Service-oriented computing (SOC) is viewed as the computing paradigm of the near future, allowing for the dynamic interaction of services provided by distributed business partners. Being a declarative knowledge representation model, ontologies serve as a foundation for SOC. Due to the heterogeneous nature of independently designed ontologies, it is problematic for partners to understand the concepts adopted in ontologies from other sources. In order for partners to achieve seamless collaboration of services, they need to reconcile their ontologies with each other. During the alignment process and the following service interactions, compatibility is an important measurement that has been neglected in most research work. We extend a vector system to encode ontology compatibility. In addition, we present a new model - probabilistic center ontology - for better recording and maintenance of ontology alignment results. Our precise and efficient approach is verified by both theoretic proofs and experimental results.