Abstract: Focused on the poor economy and weak peak shaving performance of traditional energy management strategies,an energy management strategy for energy storage system（ESS）based on distributed generation is proposed. Firstly,a demand-side optimal scheduling model aiming at energy cost diminution and load peak valley effect smooth is established combined with Pareto optimization theory. To obtain the optimal economic indicators,an improved particle swarm algorithm（IPSO）with adaptive variation and dynamic updating of inertia weights is applied during the first optimization process. Then,the scheduling interval is divided by the indicator as a constraint to achieve the complete decoupling of the target,while a penalty function is introduced to acquire an augmented objective function. Finally, the greedy algorithm（GA）is used for secondary optimiation,and the non-dominated scheduling vector which can smooth the load peak-valley effect to the greatest extent under the premise of the best economic benefit can be obtained. The simulation results show that IPSO strategy with high convergence speed and accuracy,can reduce the electricity cost by 18.6%,32.8% and17.3%,and gets 21.2%,10.3% has and 20.1% load peak-to-average rate reduction compared with the first optimization in three scenarios,respectively. The generality and reasonableness of the conclusions verified by the typical data set.