Sampling biases and mitigations in modeling shale reservoirs

Abstract

Field development of a shale reservoir is different from developing conventional reservoirs because of the tightness of formations and extensive use of horizontal wells. One critical consequence of horizontal wells is the sampling bias. Construction of a reservoir model for a shale reservoir must mitigate the sampling bias of well data for the model to be realistic, especially when horizontal wells are present. Although geostatistical modeling methods can be effective in modeling spatial distributions of reservoir properties, they generally do not account for a sampling bias. It is necessary to first decouple the debiasing of well data from the 3D reservoir modeling. In this paper, we first present methods to debias sample data from vertical or horizontal wells or a mixture of them. After sample data are debiased, we present reservoir modeling while coupling the honoring of the debiased frequency statistics and the modeling of spatial heterogeneities for shale reservoirs.