Research On Productivity Prediction Model And Influencing Factors Of Changning Shale Gas Horizontal Well Based On Data Mining

The Changning block is one of the first identified shale gas key production areas in China.It guarantees China’s energy structure adjustment.The single well production of shale gas wells in Changning block is affected by multiple reservoir and engineering parameters.The various parameters have largely different influence on the yield and the exact impact is not clear.In order to effectively select reservoir and engineering sweet points and increase single well production,it is necessary to use the actual field data to study the influence of reservoir and engineering parameters on the yield.This paper starts with the reservoir and engineering parameters affecting the shale gas test output in Changning block,selected test yield as the capacity characterization parameter.Pearson correlation coefficient indicates that the linear relationship between test yield and related parameters is not obvious The well quality analysis based on the Fuzzy Set Theory shows that porosity,gas content,actual fracturing section length and well quality are positively correlated.The grey relation analysis method is used to evaluate the importance of the parameters in test yield,demonstrating that the actual fracturing section length has the greatest impact on the test yield and that the fracturing process has a significant impact on the test yield.The shale gas well fuzzy pattern recognition model is established,and the blind well test results show more than 80% accuracy of the model identifying the poor and average wells.Finally,based on BP neural network and its optimization methods,the shale gas well test yield prediction model is established.The model training fit reaches 98% with 81% of the overall fitness.The sensitivity analysis of shale gas reservoir and engineering parameters was carried out by using the test yield prediction model.The analysis results were consistent with the actual situation,which proved the good model reliability.The research results are expected to provide guidance for the optimization of on-site reservoir and engineering sweet points selection.