Mobile　edge　computing　(MEC)　provides　high-quality　network　service　to　mobile　users.　In　a　MEC　network,　the　placement　of　edge　servers　not　only　affects　the　service　experience　of　users,　but　also　has　a　great　influence　on　the　energy　consumption　of　the　MEC　system.　In　this　paper,　we　study　the　edge　server　placement　problem　in　dynamic　MEC　scenarios,　to　minimize　the　energy　consumption　of　the　MEC　system　in　real　time,　so　as　to　achieve　the　goal　of　energy　saving　at　every　time.　We　propose　a　reinforcement　learning　(RL)　based　approach　called　ESDR.　Moreover,　in　order　to　make　RL　suitable　for　our　proposed　problem　and　enable　the　agent　to　explore　better　in　MEC　scenarios,　we　introduce　Double　Q-learning　and　delaying　policy　updates　to　improve　the　original　RL　framework.　Furthermore,　experiments　are　conducted　using　a　real-world　dataset.　The　results　show　that　ESDR　is　outstanding　in　terms　of　energy　saving　with　fewer　edge　servers　than　other　approaches　in　dynamic　MEC　scenarios.　©　2023　IEEE.