Optimizing RL Training Horizons for Renewable Grid Dispatch

This coursework explores the application of reinforcement learning (RL) in optimizing the dispatch of thermal power plants within stochastic grids impacted by variable renewable energy (VRE). The study models the grid as a finite Markov Decision Process (MDP), defining state and action spaces to focus on cost-revenue optimization. By simulating diverse dispatch portfolios and examining episode horizons, it identifies how horizon length influences agent performance. Results reveal monthly...