Calibration of Sand Production Prediction Models at Early Field Life in the Absence of Field Sanding Data

Abstract The risk of sand production must be evaluated from very early stages of field development planning to select the optimum well completion, and if required, identify the appropriate sand control options. However, during the appraisal or early development stages, sand production predictions often cannot be verified against field data because the lack of sanding observations. The analysis of predicting the onset and severity of sand production generally consists of using analytical or numerical rock mechanical models and ideally calibration with field sand production data. In this paper we show how in the absence of field sanding data, reliable sanding predictions can still be achieved by combining commonly used analytical and numerical prediction methods. This approach has been validated using a brown field dataset from South East Asia with several sand producing wells and multiple pressure depleted reservoirs. The analytical method uses a poro-elastic model and core-calibrated log-derived rock strength profiles with an empirical effective rock strength factor (ESF). The ESF should be calibrated against documented field sanding observations from wells with credible formation and drawdown pressures. The numerical method uses a poro-elasto-plastic model defined from triaxial core tests. The rock failure criterion in the numerical method is based on a critical strain limit (CSL) corresponding to the failure of the inner wall of thick-walled cylinder core tests or if existing field sanding data. In the brown oil field example, both analytical and numerical methods accurately predict the onset of sanding in the sand-prone wells. The core-based CSL defined by the numerical method match very well with the field sanding data with a very small margin of error. In contrast, the default and non-calibrated ESF values commonly used in the analytical method required considerable tuning to match with the field sanding data. This suggests that the core-based CSL can be used with a reasonable confidence for sand production prediction purposes in the field and the prediction can be as reliable as field-calibrated models. After the field validation, the predictions for planned infill wells by both analytical and numerical methods were consistent and similar. The paper will show how this approach is used in two other fields from the Asia Pacific region both at appraisal stages and with no field sanding data. Both cases show that different ESF values are required to match with the numerical simulations. In the absence of field sanding observation, particularly in the early stages of field development, the use of non-calibrated default ESF values in the analytical method could lead to erroneous sanding assessment and poor sand management decisions. Application of a combined analytical and numerical sanding evaluation enhances the reliability of sand production predictions. After a calibration is obtained, either with field sanding data or calibration against numerical simulation and core-based CSL criteria, the analytical method can be used with confidence and has the added benefits of simplicity and quick realizations of various scenarios and output plotting capabilities and inputs for sand control and well completion decisions.