Power system operation and control relies heavily on models for decision making. Topology is a critical part of producing these models and maintaining up-to-date topologies of electrical distribution networks is a resource consuming and challenging task. This paper proposes a methodology and system architecture for inference of electrical topology using process and model data from IEC 61850 compliant substation automation devices. A system of autonomous intelligent agents communicating via an overlay network is proposed where agents are capable of communicating on the IEC 61850 station bus. An algorithm for topology inference using structured exchange and comparison of process and model information is developed. The capabilities of structured information exchange and interfacing of substation automation devices enables plug-and-play operation of the topology inference requiring minimal prior knowledge of electrical network structure. Decentralized topology inference forms the basis for future work in operation and management of active distribution networks.