Planning for Resilient Power Distribution Systems using Risk-Based Quantification and Q-Learning

Grid hardening is one of the most effective approaches to improving the power distribution systems' resilience against extreme events. Unfortunately, hardening and upgrading the entire system is prohibitively expensive, mostly to protect against high-impact low-probability (HILP) events. This paper adopts a reinforcement learning (RL) algorithm to effectively search for the optimal hardening strategy to enhance power distribution systems' resilience - the resilience is quantified using a risk-based metric for the loss of load probability. The proposed Q-learning algorithm identifies the sequential optimal actions for grid hardening that minimizes the Conditional Value at Risk (CVaR) for the loss of load for a given budget. A case study is presented using the IEEE 123- bus test feeder to demonstrate the proposed approach's effectiveness in optimally allocating the budget-limited resources in resilient distribution system planning.