Research On Reservoir Heterogeneity Based On BP Neural Network

Reservoir heterogeneity is one of the most important research content in reservoir description, and most oilfield of our country has entered in high water cut period in the current, developing the potential is more difficulty. Due to the various size of reservoir heterogeneity, the remaining oil in space on a highly complex decentralized state, we need to descript and predict reservoir more precise. The research of reservoir heterogeneity has made significant progress since it was come up, research contents and direction are much wider, research methods and techniques develops well, and from single to comprehensive, from qualitative to quantitative.The Identification of the interbed and physical property parameter (such as porosity and permeability) is the key of the reservoir heterogeneity research. Conventional research method is that combining with logging data, core analysis and laboratory test data firstly, then analyzing four characters of reservoir and getting the lithology classification standards and physical interpretation equation. Lithologic recognition rate is low, and property prediction error is big when using this method, especially in predicting low resistance, high radiation and other unconventional reservoir. One of the main reasons is the actual geological conditions often exhibit nonlinear, cannot be described accurately using the linear model and processing.Artificial neural network technology is one of nonlinear processing method widely regarded, this paper mainly introduces the BP neural network algorithm, summarizes the basic principle of the algorithm and realization method, and uses the method to identify interlining and physical parameters in the study area. On the basis of geological data in the study area, establish proper nonlinear network model through training, to simulate the actual formation of study area, to predict the lithology and physical property of unknown sections.Wenmingzhai Oilfield has been explored and developed for nearly 40 years, having found many sets of oil-bearing series and multiple oil-bearing structures, and it is still one of the key blocks of Zhongyuan Oilfield although it has entered into high water cut development period. Wenmingzhai Oilfield has been developed for many years, it has formed a set of complicated fault block reservoir description and stimulation technology and realized efficient development, but with the Oilfield development degree improved, it needs to strengthen the reservoir heterogeneity research in order to learn the residual oil distribution. This study systematically collect seismic, well logging, laboratory analysis, and comprehensive use reservoir geology principle and sedimentology principle, base on sedimentary research, use BP neural network technology, focus on identifying interbed and physical properties of study area, then analyze the reservoir heterogeneity.Wenmingzhai Oilfield shahejie formation mainly develop meandering stream delta-lake facies, most of the area are in the delta front end, affected by lake, mainly develop sub-water tributary, sheet sands, interdistributary bay and other sedimentary microfacies. The overall size is fine, and mainly develop powder sandstone argillaceous sandstone, sandy mudstone and shale.According to core observation and test data analysis, the study area mainly develop arkose and lithic arkose, containing a small amount of feldspar quartz sandstone. The porosity of reservoir is 26.81% on average, the average permeability is 662.07 mD, belonging to high porosity, high permeability reservoir. The reservoir heterogeneity of study area is strong, especially in the intraformation and plane. Particle size and permeability in the vertical main show as compound rhythm, porosity and permeability in the plane heterogeneity is strong. The permeability variation coefficient is more than 1. The thickness of interbed is large and wide distribution, mainly develop calcareous intercalation and argillaceous interbed.