Unbalanced　voltages　are　one　of　the　voltage　quality　issues　affecting　customer　devices　in　distribution　systems.　Conventional　optimization　methods　are　time-consuming　to　mitigate　unbalanced　voltage　in　real　time　because　these　approaches　must　solve　each　scenario　after　observation.　Deep　reinforcement　learning　(DRL)　is　effectively　trained　offline　for　real-time　operations　that　overcome　the　time-consumption　problem　in　practical　implementation.　This　paper　proposes　a　safe　deep　reinforcement　learning　(SDRL)　based　distribution　system　operation　method　to　mitigate　unbalanced　voltage　for　real-time　operation　and　satisfy　operational　constraints.　The　proposed　SDRL　method　incorporates　a　learning　module　(LM)　and　a　constraint　module　(CM),　controlling　the　energy　storage　system　(ESS)　to　improve　voltage　balancing.　The　proposed　SDRL　method　is　compared　with　the　hybrid　optimization　(HO)　and　typical　DRL　models　regarding　time　consumption　and　voltage　unbalance　mitigation.　For　this　purpose,　the　models　operate　in　modified　IEEE-13　node　and　IEEE-123　node　test　feeders.　IEEE