With the increasing proportion of distributed photovoltaic access to the distribution network, the stochasticity and uncertainty of photovoltaic power generation output to the distribution network brings voltage fluctuations and other adverse effects are becoming more and more obvious. Therefore, this paper proposed a variable time scale optimization active-reactive co-optimization of distribution network with distributed PV access to achieve the purpose of reducing voltage fluctuation and network loss. The optimization is divided into two phases: day-ahead and real-time. The day-ahead stage is based on the improved multi-objective particle swarm algorithm to derive the scheduling results of the next devices. The real-time phase uses the device scheduling results of the day-ahead phase as the reference trajectory to derive the device scheduling results as well as the optimized voltage and network loss changes. Extensive simulations and comparisons are performed at the IEEE33 node test system. The results show that the model improves the voltage stability of the distribution network containing PV access while reducing the operating costs. The effectiveness of the proposed strategy is verified.