Application Of Multi-Kernel Function Method In Reservoir Lithology Identification

The existing main oil fields in our country,basically has entered the mid late stage of the development,the remaining recoverable oil and gas reservoir is often unconventional oil and gas formation condition complex reservoir,mining is very difficult.Accurate identification of reservoir lithology is of great significance for reservoir evaluation.It can accelerate the process of exploration and development and promote the stable production of oil and gas.In this paper,a new reservoir lithology identification model is proposed,which is based on the abundant geological data and the artificial intelligence technology,provide strong support for the exploration and development of oil and gas fields.This paper analyzes the method and principle of the current reservoir lithology identification technology,the most accurate method of lithology identification is identifying core slices,but the core cost is very high,time-consuming,and it’s not realistic for each single well,and can’t meet the requirement of the actual work.Then the clustering and principal component analysis methods came up,these methods solved the core problem and achieved good results,but the accuracy of lithology identification is not enough.To solve these problems,this paper proposes an intelligent recognition method based on artificial intelligence technology.This paper mainly uses the logging curves to identify lithology,the input and output of the conventional intelligent model are geometric points,without considering the logging curve varies with the depth of transformation,so we need the model can process the direct signal.Part of the reservoir heterogeneity in our actual work area is serious,resulting in lithology recognition accuracy is not high enough,this paper proposes a multi-kernel function method,using multi-kernel functions more accurately describe the characteristics of lithologic reservoir to improve the recognition rate.In this paper,two kinds of intelligent models are proposed,which are multi-kernel process support vector machine and multi-scale-kernel radial basis function process neural network.Multi-kernel model improved the complex signal feature representation capability by the mechanism,and the use of intelligent algorithm to optimize the parameters of the model,to achieve higher recognition accuracy.Based on the artificial intelligence technology,this thesis constructs the software prototype system which can be applied in practice,which achieved good results through the experiment with the actual data.The research results of this paper are of practical significance to the exploration and development of oil and gas fields,which can help in actual work,have theoretical value and practical application value.