Comprehensive empirical assessment of nuclear power risks

Abstract

The opposition in a number of countries to the inclusion of nuclear energy in a sustainable energy portfolio, in part due to the dread of what the “nuclear” word inspires, has limited quantitative scientific foundation of the real benefits and risks. This has been amplified by the lack of a sound operational risk estimate due to the scarcity of the relevant empirical data. Using what is by far the largest - recently constructed - open database on accident precursors, and using our in-house generic probabilistic safety assessment (PSA) models, we provide the first comprehensive statistical study of the operational risks in the civil nuclear sector. We quantify a Pareto distribution of precursor severities as well as a special runaway Dragon Kings regime for the largest events. With respect to risk assessment, our main finding is that risk is dominated by exogenous factors (95%). We calculate that, by focusing on these factors in new design concepts, the frequency of accidents of the Fukushima scale can be brought down to about one per 300 years of operation of the worldwide fleet. Our results also demonstrate the need for an international cooperation focused on the construction of full blockchains of the cascades of accident precursors.

Keywords: nuclear energy, probabilistic safety assessment, accident precursors, dragon kings, operational risks, open-source database

JEL Classification: D81, K32, Q4, C8