In the frequency response of the new power system, fused magnesium as a load side resource has great potential for frequency modulation. Due to the need to adapt to the medium and long term power market demand, the existing fused magnesium simulation methods which only take single ton energy consumption as the index are difficult to determine the declared capacity. Therefore, according to the requirement of annual time scale and the randomness of molten magnesium production process, a stochastic production simulation algorithm framework was designed to simulate the operation state of molten magnesium load. Based on the Markov chain Monte Carlo method and the idea of integer random distribution, the transition and duration of each state were simulated. According to the rule of two-step system, an optimization model for the declared capacity of fused magnesium was established, and the optimal declared capacity and enterprise income were solved by the ergodic optimization algorithm, so as to achieve the goal of frequency response on the load side. The simulation example shows that the proposed method can well characterize the production process of electrical fused magnesium load, calculate the optimal declared capacity for the scenario of power loss in the medium and long term market, which brings economic benefits to enterprises and provides fast frequency response resources for power grid.