RETRACTED ARTICLE: Research on shale gas productivity prediction method based on optimization algorithm

Shale gas, as one of the new natural gas deposits, has been widely concerned. Due to the multi-stage fracturing technology of horizontal wells used in shale gas development, frequent opening and closing of gas wells, and complicated characteristics of gas reservoirs, the problem of productivity prediction has not been well solved. At home and abroad, the empirical formula methods, analytical methods based on seepage theory, and reservoir numerical simulation methods are mainly used for shale gas productivity prediction. The common problem of these methods is that the productivity prediction accuracy is not high and it can not effectively guide shale gas development. In this paper, the traditional productivity prediction method is improved by using machine learning, the characteristics that represent the productivity change of gas wells are selected, and the optimization algorithm with strong classification ability for small sample data is introduced to establish an effective productivity prediction model. The model has been applied to the gas reservoir production prediction of a platform in Chinese Southwest Region and achieved high productivity evaluation accuracy, which proved to be a useful supplement to the traditional productivity prediction methods.