Optimizing product yield and quality of a local NGL Plant using Model Predictive Control

Abstract

Abstract The objective of any plant whether operating in the oil, gas or petrochemical industry is to add value in the process chain through converting raw materials into more valuable products while subject to many constraints including but not limited to raw material availability, product prices, safety and environmental constraints. To meet these constraints in an optimal manner, operators of these plants resort to control strategies and philosophies that would help them to better enhance production and meet quality constraints. A Natural Gas Liquid Plant (NGL) located in the Kingdom of Bahrain with its recently commissioned Train – III has a total combined processing capacity of 660 Million Standard Cubic Feet per day (MMSCFD). Currently, the plant uses a traditional regulatory control strategy, namely, the Distributed Control System (DCS) to control its key process parameters. However, such control system lacks the ability to handle multiple interactive constraints explicitly as well as the ability to address the dynamic process interactions and hence provides sub-optimal solution to the constrained control problem. This research aims to explore the potential of Model Predictive Control (MPC) as an advanced optimal control strategy to address some of the quality and production challenges currently encountered in operating the local NGL plant. In order to provide a realistic basis for this study, the studied NGL plant was simulated on three different levels which were necessary to achieve the objective of the work. Firstly, a steady state simulation of the plant has been built while incorporating design details of the plant and validated with real plant design data. Secondly, a dynamic simulation of the plant was later built on top of the steady state simulation in order to capture the transient behavior of the plant’s process in response to process changes. The dynamic simulation built was also verified by conducting test simulations with scenarios derived from actual plant data and the response of which have been compared and verified. The validated dynamic simulation was then used to generate a model of the plant which is necessary for the design of the MPC. Lastly, a supervisory MPC was built above the dynamic simulation layer in order to achieve the several objective requirements of the different processing units. A framework to formulate an optimal control problem has been presented. This study demonstrates the superiority of the MPC in iv achieving the different economic objectives of a train of distillation columns in the considered NGL plant namely – Deethanizer, Depropanizer and the Debutanizer units. Several operational scenarios – similar to what the real plant experiences, have been simulated in order to evaluate the robustness of the MPC to sustain its optimization performance. Several simulation scenarios have been considered such as changes in the feed rates and compositions. Comparative analysis showed the ability and superiority of the MPC system to constantly optimize the different objective requirements defined for each of the units. The study also presents an estimate of the quantitative benefits to be achieved for the subject plant. An annual incremental revenue of $1,971,565 has been estimated when the production of the three main products of the NGL plant i.e., Propane, Butane and Naphtha is optimized. Further economic analysis into the investment required for implementation of the MPC system for the subject plant yields a payback period of 1.25 years