An adaptive ensemble learning by opposite multiverse optimizer and its application in fluid identification for unconventional oil reservoirs

Abstract

Unconventional reservoirs are rich in petroleum resources. Reservoir fluid property identification for these reservoirs is an essential process in unconventional oil reservoir evaluation methods, which is significant for enhancing the reservoir recovery ratio and economic efficiency. However, due to the mutual interference of several factors, identifying the properties of oil and water using traditional reservoir fluid identification methods or a single predictive model for unconventional oil reservoirs is inadequate in accuracy. In this paper, we propose a new ensemble learning model that combines 12 base learners using the multiverse optimizer to improve the accuracy of reservoir fluid identification for unconventional reservoirs. The experimental results show that the overall classification accuracy of the adaptive ensemble learning by opposite multiverse optimizer (AIL-OMO) is 0.85. Compared with six conventional reservoir fluid identification models, AIL-OMO achieved high accuracy on classifying dry layers, oil–water layers, and oil layers, with accuracy rates of 94.33%, 90.46%, and 90.66%. For each model, the identification of the water layer is not accurate enough, which may be due to the classification confusion caused by noise interference in the logging curves of the water layer in unconventional reservoirs.

Nb-Ta-Sn Oxides From Lithium-Beryllium-Tantalum Pegmatite Deposits of the Kolmozero-Voronja Belt, NW Russia: Implica-Tions for Tracing Ore-Forming Processes and Mineralization Signatures

“…Indeed, the Kolmozero sample has an elevated concentration of CO 2 (0.25 wt.% vs. <0.01-0.02 wt.% in low-grade samples (see Table 1)). Due to the strong depolymerization effect of CO 2 , a high CO 2 content causes rapid separation of volatile-rich melts from the initial magma, through immiscibility [66]. At this stage, the activity of rare metals increased.…”