Speech Title: Neural network model application in prediction of organic matter richness of saline lacustrine hydrocarbon source rocks of the paleogene in western Qaidam Basin
Abstract: Aiming at the issue of salinity effects on the prediction of organic matter richness of hydrocarbon source rocks, this research adopted the Δ log R method, multiple regression method and BR-BP neural network method to simulate organic carbon content and discussed the differences among three methods' prediction. The results show that the multiple regression method has a general effect; the Δ log R model has good accuracy, but its applicability is general; the BR-BP neural network model has the best prediction effect. Therefore, the neural network model was used to predict in the middle and low salinity areas. In high salinity areas, the neural network parameters were adjusted to match the model application. The research results can improve the accuracy of source rock identification and guide the accurate evaluation of source rocks in the basin.