Cognitive Digital Twins of the natural environment: Framework and application

Abstract

Digital Twin (DT) technology offers a method of creating digital models of natural systems to enhance their ability to withstand natural disasters. Currently, DT of the natural environment is in its initial phases, lacking adaptive capabilities and relying on human-assisted modeling. The key to endowing DT of the natural environment with greater autonomy lies in the integration of expert knowledge. Knowledge graphs can efficiently arrange and structurally store expert knowledge, thereby supporting the autonomous functionality of DT. This paper introduces the concept of Cognitive Digital Twin(CDT) derived from the industrial domain and presents a framework for CDT of the natural environment. This framework is centered around knowledge graph technology, aiming to provide more insights and guidance for system development. This framework integrates human cognition by constructing knowledge graphs of objects, models, events, and scene modes. Moreover, these knowledge graphs support agents for the dynamic adjustment of processes, as well as the adaptation and parameter optimization of related models. As a use case, we utilize this framework to implement digital twin watersheds. We develop appropriate ontologies and agents to facilitate the construction of cognitive digital watersheds for various regions. Cognitive digital watersheds effectively fulfill the application needs of integrated flood forecasting and control scheduling. This application validates the framework’s effectiveness and provides a reference for constructing CDTs of other natural systems.