Optimizing reservoir features in oil exploration management based on fusion of soft computing

This paper introduces concepts and algorithms for feature selection, surveys existing feature selection algorithms for classification and clustering, groups and compares different algorithms by a categorizing framework based on search strategies, evaluation criteria, and data mining tasks and provides guidelines in selecting feature selection algorithms. Search strategies include complete ones, sequential ones and random ones. Evaluation criteria includes filter, wrapper and hybrid. Data mining tasks include classification and clustering. Then, a feature selecting platform is proposed as an intermediate step based on the data and requirement of the task. According to the platform and categorizing framework, some appropriate algorithms are compared. At last, an experiment based on data oilsk81, oilsk83, oilsk85 wells of Jianghan oil fields in China was operated by using one of the appropriate algorithms. This algorithm utilizes fusion of soft computing methods to distinguish the key features of reservoir oil-bearing formation and establishes a model with fusion of soft computing methods to forecast these key features. The following part is the process: Firstly, use genetic algorithm (GA) and fuzzy c-means algorithm (GA-FCM) to reduce well log features of oil-bearing formation and to obtain the key features that can describe oil-bearing formation of reservoir. Secondly, fuse genetic algorithm with BP neural network (GA-BP) to construct the fusion model that forecasts these key features. GA-BP searches the inputs and optimal number nodes of hidden layer of BP neural network through GA to choose the optimal structure of BP neural network forecasting model. Then test effectiveness of the forecasting model with recognition accuracy of testing samples. Finally, the optimal model for forecasting key features can be obtained.