Reducing Composition Characterization Uncertainties Through Advanced Machine Learning (ML) Techniques - Data Clustering

Abstract

Objectives/Scope Round-robin tests and laboratory audits have demonstrated that reservoir fluid compositions measurements can be systematically uncertain. To reduce compositional uncertainties, this work uses Machine Learning (ML) algorithms to update the existing measured fluid compositions and applies a compositional adjustment anchored on a decreased number of selected fluids representative of the entire compositional space. The resulting composition will reduce uncertainty in exploration and production models and diminish the risks in field development decisions. Methods, Procedures, Process Statistical error propagation analysis of reservoir fluid compositional measurements were completed to identify the parameters that most affect the overall uncertainty of the compositional analysis data. Advanced ML techniques were then used to perform clustering analysis of 800+ unique fluid compositions from legacy PVT reports. Dimensionality reduction exercise and data clustering utilizing the k-means algorithm were employed to identify parameters that best characterize the reservoir fluid properties to be corrected. The sampling candidate wells were identified to obtain new samples for the selected anchor points fluids that will be subjected to detailed fluids characterization. Results, Observations, Conclusions Systematic error propagation analysis demonstrated that, of the 140 independent parameters measured to determine a reservoir fluid composition (to C30+), the plus fraction concentration and molecular weight measurements uncertainties contributed the most towards the overall composition uncertainty for most black oil fluids. After extraction and collation of available reservoir fluid analysis data, a detailed review was completed to eliminate samples with significant mud filtrate contamination, obvious light end weathering and/or other inconsistencies. Once the dataset was validated, the dimensionality reduction task led to 5 principal dimensions space out of 57 initial possible features. The reduced dimension set included two distinct Cn- molar distribution slopes that best correlated the compositions data. Data clustering analysis was performed to identify the cluster centroids in the reduced dimension space. Wells with the closest compositions to cluster centroids were selected for sampling based on this work and the suitability of target wells for sampling. Further, extensive fluid compositions analysis with physical separation processes will be undertaken on the 50 selected anchor point fluids. Novel/Additive Information An original combination of reservoir fluids domain knowledge and advanced ML techniques have been applied to identify candidate wells to sample with the goal to significantly reduce the uncertainty in measured reservoir fluid composition. This unique workflow for composition corrections will help reduce the uncertainty in the critical techno-economic decisions that depend on fluids characterization data in the oil and gas field development domain.

Can Artificial Intelligence Accelerate Fluid Mechanics Research?

“…A great number of reviews meant that an even greater number of original research articles had been published. If we limited the results of the Scopus search to reviews related to fluids, three major fields of application emerge that attracted research interest in applying ML methods to (a) theoretical and industrial oil and gas [33][34][35][36][37][38][39][40][41][42][43][44][45][46][47][48][49][50][51][52], (b) CFD simulations [53][54][55][56][57][58][59][60][61][62][63][64][65][66][67][68], and (c) health and medical applications [69][70][71][72][73][74][75][76][77][78][79][80]…”