A Decision-tree SVM classifier is applied to the state recognition of the running pump station based on statistical learning theory(SLT). SVM is a novel machine learning method based on SLT and powerful for the problems with small sample, nonlinear and high dimension. The data of pump station system tends to have higher dimension, and the data is dimensioned down by principal component analysis. The Decision-tree SVM classifier, trained with the sampling data from the above dealing process and forming an identification model, identifies the state of the pump station. The test results show that the proposed classifier has an excellent performance on correcting ratio and training speed.