Long-term production forecast in tight and shale reservoirs: Adapting probability density functions for decline curve analysis

Abstract

Decline curve analysis (DCA) is a popular method for predicting the production of oil and gas wells over time. It involves using historical data to match a mathematical function with the production data, which can then be used to forecast future production. However, DCA in unconventional tight and shale reservoirs poses unique challenges. Unlike conventional reservoirs, these unconventional reservoirs exhibit a production decline that is both faster and more erratic. This behavior, combined with the inherent complexities of the reservoir properties and fluid flow mechanisms, makes it challenging for traditional DCA models to provide accurate predictions of future production.This study proposes a new approach for forecasting long-term production in tight and shale reservoirs, using probability density functions (PDFs) that have been adapted and evaluated for this purpose. The PDF-based DCA models were tested against traditional DCA models using data from over 80 wells in different unconventional basins across North America.Results show that the PDF-based models outperformed the traditional models, with the Dagum and Beta prime models achieving the highest performance scores on various statistical measures, especially in wells with limited production history for model training. To ensure accuracy, appropriate statistical tests should be used to validate the PDF-based decline curve analysis, such as calculating the mean absolute error (MAE) or other relevant metrics.Our research introduces a novel method of adapting PDFs for decline curve analysis, specifically tailored for tight and shale reservoirs. This adaptation offers a more accurate and comprehensive prediction of future production by effectively capturing the inherent uncertainty and variability seen in unconventional reservoirs. One of the standout features of our proposed PDF-based approach is its versatility. Unlike traditional models that may struggle with erratic data, our models can seamlessly fit production data, whether it possesses peaks or not. This approach could be valuable for producers and investors who want to make informed decisions about drilling and production in unconventional reservoirs.