Title: Prediction of reservoir gas-oil ratio based on PSO-ELM algorithm

Authors: Haibo Liang; Xia Zhang; Yi Yang; Jialing Zou

Addresses: School of Mechanical Engineering, Southwest Petroleum University, Chengdu 610500, China ' School of Mechanical Engineering, Southwest Petroleum University, Chengdu 610500, China ' Integrated Logging Service Department, CFBGC China France Bohai Geoservices Co., Ltd, Tianjin 300450, China ' School of Mechanical Engineering, Southwest Petroleum University, Chengdu 610500, China

Abstract: Fluid identification is the key to reservoir's characterisation in oil exploration, and the gas-oil ratio (GOR) is an important parameter for fluid's property. Traditional methods are mainly based on the artificial division, which can only be used for simple case. For this issue, this study proposes a predictive model of extreme learning machine (ELM) optimised by particle swarm optimisation (PSO). Firstly, the characteristic parameters are obtained by fluids logging with real-time data analysed based on improved methods, and the characteristic vectors are obtained by Pearson correlation and random forest. Then, the prediction model of extreme learning machine was established, and two decisive parameters were optimised by particle swarm optimisation. Finally, the GOR of the oil layer and condensate gas layer were predicted, and a comparison of PSO-ELM and some other algorithms was performed. The result shows, PSO-ELM model can predict the GOR with high precision, providing theoretical support for complex fluid identification.

Keywords: gas logging; random forest; RF; extreme learning machine; ELM; particle swarm optimisation; PSO; gas-oil ratio; GOR.

DOI: 10.1504/IJPE.2024.142131

International Journal of Petroleum Engineering, 2024 Vol.4 No.2, pp.149 - 173

Received: 31 Oct 2023
Accepted: 16 Feb 2024
Published online: 08 Oct 2024 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article