Monte-Carlo tree search algorithm that simulate ahead into the future, evaluate future states, and back-up those evaluations to the root of a search tree. We use the formalism of POMDPs (Partially Observable Markov Decision Processes) as the core of an intelligent assistant for power system control and dispatch. We demonstrate the feasibility of the approach to resolve the voltage and reactive power control in substation.