Asphaltene deposition prediction using adaptive neuro-fuzzy models based on laboratory measurements

Deposition of asphaltene is recognized as a well-known severe problem, which can significantly affect oil production and enhanced oil recovery processes through mechanism of wettability alteration and blockage. The natural mechanism is not fully comprehended until now due to impossibility to carry out actual field experiments. In this work, different flow dynamic test scenarios are organized to perform on sandstone as well as carbonate rocks to practically explore process of asphaltene deposition. Ordinary optimized methods are not applicable to asphaltene deposition due to its dependency on the involved parameters and complexity of process. The permeability impairment data is monitored through analysis of recorded pressures during the test experiments. Then, a new adaptive neuro-fuzzy inference system is developed to predict asphaltene deposition in terms of permeability (K/K0) and pressure drop (DP), considering pore volume injection (PVI) and time data as input variables. Accordingly, two adaptive neuro-fuzzy models are sequentially developed in a nonlinear affine-type configuration to investigate the effect of multiple variables and parameters on asphaltene deposition on the basis of the most recent input–output data. es of test studies has been conducted to demonstrate the efficient capabilities of the proposed algorithm to automatically predict asphaltene deposition for different prediction horizons using the online affine-type identified models. Eventually, acceptable agreement between experimental and estimated permeability and pressure drop is investigated to demonstrate superiority of the proposed approach to monitor future asphaltene status that will be useful to prediction of field production under natural depletion process.