Multi-Objective Reinforcement Learning with Shielding for Control and Safety in Power Grids | Synapse