As the basis of power dispatching, load forecasting requires fast response speed and high accuracy.A Support Vactor Regression (SVR) algorithm based on RBF kernel function was proposed for data Regression and simulation, and Particle Swarm Optimization (PSO) was used for evolution of the penalty factor C and kernel gamma of SVR.The load data of a light industry enterprise in an industrial park in Xi 'an is simulated and analyzed. The experimental results show that the proposed algorithm has better accuracy in short-term load prediction than the traditional prediction algorithm.