Study On Well Logging Prediction Methods Of Enping And Wenchang Formation Source Rocks In L Sag,Pearl River Mouth Basin

The source rock of Enping and Wenchang Formation in the deep-water area of the Pearl River Mouth Basin are buried deep,and there are few wells to drill those source rocks.The measured organic carbon content(TOC)data is difficult to meet the need for precise evaluation of the source rocks.Therefore,it is necessary to use geophysical data to predict geochemical parameters for source rock evaluation.Studies have shown that there is a certain relationship between well log data and TOC of source rocks,which can be used to predict TOC.We used resistivity curve(RT,?·m),acoustic curve(AC,?s/ft),neutron porosity curve(CNL,%),natural gamma curve(GR,API)and density curve(DEN,g/cm3)in this paper to establish curve overlay model,multiple regression model and BP neural network model,discussed the prediction effect and compared the advantages,disadvantages and regional applicability between the three models.The results show that the multiple regression model is superior to the curve overlay model and neural network model in predicting the delta front subfacies of the Enping Formation.The correlation coefficient(R)between measured TOC and predicted TOC of the multiple regression model is 0.677,the prediction result R of the BP neural network model is 0.6543,and the result of the single-well curve overlay model does not exceed 0.65.In Wenchang Formation,the multiple regression model predicts result of the delta front(R=0.831)and the middle-deep lake subfacies(R=0.862)are superior to the shore shallow lake subfacies(R=0.665).That of BP neural network model is better than the multiple regression model,and the overall prediction result of the curve overlay model is poor.It is considered that the neural network model is more suitable for the prediction of TOC in the Wenchang Formation in the study area,and the multiple regression model is more suitable for the prediction of TOC in the Enping Formation.Based on the analysis of the above model,it can be applied to other secondary sags in the depression.Compared with the prediction results of the neighboring sags,it is believed that the BP neural network model is suitable for the case where the logging parameters and TOC are difficult to express with an explicit function,and there is a sufficiently large sample size;the multiple regression model is suitable for logging data and TOC are significantly related in some geological conditions;the curve overlay model is suitable for formations without potassium feldspar,and the gamma curve responds significantly to clay and organic matter content.