A decision theoretic approach to an accident sequence: when feedwater and auxiliary feedwater fail in a nuclear power plant

Abstract

This study applies a decision theoretic perspective on a severe accident management sequence in a processing industry. The sequence contains loss of feedwater and auxiliary feedwater in a boiling water nuclear reactor (BWR), which necessitates manual depressurization of the reactor pressure vessel to enable low pressure cooling of the core. The sequence is fast and is a major contributor to core damage in probabilistic risk analyses (PRAs) of this kind of plant. The management of the sequence also includes important, difficult and fast human decision making. The decision theoretic perspective, which is applied to a Swedish ABB-type reactor, stresses the roles played by uncertainties about plant state, consequences of different actions and goals during the management of a severe accident sequence. Based on a theoretical analysis and empirical simulator data the human error probabilities in the PRA for the plant are considered to be too small. Recommendations for how to improve safety are given and they include full automation of the sequence, improved operator training, and/or actions to assist the operators' decision making through reduction of uncertainties, for example, concerning water/steam level for sufficient cooling, time remaining before insufficient cooling level in the tank is reached and organizational cost-benefit evaluations of the events following a false alarm depressurization as well as the events following a successful depressurization at different points in time. Finally, it is pointed out that the approach exemplified in this study is applicable to any accident scenario which includes difficult human decision making with conflicting goals, uncertain information and with very serious consequences.