文献类型：会议论文

英文题名：Optimization of Exploration Prospects Based on Ant Colony Algorithm and XGBoost Combined Optimization Model

作者：Song, Mengxin[1]; Xu, Bingxin[2]; Feng, Mei[1]; Fu, Xinxi[1]

第一作者：Song, Mengxin

机构：[1] Research Institute of Petroleum Exploration and Development RIPED, CNPC; [2] Beijing Key Laboratory of Information Service Engineering, Beijing Union University, China

第一机构：Research Institute of Petroleum Exploration and Development RIPED, CNPC

会议论文集：Society of Petroleum Engineers - Abu Dhabi International Petroleum Exhibition and Conference, ADIP 2021

会议日期：November 15, 2021 - November 18, 2021

会议地点：Abu Dhabi, United arab emirates

语种：英文

外文关键词：Gasoline - Learning systems - Classification (of information) - Ant colony optimization - Decision support systems - Forecasting - Geology - Petroleum prospecting

摘要：Traditional exploration prospect optimization is uncertain due to human factor, the primary reason of that problem is the complex nonlinear relationship between trap quality and related geological factors. Some researchers proposed use artificial neural network (ANN) to solve the problem of the comprehensive geological evaluation of traps, because ANN can describe the nonlinear relationship of multiple geological factors. Considering ANN has some drawbacks, such as it is need lots of parameters for training, and the learning process can not be observed. In this paper we proposed a combined optimization model to accomplish optimization of exploration prospects, and express the affinity order between the prospects and its related geological factors, also can provide the data support for exploration. Based on trap data of an oilfield in Africa, there are 12 geological factors related to trap quality, including trap coefficient, trap depth, trap scale, trap area, Reservoir coefficient, Preservation coefficient, hydrocarbon source coefficient, resources etc.. The ant colony algorithm is used for feature selection, and irrelevant and redundant features are eliminated through multiple iterations, making it suitable for model processing and improving training speed. Based on ant colony algorithm, we get the key parameters for XGBoost model training, namely trap area, reservoir coefficient, preservation coefficient, resource, and the key features are used in XGBoost model for training and prediction. Finally, we compared our prediction results with expert prediction, the error is 0. In this paper, we proposed a combined optimization model based on ant colony algorithm and XGBoost for exploration prospect optimization. We recognized the key geological factors and different characteristic rules for exploration prospect optimization, in the process of optimization, ant colony discards the bad features that interfere with classification and recognition, and retains the features that contribute greatly to classification. In comprehensive geological evaluate of trap, the proposed combined optimization model is suitable for complicated nonlinear geological relationship, and express the affinity order between the prospects, the proposed method can work as an auxiliary way in petroleum exploration, also the proposed method can provide decision support for exploration prospect optimization, and finally can fulfill cost decreasing and benefit increasing. ? Copyright 2021, Society of Petroleum Engineers