Abstract: The surface gathering and transportation system for coalbed methane (CBM) is prone to liquid accumulation in pipelines due to combined effects of terrain undulation and multiphase flow, leading to significantly increased system pressure drop, reduced transportation efficiency, unscheduled well shutdowns, and other operational risks that hinder efficient development of CBM. Existing approaches primarily rely on empirical formulas based on steady-state assumptions or numerical simulations requiring manual post-processing, making it difficult to automatically identify liquid accumulation locations and classify associated risks. To address these limitations, this study focuses on 15 field-measured pipeline segments from a block in Huabei Oilfield and proposes a three-stage technical framework integrating mechanism modeling, numerical simulation, and automatic identification. Firstly, a simulation database comprising 320 operating scenarios was established using OLGA software, incorporating key parameters such as pipe diameter, flow rate, and elevation difference. Secondly, a Liquid Accumulation Risk Index (IRI) was proposed to quantitatively characterize liquid accumulation risk by weighted fusion of liquid holdup, pressure gradient, and condensation conditions. Finally, an engineering-oriented automatic identification method for liquid accumulation was developed. Field blind-test validation demonstrated that, among five independent pipelines not involved in model development, the method successfully identified 15 field-confirmed liquid accumulation points, achieving an average localization error of less than 15 meters and an overall accuracy of 88.2%.Field application demonstrates that the proposed method effectively reduces unscheduled well shutdowns, lowers compressor energy consumption and manual maintenance costs, and enhances gas recovery efficiency, thereby validating the technical and economic feasibility of an active prevention-and-control paradigm based on “risk identification–dewatering.” The method outperforms conventional empirical formulas and data-driven models in terms of both identification accuracy and engineering practicality, making it particularly suitable for real-world CBM gathering and transportation systems.