摘要：W ith the continuous development of gas fields, the increasing number of liquidaccumulating gas w ells has a greater im pact on the gas production of gas reservoirs. How to predict the liquid-accum ulating state of gas w ells efficiently and quickly is of great significance for the rational development of gas reservoirs and the improvement of gas reservoir recovery. M any prediction models of gas w ell flu id accumulation based on critical flow theory and em pirical form ula have been proposed by the previous, but in the actual application process, the calculation results of different models have large deviations, and the general applicability of the calculation method is not strong, which brings different degrees of flaws to the actual production operation and management. This paper presents a new method based on machine learning for predicting liquid accumulation in gas wells. Based on the analysis of influencing factors and prediction model of liquid accumulation, this method makes use of open experimental data and actual production data, and compares it w ith main prediction models of liquid accumulation. The calculation results show that the method not only sim plifies the traditional complex mechanism study, but also has a high prediction result of gas w ell flu id accumulation, which provides effective theoretical support and reasonable guidance for the efficient development of gas fields and decision-m aking deployment.
关键词：O il & gas development;integrated learning environment;gas well;model training;fluid accumulation prediction
