Abstract
Oil and gas exploration and production are critically dependent on well logging to gather essential subsurface data for effective reservoir characterization. Traditional well logging, while invaluable, incurs substantial costs and is typically confined to specific depths, thereby restricting a thorough understanding of reservoir properties across the entire wellbore. Addressing this challenge, this study harnesses the potential of various time series models, including ARIMA, LSTM, bi-directional LSTM, GRU, and simple RNN, to predict porosity in areas of the wellbore not yet reached by logging tools.
This research introduces a novel approach that involves predicting real-time reservoir porosity during drilling, using time series models to extend our knowledge into unlogged intervals. This method leverages existing log data to forecast porosity ahead of current drilling, allowing for real-time data-driven decision-making that can influence drilling directions and strategies. The methodology consists of preprocessing the data, segmenting it into training and testing sets, and training distinct models on porosity with depth serving as the time component. The efficacy of each model is rigorously evaluated using metrics such as mean squared error (MSE) and mean absolute error (MAE), with the most accurate model selected for real-time forecasting.
Results from the study highlight that while all tested models demonstrate capability in predicting subsurface properties, LSTM and bi-directional LSTM, in particular, show superior performance in modeling the complex patterns inherent in geological data. The introduction of a prediction strategy during drilling not only refines the model's accuracy but also revolutionizes how geological data is integrated into operational practices.
In conclusion, this study transcends traditional logging limitations by applying advanced time series forecasting models to predict porosity in the unlogged sections of the wellbore. The innovative approach of real-time forecasting of subsurface porosity during drilling could significantly alter operational dynamics in the oil and gas industry, facilitating more precise reservoir characterization and informed decision-making in field development and management.
