Deterministic sampling for propagating epistemic and aleatory uncertainty in dynamic event tree analysis

Abstract

Dynamic Event Tree (DET) analysis allows for integrated deterministic and probabilistic safety assessment by coupling thermal-hydraulic system models with safety system and operator response models. It is a realistic but computationally challenging approach for risk quantification in a nuclear power plant. DET can also provide a two-loop nested framework to quantify uncertainty arising from aleatory and epistemic parameters of the risk assessment model. However, the propagation of uncertainties in a DET is a challenge, since the set of uncertain parameters is often very large and the computational cost of each run can be significant (e.g. prolonged station-blackout scenarios). In this case, the intensive calculation required to propagate epistemic and aleatory uncertainty in two-loop approaches with usual Monte Carlo sampling makes the DET computationally impractical for uncertainty quantification in many complex nuclear power plant transient applications. To overcome this computational burden, a sampling approach called Deterministic Sampling (DS) is adapted and evaluated in this work as a potentially more efficient alternative to Monte Carlo sampling. The application and performance of DS are first tested by quantifying the system failure probability for an illustrative problem, including the propagation of uncertainties. Subsequently, DS is applied to a DET analysis of a realistic nuclear power plant transient, namely, a Station Blackout with feed and bleed sequence. The impact of epistemic and aleatory uncertainty on the core damage frequency contribution from the accident sequence of Zion power plant is evaluated using discrete DET and deterministic sampling based DET approaches. The comparison and analysis of the results reveal that the DS-based approach is computationally efficient and practical.