Research On Key Technology Of Reservoir Geological Evaluation Based On Machine Learning

In recent years,with the increasing exploration degree of oil and gas resources,the exploitation process is becoming more and more complex,and the exploration focus is gradually changing from conventional oil and gas reservoirs to complex oil and gas reservoirs.The geological conditions of complex oil and gas reservoirs are more complex,the amount of data collected is large and complex,the traditional reservoir geological evaluation technology is difficult to accurately analyze these valuable data,and the current oil and gas market competition intensifies,the requirements of exploration efficiency and technical means are also constantly improving.Faced with the rapid upgrading of exploration and development problems,researchers gradually break the inherent mode of thinking and begin to use machine learning and other new artificial intelligence technologies and methods to guide the exploration and development of oil and gas resources.Papers in the research and analysis of the various machine learning method based on oil and gas resources in the field of application,combining research status at home and abroad,for single model generalization performance is poor and low accuracy problem,using the ensemble learning combined random forest(RF)algorithm,light gradient boosting machine(Light GBM)algorithm,improved grey relation analysis and TOPSIS(GRA＿TOPSIS)algorithm built a parallel heterogeneous integration based on CAPSO algorithm learning model for comprehensive evaluation of reservoir.Experimental results show that this model has a better evaluation effect.Considering the adverse effects of redundant features on the evaluation results,aiming at the problems that traditional feature selection methods can not effectively remove redundant features and ignore the interaction between features,a hybrid feature selection algorithm based on interactive information is proposed to screen geological attributes by combining the advantages of filter method and wrapper method.Experimental results show that the algorithm can not only remove redundant features effectively,but also has strong stability.Then,considering that the parameters and weight values involved in the model will have a great impact on the performance of the model,aiming at the problems of poor accuracy and easy to fall into local optimum when particle swarm optimization algorithm is used for optimization,combined with chaos theory,a chaos adaptive particle swarm optimization algorithm is proposed for model parameters and weight optimization.The experimental results show that the algorithm has strong stability,high convergence accuracy and is not easy to fall into the local optimal solution.The results of geological evaluation in Dai 1 member of Yong’an area show that the accuracy of the evaluation results is 87.5%,which is basically consistent with the actual drilling results,and has good practical application value and prospect.