Logic-based data-driven operational risk model for augmented downhole petroleum production systems

Abstract

This paper presents an operational risk model for a pressure-augmented downhole petroleum production system. The model is built by integrating multilayer perceptron (MLP) and early warning index system (EWIS) with Bayesian network (BN). The introduced model employs its evidence-based dynamic risk features to monitor the associated operational risks of downhole pump discharge pressure, downhole pump intake pressure, downhole pump pressure difference, drawdown, and bottom-hole pressure. The evidence-based mechanism enables the proposed model to accurately predict the resultant real-time production risks as the wells are being produced from the reservoirs. Hence, the model facilitates management decisions to make operational adjustments to avert downtime or “no flow”. The model captures the temporal and spatial dependence of the variables. The failure probabilities of the downhole pressure system are modelled as a function of time while using the evidence-based risk model. The results demonstrate downhole process system's contribution to the overall risk and its vulnerability to the overall production scenarios. The proposed novel strategy can simulate the progressive cavity pump (PCP) impacts on reservoir systems during production. The introduced model serves an important tool for operational decision-making to manage risks of reservoir systems equipped with downhole pressure pumps for production.