To directly reflect the mapping relation between controllable boundary parameters and heat rate of steam turbine, a regression model based on υSVM was established by taking the controllable boundary parameters with strong relevance with heat rate as the input parameters, which was subsequently verified using gray correlation degree model. The process of establishing υSVM regression model based on Libsvm software was described in detail, and the new model was compared with that of the BP neural network. Results show that under smallsample circumstances, the υSVM model has higher regression precision and better generalization capability; whereas under slight fluctuation of input parameters, the outputs of υSVM model are basically stable, indicating that the model has good robustness and can meet precision requirement of actual applications.