AN ILLUSTRATIVE, INTERVIEW-BASED RISK FRAMEWORK FOR TREATMENT OF HIGH-STRESS HUMAN ACTIONS IN MULTIUNIT NUCLEAR POWER PLANT ACCIDENTS

摘要

Multiunit risk assessment has drawn growingattention after the 2011 Fukushima Accident. Based on ourreview of operation experiences, we’ve identified sharedsystems, shared human management, physical proximity ofunits and electricity disturbance as significant multiunitrisk contributors. In order to further understand multiunitdependencies in Fukushima Accident, we conductedinterviews with Tokyo Electric Power Company (TEPCO)engineers about their experiences with the Fukushimaaccident. The evidence from the interviews shows thataccident propagation and human related events were themost important risk contributors. In order to mitigate theaccident, the operators need to evaluate the urgency ofmultiple units, diagnose the working status of the units andplant and implement response strategies. All of these taskshave to be finished under high mental stress and highsituational uncertainty. Moreover, restored portableelectricity and cooling systems are vulnerable to accidentpropagation, as due to hydrogen explosion fromneighboring units. Given the time-dependent nature ofthese interactions, simply extending a single unitProbability Risk Assessment (PRA) to the whole site is notsufficient; new methods to quantify multiunit risk areneeded.In this paper, we provide an illustrative example of aninterview-based multiunit PRA illustrative simplifiedframework that addresses accident propagation andhuman-related events. A two-unit site with simplified safetysystems is used to illustrate the approach. In the accidentscenario of this two-unit site, all cooling methods exceptfor portable equipment are assumed to be unavailable forboth units right after the accident occurs. Operators needto restore portable cooling in order to protect the units.Meanwhile, the restoration process of portable cooling forone unit can be disrupted by debris from a hydrogenexplosion of the other unit. Except for physical interactionof two units due to propagation of hydrogen explosiondebris, human-related dependencies of units are modeledas well. Specifically, both units are assumed to experiencethe same restoration delay because of the sharedmanagement of the site. Meanwhile, onsite resources areassumed to be limited. The effects of resource distributionamong units on success probabilities are analyzed in thispaper. It turns out that there exists a resource distributionthat maximizes the probability of both units beingsuccessful. Results of this work provide a theoretical basisfor modeling and decision making in future severe accidentmitigation.