Uncertainty propagation in the simulation of a nuclear power plant operational transient

With the continuous deployment of various intermittent energy sources on the grid there is an increasing opportunity for nuclear reactors to perform power maneuvers which can help alleviate issues related to supply oscillations during some periods of high renewable output. While utility specific codes can model such transients for safety assessments, there is a need to increase confidence in the application of these codes and to quantify the capabilities and accuracies of models in predicting these transients. In this work a power maneuver from 100 % full power to 59 % full power for a large CANDU reactor was studied using the TRACE-PARCS toolset. The work was computationally expensive as it examined both the effects of i) nuclear data library uncertainty which requires re-generation of the few-group cross sections and regeneration of an equilibrium core using 4 years of synthetic fuelling, and ii) initial core burn-up distribution involving the simulation of online fuelling operations over a period of 1400 days. The figures of merit considered in this work include changes in the average liquid zone controller fill level, individual channel powers, and the timing of adjuster bank withdrawals and reinsertions. Amongst the 100 s of simulated transients, the average time between adjuster bank withdrawals to override the Xenon transient and absolute standard deviations in these timings were found to be consistent between different nuclear datasets. The observed standard deviations were small, suggesting that the sensitivity to both nuclear data and initial core configuration is relatively small for both the adjuster withdrawal phase and the reinsertion phase. The contributions to total uncertainty due to both the core initial conditions and the nuclear data uncertainty were determined. During the early phase of the transient (i.e., adjuster withdrawals) nuclear data uncertainty and initial condition uncertainties were found to contribute nearly equally to total uncertainty, with slightly larger contributions due to nuclear data uncertainty. For adjuster bank reinsertions later in the transient it was observed that the contribution to total uncertainty by initial condition uncertainty was on average twice that of nuclear data uncertainties. The transients were compared to an actual power maneuver in a ∼900 MW CANDU reactor. In all cases, the model’s behaviour, while qualitatively correct, systematically over-predicted the time durations between adjuster movements and this difference exceeded the envelope of simulations that account for nuclear data and initial core configuration. Sensitivity studies performed to investigate the discrepancies suggest that additional refinements to the RRS emulator used in this work are required to match the station RRS more closely and that the transient evolution is highly sensitive to the initial excess reactivity in the core, which was observed to be lower in the station than in the TRACE simulations. Based on the simulation results the combined impact of the emulator responses, initial average zone levels, and incrementals modelling have a larger uncertainty than those related to nuclear data and initial core burnup distribution.