Spontaneous imbibition or capillary driven flow is important for geological CO2 sequestration, hydrogen storage in geological formation, diagnostic devices, capillary pumped loops etc. Generally, these porous media are heterogeneous and the heterogeneity is due to the layering of the porous medium. Interplay of viscous, gravity and capillary forces govern the location of the leading front. Earlier studies showed that leading front is in the narrowest pore due to high capillary forces. However in our research we have shown that the leading front is strongly dependent on the layering and in many cases with high heterogeneity the front in the narrow pores will lag. For more information, please see our publications [J7, J9, J10, J11].
Quantification of wettability inside a porous medium is challenging. Most of the currently available methods use either a proxy surface of the porous solid or a small sample of the porous medium for wettability quantification. Recognising that more wetting fluid has more interfacial area with porous solid, we develop a tracer method to quantify the solid-liquid interfacial area for a porous medium. We use two tracers for a fluid-phase, an ideal tracer, and an adsorbing tracer. The ideal tracer follows the flow path, and the adsorbing tracer dynamically adsorbs on the porous surface. We use the tracers’ mean residence time to quantify the solid-liquid interfacial area during multiphase flow in a porous medium. The figure on the left shows the contact area of wetting phase is more than the non-wetting phase at same saturations. [J14]
We use simulations at geological reservoir scale to understand the flow behaviour. Currently we are working on gas hydrate reservoirs. Methane hydrates are ice-like compounds, which store methane in huge volumes at high pressure and low temperature. Naturally occurring reservoirs of methane hydrates can become potential energy resource. We use an in-house methane hydrate reservoir simulator to explore the production methodologies for heterogeneous methane hydrate reservoirs. Using our simulations we could prove the productivity difference between different kinds of hydrate reservoirs as shown in figure on the left. [J4, J5, J17]
The data generated from the reservoirs is huge, for example seismic data. For interpretation of these data sets, we recently started exploring machine learning techniques to find faults and horizons. This project was funded by Schlumberger. [J13, C12]. Figure on the left compares the horizons from our prediction from machine learning with the ground truth. We are now working on using the data science techniques for speeding up the simulations process.
The uncertainty in measuring reservoir parameters like permeability, porosity is quite high. Using deterministic simulations to forecast the production can lead to over-prediction or under-prediction of oil reserves using reservoir simulation. Monte-Carlo simulations are computationally expensive, while Design of Experiments method, which is generally used, neglects the probability distribution of the reservoir parameters. We use Polynomial Chaos Expansion to quantify uncertainty in the reservoir simulation predictions given the prior probability of the reservoir parameters. For history matching, bayesian inference can be used to quantify the priors once the actual production data becomes available. Figure on the left shows the predictions from our method. For more details see publications. [J12, J16, C7]
Asphaltenes are heavy components in crude oil which are difficult to remove from sand particles in reservoirs, pipelines outside the reservoir and in refineries. We use AFM to understand the interaction of asphaltenes with these inorganic substances. We show that shear forces can be used to remove the asphaltenes. For more details please see publications [J15]