Research On Modeling Methods For Real-time Risk Monitoring Analysis At NPPs

The nuclear power plant(NPP)is a huge complex system,which can be operated in a variety of conditions.The possible changes of system configuration and component state are inevitable during the life cycle of nuclear power plants.It is one of the keys to ensure the operation safety of NPP to timely and accurately evaluate the impact of these changes on the risk level of NPP,identify important risk contributors,and realize the correct operation conditions.Risk Monitor(RM)is an effective means of achieving the purpose that is to timely give the risk reflecting the current plant configuration in terms of the known status of the various systems and components,which is used for the risk-informed operation and maintenance(O&M)decision-making process,and play an important role in ensuring the operation safety and improving the economy of NPP.Existing risk monitoring models are developed mainly based on stage 3 Living Probabilistic Safety Assessment(Living PSA),which still have some deficiencies:a)RM models have a burdensome scale and a poor flexibility of automatic updating;b)assuming that all component life are follow the exponential distribution and ignoring the impact of the component cumulative running time on its failure probability,which results in the underestimated risk value;c)ignoring some dependencies in the system,which results the potential deviation or even error in the calculation results of the risk monitoring model.While,"time dependency" is introduced in the real-time risk monitoring model based on stage 4 Living PSA.The risk result deviation caused by the above problems will be more and more significant over time,the accuracy and conservatism of the results cannot be ensured,and some analytical models will no longer be applicable to time dependent phenomenons.Therefore,the real-time risk monitoring models and analysis methods of NPPs will be further studied in this dissertation to improve the accuracy and credibility of the results.The introduction of the "time dependency" for the real-time risk monitoring model faces two challenges:①the construction of time-depended real-time risk monitoring model;②the unavailability of the system dependencies analysis models/methods in the risk monitoring application due to the introduction of "time dependency".The purpose of this dissertation is to solve some key technical difficulties involved in the above two parts.The detailed works completed in this study are listed as follows,in which works(1),(2)correspond to the challenge ①and works(3)-(5)correspond to the challenge ②.(1)A flexible automatic update mechanism for real-time risk monitoring model based on component modular fault tree(CMFT)is proposed,which solve the problem including massive scale and poor automatic update ability of the existing risk monitoring model.And in comparing with the existing risk monitoring modeling method,the proposed method can effectively reduce the model size,and the higher the system redundancy,the more significant the advantages.(2)The time-dependent phenomenons in real-time risk monitoring evaluation are systematically analyzed,and the time factors involved in real-time risk monitoring models are fully understood.And the component time-dependent unavailability models are developed to introduce "time-dependency" into the bottom event probability of CMFT.Moreover,the existing average models only for exponential distributions are extended to the time-dependent models for a variety of common component life distributions.By comparison with the existing models/methods,it shows that the component time-dependent unavailability models established in this study are more general,which improve the consistency with the current status of the system/component and accuracy of the real-time risk results.(3)The state dependency and sequential dependency due to dynamic behaviors of reserve systems are analyzed,then the analytical models for system dynamic unavailability considering the above dependencies are developed.In contrast with the models that simplifies these dependencies,the proposed models are not limited to the exponential distribution and are suitable for warm/thermal/cold standby reserve systems,which verifies the effectiveness and universality of the model.And an implicit analysis method of the extended sequential logic model based on the CMFT is proposed to avoid adding new dynamic logic gates and reduce a large number of model structure increase and modification work.(4)The time-dependent common cause failure(CCF)analytical models are proposed,and the CCF models under the exponential distribution are extended to that under the weibull distribution;subsequently,the time-dependent CCF models and analysis methods under the case of "asymmetry" are created to solve the invalidation of the classical CCF analysis models due to the introduction of "time dependency".And,the update rules for the CCF models mapped to the change process of system/component status,in which the "time dependency"and "asymmetry" are comprehensively taken account,are studied.The effectiveness and necessity of the proposed methods are demonstrated by a series of comparative studies with existing treatments.(5)For the first time,the multi-time period concept in real-time risk monitoring analysis is presented,and the multi-time period analytical models that can describe the hybrid effects from multiple dependencies including sate dependencies,sequential dependencies and CCF.The update methods of the multi-time period models with hybridized dependencies are developed to extend the model and update methods describing only a single dependency.Finally,the comprehensive application cases compared with the existing models and methods are studies based on the modelling and update methods for the real-time risk monitoring analysis proposed in the previous chapters,which show the methods proposed in this dissertation play an important role in improving the accuracy and credibility of the real-time risk monitoring results of NPPs and demonstrate the effectiveness and necessity of the proposed methods.