Development of Brittle Shale Fracture Network Model - Part 2: What is the Value of SRV?

Abstract

Abstract This paper discusses the workflow for developing a model to predict gas production for hydraulic fracture stages in a brittle shale environment. In our last paper (SPE 163829), we developed a model to predict fracture network geometry (width, length, and height) obtained from microseismic data for various fracture treatment designs. In this study, using both microseismic data and fracture/completion data; frac treatment production is estimated. The well database used for this purpose is comprised of fracture design parameters including treatment volumes, rates, proppant mass and size; well properties include perforation interval length, and perforation depth. The goal of this study is to provide insight into factors affecting well production in a brittle shale environment. Initial data screening demonstrated massive data scattering and as a result data mining techniques are employed to find possible hidden relationships to explain the nature of the data. Also, using sensitivity analysis on the predictive model, improvements in the current fracture designs and completion schemes in Barnett shale are made. The database is examined from different aspects using various data mining approaches. After screening and preprocessing the data, non-process affected outlier wells are removed from the dataset. Then, a forward predictive neural network model is trained with fracture design and well data parameters as inputs and well cumulative gas production per stage as outputs. Neural networks are trained with the help of genetic algorithm (GA). The sensitivity study on the trained network provided many insights about well completion and stimulation strategies. Recommendations on how to improve fracture designs and well completion schemes are provided based on sensitivity analysis on the neural networks. Results of neural network modeling in Barnett shale are compared to other gas producing shale assets such as Fayetteville and Haynesville shale to compare the findings. The results of this work potentially help understanding of completion and fracture treatment designs on well productivity in gas producing shale assets. This will potentially help operators understand how to more effectively design frac treatments and/or reduce the operational costs associated with well completion in a brittle shale environment. Considering the fact that the relationship between stimulated reservoir volume and production is not entirely understood, this work may shed some lights on the aforementioned issue.