Abstract: The Web service composition system with static configuration can not adapt to the failure-prone environment and the variable Quality-of-Service (QoS) of component services. Therefore, a dynamic configuration method of Web service composition, Service Selection Constraint Model, is presented in this study. The candidate component services with same functionality are organized as a service class. The functional dependency relationships between component services are reflected as the service selection constraints. The optimal configurations conforming to multi-objective QoS constraints, known as the Pareto optimal solutions, are searched by a special ant colony optimization algorithm for Web service (ACO4WS). The feasibility and soundness of the method are proved by simulation experiments and corresponding analysis. By using the presented method, not only the QoS of service composition system is greatly improved, but also the multiple functional and non-functional constraints are satisfied.