Research On The Identification Of Dominant Seepage Channels Based On Logging Curves

Daqing Sazhong Development Zone is a typical continental sandstone reservoir.The long-term water flooding of crude oil has caused the injected water to continuously erode the oil layer in the well,causing varying degrees of damage to the pore structure in the oil well reservoir.The dominant seepage channel zone is formed in the reservoir.Part of the injected water will do useless work along the dominant seepage channel,resulting in a decline in the oil production rate of the oil well and a rapid increase in the liquid/oil ratio,which severely restricts the oilfield from extracting the remaining oil.Through the research and analysis of logging principles,conventional logging curves,and the formation mechanism of dominant seepage channels,it is found that the shallow,medium and deep three lateral curve amplitudes decrease,the spontaneous potential curve amplitude increases,and the microelectrode curve decreases in the interval that forms the dominant channel.In response to the characteristics,eight types of logging curves(AC,CAL,R25,RLLD,RLLS,RMG,RMN,SP)are finally selected as the actual experimental data set samples.The composition parameters of the learning sample set and the test sample set required for the experiment.Based on the research of machine learning recognition technology at home and abroad,this subject proposes a method based on improved particle swarm support vector machine to identify the dominant percolation channel.The improved particle swarm algorithm is used to optimize the parameters of the support vector machine to improve the generalization ability of the model and Accuracy.First,select the experimental data from the oilfield measures database,and then preprocess the experimental data.For the corrected logging curve data,use Gaussian distribution to expand the sample,and then use eight feature extraction methods to extract features of the logging curve,and perform linear discriminant analysis on the features of the data to reduce the dimensionality to obtain the experimental data set.The input is the characteristic vector of standardized logging curve data,and the output is whether there is a dominant seepage channel.Use the data set to conduct four different experiments and record the data of the four experiments.The particle swarm support vector machine discriminant model obtained through standardized training was compared with the training model obtained by cross-validation and the particle swarm training model obtained by logging curve.Through the comparison,it was found that the performance of the particle swarm support vector machine discriminant model was obvious.Better than the other two models.Apply the actual results to the oil field,complete the design and development of the oil field related system function modules,and apply them to actual production.