Injection Impairment by Suspended Solids: Well Models, Filtration Parameters, and Models’ Validation with 3-D Injection Experiments



Journal Title

Journal ISSN

Volume Title



A substantial portion of water is produced along with oil production. Due to environmental and economic constraints, produced water is re-injected into formation for reservoir pressure maintenance and disposal. However, suspended particles in produced water cause formation plugging, resulting in reduced permeability and loss of injectivity. Water injection, without solids suspension, at high velocity generates a loss of injectivity as a non-Darcy flow effect. The compounding impacts of high injection fluid velocity with suspended solids are expected to produce more severe injectivity loss than water alone. The problem is complex, and few experiments and models are available in the literature. Among currently available models, the solids deposition profile is based on a linear relationship between pressure drop ratio and amount of solids deposition. We propose a second-order relationship that is more accurate in capturing the effects of high injection velocity by incorporating the effects of non-Darcy flow. This model predicts injection decline profile as a function of deposition amount for radial flow geometry. Another way to maintain a high and sustained injection rate is by using a horizontal well. The models for predicting horizontal well injectivity due to particle plugging are also limited. The complex or non-radial flow geometry of horizontal wells (especially for reservoirs with small net pay and large drainage area) is a reason. Based on Furui’s model, we treat the fluid flow as radial around the wellbore region and linear after flows reach the reservoir’s upper and lower pay boundaries. Both vertical and horizontal wells have been validated with 3-D sand pack tests. Filtration coefficient and damage factor are required for modeling internal formation plugging from suspended solids particles during water injection. This is obtained experimentally and depends on flow velocity, particle slurry concentration, particle size, and formation grain size (or pore throat size). Such experiments are complex and numerous to model different injection scenarios. This study collects experimental data from open literature and employs machine learning regressions to help predict the parameters. The collected dataset ranks the XGBoost regression at the top, and concentration is the most significant factor.



Formation damage, Horizontal well, Non-Darcy flow, Solid suspension