Virtual sensors for robust on-line monitoring (OLM) and Diagnostics
Unscheduled shutdown of nuclear power facilities for recalibration and replacement of faulty sensors can be expensive and disruptive to grid management. In this work, we present virtual (software) sensors that can replace a faulty physical sensor for a short duration thus allowing recalibration to be safely deferred to a later time. The virtual sensor model uses a Gaussian process model to process input data from redundant and other nearby sensors. Predicted data includes uncertainty bounds including spatial association uncertainty and measurement noise and error. Using data from an instrumented cooling water flow loop testbed, the virtual sensor model has predicted correct sensor measurements and the associated error corresponding to a faulty sensor.
- Research Organization:
- Pacific Northwest National Laboratory (PNNL), Richland, WA (US)
- Sponsoring Organization:
- USDOE
- DOE Contract Number:
- AC05-76RL01830
- OSTI ID:
- 1440688
- Report Number(s):
- PNNL-SA-124503; NT0107000
- Country of Publication:
- United States
- Language:
- English
Similar Records
Extending Sensor Calibration Intervals in Nuclear Power Plants
High Accuracy Signal Validation Framework for Sensor Calibration Assessment in NPPs