Application of Machine Learning to Predict the Organic Shale Sweet-Spot Quality Index

Abstract

Abstract Shale plays -- that have challenging characteristics, such as deep formation or a shortage of freshwater needed for hydraulic fracturing -- present a special challenge for determining the feasibility of economic development. A new graphic technique has been developed to visualize and evaluate shale sweet-pots, taking into consideration their characteristic properties and showing them using only a single index: the "Sweet-spot Quality Index" (SSQI). This index reflects the grade and scale of the sweet-spot. SSQI is a function of the rock quality, the hydrocarbon in place, completion, and operation quality. Data sets were collected from different shale basins including well logging and operation conditions. Well logging data provide a considerable amount of information and data for unconventional reservoirs related to a formation's inplace and geomechanical properties. Moreover, the field operations cost and availability of material and equipment, hydrocarbon prices, and HSE issues are important to define the operation quality. Machine learning (ML) techniques are used to calculate the SSQI for different shale basins in the USA and Africa. Results showed the capabilities of the ML models to SSQI from input data. The models were used to conduct a sensitivity analysis to rank the effect of different input parameters on the SSQI. The SSQI enables the comparison of sweet-spots and is calculated from four indices: The Reservoir Quality Index (RQI), the Completion Quality Index (CQI), the Conventional Behavior Index (CBI), and the Operation Index (OI). Calculating the SSQI using different machine learning techniques provides an improved approach to visualization, evaluation, comparison, and recommendations. It lends itself to becoming a standard evaluation technique for both E&P and service companies.