Just as the Web is shifting its focus from information and communication and emphasizes the need to reuse of knowledge as a huge distributed knowledge base, the Semantic Grid in which information and services are given well-defined meaning extends the current Grid to enable software agents, users, and programs to work in cooperation. How to process the different sources of information, in particular, to combine a variety of sources of knowledge to assist users with effective reasoning or collaborative problem-solving, is one of great barrier in realizing the above vision. Against this background, we propose a knowledge fusion and integration approach based on statistical decision theory and Bayesian analysis for the Semantic Grid.