Prediction by reservoir porosity using micro-seismic attribute analysis by machine learning algorithms in an Iraqi Oil Field.

Abstract

The current work deals with an intelligent application that uses machine learning to analyze and attribute the
resulting seismic data to improve and predict the locations of exploration, drilling, and production operations in petrophysical
oil. Statistical analyzes of exploratory data analysis (EDA) were used to extract seismic features. This follows the application
of two intelligent approaches of Recurrent Neural Networks (RNN) and K-Nearest Neighbors (KNN) to predict porosity. The
parameters of seismic data with different seismic attributes were detected at the site by the seismic time-series method. The
long-term memory (LSTM) algorithm is the most appropriate way to handle serial data. LSTM has a high capacity for data
structure manipulation, which is applied for porosity prediction. Which depends a lot on choosing the best attributes? The two
approaches evaluate by absolute error (MAE) and root means square error (RMSE). The results for both models used showed
that using (LSTM) is more effective than using (KNN) in predicting porosity through seismic data, where the mean absolute
error was obtained. (MAE) 0.017, while with KNN the mean absolute error (MAE) is 0.260 and the results showed that the
model used can predict porosity very effectively.