Intelligent Representation Of Sedimentary Facies Based On Ensemble Learning

Sedimentary microfacies is one of the core contents of reservoir characterization.The types and distribution of sedimentary microfacies often determine the reservoir quality difference and oil and gas distribution.Traditional single-well sedimentary microfacies division is based on drilling coring and analysis of laboratory data to establish sedimentary microfacies model and logging facies markers,and on this basis,the types and vertical evolution sequence of sedimentary microfacies can be determined by logging curve response characteristics such as amplitude,shape and smoothness.However,as exploration and development progresses and the number of Wells drilled increases,the labor cost and time cost of this method will increase simultaneously.At the same time,it is difficult for different researchers to establish a unified interpretation standard due to the large number of human factors affecting the interpretation of the conclusion.Therefore,based on the drilling coring and analysis of laboratory data,logging data,etc.,and through the deep integration of artificial intelligence and sedimentology theory,this paper generates a machine learning model training sample base and builds an intelligent characterization model of single well sedimentary microfacies based on BILSTM network.The intelligent characterization of single well sedimentary microfacies can be divided into three steps:(1)through the improvement of u-net network in semantic segmentation field,the intelligent classification model of sedimentary units based on logging data form is constructed to realize the vertical division of sedimentary units;(2)Considering the vertical sequence of logging response and the correlation information of adjacent data,a lithology discrimination model based on TCN time convolution network and CRF conditional random field was constructed to realize lithology intelligent identification;(3)Using the integrated learning idea,the Bi-LSTM bidirectional long and short-term memory neural network model architecture is modified to realize the intelligent characterization of single well sedimentary microfacies.Taking The Karamay Formation of Triassic in Lukeqin Oilfield as the research object,the intelligent characterization method of single well sedimentary microfacies was tested based on Bi-LSTM network,and the distribution law of sedimentary microfacies in target strata was analyzed.The results show that this method not only improves the accuracy of sedimentary facies identification by logging data,but also improves the interpretability and reliability of the discriminant model by introducing geological prior knowledge.