关键词: BP 神经网络, 改进BP 算法, 网络仿真训练, MATLAB, 渗透率预测

Abstract:

Low permeability reservoirs always have a thicker oil-water transitional zone, while mobile fluid usually is in form of tow-phase flow. Therefore, water and oil saturation will decrease undergoing the release of pressure and outgassing as the process of sealed coring, which results in the difficulty in acquiring oil saturation of sealed core. In practice, evaporation experiments were used to correct it, however, data of laboratory of sealed core will be wasted when there is no evaporation experiment. There is a method which can correct the sealed core saturation of low permeability reservoirs with no evaporation experiment. First of all, classify the reservoir types according to reservoir quality index based on fairly well relationship between reservoir quality index and the fluid loss rate of sealed coring. Then study the correction factor of different types of reservoirs to obtain different correction factors. According to the analysis of laboratory and production data, it is proved that the correction method for obtaining initial saturation of low permeability reservoirs based on pore structure classification is effective. So the valuable sealed coring data with no evaporation experiment can be made use for obtain initial saturation, and provide criterion for logging interpretation.

Key words: BP neural network, improved BP algorithm, network simulation training, MATLAB, permeability prediction