Multi-Agent Reinforcement Learning for Active Voltage Control on Power Distribution Networks: Supplementary Material