Abstract: In this paper, we propose a safe deep reinforcement learning (SDRL) based method to solve the problem of optimal operation of distribution networks (OODN). We formulate OODN as a constrained ...