Real-time fault detection and transient identification using extended Kalman filters - 254
- Department of Mechanical Engineering and Materials Science University of Pittsburgh 3700 O'Hara Street, Pittsburgh, PA (United States)
In this paper, real-time fault detection and transient identification are studied with adaptive estimation methods, in particular parallel processing methods by implementing a bank of extended Kalman filters. Adaptive estimation is combined with probabilistic risk assessment to allow for real-time continuous monitoring of initiating events. The approach provides a probability of an initiating event happening. Real-time monitoring of initiating events can reduce labor demand. Each extended Kalman filter is configured to operate at a single configuration, either at normal operation or a faulted state. Fault detection is based on calculating a likelihood function for each extended Kalman filter to determine the posteriori conditional probability that the system configuration is the true configuration. The transient identification is determined by performing a multivariate statistical analysis on each extended Kalman filter to provide a posterior probability assessment that the reactor states and its dynamics are within the allowable limits. The probability distribution being outside the allowable limits indicate a transient has occurred. The combination of parallel processing methods with the multivariate statistical analysis for transient identification allows for the operator to perform transient identification with having a high level of certainty of the current system configuration. We demonstrate the real-time fault detection and transient identification on a simple reactor example to determine the reactor power, reactor temperature, and coolant temperature in the presence of various faults. (authors)
- Research Organization:
- American Nuclear Society - ANS, 555 North Kensington Avenue, La Grange Park, IL 60526 (United States)
- OSTI ID:
- 23035367
- Country of Publication:
- United States
- Language:
- English
Similar Records
A Multi-Model Adaptive Kalman Filtering Approach to Power System Dynamic State Estimation
Boiling water reactor stability analysis via Kalman filter-based state estimation and maximum A posteriori detection