Data-Driven Productivity Regularity Research of Fractured Well in Tight Gas Reservoir

Abstract

AbstractThe seepage law of unconventional fractured gas well is relatively complex, and tight gas reservoir is also affected by pressure-sensitive effect and start-up pressure, so it cannot be deduced by conventional seepage theory. Therefore, the results obtained by conventional methods such as mathematical model solving cannot completely accord with the particularity of unconventional gas reservoir. In view of the difficulties and deficiencies in the field production prediction of fractured well productivity in tight gas reservoirs, this paper combined with machine learning algorithm to establish a data-driven prediction model of fractured well productivity in tight gas reservoirs. First, based on the theory of tight gas reservoir seepage, a sample set of factors affecting the productivity of fractured wells was established. Machine learning regression algorithm BP neural network, support vector machine method and GM(1,1) model were used to predict and compare, and a comprehensive mathematical model was established to dig the potential rules between the data. The results show that the accuracy of the data-driven machine learning algorithm is better than that of the traditional methods, and the support vector machine method does not need a large amount of sample data, but has a higher accuracy. This method has been successfully applied in the development of a tight gas field in China, and the research results are of great significance for the rational and effective development of unconventional oil and gas.KeywordsTight gas reservoirProductivity predictingBP neural networksSupport vector machineGM (1,1) model