Reinforcement learning and A* search for the unit commitment problem

Previous research has combined model-free reinforcement learning with model-based tree search methods to solve the unit commitment problem stochastic demand and renewables generation. This approach was limited shallow depths suffered from significant variability in run time across instances varying complexity. To mitigate these issues, we extend this methodology more advanced algorithms based on A* search. First, develop a problem-specific heuristic priority list apply Guided search, reducing by up 94% negligible impact operating costs. In addition, address issue employing novel anytime algorithm, IDA*, replacing fixed depth parameter budget constraint. We show that IDA* mitigates of previous guided enables further cost reductions 1%.