Under the trend of large-scale new energy grid-connected consumption, the random characteristics of power grid new power generation and system peaking margin need to be considered. Firstly, according to the non-parametric estimation theory, the stochastic characteristics of the new energy output are processed. Secondly, based on the deep peak-shaving process of the thermal power peak-shaving unit, the operating cost of the unit is analyzed and modeled. Then, a multi-objective optimal scheduling model considering the economic stability and peak shaving flexibility of the power grid is constructed, and the multi-objective power grid deep peaking operation optimization scheduling strategy considering the stochastic characteristics of the new energy output is proposed. Finally, based on the multi-objective solution method, the power system of 10 thermal power units and the actual grid operation data of a certain area are simulated and analyzed, and the effectiveness of the multi-objective scheduling strategy is analyzed. This work is supported by National Natural Science Foundation of China (No. 61633016) and Project of State Grid Anhui Electric Power Company Ltd. (Key Technologies Research of Corporation of New Energy Consumption and Coal-fired Unit Peak Regulation).