Optimized Inversion Of Inter-well Connectivity Based On Production Logging Data

In the process of oilfield development,the connectivity between reservoir wells is an important indicator of reservoir evaluation and plays a key role in the design of oilfield management plans.Quantitative interpretation of inter-well connectivity can improve the efficiency of reservoir injection-production and development,which is of great significance.The traditional static and dynamic methods of inter-well connectivity in oil reservoirs are expensive and complicated to operate.This research paper is based on three non-linear data regression analysis algorithms,and fully combines reservoir production logging data to provide reservoir production and development enterprises with an efficient and easy-to-use inter-well oil and gas connectivity evaluation and analysis method.It has very important technical theoretical research value and important practical significance.This dissertation firstly combines the tracer method to complete the preliminary inter-well connectivity exploration of the Yue 5552 Xiang well group,and then uses the multiple linear regression method to verify and analyze the tracer monitoring results in combination with the production logging data.The analysis results show that the conclusion of the multiple linear regression model is consistent with the conclusion of tracer monitoring.Based on the consistency of these two conclusions,he kernel ridge regression model,the random forest model,and the GRNN model(generalized regression neural network)were established respectively.The oil production prediction of the Yue 5552 Xiang well group was completed through the model solving,and the three models The prediction result is excellent in fitting the actual oil production,which provides a guarantee for the follow-up development of the well group.The analysis and research results of the inter-well connectivity inverted by the multiple linear regression model show that the water-driven reservoir signal injection and production system itself is a linear system.When the injected reservoir signal propagates in the linear system,it will be affected by well spacing,The influence of parameters such as compressibility,porosity,permeability,water viscosity of crude oil,the specificity of each vector in the layer,and effective thickness.Finally,the influence of various parameters on the system will be reflected on the output signal.Therefore,the process of solving the inter-well connectivity coefficient based on the multiple-choice linear regression method is the process of solving the coefficient between the input and output of the linear system.The research on the oil production prediction model of the reservoir shows that the conventional linear regression model is no longer suitable for oil production prediction,and the nonlinear relationship between water injection and oil production needs to be considered;the occurrence of the nuclear ridge regression model is to change the variable Through a nonlinear mapping,the nonlinear relationship between becomes a linear relationship and then solved;the random forest model establishes a nonlinear injection-production model of water injection and oil production from the perspective of decision trees;the GRNN model uses nerves The network simulates this non-linear relationship.In theory,he model can simulate any non-linear decision surface.