Predict Continuous Total Organic Carbon Content by Support Vector Regression
J. Rui, H. Zhang and Q. Guo
Event name: 81st EAGE Conference and Exhibition 2019
Session: Poster: Unconventional A
Publication date: 03 June 2019
Info: Extended abstract, PDF ( 707.44Kb )
Price: € 20
Total organic carbon (TOC) is a significant factor to evaluate hydrocarbon potential of the unconventional reservoir, which is difficult to obtain the continuous and accurate prediction result by traditional methods. Support vector machine is an efficient supervised learning artificial intelligence algorithm, which performs well in classification and regression. In this paper, we use the support vector regress machine to predict TOC content through wireline log data. Through the particle swarm optimization algorithm, we get reliable model parameters judged by cross validation, and then we utilize the SVR and model parameters selected by PSO to predict continuous TOC content. Finally, an improved ∆logR method is used for TOC content prediction, whose result is worse than PSO-SVR-derived TOC content result. This study shows that SVR is efficient and reliable for TOC content prediction whose result has high correlation coefficient compare to traditional ∆logR method.