document

Summary

As gas storage becomes increasingly important in managing the nation's gas supplies, there is a need to develop more gas-storage reservoirs and to manage them more efficiently. Using computer reservoir simulation to rigorously predict gas-storage reservoir performance, we present specific procedures for efficient optimization of gas-storage reservoir performance for two different problems. The first is maximizing working gas volume and peak rates for a particular configuration of reservoir, well, and surface facilities. We present a new, simple procedure to determine the maximum performance with a minimal number of simulation runs. The second problem is minimizing the cost to satisfy a specific production and injection schedule, which is derived from the working gas volume and peak rate requirements. We demonstrate a systematic procedure to determine the optimum combination of cushion gas volume, compression horsepower, and number and locations of wells. The use of these procedures is illustrated through application to gas-reservoir data.

Introduction

With the unbundling of the natural gas industry as a result of Federal Energy Regulatory Commission (FERC) Order 636, the role of gas storage in managing the nation's gas supplies has increased in importance. In screening reservoirs to determine potential gas-storage reservoir candidates, it is often desirable to determine the maximum storage capacity for specific reservoirs. In designing the conversion of producing fields to storage or the upgrading of existing storage fields, it is beneficial to determine the optimum combination of wells, cushion gas and compression facilities that minimizes investment. A survey of the petroleum literature found little discussion of simulation-based methodologies for achieving these two desired outcomes.

Duane¹ presented a graphical technique for optimizing gas-storage field design. This method allowed the engineer to minimize the total field-development cost for a desired peak-day rate and cyclic capacity (working gas capacity). To use the method, the engineer would prepare a series of field-design optimization graphs for different compressor intake pressures. Each graph consists of a series of curves corresponding to different peak-day rates. Each curve, in turn, shows the number of wells required to deliver the given peak-day rate as a function of the gas inventory level. Thus, the tradeoff between compression horsepower costs, well costs, and cushion gas costs could be examined to determine the optimum design in terms of minimizing the total field-development cost. Duane's method implicitly assumes that boundary-dominated flow will prevail throughout the reservoir.

Henderson et al. ² presented a case history of storage-field-design optimization with a single-phase, 2D numerical model of the reservoir. They varied well placement and well schedules in their study to reduce the number of wells necessary to meet the desired demand schedule. They used a trial-and-error method and stated that the results were preliminary. They found that wells in the poorest portion of the field should be used to meet demand at the beginning of the withdrawal period. Additional wells were added over time to meet the demand schedule. The wells in the best part of the field were held in reserve to meet the peak-day requirements, which occurred at the end of the withdrawal season.

Coats³ presented a method for locating new wells in a heterogeneous field. His objective was to determine the optimum drilling program to maintain a contractual deliverability during field development. He provided a discussion of whether wells should be spaced closer together in areas of high kh or in areas of low kh. He found that when φ h is essentially uniformly distributed, the wells should be closer together in low kh areas. On the other hand, if the variation in kh is largely caused by variations in h, or if porosity is highly correlated with permeability, wells should be closer together in areas of high kh. Coats' method assumes boundary-dominated flow throughout the reservoir.

Wattenbarger⁴ used linear programming to solve the problem of determining the withdrawal schedule on a well-by-well basis that would maximize the total seasonal production, subject to constraints such as fixed demand schedule and minimum wellbore pressure.

Van Horn and Wienecke⁵ solved the gas-storage-design optimization problem with a Fibonnaci Search algorithm. They expressed the investment requirement for a storage field in terms of four variables: cushion gas, number of wells, purification equipment, and compressor horsepower. They chose as the optimum design the combination of these four variables that minimized investment cost. The authors used an empirical backpressure equation, combined with a simplified gas material-balance equation, as the reservoir model.

In this paper we present systematic, simulation-based methodologies for optimizing gas-storage reservoir performance for two different problems. The first is maximizing working gas volume and peak rates for a particular configuration of reservoir, well, and surface facilities. The second problem is minimizing the cost to satisfy a specific production and injection schedule, which is derived from the working gas volume and peak rate requirements.

Constructing the Reservoir Model

To optimize gas-storage reservoir performance, a model of the reservoir is required. We prefer to use the simplest model that is able to predict storage-reservoir performance as a function of the number and locations of wells, compression horsepower, and cushion gas volume. Although models combining material balance with analytical or empirical deliverability equations may be used in certain situations, a reservoir-simulation model is usually best, owing to its flexibility and its ability to handle well interference and complex reservoirs accurately.

It is important to calibrate the model against historical production and pressure data; we must show that the model reproduces past reservoir performance accurately before we can use it to predict future performance with reliability. However, even calibrating the model by history matching past performance may not be adequate. It is our experience that information obtained during primary depletion of a reservoir is often not adequate to predict its performance under storage operations. Primary production over many years may mask layered or dual-porosity behavior that significantly affects the ability of the reservoir to deliver large volumes of gas within a 4- or 5-month period. Wells and Evans⁶ presented a case history of the Loop gas storage field, which exhibited this behavior. It may be necessary to implement a program of coring, logging, pressure-transient testing, and/or simulated storage production/injection testing to characterize the reservoir accurately.