A New Approach To Estimating Ultimate Recovery for Multistage Hydraulically Fractured Horizontal Wells by Utilizing Completion Parameters Using Machine Learning

Reserves estimation is an essential part of developing any reservoir. Predicting the long-term production performance and estimated ultimate recovery (EUR) in unconventional wells has always been a challenge. Developing a reliable and accurate production forecast in the oil and gas industry is mandatory because it plays a crucial part in decision-making. Several methods are used to estimate EUR in the oil and gas industry, and each has its advantages and limitations. Decline curve analysis (DCA) is a traditional reserves estimation technique that is widely used to estimate EUR in conventional reservoirs. However, when it comes to unconventional reservoirs, traditional methods are frequently unreliable for predicting production trends for low-permeability plays. In recent years, many approaches have been developed to accommodate the high complexity of unconventional plays and establish reliable estimates of reserves. This paper provides a methodology to predict EUR for multistage hydraulically fractured horizontal wells that outperforms many current methods, incorporates completion data, and overcomes some of the limitations of using DCA or other traditional methods to forecast production. This new approach is introduced to predict EUR for multistage hydraulically fractured horizontal wells and is presented as a work-flow consisting of production history matching and forecasting using DCA combined with artificial neural network (ANN) predictive models. The developed workflow combines production history data, forecasting using DCA models and completion data to enhance EUR predictions. The predictive models use ANN techniques to predict EUR given short early production history data (3 months to 2 years). The new approach was developed and tested using actual production and completion data from 989 multistage hydraulically fractured horizontal wells from four different formations. Sixteen models were developed (four models for each formation) varying in terms of input parameters, structure, and the production history data period it requires. The developed models showed high accuracy (correlation coefficients of 0.85 to 0.99) in predicting EUR given only 3 months to 2 years of production data. The developed models use production forecasts from different DCA models along with well completion data to improve EUR predictions. Using completion parameters in predicting EUR along with the typical DCA is a major addition provided by this study. The end product of this work is a comprehensive workflow to predict EUR that can be implemented in different formations by using well completion data along with early production history data.