In dynamic and heterogeneous Grid environments, the majority of the scientific applications require to be expressed as complex workflows which assemble multiple Web Services to implement complex scientific tasks. In addition, most workflow management systems are bound to concrete services distributed in different physical domains or concrete environments. Therefore, scientists still need to discover resources manually and schedule the jobs directly onto the Grid, and it could not meet users' requirement of scheduling workflow with large scale services. In this paper, we provide a comprehensive QoS (Quality of Service) model to support the possibility that a service instance is capable of offering to satisfy the user's requirements. Further we propose a complex workflow scheduling approach using dynamic programming, which focus on how to select a global optimal path for workflow scheduling based on QoS, and addressing services binding problem. At last, we demonstrate the efficiency and accuracy of the proposed approach by simulating experiments in heterogeneous and dynamic grid environment.