Kolmogorov–Arnold Network‐Enhanced Timeseries Networks for Dynamic Production Prediction in Carbon Capture, Utilization, and Storage‐Enhanced Oil Recovery Projects
Year: 2025
Type: article
Abstract: ABSTRACT Obtaining fast, reliable, and low‐cost predictions in the petroleum industry is an important task in reservoir engineering that can help develop future development plans efficiently and increase recovery rates. In this paper, the Kolmogorov–Arnold Network enhanced time‐series net (KAN) is designed for predicting oil production in Carbon Capture, Utilization, and Storage‐Enhanced Oil Recovery scenarios. By comparing algorithms of the same type, the study revealed significant advantages of KANs, such as improved prediction accuracy and improved parameter efficiency. Targets demonstrate that KANs consistently surpassed other techniques by exhibiting lower error metrics, indicating more accurate predictions. The study concludes that KANs' effectiveness and efficiency position them as a viable alternative to conventional networks.
Source: Energy Science & Engineering
DOI: 10.1002/ese3.70357
Authors: Mingguo Peng, Quan Shi, Qiu Li, Song Deng, Chengguo Liu, Guo‐Dong Wang, Ya‐Li Liu, Ruitong Wei
Language: English
Subfield: Ocean Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $2,780
Funder: Government of Jiangsu Province
OpenAlex ID: W4416063193