Research On Super-resolution Method Of Logging Curves Based On Structural Information Embedding

The continuous exploitation of conventional oil and gas resources over a long period of time has resulted in reserves that can no longer meet the current global energy demand.Improving the level of geological characterization and analysis is a top priority for countries around the world in their unconventional reservoir development efforts.Obviously,since logging data is one of the most important data for obtaining fine reservoir models,obtaining logging data with sufficient vertical resolution has been an important problem and challenge.Therefore,the use of low cost and less time consuming signal processing technology to improve the vertical resolution of logging has been a research hotspot in related literature.To this end,this paper focuses on researching a super-resolution method for logging curves by embedding structural information and using signal processing techniques,machine learning,and other approaches to exploit the self-similar fractal features of logging curves.In this paper,a multi-view and multi-scale logging super-resolution method based on fractal theory is proposed to solve the problems that traditional methods to improve logging resolution do not consider the heterogeneity of geological reservoir and some machine learning methods ignore the fractal structure characteristics of logging data.On the basis of improving the resolution of logging curve by traditional fractal interpolation method,this method constructs different iterative function systems through piecewise fractal interpolation to mine the structural information of each section of logging curve,uses different filtering methods to process the interpolation results at different scales,and combines the longitudinal semantic information mining capabilities of the Long and Short Term Memory(LSTM)network to achieve the utilization of logging curve structural information and timing information.Finally,a new super-resolution method for logging curves was implemented.Through the super-resolution experiment of actual logging data,multiple logging curves of different Wells are tested with 4-fold super-resolution.And this method exhibits higher accuracy than other methods,as evidenced by the results.Since the value of the vertical scale factor in fractal interpolation directly affects the accuracy of the interpolation results,the above method generates the vertical scale factor randomly.Although the interpolation results have the same trend as the measured data,excessive noise requires subsequent filtering processing.Therefore,a fractal multiscale well logging data super-resolution method based on Grey Wolf Optimizer is proposed.The Grey Wolf Optimizer has the advantages of self-adaptation,strong global search ability,good robustness,etc.It can find the optimal solution of vertical scaling factor within the solution range of the problem.This method not only improves the accuracy of interpolation results,but also eliminates the filtering process and the structural information loss caused by this step.The actual logging data is used to verify this method’s effectiveness through 2-fold and4-fold super-resolution experiments.From the quantitative evaluation results,it can be seen that the indexes of this method are significantly improved compared with those of the above methods,and it is applicable to other logging curves.