As the utilization of the natural gas all over the world has been increased, it has become more important to increase the gas production worldwide. Accordingly, many efforts have been expended to improve the production techniques such as using multi-lateral well techniques, which are capable to accomplish higher production rates. The utilization of the multi-lateral well techniques increases the cumulative fluid recovery, decreases the environmental impacts and decreases the drilling and completion costs.The objective of this work is to generate artificial neural network tools that are capable of providing the necessary knowledge to evaluate the utilization of dual horizontal well technique in tight gas reservoirs. The predicted data include flow rate profile, gas recovery profile, dual horizontal well configuration and pattern size. Two artificial neural networks, forward and inverse, were developed to achieve the main objective. The forward ANN was developed to predict the flow rate and the gas recovery profiles for the chosen dual horizontal well configuration and pattern size over a specified production period of 50 years. The inverse ANN was developed to predict the dual horizontal well configuration and pattern size that can be used to achieve a desired gas recovery over a specified production period of 50 years. The Monte Carlo simulation has been implemented to quantify the probability of meeting the desired recovery.The ANNs were trained, validated and tested using the training data generated by a commercial simulator. The developed forward tool was tested and gave a mean square error of 7.5%, while the inverse tool was tested yielding a mean square error of 9.8%. The developed tools are capable of comparing thousands of different input combinations in a much more rapid manner as compared to commercial simulator.
All Science Journal Classification (ASJC) codes
- Fuel Technology
- Geotechnical Engineering and Engineering Geology