Adaptive governance is a new choice to solve the problem of water resource' governance which caused by the increasing complexity and uncertainty. It offers a great feasible approach with the learning capacity to settle issues caused by climate change, etc. This paper proposed the experience-weighted attraction (EWA) transformational learning based on the EWA learning theory, social learning theory and behavioral game theory. The EWA transformational learning has characteristics of multi-level and triple loop characters to promote the knowledge sharing and communication between individuals. Meanwhile, the paper analyzed the factors which may affect the human behavior pattern by system dynamics. Then, the paper modeled the dynamic system of water transfer in Hami region to analyze the law of the human behavior pattern, identify the key factors which affected human decision-making, and offer the plan for policy adjustment. The outcome of simulation proved that the learning rule of adaptive governance could improve the effective and the continuity of water transfer.