Research On Oil Layer Identification And Reservoir Parameter Prediction Based On MPSO-BP Neural Network

In the future intelligent steerable drilling process,how to accurately identify oil layers and predict reservoir parameters is the key to successful intelligent steering.The conventional method of identifying oil layers and predicting reservoir parameters is to use experiments on the ground to obtain or interpret the potential information of relevant data.The hysteresis of this method cannot identify oil layers and predict reservoir parameters in real time downhole.Therefore,logging while drilling technology is mainly used to directly measure downhole logging while drilling data,establish downhole real-time model of logging while drilling data and oil layer identification and reservoir parameter prediction,and provide wellbore trajectory design for intelligent steerable drilling tools downhole Based on this,increase the drilling rate.In view of the above problems,this paper combines neural network with oil layer identification and reservoir parameter prediction,and proposes a model based on MPSO-BP neural network for oil layer identification and reservoir parameter prediction research.Based on extensive research on the application of artificial neural networks and methods for oil layer identification and reservoir parameter prediction,this paper carries out specific research work.The distribution law of faults determines the change of lithology and the distribution law of oil layers.After that,based on the research of BP neural network theory,with the constraints of lithology characteristics and oil layer distribution rules,the MPSO-BP neural network model is established to identify oil layers and predict reservoir parameters.Finally,according to the classification problem of oil layer identification and the regression prediction of reservoir parameters,different MPSO-BP neural network parameter combinations are designed,and various types of RMN,RLLS,GR,DEN,SP,AC are basically available for LWD data.The nonlinear relationship between logging data and reservoir identification and reservoir parameters was tested,and the performance of BP neural network,PSO-BP neural network and MPSO-BP neural network model was tested.The results showed that the MPSO-BP model achieved high prediction accuracy in the test set prediction,with 95% accuracy in the oil layer recognition classification,which improved by 6.3% and 14.7% over the PSO-BP and BP neural networks,respectively.The accuracy is about 98%,BP is 2.6% and 13.4% compared to PSO-BP,and the accuracy is 98%,4.3% and 17.8%,respectively.The results show that the MPSO-BP model is stable and can be used to identify oil layers and predict reservoir parameters in other oil Wells in Block F.It is an efficient and reliable oil layer identification and reservoir parameter prediction method,which provides some help for realizing intelligent downhole drilling.