SVM and Simulated Annealing Algorithms were applied to grid resources prediction. In order to build an effective SVR model, the parameters must be selected with attention. Here, a simulated annealing algorithm-based SVR (SA-SVR) model has been developed to determine the optimal parameters of SVR. The performance of the hybrid model, the back-propagation neural network and traditional SVR model whose parameters are provided by trial-and-error procedure have been compared with benchmark data set. Experiments validate that SA-SVR model works better than the other two models.