Application Of BP Neural Network In Different Lithology Identification Of Logging Interpretation

The lithology identification of the reservoir, is an important work in the petroleum exploration and development. Conventional logging measures the rock physical parameters of formation, such as the radiation, resistivity, density, porosity, acoustic travel time and so on. These physical parameters of rocks is a single response or comprehensive response of lithology, physical properties and oil-bearing of the formation. Therefore, it is feasible in theory and effective in practice using logging data to classify formation lithology. Compared with other exploration methods include coring, cross-plot. Using logging data (the radiation, resistivity, density, porosity, acoustic travel time and so on) to classify formation lithology is faster and more low-cost. As one of nonlinear processing methods, BP artificial neural network method provides a new method for identify identification. However, presented BP models in the literatures were only tested for a kind of lithology and in a few wells. In this paper, three BP network model for classifying the lithology of sand shale reservoirs, carbonate reservoir and igneous rock reservoir respectively are built, and have been applied in the real logging exploration workThe main content of this paper is shown as follows:Firstly, the logging data and coring data of sand shale reservoirs, carbonate reservoir and igneous rock reservoir are analyzed, and four characteristic parameters are obtained as the inputs of BP neural network model.Secondly, three BP neural network models are designed for identifying different properties reservoir. The results show that the models are effective.Thirdly, the presentd models have been applied to logging interpretation of real data, and the application results show that the presented method can be used in real work.