Research On Well Logging Evaluation Of Fractured Carbonate Reservoir In Tahe Oilfield

To study the well logging evaluation method of fractured carbonate reservoir in Tahe oilfield,the Ordovician target reservoir in Tahe oilfield was selected as the research area to carry out the research.Based on the thought of ‘Core Scale Logging' and ‘Imaging Calibrate Conventional',the logging response characteristics of fractured reservoir was studied.With the crack sensitive logging information was selected,the cross-plot method,crack comprehensive probability method and BP neural network method was used to establish fracture identification model.On the basis of core description data and electric imaging data,the cross-plot method,principal component analysis and Bayes discriminant method,and GA-BP neural network was used to establish fillings identification standard after the fillings sensitive logging information was selected and enlarged.Synthesizing the well logging data,core data,and testing data,and on the basis of logging response characteristics of oil,gas,and water,the well logging information crossing,normal probability distribution,fuzzy clustering algorithm improved by particle swarm optimization(PSO-FCM)was used to build the regional fluid identification standard.On the basis of synthesis of the identification results of fractures and fillings,the reservoir types were divided,and the calculation model of reservoir parameters was established separately based on the dual pore component model,and optimization methods,and then the quickly identify standard of the target reservoir classification was established.The results of the study showed that,the response characteristics of fracture reservoir was as follows: higher GA,higher SP,lower resistivity,higher AC in reservoirs with low angle fracture developing,and FMI image could be used to calibrate the occurrence and effectiveness of cracks.The response characteristics of fracture reservoir provided instructions for the applicability of the intersection chart,while the coincidence rate of fractures identification using comprehensive probability and the BP neural network method were higher,and the recognition accuracy of neural network reached 89.65%.It was difficult to accurately identify fillings for cross-plot method,and using GA_BP neural network method to recognize fillings had a certain effect.In addition,the identify result of multivariate analysis method was best with the coincidence rate of identification being 91.7%.Logging information crossing method for fluid identification was easily affected by the influence of the hole,and so on.On the other side,the scopes of normal probability distribution of oil and water layer tended to be different.Besides,the effect of PSO-FCM method combining the probabilistic threshold cross plot of fluid identification was better,and it provided reference for carbonate fluid identification.There were three kinds of effective reservoir in study area,and the results of classification parameters modeling showed that,the calculate results of double-pore components porosity model and optimization method were consistent,and the calculation results of dual-pore permeability model was more close to the actual situation of the formation,while the double-pore components water saturation model made up the deficiency of the Archie formula.There were four levels of reservoirs in study area,and the division for reservoir level proved instructions to evaluate reservoir quickly.