In order to formulate the peak-valley time-of-use power price strategy which is beneficial to each participant in DSM, this paper not only analyses the cost and income change of each participant but also proposes a reasonable cost-benefit analysis model. On this basis, an optimized model of peak-valley time-of-use power price with ratio of peak to valley price and pull-apart rate as the decision variables is built. Next, the tolerant stratified sequence is used to simplify the original model from a multi-objective problem to a single objective problem whose satisfactory decision variables can be obtained by the genetic algorithm. Finally, the optimized peak-valley time-of-use power price strategy is formulated according to the obtained decision variables. The example analysis shows that the optimal strategy has a good performance on peak shaving and valley filling. Additionally, it also brings more profits for Power Grid Corp, customers, and all society, thereby improves the stability and economy of power system effectively. This work is supported by National Natural Science Foundation of China (No. 61601414).