In order to solve the influence of uncertainties in the operation process of demand response, the dispatching cost and energy consumption before demand response day are reduced. In this paper, the interruptible load is taken as the research object for the incentive demand response project in China. By calculating the user’s baseline load and forecasting the actual load, the load reduction can be obtained. An improved particle swarm optimization (PSO) algorithm based on cloud model is used to establish a mathematical model with the objective of minimizing the scheduling cost of demand response day before. Considering the influence of risk factors in the process of demand response, forward cloud generator is used to transform from uncertain space to specific space, which alleviates the influence of uncertain factors on demand response scheduling. At the same time, the improved particle swarm optimization algorithm of cloud model overcomes the disadvantage that the algorithm is easy to fall into local optimum, and achieves more accurate results. By comparing the average fitness values of all the particles in the current PSO, the PSO is divided into three sub-groups. The normal cloud model is used to dynamically adapt, adjust the inertia weight of the PSO, and then optimize the dayahead scheduling cost. Comparing with the PSO algorithm, the experimental results show that the proposed method can effectively reduce the scheduling cost before demand response day. Reasonable demand response scheduling strategy can effectively alleviate the contradiction between power supply and demand, save energy and improve the stability of smart grid operation.