A framework for probabilistic risk assessment of nuclear power plants coupled with data centers using causal analysis

The rapid growth of Artificial Intelligence (AI) is driving an unprecedented increase in electricity demand from Data Centers (DCs). Nuclear Power Plants (NPPs) are increasingly being considered as reliable low-carbon energy sources capable of supplying the required baseload power. However, the coupling introduces new operational interdependencies that may affect the overall plant safety and reliability, which are not captured by traditional Probabilistic Risk Assessment (PRA) frameworks. This work proposes a Bayesian Network (BN)-based framework for the risk assessment of NPPs coupled with DCs. The BN structure explicitly captures the causal relationships from component failures and cyber-attacks to initiating events and their progression to core damage, consistent with standard PRA logic. The BN parameters are estimated through both literature-informed engineering judgment and data-driven learning from data generated with a system simulator integrated within a Monte Carlo Simulation, while epistemic uncertainty is propagated with probabilistic parameter modeling. The framework enables the estimation of the Core Damage Frequency (CDF) increase due to coupling, identification of the most impactful failures and cyber-attacks, and extraction of the most probable causal paths that lead to core damage. Application to multiple coupling configurations involving Pressurized Water Reactors (PWRs) and Small Modular Reactors (SMRs) demonstrates the capability of the approach to provide both quantitative risk estimates and interpretable insights, supporting risk-informed design and operation of integrated NPP-DC systems.