Calibration of Stimulated Reservoir Volume (SRV) Estimation Using Continuous Wavelet Transform (CWT) and Advanced Deep Learning with Rate Transient Analysis (RTA): A Case Study from the Marcellus Shale

Estimating the Stimulated Reservoir Volume (SRV) is vital for optimizing well performance and enhancing hydrocarbon recovery in unconventional reservoirs. Accurate SRV estimation offers crucial insights into fracture networks and the extent of reservoir stimulation, which are fundamental for effective reservoir management. Traditional methods, such as microseismic monitoring, rate transient analysis (RTA), and fracture simulation models, provide different perspectives on fracture propagation but are often hindered by high costs and limitations in resolution. This paper introduces a novel approach for modeling dynamic fracture propagation. The proposed method leverages Continuous Wavelet Transform (CWT) to analyze fracture-treating pressure signals, functioning as a "mathematical microscope" to detect subtle pressure changes and trends during the hydraulic fracture propagation, thus offering a more nuanced view of fracture evolution. By continuously convoluting the complex Morlet wavelet with the hydraulic fracture treating pressure data, CWT effectively captures the nuances of fracture dynamics with high resolution, allowing for real-time monitoring and analysis. A unique representation, the normalized CWT scalogram, has been developed for each fracture propagation mode observed during hydraulic fracturing. The proposed framework transforms treating pressure data into a dimensionless, normalized CWT scalogram, providing a detailed analysis of fracture propagation across varying wavelet scales, capturing even the most subtle fracture behaviors. This scalogram is utilized to train a deep learning model with microseismic event data, enabling precise prediction of the microseismic event cloud corresponding to each hydraulic fracturing treatment, particularly within the same geological formation. The method's effectiveness is validated through a comparison of SRV estimates with results from RTA based on three years of production data. The complete dataset of treating pressure, and production data are publicly accessible via the Marcellus Shale Energy and Environment Laboratory (MSEEL). Integrating CWT with deep learning offers a comprehensive and cost-effective tool for enhancing the understanding of dynamic fracture events, leading to more accurate SRV estimations for the Marcellus shale formation. This methodology is scalable and can be extended to other unconventional reservoirs, promising broader applicability in unconventional reservoir management.