Presentation on theme: "Pattern Recognition Techniques in Petroleum Geochemistry L. Scott Ramos and Brian G. Rohrback Infometrix, Inc. Daniel M. Jarvie Humble Instruments & Services,"— Presentation transcript:
Pattern Recognition Techniques in Petroleum Geochemistry L. Scott Ramos and Brian G. Rohrback Infometrix, Inc. Daniel M. Jarvie Humble Instruments & Services, Inc.
InfoMetrix Computer-Assisted Geochemistry The emphasis in production geochemistry is to match oils to source rocks and to correlate one crude oil to others. We do this to trace migration or to assess the degree of communication among reservoirs. Computerized pattern recognition (aka chemometrics) is an efficient way to exploit the information richness of the data without sacrificing speed or accuracy.
InfoMetrix An Overlay of Chromatograms By overlaying chromatograms we can look both at the similarities and the differences in the crude oils. Software can use this underlying pattern to build quantitative and objective models.
InfoMetrix Example: Automation of Geochemical Evaluations Source rock typing can be done by using GC, GC/MS and stable isotopes on crude oils. We employ a series of chemometric models to first separate the samples based on gross characteristics (I.e., lacustrine versus marine) and then use fine tuning models to further characterize samples.
InfoMetrix GC/MS Mass Chromatograms Tricyclic Terpanes m/z=191
InfoMetrix GC/MS Mass Chromatograms Steranes m/z=217
InfoMetrix Source Rock Type# of Oils Marine Shale146 Paralic/Deltaic Marine Shale 26 Marine Carbonate/Marl157 Evaporite/Hypersaline Marls 11 Coal/Resinitic Terrestrial Source 29 Lacustrine, Fresh 35 Lacustrine, Saline 20 Construction of a Geochemical Library The issue here is to assemble data on a sufficient number of oils to make the library valuable.
InfoMetrix Assembly of a Library x 11 x 12 x x 1m x 21 x 22 x x 2m x n1 x n2 x n3... x nm A data matrix is constructed based on geochemically significant ratios drawn from the GC, GC/MS and stable carbon isotopes (saturate and aromatic).
InfoMetrix KNN Method to Classify Unknown MarineLacustrine
InfoMetrix SIMCA Method to Qualify Marine Lacustrine Unknown
InfoMetrix Oil Classification Schematic Oil Sample Aquatic Terrestrial Marine Lacustrine Paralic/Deltaic Coal/Resinitic Fresh Water Saline Water Shale Marl/Carbonate Evaporite
InfoMetrix elseif All == 3 load knn model from aquatic.mod load knn model from aquatic.mod G3 = predict G3 = predict if G3 == 1 if G3 == 1 load knn model from marine.mod load knn model from marine.mod predict predict elseif G3 == 2 elseif G3 == 2 load knn model from lacustr.mod load knn model from lacustr.mod predict predict end end Automation of a Hierarchical Classification
InfoMetrix Example: Reservoir Oil Fingerprinting Chromatography allows us to determine if one reservoir is linked to another by looking at marker peaks that show between the normal alkanes. This process can be done either by choosing an appropriate set of marker peaks ahead of time or by evaluating the whole chromatographic pattern. GC is usually the technique of choice due to the lower cost of analysis and faster turnaround time.
InfoMetrix Crude Oils from Two Reservoir Systems Pr Ph n-C 15 n-C 17 n-C 19 n-C 12
InfoMetrix Marker Compounds Between n-C 15 and n-C 16
InfoMetrix Normalizing the Chromatograms to Accentuate Differences
InfoMetrix Example: Monitoring Yield from Multiple Reservoirs in Open Hole Completions We can use chromatographic patterns to determine the relative yield from more than one reservoir even where there is no casing. In this example, the field is undergoing water flood to drive the oil to producing wells. One of the producing zones is significantly more porous than the other. Because pumping water is the primary cost, knowing the relative yields from each reservoir is important. Pattern recognition also can flag the unusual...
InfoMetrix Production Well Well Stimulation Production in the latest 30 production intervals (bbl/day) After closing Well 696 in and pressurizing the reservoir system, an increase in production was noted.
InfoMetrix Well Chromatograms 1994 Production Pre-Stimulation 1995 Production Post-Stimulation Are the differences in hydrocarbon distribution significant?
InfoMetrix Well Oil Profile Production in Well 696 has changed in composition significantly since stimulation work was done. The interpretation is that the well is now producing from a new zone, deeper than the A or B zones already characterized. Zone B Zone C Zone A Some other wells also seem to show Zone C input.
InfoMetrix Zone Apportionment Well 696 Well Stimulation Zone C Zone A Zone B Yield by Zone in the latest 30 production intervals (bbl/day) We have an implied interpretation based on the geochemical differences in the chromatograms.
InfoMetrix B Zone Dominates C Zone Significant Injection Wells A Zone Dominates Field Production Characteristics Well 696, Region 4 Production 23 bbls/day Water 85% 13% Zone A; 17% Zone B; 70% Zone C Perhaps the best way to display the interpretation is by color-coding a map.
InfoMetrix ConclusionsConclusions Source of a crude oil: Chemometric pattern matching is effective in routine geochemical evaluations and multi-step classification procedure is preferable (minimizes classification errors) uGC, GC/MS, GC/MS plus isotopes Reservoir fingerprinting: The techniques can determine if a reservoir is connected to its neighbors, evaluate reservoir mixing and flag unusual samples uGC on peak tables or whole chromatograms