Nowadays, distributed photovoltaic (PV) is integrating into the distribution system with large-scale capacity. In order to solve the problems of voltage over-limit and PV generators hosting capacity, a dual time-scale voltage optimization control model for medium- and low-voltage distribution networks is proposed. In the day-ahead stage, the centralized optimization model for medium-voltage distribution networks is proposed, setting the tap positions of discrete on-load tap changers (OLTC) and the reactive power output of shunt capacitor banks (CBs) as decision variables. In the intraday optimization stage, a distributed optimization model for medium- and low-voltage distribution networks is constructed, which makes full use of the reactive power capacity of distributed PVs and the fast regulation capability of energy storage systems (ESSs). Finally, the proposed method is verified on the actual distribution system in Tangshan city, China.