AI-Enabled Predictive Maintenance Digital Twins for Advanced Nuclear Reactors
Advanced reactors must be designed to be financially competitive with fossil fuel power plants to gain a foothold in future energy markets. The GE Research team aims to reduce operations and maintenance (O&M) costs by moving from a time- to condition-based predictive maintenance framework, using GE Hitachi's BWRX-300 boiling water reactor as the reference design. GE will develop operational, health, and decision predictive maintenance digital twins (PMDTs) to enable continuous monitoring, early warning, diagnostics, and prognostics for the reactor systems. The team will develop a “Humble AI” framework—which defaults to a known safe operation mode when there is a situation the algorithms do not recognize—to ensure systematic handling of uncertainties, data and model assurance, and continuous learning for these twins. The project will use system-wide risk information for decisions on plant operations, system reconfiguration, and maintenance planning to optimize cost while maintaining safety margins. PMDT-driven continuous monitoring will enable optimized use of a centrally located crew across distributed AR fleets, further reducing overall staffing costs.
The program goal is to reduce fixed O&M costs from ~13 $/MWh in the current fleet to ~2 $/MWh in the advanced fleet. Benefits include: