Nuclear power plants must identify and analyze diverse transient events from long-term (often decades of) historical operating data to support unit life-extension and decommissioning assessments. Traditionally, this work relied on analysts manually labeling events against empirical rules, labor-intensive, slow, and prone to subjective misjudgment.
Our approach. We reformulated transient-determination rules as algorithmic logic and implemented them in MUUSE, our in-house Modelica-based graphical simulation and modeling platform. Leveraging Modelica's mature libraries, hierarchical modeling paradigm, and standardized encapsulation, we built, tested, and optimized the logic using standard and custom libraries, realizing identification, classification, key-event-information extraction, and alarming. To integrate with heterogeneous platforms, the logic was packaged as a standalone FMU for unified scheduling by the algorithm scheduling platform. Because the logic is authored in MUUSE on Windows yet must run on a Linux-based algorithm scheduling platform, the FMU is repackaged for the Linux host.
Takeaways. The deployment markedly improves transient-determination efficiency and consistency of results, demonstrating how Modelica and FMI enable industrial-scale operational-data mining. Cross-platform validation shows consistent trajectory agreement between the Windows-host and Linux-host FMU across representative transients, demonstrating that the open FMI sources mechanism enables reproducible, cross-platform deployment of Modelica models on domestic Linux stacks. The talk will present the modeling methodology, FMU encapsulation workflow, and lessons learned from deployment.
Keywords: Modelica; FMU; transient identification; operational data mining; nuclear power