Shale gas well production evaluation and prediction based on machine learning

Well production evaluation and prediction are of great significance to the efficient development of shale gas. The assumptions of existing analytical models are more different from actual shale gas wells, and the numerical model has the problems of high calculation difficulty, low efficiency and high uncertainty, which makes the prediction of shale gas well production more difficult. In this paper, the production rates of the gas wells in A shale gas reservoir were evaluated and predicted by integrating the geological, drilling, fracturing and production data in the whole cycle of shale gas production and considering geological factors and engineering factors comprehensively based on machine learning methods. Firstly, the initial data was processed based on themissing value interpolation, correlation analysis, outlier processing and principle component analysis to reduce data noise. Then, the production rates of shale gas wells were evaluated using the cluster analysis method, and the dominated factors influencing the production rates of the gas wells in A shale gas reservoir were analyzed. Finally, the production rates of the gas wells in A shale gas reservoir were predicted by using the random forest method. It is indicated that high, moderate and low production wells in A shale gas reservoir account for 36.4%, 37.8% and 25.8%, respectively, and fracturing factor has the greatest influence on the evaluation results of the production rates of the gas wells in A shale gas reservoir. The prediction accuracy of shale gas well production after parameter adjustment is up to 90%, indicating better prediction accuracy. In conclusion, this proposed model can be used to predict production rates of shale gas wells.