An Approach To Optimize Economics in a West Texas CO2 Flood

Abstract

Summary EOR projects, notably CO2 floods, are the next generation of recovery methods in the more mature west Texas waterfloods. Installing and operating a CO2 flood can be extremely expensive. In this paper, we discuss methods used to make several active CO2 floods more profitable by reducing operating costs and deferring investments. Introduction Our goal in studying several active west Texas CO2 floods was to determine the optimum near-term cash flow, overall project economics (rate of return, present woth, etc.), and oil recoveries. We developed various CO2 flood designs with a reservoir simulator by altering specific operating parameters, including the half-cycle slug size, gas/water ratio (GWR), injection schemes, and total CO2 slug sizes. The resulting injection and production rates were then entered into an economic simulator to determine the most economical set of operating conditions. We also assessed the impact of various economic conditions - oil prices, CO2 prices, gas processing fees, and operating costs - on the optimization process. This was done to ensure that the optimum set of operating conditions did not change drastically should the economic environment change. This study shows how optimizing operations can significantly improve the economics of existing CO2 floods. An additional benefit of the study was insight gained on plant sizing and adjusting injection schemes to minimize or defer investments while maximizing the economics of new projects. Background During the late 1970's and early 1980's, Amoco Production Co. committed significant manpower to evaluate the feasibility of full-scale CO2 flooding in west Texas. Before the feasibility work, Amoco undertook numerous CO2 pilots in various fields.1 Because of high crude prices, optimistic price forecasts (Fig. 1), and successful pilot recoveries, the future of CO2 flooding looked promising. The approach taken at the time was first to history match waterflood performance with an in-house black-oil simulator.2 The resulting reservoir description then was transferred to a streamline generating model and finally to a miscible streamtube model. The CO2 prediction was done at a single GWR with the CO2 injection slug size changed to find an optimum based on present worth. The reservoir simulation provided oil, gas, and CO2 production rates and CO2 injection rates for input into an economic model. Because of the man-hours and computer time involved in simulating the CO2 process, only a few GWR's were tried, with slug size as the main variable. In these early CO2-flood designs, CO2 reinjection was planned and a plant design was needed to remove H2S and hydrocarbons and to handle the peak predicted total gas production. Because of the single GWR being used, the gas rate was expected to increase to some peak and then decline, leaving excess plant capacity. This type of injection scheme impaired the economics because a large plant investment was needed at the beginning of the project. Unless there was an offset property feasible for CO2 flooding, the excess plant capacity went unused once the peak was past. The costs of plants and other investments were relatively high compared with today's environment because of high oil prices and high inflation rates. Most projects were still economical, however, because oil price forecasts were optimistic and tax relief from the Windfall Profits Tax was granted for EOR projects. One shortcoming of this early approach was the optimistic nature of the miscible streamtube model. Although this model adequately handled loss of miscibility, it was unable to regain it should the pressure rise above the miscible pressure. The model also was too efficient in areal sweep: the displacement was piston-like through the streamtube and all streamtubes were eventually swept for nearly 100% areal sweep efficiency. Injectivity losses also were handled inadequately in the streamtube model. Water hysteresis was accounted for, but solvent relative permeability effects3,4 were not. The miscible streamtube model also had limited ability to make changes in the pattern configuration and injection schemes. If such changes were to be made, the streamline generator had to be run again for entry into the miscible streamtube model. This was both time-consuming and costly. As a result of these model shortcomings, many economic sensitivities were run with various rate and reserve scenarios to become comfortable with the final CO2-flood design.