Assessment of Unconventional Resources Opportunities in the Middle East Tethyan Petroleum System in a Transfer Learning Context

Abstract

Decision making in new fields with little data available relies heavily on physics-based simulation models. However, due to a lack of full understanding of the physical processes governing flow in the unconventional resources, data-driven modeling has emerged as an alternative and complimentary tool to create recovery forecasts that honor the available data. Transfer Learning provides an opportunity to start early-stage analysis of the asset before adequate data becomes available. New challenges in the energy industry as well as shifting dynamics in both domestic and global supply and demand has encouraged some of the petroleum exporting countries in the Middle East to strategize the development of unconventional resources. In this research we have developed a data-driven Transfer Learning framework that allows the basin-wide assessment of new shale gas and tight oil prospects. The proposed Transfer Learning method is developed on real-world data from several thousand horizontal multistage wells in the Eagle Ford super-basin in South Texas. In this method we have integrated reservoir engineering domain expertise in the data pre-processing and feature generation steps. We have also considered the temporal and spatial balancing of the training data to assure that the predictive models honor the real practice of unconventional field development. Our full cycle Transfer Learning workflow consists of dimensionality reduction and unsupervised clustering, supervised learning, and hyperparameter fine-tuning. This workflow enables reservoir engineers to experiment with multiple hypothetical scenarios and observe the impact of additional data in the learning process. We use the developed workflow to examine the performance of a data-driven model of the Eagle Ford Basin on potential plays in the Middle East. Existence of all liquid types of oil, condensate and dry gas in the Eagle Ford has resulted in training a model flexible enough to be tested on various types of assets in a new location. We first present the successful deployment of our model within the Eagle Ford. Next, we use the information from major formations such as Tuwaiq Mountain and Hanifa and show the value of a pre-existing model from a fully-developed shale play on achieving acceptable accuracies with minimal information available in a new field. Our model is developed by data types with relatively low resolution that minimizes overfitting effects and allows generalization to different geologies with basin-wide accuracy. This approach allows conducting accelerated assessment of various sections of a large asset to enhance field development planning processes. This is a first example of such an effort on a basin scale that examines the effectiveness of Transfer Learning on some of the major unconventional plays in the Middle East region. This workflow allows investigating the relationship among geologic and petrophysical variables, drilling and completion parameters, and productivity of a large group of wells in a new asset.