Universities play an important role in the national innovation system, and the research on the university S&T innovation capability evaluation is of great significance. Kernel principal component analysis(KPCA) method is proposed for the evaluation of university S&T innovation capability. This method, in comparison with Principal Component Analysis (PCA) method, can solve nonlinear correlation problem of evaluation index in the analysis of the S&T data released in 2002 from 15 universities directly subordinate to Chinese Ministry of Education. In particular, the result shows that contribution of the first and second principle components are more concentrated by KPCA than by PCA, and KPCA has better evaluation performance than PCA.