K-Mean Cluster Analysis for Better Determining the Sweet Spot Intervals of the Unconventional Organic-Rich Shale: A Case Study

Abstract

The petrophysical analysis is the crucial task for evaluating the quality of unconventional organic-rich shale and tight gas reservoirs. The presence of organic matter and the ultra-tight with over complex pore system have remained a lack of understanding of how to evaluate the extensive parameters of porosity considering organic content, gas saturation, organic richness, brittleness index, and sweet spot interval by only using conventional log. Therefore, this study offers effectively applied techniques and better analysis for interpreting these parameters by maximizing and integrating geological, geochemical, rock mechanical and engineering data. In general, the field data used in this study are from the first dedicated well for source rock exploration in the North Sumatra Basin, Indonesia. The developed method was derived by using conventional log. All interpretation results were validated by laboratory data measurements of routine and special core analysis, petrography, total organic carbon (TOC) and organic maturation, and brittleness index (BI) calculation. Moreover, the high quality of NMR log data was used as well to ensure our developed techniques present good estimations. Briefly about the methods, we started to determine the total and effective porosity based on the density log by including the presence of organic matter and multi-mineral analysis in these estimations. Then, we used the revised water saturation-TOC of water saturation while the TOC was predicted in advance by averaging three results from the correlation of TOC-Density, modified CARBOLOG and Passey’s ΔlogR methods. Equally important, in order to obtain the reliable gas saturation prediction, we used saturation exponent (n), cementation factor (m), and the tortuosity factor (a) parameters which obtained from laboratory measurement of formation resistivity factor and resistivity index (FFRI). In addition, the brittleness index was predicted based on sonic log data. Finally, all parameters needed for determining gas shale sweet spot have been made. Then, we developed a way to evaluate the sweet spot interval by using K-mean clustering. In conclusion, this clustering result properly follows the shale quality index parameters which consist of organic richness and maturation, brittleness index, the storage capacity of porosity and gas saturation. This study shows that these petrophysical applied techniques leads us to interpret the best position of shale interval to be developed with a simple, fast, and accurate prediction way. Furthermore, as a novelty, this method can be used as rock typing method and obviously can reduce uncertainty and risks in organic-rich shale exploration.

Implementation of Data Science and Decision Analysis to Determine Shale Gas Sweet Spot Depth Interval

“…Based on a literature study of similar previous studies, there are several advantages and disadvantages in each study that can be used as guidelines in this study. The research entitled K-Mean Cluster Analysis for Better Determining the Sweet Spot Intervals of the Unconventional Organic-Rich Shale: A Case Study by Akbar and Nugraha (2018) used the K-Mean Clustering method in classifying data from hydrocarbon exploration wells. This method is considered to be able to determine the desired number of clusters based on the number of parameters that depend on the distance between the data or the centroid.…”

Prediction of Rock's Brittleness and Dynamic Properties Utilizing Effective Artificial Intelligence Approaches

“…Altindag and Guney [14] presented some relationships to estimate brittleness based on UCS, Schmidt hardness number, and Ts. Akbar and Musu [15] studied the effect of mineralogy on the brittleness. Studies have revealed that MFFNN has high precision to estimate the Vs rock samples [16,17].…”

On the Pore Geometry and Structure Rock Typing

“…Samples in each class seem indistinguishable due to similar geological depositions and diagenetic alterations . Many methods have addressed rock typing, for example: Based on mechanical properties, mineralogy, and organic geochemistry The use of permeability, porosity, and irreducible water saturation data either empirically ,− or with a hydraulic flow unit (HFU) approach; ,, Involvement of capillary pressure data and J -function − and combined with radius; Consideration of thin section descriptions and interpretations such as rock fabrics, depositional facies, and rock textures; , Geostatistics and machine learning implementation such as clustering, , ANN, self-organizing map, , and fuzzy logic; Grouping based on the dimensionless form of absolute permeability, relative permeability, porosity, and phase viscosity, the so-called true effective mobility TEM function; , The use of resistivity data and porosity to yield in electrical flow unit; Further development of analytical models, such as the pore geometry and structure (PGS) method. − …”

“…Geostatistics and machine learning implementation such as clustering, , ANN, self-organizing map, , and fuzzy logic;…”

“…• Based on mechanical properties, mineralogy, and organic geochemistry 7 • The use of permeability, porosity, and irreducible water saturation data either empirically 5,8−10 or with a hydraulic flow unit (HFU) approach; 1,3,11 • Involvement of capillary pressure data and J-function 12−14 and combined with radius; 5 • Consideration of thin section descriptions and interpretations such as rock fabrics, 15 depositional facies, 16 and rock textures; 17,18 • Geostatistics and machine learning implementation such as clustering, 19,20 ANN, 21 self-organizing map, 22,23 and fuzzy logic; 24 • Grouping based on the dimensionless form of absolute permeability, relative permeability, porosity, and phase viscosity, the so-called true effective mobility TEM function; 25,26 • The use of resistivity data and porosity to yield in electrical flow unit; 27 • Further development of analytical models, such as the pore geometry and structure (PGS) method. 28−31 The PGS method is built on the basis of the Kozeny−Carman and capillary bundled model equations, ending up with a plot between y k/ and x � k/ϕ 3 within a power function y = ax b28, 29 such that The inspiration came from the facts that many physical phenomena in nature are in a power law scaling relationship.…”