Machine Learning Assisted Prediction of Dissolution Spatial Distribution in Volcanic Weathered Crust Reservoirs: A Case Study presented by Da Kun Xiao

Abstract

Characteristics of volcanic reservoirs are quite different from unconventional reservoirs. Especially in northern China, volcanic rocks of the Carboniferous age went under an eruption-weathering-burial process, which led to the formation of weathered crusts. Favorable reservoirs formed in the crust due to the existence of secondary corrosion pores. The main control factor for the spatial distribution of weathered crust reservoirs is the extent of volcanic rock corrosion rather than lithology. We established a correlation between dissolution index, namely a new artificial reservoir parameter interpreted by using image logging data to indicate the extent of corrosion in volcanic rocks, and well test interpretations along well paths. Oil-bearing intervals proved by test results show relative higher dissolution index values. According to this relationship, dissolution index can be a key attribute to help us do the volcanic weathering crust reservoir characterization and prediction. Further more, cross-borehole attributes simulation have to depend on 3D seismic data. However, one single seismic attribute respond to the dissolution index quit bad. There are quit various methods like seismic inversion and so on to get 3D volume of logging interpretation properties using muti-seismic attributes. Seismic-logging integration is an effective approach to ameliorate lateral resolution of reservoir attributes among wells. As a replacement of conventional seismic geostatistical inversion, we developed an alternative artificial intelligent synthetical method to estimate reservoir qualities through geological and seismic attributes. Associate seismic attributes extracted from volcanic rock bodies segmented by geological boundaries first. Then the high accuracy 3D dissolution index volume calculated through deep MLP network (Multi-layer Perceptron) show us that, dissolution spatial distribution has great consistency with well test and well production data.

AAPG Datapages/Search and Discovery Article #90332 © 2018 AAPG International Conference and Exhibition, Cape Town, South Africa, November 4-11, 2018