ICONE23-1952 Co-Simulation for Real Time Safety Verification of Nuclear Power Plants

Abstract

Small and major accidents and near misses are still occurring in nuclear power plants (NPPs). Risk level has increased with the degradation of NPP equipment and instrumentations. In order to achieve NPP safety, it is important to continuously evaluate risk for all potential hazard and fault propagation scenarios and map protection layers to fault/failure/hazard propagation scenarios to be able to evaluate and verify safety level during NPP operation. There are major limitations in current real time safety verification tools, as it is mainly offline and with no integration to NPP simulation tools. The main goal of this research is to develop real time safety verification with co-simulation tool to be integrated with plant operation support systems. This includes the development of static and dynamic fault semantic network (FSN) to model all possible fault propagation scenarios and the interrelationships among associated process variables. Safety and protection layers along with their reliability are mapped to FSN so that safety levels can be verified during plant operation. Errors between multiphysics models and real time data are modeled to accurately and dynamically tune FSN for each fault propagation scenario. The detailed methodology will show how to integrate process models, construction of static FSN with fault propagation scenarios, and evaluation and tuning of dynamic FSN with probabilistic and process variable interaction values. Principle Component Analysis method is used reduce dimensionality and reduce process variables associated with each fault scenario. Then map independent protection layers (IPL) to FSN with estimated reliability measures of each protection layer to accurately verify safety for different operational scenarios. Intelligent algorithms is used with multivariate techniques to accurate define the interrelation among process variables, in terms of signal strength and time delay, using Genetic Programming (GP), which will provide basis for fault detection and tuning of FSN, as well as fault diagnosis to understand the closest state of fault scenario. Intelligent algorithm for Bayesian Believe Networks (BBN) is developed to estimate probabilities associated with dynamic FSN with priori and posteriori probabilities. This will dynamically tune FSN with probabilities and real time and simulation data. Probabilistic risk are estimated for each propagation scenario along with the reliabilities of associated IPLs. This will accurately verify safety for all propagation scenarios during plant operation and maintenance. And in order to fine tune propagation scenarios within FSN, rules are synthesized using fuzzy logic using real time and simulation data.