Forward Integration of Dynamic Data into 3-D Static Modeling Significantly Improves Reservoir Characterization

Abstract

Geo-modelling is usually done to honor static data such as core, well logs and seismic acoustic impedance (AI) map where available. Once the static geo-model is complete, history matching is carried out by tuning the static model properties until the model reproduces observed dynamic behavior. The objective of this paper is to showcase how a systematic a priori integration of dynamic elements into geo-modelling eliminated the need for history matching. These dynamic elements are; connected reservoir regions CRR (Kayode et.al 2018) and permeability-thickness (kh) interpretation from Pressure Transient Analysis PTA. This paper also introduces the concept of CRR based permeability modeling. CRRs were defined based on time-lapse shut-in pressure trend groups. Core and log data were grouped on the basis of the identified CRR and used to build CRR-based Neural Network models for predicting permeability logs of non-cored wells within each CRR. The geo-modeler then created two geo-realizations by using the permeability logs within each CRR to distribute permeability within the CRR using two assumptions of variogram lengths (i) variogram range obtained from analysis of limited core data, (ii) variogram range required to ensure intra-CRR connectivity. Pressure transient was simulated for wells with observed PTA data using the two realizations, and a comparison of the log-log plots of simulated pressure transient derivative and observed pressure transient derivative were used to determine the quality of each realization for each well. The realization that provided the least squares of error across all the wells was selected as base-case geo-model. Permeability correction coefficients were applied on the base-case geo-model until PTA kh were acceptably matched. The resulting permeability log at the PTA well is referred to as PTA-corrected permeability log. Some cored wells were originally exempted from the neural-network permeability modelling because they didn't have logs (sonic, density and neutron logs). Hybrid permeability logs were derived from a combination of the predicted permeability logs and core permeability at these well locations. All permeability correction logs (i) PTA-corrected permeability logs and (ii) Hybrid permeability logs were then fed back into the geo-modeling workflow to generate an improved permeability distribution which respects core data, PTA kh, and CRRs. The do-nothing simulation run has more than 80% of wells’ pressure data acceptably history matched. This application demonstrates that a priori integration of dynamic elements like CRR, PTA kh, and the use of CCR-based permeability modeling results in a better characterized geo-model with potential for eliminating the need for history matching.