Application Of Data Mining In The Characterization Of Dominant Seepage Channels

Sandstone reservoir water flooding technology is the most common oil recovery method at present.During the long time of water flooding development,injecting water will affect the reservoir pores,skeleton particles,cements and reservoir fluids.The changes in the temperature and pressure of the oil reservoir result in great changes in the pore structure of the reservoir.The pore throat radius increases as well as the porosity and the permeability increaseing.So,in the reservoir,the high permeability zone and hypertonic belt is easy to produce namely,the large pore channel.In the strata of large pore channels,it has become the predominant seepage channels for injecting water,injecting water along the large channel channeling,resulting in inefficient or ineffective circulation of injected media.The efficiency of injected water is low,water flooding is small and the well is hydrated quickly and seriously affect the displacement efficiency,resulting in significant differences in the remaining oil saturation on the plane,affecting oil recovery and development of improved efficiency.This paper systematically studies the geological mechanism of seepage channel developmen,the change characteristics of logging curve after formed,the distribution of reservoir heterogeneity,and the development of dynamic data identification responding to the seepage channel advantages.This paper studied the corresponding data mining algorithm.The analysis experiment is carried out combined with the actual geological conditions of the research area.the main results and recognition are achieved as follows:(1)Thin section data is processed based on seeds point generation and region growing algorithm.It can quickly identify microscopic characteristics of the reservoir before and after water injection development.It has been clear about the development mechanism of the seepage channel advantages.(2)According to the geological law of different water flooding characteristics,the logging response characteristics of watered-out reservoirs with dominant seepage channels are studied.By using the ratio method and the neural network algorithm,the method of obtaining water saturation in different water flooding levels was tested.Through the analysis of the test results,the water saturation is consistent with the geological understanding.(3)Based on the development of production dynamic data,the development response characteristics of the dominant seepage channel were summarized.According to the production index chart of oil Wells and water injection Wells,it has obvious response characteristics after the formation of the dominant seepage channel.(4)Development of seepage channel advantages is closely related to the reservoir heterogeneit.PCA dimensionality reduction algorithm is used to extract multiple heterogeneous parameters contained reservoir information.The comprehensive index of reservoir heterogeneity is calculated.(5)Thick oil reservoir of Nan3 PuⅠ group,Shaertu Field in Daqing is selected to test data mining algorithm.On the basis of fine geologic research results,the prediction method of seepage channel was tested.The results and the theoretical results fit well with the current development situation.It is shown that the application of data mining method to predict the seepage channel advantage has a certain application value.