Shale gas well productivity potential evaluation based on data-driven methods: case study in the WY block

Abstract

Evaluating the productivity potential of shale gas well before fracturing reformation is imperative due to the complex fracturing mechanism and high operation investment. However, conventional single-factor analysis method has been unable to meet the demand of productivity potential evaluation due to the numerous and intricate influencing factors. In this paper, a data-driven-based approach is proposed based on the data of 282 shale gas wells in WY block. LightGBM is used to conduct feature ranking, K-means is utilized to classify wells and evaluate gas productivity according to geological features and fracturing operating parameters, and production optimization is realized through random forest. The experimental results show that shale gas productivity potential is basically determined by geological condition for the total influence weights of geologic properties take the proportion of 0.64 and that of engineering attributes is 0.36. The difference between each category of well is more obvious when the cluster number of well is four. Meanwhile, those low production wells with good geological conditions but unreasonable fracturing schemes have the greatest optimization space. The model constructed in this paper can classify shale gas wells according to their productivity differences, help providing suggestions for engineers on productivity evaluation and the design of fracturing operating parameters of shale gas well.