(696b) Economically Optimal Steam Injection and Power Generation for Enhanced Oil Recovery Using Predictive Model-Based Optimization Technology

Enhanced Oil Recovery (EOR) is a commonly adopted practice in oil production where petroleum is recovered from an underground source by injecting an additive, such as steam, into the well. The use of steam injection has proven to be a cost effective technology for improved efficiency of primary recovery operations as well as additional oil recovery from older or abandoned oil fields. Power generation equipment is often additionally coupled with EOR both to satisfy site electricity requirements and to provide added efficiency with heat recovery.

To enable steam injection, a large number of steam generators are strategically located throughout the oil field and are banked together to produce the steam that is fed into an often complex distribution network that is designed to ensure the delivery of the right amount of steam with pre-determined specifications to the injection sites. Planning and ensuring the proper delivery of steam to injection sites is a key operational challenge with significant impact on operation cost as well as the productivity of the oil field. Poor instrumentation often further complicates this challenge.

In this paper, we use a representative simulated network to demonstrate that a model-based distributed optimization approach can be used to schedule and manage the operation of steam and electricity generation in an economically optimal manner. Steam generation units, gas turbines, and the distribution network are modeled first. The steam demand at various injection sites is input to the optimization problem. A mixed integer nonlinear programming (MINLP) problem uses unit operation models, the topology of the distribution network, the operational constraints, and the cost information (e.g. fuel cost) to determine the optimal assignment of steam and power generation units as well their dispatch level to meet the given demand at minimal cost. Using an illustrative example we will demonstrate that even in relatively simple cases the intuitive operation strategies could prove suboptimal.