Artificial Intelligence Recovery Modeling of <5 API Unconventional Next-Generation Heavy Oil Using CHNSO Technology

Abstract

Extreme heavy oil <5 °API is considered a type of unconventional tight oil, which will require a challenging petroleum production system for future new-generation extreme heavy oil or bitumen carbonate reserves. This oil is abundant in great amounts around the globe, yet is extremely difficult to produce due to its solid-like physical state locked deep underground. The world strategy eventually will shift focus to this type of oil since conventional and other less-quantitative-difficult reservoirs are continuously depleting. The interest of this study is directed towards a specific type of unconventional oil, which is available in tight carbonate reservoirs. Extreme heavy oil <5 °API exists in large quantities in Kuwaiti fields. This study presents a novel heavy oil classification especially for <5 °API crude oil types as well as their potential recoveries. All recoveries considered for this study are bench-scale laboratory physical experiments with toluene, de-ionized water and water-aided surfactants augmented with applied field thermal 25 °C, 135 °C, 225 °C and 315 °C heat treatments. The main objective for this research is to model five signature atoms available in almost all heavy crude oils: carbon, hydrogen, nitrogen, sulfur, and oxygen (CHNSO). These CHNSO fingerprints determine qualitatively and quantitatively the potential amount and quality of future extreme heavy crude oil recovery. An Artificial Intelligence (A.I.) neural network algorithm is developed for all possible conjecture atoms. A Multiple Layer Forward Feed (MLFF) learning system is designed, trained and applied for developing the A.I. neural network. Forty-one recovery models are manifested in this study, clustered in possible atom conjecture operational base-function domains, which are unary (one atom), binary (two atom), ternary (three atom), quaternary (four atom) and quinary (five atom) approach models. The main technological motivation for CHNSO research is finding the optimized conventional EOR recovery efficiency factor that will extract <5 °API oil. The model predicts the recovery potential factor in a classic, optimum and conventional economic scenario, considering the unconventional environmental impact, crude oil subsurface-mobility issues and technology limitations used as current economic challenges. The general summary of results suggest that CHNSO models are useful in better understanding and better predicting <5 °API oil recoveries. The three-atom nitrogen-sulfur-oxygen (NSO) ternary conjecture model has a significant impact regarding heavy crude oils, maximizing recovery in general, and extreme heavy oil potential recovery in particular, in regards to the difficult mobility of this type of crude oil.