Most of the existing multi-source fusion methods are to choose the most reliable information from the multi-source information system to form a single-source information system. Obviously, this process is accompanied by information loss. In order to solve this problem, the multi-granulation method of information fusion in multi-source decision information system is studied in this paper. Firstly, decision support characteristic function and decision related characteristic function are constructed. Secondly, a pair of aggregation operators, including fixed aggregation operator and possible aggregation operator, is defined through two characteristic functions. Meanwhile, the two cases when thresholds α and β take special values are discussed. Finally, the relevant properties of aggregation operators in different situations are proposed and proved. What is more, two groups of comparative experiments are carried out to illustrate the effect of the aggregation operators. The experimental results show that the proposed multi-source fusion method can always find a set of thresholds (α,β), which makes the fusion effect better than the mean fusion.