Monte Carlo Simulations to Evaluate Error Propagation in Computation of Thermal Power

Data reconciliation is a commonly used technique for correcting random errors in measurement data in the process industry. The technique uses models describing the mutual relationships of process variables related to available measurements. These models are based on knowledge of process physics. Measurement readings are adjusted so that especially mass and energy balances described by the model match. The technique has proven effective in reducing measurement uncertainties. The paper presents a Monte Carlo study of error propagation in data reconciliation of the turbine section of a VVER 440 nuclear power plant. Uncertainties in model parameters describing turbine dry efficiencies and the quality of steam exiting the steam generators are considered in addition to measurement noise. The impact of these factors on estimated reactor thermal power is evaluated, both individually and as joint impacts. For both the measurement signals and the plant parameters, the resulting effect on the uncertainty of thermal power is lower than the 2% uncertainty for reasonable levels of added noise. These results support the use of data reconciliation for reducing the uncertainty in thermal power.