Real time continuous reservoir strength prediction from grain size distribution information using neural network

Gbenga Oluyemi, Babs Oyeneyin, Chris MacLeod

Research output: Chapter in Book/Report/Conference proceedingConference contribution

1 Citation (Scopus)

Abstract

More than 70% of the world oil and gas are domiciled in unconsolidated reservoir rocks with a high risk of sand production. The nature of these rocks imposes limitation on production rate from highly productive fields, and makes development of marginal fields uneconomic. This, coupled with the harsh business environments which necessitate an agitation for quick returns on investments, poses a great threat to the World economy and particularly Nigerian economy, which is solely dependent on petroleum resources. It is therefore imperative that an intelligent approach be adopted in analysing the potential of a reservoir rock to produce sand at any stage of its life in order to face this challenge headlong. Intelligent sanding potential analysis requires continuous real-time prediction of Unconfined Compressive Strength (UCS) and versatility of prediction techniques. Most of the widely used models for predicting UCS have been reviewed and analysed and found deficient in these regards: they either under-predict or over-predict UCS and are not capable of predicting UCS in real time. The review and analysis was based on the number of parameters, already established as having great influence on UCS, that the selected models account for, and the real time measurability of the parameters. As a solution to this problem, a new neural network model has been developed. The model is based on five input parameters of bulk compressibility, stress path, porosity, median grain size, and sorting; which, on the basis of the review and analysis carried out, have greater influence on UCS, the output parameter. The parameters considered are also measurable and can be obtained easily from logs or empirical relations. The model has been tested and found to perform better; the results of comparison studies are included in this paper as a proof. The model can be used to predict UCS on a continuous basis in real time for the evaluation of sanding potential in any reservoir rock.

Original languageEnglish
Title of host publicationSociety Petroleum Engineers - Nigeria Annual International Conference and Exhibition 2006, NAICE 2006
DOIs
Publication statusPublished - 2006
Event30th Nigeria Annual International Conference and Exhibition 2006, NAICE 2006 - Abuja, Nigeria
Duration: 31 Jul 20062 Aug 2006

Publication series

NameSociety Petroleum Engineers - Nigeria Annual International Conference and Exhibition 2006, NAICE 2006

Conference

Conference30th Nigeria Annual International Conference and Exhibition 2006, NAICE 2006
Country/TerritoryNigeria
CityAbuja
Period31/07/062/08/06

Fingerprint

Dive into the research topics of 'Real time continuous reservoir strength prediction from grain size distribution information using neural network'. Together they form a unique fingerprint.

Cite this