The Application Of BP Neural Network In Prediction Of Strata

In the offshore oil extraction operations,the selection and establishment of drilling pos ition is a very complicated process.In this process requires a certain understanding of the fo rmation,in order to ensure the oil out of the drilling,and platform security.In the field of st ratigraphic data acquisition,the present method for stratigraphic information is text descript ion,which is not conducive to data analysis.Therefore,in the data analysis,is still using art ificial connection method,which for the operator's experience and technical requirements ar e very high,resulting in the researchers in the work process inefficiency.In data prediction,despite the large amount of data,but not the potential value of its in-depth mining.In order to improve the efficiency of China's offshore oil production,this paper designs the geologic al data management platform,using BP neural network to predict the geology:First of all,in view of the development of geological prediction and the status quo don e a lot of research.It is understood that early mathematical statistics have great limitations i n geological prediction.The advantage of BP neural network in geological prediction is dee ply understood.Secondly,the paper deeply studies the network structure,the mathematical principle a nd the working method of BP neural network.The superiority of BP neural network in form ation prediction is analyzed.And the BP neural network algorithm is improved.The realizat ion process includes structure module building,initialization module design,training modul e design forward propagation class and weight adjustment design,and using XOR algorith m to design the network reliability test.Thirdly,the digitalized data of the soil layer is described.The processing method inclu des two types: large sorting and sorting.After the data were digitized,a neural network with hidden layer 3,input node 3,output node 1 and hidden layer node 7 was designed.Then th e original data are input to the network,and the output results are compared with the origina l data to determine the prediction precision of the network.Finally,we construct the 3D space by OpenGl,visualize the formation data,mark the k nown data in the space,and predict the unknown area by the trained BP network,and fill th e prediction result into the space.And for the three-dimensional vertical and horizontal screenshots for the operator to provide a richer data reference,greatly improving the efficiency of the researchers.