Investigation of Ball-Sealer Transport and Diversion in a Horizontal Wellbore Based on Machine Learning

The coupling method of CFD-DEM is used to simulate the transportation and steering blocking process of temporary plugging ball in horizontal wells, to investigate the influence of the position of temporary plugging ball, injection flow rate and injection displacement on the blocking law of temporary plugging ball, and to clarify the sealing mechanism of temporary plugging ball in horizontal wells. The results show that the position of temporary plugging ball is the dominant factor controlling the success of temporary plugging ball sealing, and the closer to the borehole, the higher the probability of sealing. Under high flow rate, the drag force toward the injection hole is larger, and overcoming the axial inertia force and gravity, the temporary plugging ball will have a more obvious tendency to turn, and the probability of seating is higher. Under large displacement, the temporary plugging ball is subjected to stronger drag force in the horizontal direction, and the steering tendency is slightly slower than that of small displacement. Based on SVM and RF two machine learning models to predict the blocking results, analyze and predict different temporary plugging ball position, displacement, injection flow rate under the sealing law, compare the accuracy of SVM and RF two models on the prediction of the blocking results, the results show that the RF model prediction results are better. The advantage of machine learning is that it can learn from the data and automatically adjust the model to improve the prediction accuracy and generalization ability. 1. INTRODUCTION Unconventional reservoirs, such as shale and tight sandstone, are characterized by exceptionally low porosity and permeability. Elevating the intricacy of hydraulic fracturing holds paramount importance in augmenting reservoir productivity. In the context of unconventional reservoir transformation, horizontal well multi-cluster fracturing emerges as a pivotal technique. However, the influence of reservoir heterogeneity often results in disparate expansion among multiple fractures, leading to minimal production contributions from the majority of perforation clusters(X. Li, Chen, Chaudhary, & Novotny, 2005; Liang et al., 2016). Despite the widespread use of temporary plugging and reorientation techniques to encourage uniform fracture expansion, the precise sealing dynamics of temporary plugging balls remain elusive, with on-site applications yielding suboptimal outcomes.(Yuan et al., 2022)through meticulous laboratory experiments, delved into the intricate interplay of factors such as temporary plugging ball density and injection rate on seat sealing.(Tan, Weng, Ahmed, Bukovac, & Enkababian, 2018),employing theoretical model constructs, distilled fundamental principles governing the seat sealing behavior of temporary plugging balls. In the realm of petroleum engineering,(Baki & Dursun, 2022; Peng et al., 2012; Wang et al., 2021) leveraged artificial intelligence to address challenges encompassing the continuous tubing lifespan, bottomhole pressure in multiphase flow, and the prediction of intricate parameters. Notably, machine learning has yet to find application in forecasting the seat sealing behavior of temporary plugging balls. Thus, this paper introduces the avant-garde concept of machine learning to predict the seat sealing dynamics of temporary plugging balls in horizontal wellbores. Adopting the coupled CFD-DEM method, a sophisticated three-dimensional model of a horizontally drilled well with helical perforations is crafted to meticulously simulate authentic field conditions. The study meticulously investigates and analyzes the intricate sealing and reorientation mechanisms of temporary plugging balls in horizontal wellbores, encompassing factors such as ball-sealer position, injection displacement, and perforation flow ratio. The outcomes derived from machine learning numerical simulations are harnessed to prognosticate the seating conditions of the ball-sealers under diverse influencing factors, furnishing a theoretical underpinning for the pragmatic design of temporary plugging applications.