Well Performance from Numerical Methods to Machine Learning Approach: Applications in Multiple Fractured Shale Reservoirs

Horizontal well fracturing technology is widely used in unconventional reservoirs such as tight or shale oil and gas reservoirs. Meanwhile, the potential of enhanced recovery (EOR) methods including huff-n-puff miscible injection are to further increase The complexities hydraulic fracture properties multiphase flow make it difficult time-consuming understand performance (i.e., production) fractured reservoirs, especially when using conventional numerical methods. Therefore, this paper, two developed bridge gap by machine learning technique forecast production on EOR pilot projects. first method artificial neural network, through which we can analyze big data from underlying patterns relationships. A bunch factors contained parameters, completion, data. Then, feature selection performed determine key factors. Finally, network relationship between performance. second time series analysis. Since reservoir function fluid pressure, quite suitable apply analysis Training test over 10000 wells different Bakken, Eagle Ford, Barnett. results demonstrate that there a good match available predicated overall R values both above 0.8, indicating provide reliable for prediction Especially, dealing with field cases, injection, Time more accurate than network. This paper presents thorough feasibility multiple Instead methods, also provides robust way meaningful reference evaluation