Zvučna karotaža simulirana korištenjem metoda linearne regresije i neuronskih mreža

master's thesis

• Title: Zvučna karotaža simulirana korištenjem metoda linearne regresije i neuronskih mreža : diplomski rad
• Title (english): Acoustic log simulated using linear regsression and neural network methods : master's thesis
• Author: Stijepo Grljević
• Mentor: Jasna Orešković (mentor)
• Committee member: Jasna Orešković (predsjednik povjerenstva)
• Committee member: Marko Cvetković (član povjerenstva)
• Committee member: Ivan Medved (član povjerenstva)
• Granter: University of Zagreb; Faculty of Mining, Geology and Petroleum Engineering; (Department of Geophysical Exploration and Mine Surveying); Zagreb
• Defense date and country: 2022-07-15, Croatia
• Scientific / art field, discipline and subdiscipline: NATURAL SCIENCES; Geology

Abstract

Karotažna mjerenja daju bitne podatke potrebne za karakterizaciju nabušenih stijena, kao i povezivanje bušotinskih i površinskih seizmičkih mjerenja. Zvučna karotaža (AC) pri tome daje podatke potrebne za procjenu poroznosti, izračunavanje akustičnih impedancija potrebnih za konstruiranje sintetskih seizmograma i seizmičku inverziju. Često, a posebno u starim bušotinama, zvučna karotaža nije izvedena ili je izvedena samo na interesantnim intervalima. Softver Interactive Petrophysics sadrži modul Machine Learning u okviru kojeg su alati Neural Network i Multi-Linear regression koji omogućuju predviđanje krivulja karotažnih mjerenja. U ovom radu korišteni su za predviđanje zvučne karotaže na temelju drugih izmjerenih karotažnih krivulja (neutronske karotaže, karotaže gustoće, prirodne gama-radioaktivnosti, karotaže otpornosti). Zvučna karotaža je predviđana u bušotinama polja u Savskoj depresiji (naftno-plinsko polje Stružec), a zatim je uspoređena s izmjerenim krivuljama zvučne karotaže. Rezultati dobiveni objema metodama relativno se dobro podudaraju sa stvarnima, uz manja odstupanja djelomično uzrokovana zasićenjem ugljikovodicima, tankim slojevima ililošom kvalitetom ulaznih podataka i u nekim bušotinama nedostatkom karotažnih krivulja. Također su izračunati sintetski seizmogrami koristeći mjerenu zvučnu karotažu te zvučne karotaže simulirane na dva načina. Dobiveno je vrlo dobro podudaranje u oba slučaja, no nešto bolje poklapanje ostvareno je kod sintetskog seizmograma izračunatog pomoću zvučne karotaže simulirane neuronskom mrežom.

Abstract (english)

Well logging produces essential data necessary for characterization of drilled rocks as well as for comparison of borehole and surface geophysical parameters. Acoustic logging (AC) gives the data needed for porosity estimation, synthetic seismogram calculation and seismic inversion. Acoustic well logging is often omitted, especially in older wells or performed only on interesting depth intervals. Interactive Petrophysics software offers Neural Network and Multi-Linear Regression tools through its Machine Learning Module which provide the prediction of well-logs. In this research, they were used to simulate the AC logs using other well-log curves (neutron log, density log, natural gamma-ray log and resistivity log). Acoustic log prediction was performed for wells in a field located in Sava depression (hydrocarbon field „Stružec“) and compared to actual, measured acoustic well logs. Both methods produced relatively good results, i.e. they show good fitting of the simulated and measured AC-logs. Minor offsets occurred due to the hydrocarbon saturation, thin layers, bad input data quality and the lack of well logs in some wells. Synthetic seismograms were calculated using the measured AC log along with the two simulated ones. Good fitting was obtained in both cases, although the best fit was observed on the synthetic seismogram calculated using the Neural Network-simulated AC-log.

• Keywords: zvučna karotaža; neuronske mreže; linearna regresija; sintetski seizmogram; Interactive Petrophysics
• Keywords (english): well-logging; neural networks; linear regression; synthetic seismogram; Interactive Petrophysics
• Language: croatian
• URN:NBN: urn:nbn:hr:169:110131
• Study programme: Title: Geology; specializations in: Geology of mineral resources and geophysical explorations; Course: Geology of mineral resources and geophysical explorations; Study programme type: university; Study level: graduate; Academic / professional title: magistar/magistra geologije (magistar/magistra geologije)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Open access
• Created on: 2022-10-04 10:49:51