Methodology for characterization of representativeness uncertainty in performance indicator measurements of thermal and nuclear power plants

Abstract

In this thesis, a general Methodology framework to characterize, assess and quantify the representativeness uncertainty in performance indicator measurements in thermal and nuclear plants is presented. The representativeness uncertainty arises from the inherent heterogeneity or the variability of the quantity being measured or from the inadequacy of the physical models used to simulate the measurement. The main objective of the Methodology is to gain a deeper physical understanding of the Representativeness uncertainty of the measurement by using numerical simulation tools such as Computational Fluid Dynamics (CFD) and to quantify various sources of representativeness uncertainty. First, the components of the Methodology are expressed using the normal probability distribution for the uncertainty sources. Second, a non-parametric formulation of the Methodology framework is developed and demonstrated. The use of non-parametric techniques allows the quantification and integration of uncertainties that are not expressed by the normal probability distribution. The Methodology is developed based on the analysis of four industrial Case Studies involving uncertainties in performance indicator measurements to structure the analysis. They are: Mass flow rate measurement by an orifice plate (Case Study 1), Steam Generator recirculation ratio measurement using chemical tracers (Case Study 2), The simulation of cooling tower deformation using a Photomodeler (Case Study 3) and the NOx emission measurement from a Combined Cycle Gas Turbine (Case Study 4). In Case Study 1, the non-parametric bootstrap method was used to quantify sampling, iterative and discretization uncertainties thus demonstrating its applicability to CFD uncertainty analysis. In Case Studies 2,3 and 4, the parametric formulation of the Methodology is used to structure the technical analysis.