Risk Assessment of Spent Nuclear Fuel Facilities Considering Climate Change

Abstract

Natural hazards have the capability to affect technological installations, triggering multiple failures and putting the population and the surrounding environment at risk. Global climate change introduces an additional and not negligible element of uncertainty to the vulnerability quantification, threatening to intensify (both in terms of frequency and severity) the occurrence of extreme climate events. Sea level extremes and extreme coastal high waters are expected to change in the future as a result of both changes in atmospheric storminess and mean sea level rise, as well as extreme precipitation events. These trends clearly suggest a parallel increase in the risks affecting technological installations and the subsequent need for mitigation measures to enhance the reliability of existing systems and to improve the design standards of new facilities. In spite of this situation, the scientific research in this field lacks robust and reliable tools for this kind of assessment, often relying on the adoption of oversimplified models or strong assumptions, which affect the credibility of the results. The main purpose of this study is to provide a novel and general model for the evaluation of the risk of exposure of spent nuclear fuel stored in a facility subject to flood hazard, investigating the potential and limitations of Bayesian networks (BNs) in this field. The network aims to model the interaction *

Vulnerability of Hydropower Installations to Climate Change: Preliminary Study

“…Similarly, the methodology adopted integrates the use of Credal Networks (CNs), which can be regarded as a generalization of BNs able to include imprecise probabilities in the framework, with cutting-edge and robust SRMs. The choice of this particular methodology is justified by its large potential in the representation of the interaction between weather events and technological installations, as proved in former studies [19] [20]. Indeed the approach allows to embody the aleatory character of natural events as well as the epistemic uncertainty associated (in particular in the case of climate projections), through the use of probabilistic models, intervals or imprecise random variables.…”

Robust vulnerability analysis of nuclear facilities subject to external hazards

“…This limitation, as highlighted in previous studies (Tolo et al 2016), strongly straitens the ability of the method to model, with equal effectiveness, the information related to both natural hazards and technological installations.…”

“…The purpose of the model implemented in this work is to quantify the probability associated to several accident scenarios and failure events involving a fuel pond subject to the threat of flooding events, overcoming the limitations highlighted in a previous study and associated with the adoption of traditional BNs (Tolo et al 2016). The drawbacks of such an approach were mainly linked to the use of crisp probabilities, which cannot fully represent the aleatory and epistemic uncertainty affecting the variables (a crucial aspect for climate variables and projections); a further restriction is the impossibility to take into account correlation among nodes when causal models are not available for the graphical representation of the dependencies but this information is available only numerically (e.g., through the estimation of correlation factors from experimental data).…”

“…First, a brief description of the facility is proposed, followed by a detailed description of the natural-technological interaction and internal failure sections. For further details on the human error section please consult the previous study (Tolo et al 2016). Since each node of the network is designed to represent a specific event, if not differently specified, the terms defining each node and the related event are considered interchangeable from now on.…”

Bayesian networks with imprecise datasets: Application to oscillating water column

“…This concept was proposed by Judea Pearl reasoning is limited to the use of only crisp probabilities (Spiegelhalter 1987). This type of probabilities leads to discretization methods or hard assumptions, impoverishing the quality of the analysis (Tolo et al 2016a). In order to work with continuous probabilities that can take into account the uncertainty of the variables in the network and avoid discretization of the input information, Daniel Straub and Armen Der Kiureghian (Straub and Der Kiureghian 2010) proposed to enhance BNs with structural reliability methods since these techniques support the use of continuous random variables.…”