Data-driven model for fracturing design optimization: focus on building digital database and production forecast A.D. Morozov a , D.O. Popkov a , V.M. Duplyakov a , R.F. Mutalova a , A.A. Osiptsov a , A.L. Vainshtein a , E.V. Burnaev a , E.V. Shel b , G.V. Paderin b a Skolkovo Institute of Science and Technology (Skoltech), 3 Nobel Street, 143026, Moscow, Russian Federation b Gazpromneft Science & Technology Center, 75-79 liter D Moika River emb., St Petersburg, 190000, Russian Federation Abstract Growing amount of hydraulic fracturing (HF) jobs in the recent two decades resulted in a significant amount of measured data available for construction of predictive models via machine learning (ML). In multistage fractured completions, post-fracturing production analysis (e.g., from production logging tools) reveals evidence that different stages produce very non-uniformly, and up to 30% may not be producing at all due to a combination of geomechanics and fracturing design factors. Hence, there is a significant room for fracturing design optimization. We propose a data-driven model for fracturing design optimization, where the workflow is essentially split into two stages. As a result of the first stage, the present paper summarizes the efforts into the creation of a digital database of field data from several thousands of multistage HF jobs on vertical, inclined and near-horizontal wells from about twenty different oilfields in Western Siberia, Russia. In terms of the number of points (fracturing jobs), the present database is a rare case of an outstandingly representative dataset of about 6000 of data points, compared to typical databases available in the literature, comprising tens or hundreds of points at best. Each point in the data base is based on the vector of 92 input variables (the reservoir, well and the frac design parameters). Production data is characterized by 16 parameters, including the target, cumulative oil production. The focus is made on data gathering from various sources, data preprocessing and development of the architecture of a database as well as solving the production forecast problem via ML. Data preparation has been done using various ML techniques: the problem of missing values in the database is solved with collaborative filtering for data imputation; outliers are removed using visualisation of cluster data structure by t-SNE algorithm. The production forecast problem is solved via CatBoost algorithm. Prediction capability of a model is measured with the coefficient of determination (R 2 ) and reached 81.5%. The inverse problem (selecting an optimum set of fracturing design parameters to maximize production) will be considered in the second part of the study in the following paper, along with a recommendation system for advising DESC and production stimulation engineers on an optimized fracturing design. Keywords: bridging, fracture, particle transport, viscous flow, machine learning, predictive modelling, data collection, design optimization 1. Introduction and problem formulation 1.1. Introductory remarks Hydraulic fracturing (in what follows referred to as HF for brevity) is one of the most widely-used techniques for stim- ulation of oil and gas production from wells drilled in the hydrocarbon-bearing formation [1]. The technology is based on pumping at high pressures the fluid with proppant particles downhole through the tubing, which creates fractures in the reservoir formation. The fractures filled with granular mate- rial of closely packed proppant particles at higher-than-ambient permeability provide highly conductive channels for hydrocar- bons from far field reservoir to the well all the way to sur- face. The technology of HF is used commercially since 1947 in the US and since then the technical complexity of the stimu- lation treatment has made a significant step forward: wells are Email address: a.osiptsov@skoltech.ru (A.A. Osiptsov) drilled directionally with a near-horizontal segment and multi- stage fractured completion. The global aim of this study is to structure and classify ex- isting machine learning (ML) methods and to highlight the ma- jor trends for HF design optimization. Gradual development of fracturing technology is based on the advance in chemistry & material science (fracturing fluids with programmed rheology, proppants, fibers, chemical diverters), mechanical engineering (ball-activated sliding sleeves for accurate stimulation of se- lected zones), and the success of fracturing stems from it being the most cost effective stimulation technique. At the same time, fracturing may be perceived as yet not fully optimized technol- ogy in terms of the ultimate production: up to 30% of fractures in a multi-stage fractured completion are not producing [2, 3]. For example, [4] analyzed distributed production logs from var- ious stages along the near-horizontal well and concluded that al- most one third of all perforation clusters are not contributing to production. The reasons for non-uniform production from var- ious perforation clusters along horizontal wells in a plug-and- Preprint submitted to Journal of Petroleum Science & Engineering. Special Issue: Petroleum Data Science January 20, 2020 arXiv:1910.14499v2 [eess.SY] 17 Jan 2020