基于机器学习可解释性算法的石嘴山矿区煤层气井排采主控因素分析

Analysis of Main Control Factors of Coalbed Methane Well Drainage in Shizuishan Mining Area Based on Machine Learning Interpretability Algorithm

文章基于宁夏石嘴山矿区3口煤层气直井排采资料,利用机器学习算法分析了排采日报记录的8个变量对日产气量的影响。结果表明:按照对煤层气井日产气量影响强弱排序,10个变量依次为井底流压、套压、液位、冲次、电流、泵效、日产水量、累计产水量、累计产气量和日流压降幅,累计产水量、累计产气量与冲次对煤层气井日产气量存在正效应影响,井底流压和液位对煤层气井日产气量存在负效应影响,其余参数对煤层气井日产气量影响微弱。

Based on the drainage data of three coalbed methane vertical wells in Shizuishan Mining Area in Ningxia,this paper employs machine learning algorithm to analyze the impact of eight variables recorded in the daily production reposts on daily gas production.The results indicate that,according to the strength of the influence on daily gas production from coalbed methane wells,the 10 variables are bottom-hole flow pressure,casing pressure,liquid level,stroke times,electric current,pump efficiency,daily water production,cumulative water production,cumulative gas production and daily flow pressure reduction.Besides,cumulative water production,cumulative gas production and stroke times have a positive effect on the daily gas production of coalbed methane wells,while the bottom-hole flow pressure and liquid level have a negative effect.The remaining parameters have a weak effect on the daily gas production of coalbed methane wells.