Parametric Analysis and Uncertainty Quantification for Flow Assurance

Abstract

Abstract Currently, decisions involving a huge amount of money and resources are based on the outcome of simulations. Consequently, an increase in the computational resources has attracted engineers and scientists to increase the complexity of the mathematical models and their numerical algorithms to design or analyze various systems. On the other hand, the transparency of these models has decreased and the sensitivity of the outcome of these simulations on, for instant, certain assumptions or input parameters is often not known a priori. Furthermore, validation of the models for complex systems in which full scale measurements are limited or even impossible, gives more challenges to the decisions made based on these numerical simulations. Additionally, robust design to consider off-design conditions in a systematic manner is of great importance in oil and gas industry. Therefore, it would be essential to develop a systematic procedure to quantify the sensitivities and errors associated with each set of simulations in an objective manner. Uncertainty quantification (UQ) could be used as a tool to provide the validity of a performed simulation. This study describes the current research and opportunities for parametric analysis and uncertainty quantification in the flow assurance such as in performance prediction and validation, risk assessment, decision making and robust design. Additionally, we describe the importance and benefits of probabilistic analysis instead of deterministic analysis. Uncertainty quantification increases the computational expenses and effort of performing deterministic simulations. Therefore, an efficient method to quantify the uncertainties such as probabilistic collocation or surrogate models are required to reduce the computational costs without losing the accuracy of the analysis. Parametric analysis and uncertainty quantification will give more insights in the validity of simulation data than traditional methods. For instance, using the newly proposed methods will expose the sensitivity of the simulation results to off-design conditions and consequently leads to a systematic decision making based on the probability of an event occurrence. These methods could be used to quantify the most sensitive parameters in the system and the effect of their propagation to the outcome of the model. It is shown that performing non-deterministic analysis in some cases is a necessity.