The uncertainty of load in integrated energy system (IES) and the different time scales of multi-energy response load fluctuation bring challenges to the reliability of optimization results. In order to overcome these challenges and increase the renewable energy consumption of the IES. Taking rural integrated energy system (RIES) as the research object, this paper proposed a two-stage optimization model with shared energy storage (SES) on multiple-time scales, and uses particle swarm optimization (PSO) algorithm to solve it. Furthermore, in order to accurately predict power variations of the source/load, this paper employed the long short-term memory (LSTM) method, which can fulfill the requirement for the source/load prediction. A real RIES case analysis from Shandong Province was conducted, demonstrating that the proposed method held economic and environmental significance. Moreover, this paper provided valuable guidance for future IES research endeavors.