No abstract

No abstract

In the numerical modeling of a steamflood process in a highly stratified, thick reservoir with thin interbedded discontinuous permeability barriers, the vertical grid size must be comparable to the length scale of permeability variation in the vertical direction. In a thick reservoir, the smaller length scale of vertical heterogeneities results in an impractically large number of layers for the modeling of a computationally intensive process such as steamflooding. In the first part of this study a series of simulation runs were conducted for different geostatistically derived cross sectional models to study the degree of heterogeneity that is required to properly model steamfloods in the presence of thin diatomite barriers in both the dipping and non-dipping sections of a heavy oil, thick reservoir. These cross sectional models each had a different number of layers and were developed by sequential indicator simulation of log traces. The most detailed models contained 300 layers. In the second part of this work we applied different methodologies for coarsening the most detailed model while still capturing the effects of the geologic features. Two different methods were examined to coarsen the detailed models. In one method, we maintained the impermeable layers and coarsened the sands by averaging each 10 sand layers. In the other method, the 300 layer models were coarsened using a general scale up method. In this method the dominant flow paths in the cross section are first identified through solution of a single phase flow problem. This information is then used to selectively scale up the reservoir properties, leaving detail in regions where required and coarsening in other regions. The results show that the coarse models developed directly from the sequential indicator simulation underpredict the recovery in the non-dipping cross sections and overpredict the recovery in the dipping cross sections. The results also show that the scaled up coarse models predict recoveries in good agreement with the detailed model in both the dipping and non-dipping sections of the reservoir. The results also show that the coarse models which were developed by keeping the shale and averaging the sand layers only provide accurate results in the non-dipping section of the reservoir. Introduction As heat and reservoir management become more important in steamflood operations improved engineering tools are required to cost-effectively manage steam injection projects. Application of reservoir simulation as a tool for steamflood management is growing. However, many steamflood reservoirs have a high amount of geologic heterogeneities, which may impact both actual and simulated project performance. The importance and impact of heterogeneities on the performance prediction of steamflood processes in highly stratified systems have been recognized in the literature.

Summary Systematic time-lapse pulsed neutron capture (PNC) logging was conducted during a two-year steam foam mechanistic field trial in a previously steamflooded reservoir. The producing reservoir tested was the Monarch Sandstone, which is composed of thinly inter-bedded sandstones, siltstones, and diatomaceous mudstones at the base, coarsening-upward to massive conglomerates and sandstones with a few interbedded mudstones at the top. Observation wells were drilled in the study area, one of which was continuously cored through the reservoir interval (578 feet). Two of these cased wells were emptied to facilitate better temperature logging. Uncalibrated PNC log responses in the air-filled wells were normalized (calibrated) to water-filled conditions with transform functions empirically determined in the same wells. The normalized logs were used to generate a series of steam/gas saturation profiles that indicate clearly the dynamics of the fluid and foam in the reservoir during the experiment. Enhanced precision and bed resolution of the PNC logs were necessary for interpretations in the thinly bedded part of the reservoir. The time-lapse profiles were capable of delineating flow units down to a few feet in thickness and showed excellent consistency with lithologic variations described in the core. In addition, the PNC log detected mudstone layers as thin as six inches. Correlations between neutron capture cross sections (sigma), mineralogy, and rock chemistry, all from analyses of core samples, were examined to investigate additional use of the PNC logs for steamflood reservoir characterization. As expected, sigma increased with increasing clay/mica content. However, a more significant correlation was noted between sigma and diatomite content, which is due to the boron in the diatomite. This finding may result in improved stratigraphic correlations of the more laterally continuous diatomaceous mudstones based on PNC logs, with significant impact on the characterization of reservoirs of this type. The PNC log data from the field trial, when used in conjunction with core, temperature, and pressure data, was critical in developing a better understanding of foam generation and propagation in the reservoir. Furthermore, it minimized the time and cost required to successfully complete the field trial. Introduction In the management of a steamflooded reservoir, it is important to both monitor and predict steam movement in order to optimize steam sweep efficiency and heating energy usage. To these ends, the pulsed neutron capture log (PNC) has been providing useful information in steamfloods in the San Joaquin Valley and elsewhere, along with other logging methods. In most steamflood operations, steam movement and distribution are primarily controlled by gravity and reservoir geology - vertically by barriers, and horizontally by sand geometry. The more complex the geology, the more difficult it is to predict steam movement Cores and openhole logs have been providing geologic information before and during a steamflood.

No abstract