Abstract: The decentralized and random characteristics of high-proportion distributed photovoltaics aggravate the uncertainty of power flow distribution, which makes the safe and economic operation of the distribution system face severe challenges. In this paper, a multi-time scale optimization scheduling method for active distribution network considering the three-dimensional characteristics of conservation voltage reduction and distributed photovoltaic prediction error is proposed. Firstly, the sensitivity relationship between load power and feeder voltage is analyzed, and a comprehensive load modeling method based on step-down energy-saving regulation method is proposed. Then, from the perspective of time-space-power, a three-dimensional analysis model of distributed photovoltaic prediction error is established to provide more accurate error analysis results for scheduling. Finally, based on the conservation voltage reduction method and the three-dimensional analysis model of prediction error, a multi-timescale optimal scheduling model of active distribution network that can effectively coordinate the slow adjustment resources and the rapid response resources is constructed. Finally, taking the improved IEEE 33-node system as an example for analysis, the case study shows that the proposed error modeling method has a good following effect on photovoltaic power, and the optimized scheduling strategy based on this model can reduce the influence of error randomness on scheduling decision-making and improve the safety and economy of distribution system operation.