Research On Prediction Method Of Excess Adsorption Amount Of Methane And Carbon Dioxide On Shale Based On Machine Learning

With the development of big data era,artificial intelligence has been widely used in various disciplines,bringing new exploration methods for classification and regression problems.In the field of shale gas exploitation,carbon dioxide displacement of shale gas(mainly composed of methane)is a green and economical technology,which can not only extract clean energy,but also alleviate global warming by sequestration of carbon dioxide.Among them,it has always been a research hotspot for researchers to make clear the methane reserves and carbon dioxide solid stocks in shale system.For the methane excess adsorption data set,the order of R~2 is XGBoost gradient lifting tree regression(0.999 and 0.985)>multivariate nonlinear regression(0.994 and0.981)>artificial neural network(0.929 and 0.873)>support vector machine regression(-0.328 and-1.09).In XGBoost gradient lifting tree regression algorithm,the value of n＿estimators is 62 andλis 1.2.For carbon dioxide excess adsorption data set,the order of R~2 is XGBoost gradient lifting tree regression(0.999 and 0.966)>artificial neural network(0.991 and 0.955)>multivariate nonlinear regression(0.967and 0.936)>support vector machine regression(0.621 and 0.540).In XGBoost gradient lifting tree regression algorithm,the value of n＿estimators is 62 andλis 1.2.Machine learning algorithm has a wide range of applications,and can consider a variety of geological and environmental factors.The prediction model of excess adsorption capacity of methane and carbon dioxide by shale established in this paper has high accuracy and certain feasibility,which can provide some insights for the actual adsorption capacity of methane and carbon dioxide in geology.