ISI Logo

Shale Production Optimization

Problem Statement

Large number of wells are being completed in Shale assets in several locations throughout the United States. Realizing that the "One Size Fit All" prescription cannot possibly be the right way to continue completion operation, the operator was interested in taking full advantage of the existing data generated from large number of wells that had already been completed. The operator selected to use the field measurements in order to develop and validate a data driven predictive model for the completion operation in its shale asset. The predictive model should be able to assist the operator in optimizing production by identifying best hydraulic fracturing practices in the asset.

Methodology

Intelligent Solutions, Inc. used its IMprove technology (Data-Driven Analytics for Production Optimization) in order to train, calibrate, and validate a data driven predictive model. Type curves based on the data driven model were generated in order to further validate that the model is understanding the underlying physics. Furthermore, the type curves would assist in characterization of the completion operation for optimization purposes. The model was then used in inverse mode in order to be used as an objective function for optimization purposes.

Results

By integrating well construction data with reservoir characteristics (well logs), completion and hydraulic fracturing characteristics, and finally production from each well, a data driven predictive model was developed.

Showing the results from the data driven predictive model for all the wells in the asset.

The wells in the asset are divided into three segments. The first segment include the wells used to train the data driven predictive model. The R2 for the training is about 0.98. The other two segments (calibration and validation) are blind. These wells are not used during the training process. These wells are used to check the predictive capabilities of the data driven predictive model. The R2 for the calibration wells is 0.80 and for the validation wells is 0.82.

Results of the data driven predictive model for training wells.

Results of the data driven predictive model for calibration wells.

Results of the data driven predictive model for validation wells.

Upon validation the predictive model can be used for uncertainty analysis. While keeping all the reservoir characteristics and well construction parameters constant the design (completion) parameters can be varied in order to identify the quality of the hydraulic fracturing practices in any given well. .

Monte Carlo Simulation performed on the predictive model.

Data driven predictive model can be used to generate type curves. The well behaved nature of the type curves that are completely data driven attest to the quality of the model and the fact that it has understood the underlying complex physics of the fluid flow and production from shale in presence of massive multi-stage hydraulic fractures.

Type Curve showing the change in 180 days Cum. production as a function of stimulated lateral length and average injection pressure.

Type Curve showing the change in 180 days Cum. production as a function of stimulated lateral length and Total Clean Volume injected.

The data driven predictive model can be used in inverse mode in order to optimize the hydraulic fracturing practices in a given asset. The computational footprint of the resulting data driven predictive model is small enough that it can be deployed on a smart phone or a tablet for real-time analysis.

The data driven predictive model can be used to design new hydraulic fractures for a well.

The data driven predictive model can be deployed on a smart phone or a tablet for real-time analysis and frac design.