MultiFracSimPPM: A Data-Driven Probabilistic Predictive Model for Hydraulic Fracture Growth from Uniformly and Non-Uniformly-Spaced Perforation Clusters

The thrust of this study is the assessment of using non-uniform-spaced perforation-clusters (as opposed to uniformly-spaced perforation clusters) in order to promote improved multi-frac growth geometries, by reducing stress-induced interference during simultaneous-HF growth. A probabilistic predictive model (PPM) that integrates empirical data obtained from laboratory-scale experiments is overviewed. MultifracSimPPM employs published results from multi-frac growth experiments on transparent gelatin blocks with single-phased perforations and various perforations-spacing arrays. A realistic and dependable expectation for multi-frac geometry was generated through the implementation of a Monte-Carlo simulation scheme, assessing such "limited-entry" technique. Statistical bias is introduced into random probabilistic distributions using this empirical dataset by MultifracSimPPM, resulting in geometric profiles that correspond to experimental findings that disprove the initial hypothesis (non-uniformly-spaced perforation clusters did not perform better than uniformly-spaced perforation clusters). Integrating economic principles could further enhance HF-treatment design, leading to higher post-stimulation hydrocarbon production and improved economics. 1. INTRODUCTION Multistage hydraulic fracturing (HF) operations in horizontal wells enables hydrocarbon production from unconventional, low-permeability shale formations. The key element of HF modelling remains the prediction of the generated HF growth. When it comes to reservoir stimulation using multistage HF treatments, where many HFs are being "pumped" simultaneously within each frac stage, the problem's complexity increases exponentially. Standard treatment design software do not incorporate "real-world" phenomena, such as dominant HF creation and inactive perforation clusters (Michael, 2022). This work presents a probabilistic predictive model (PPM) that incorporates empirical inputs from laboratory-scale experiments. HF geometry is arguably the most critical element of HF-treatment design, with a number of considerations made for influencing multi-frac growth to optimize production and economics. Employing random sampling, the algorithm, which is executed in MATLAB, populates output arrays that contain vectors of the expected HF heights and lengths. The statistical probability of occurrence of the corresponding HF-growth dimensions (length and height) in our laboratory experiments is utilized to populate these input arrays. The width of each HF is determined using the "Perkins-Kern-Nordgren" (PKN) geometry model. Subsequently, the calculated HF length, height, and width are applied to the formation of semi-ellipsoid-shaped HF-growth profiles that illustrate the anticipated final multiple-HF-growth geometry. The homogeneity of multiple-HF growth is likewise quantified by using our novel fracture-geometric-homogeneity index (FGHI), as a metric.