Research On Key Parameters Of Drilling Leakage Prevention And Plugging Based On BP Neural Network

Well leakage is a difficult problem which needs to be solved immediately and which is not solved yet.One of the main reasons is that the knowledge of the characteristics of underground fractures are unclear and the main reasons is that the knowledge of the characteristics of underground fractures are unclear and the acquisition of key parameters such as leakage position and crack width is the premise of the success of f-ormation leakage prevention and sealing.However,due to the complexity of the stratum,the randomness,and the interference of engineering factors,using the existing test calculation methods,it's not easy to detennine and describe the real situation of these parameters under the original condition,which greatly affects the plugging efficiency.In the paper,the well history date of different different wells in the same block are used as the training set.After the training,the appropriate neural network model is established and applied to the prediction of the missing key parameters and the crack width of the missing block in the fluid system which is suitable for different crack widths is obtained.The main research contents are as follow:(1)By investigating and analyzing the influence factors of well leakage,I preferably chose the drilling fluid density,flow rote,ROP,drilling fluid Gel strength,drilling fluid viscosity,drilling fluid displacement,pump pressure.The leakage channel properties and formation pore pressure are used as input feature quantities for predicting the location of the drain layer.Based on the actual location of well leakage which is analyzed by the well history record,the BP neural network prediction model based on the BP fault location of the drain layer.Based on the actual location of well leakage which is analyzed by the well history record,the BP neural network prediction model based on the BP fault location is established.The BBO-Biogeography-based algorithm is used to optimize the BP neural network.The predicting result is very close to the actual well leakage position,and the overall error does not exceed 10%.(2)Research and analysis of adjacent well data,selecting plastic viscosity,loss rate,leakage loss,drilling fluid Gel strength,ROP,well depth,pump pressure and flow rate as input characteristics of predicted crack width,based on BP neural network.It is suitable for the prediction model of crack width in this block,and genetic optimization algorithm is used to optiinize BP neural network.After optimization,the error is reduced by 7.21%.Compared with the actual leakage crack width,the prediction error value is less than 10%,which achieves the purpose of predicting the crack width.(3)A series of rigid mineral particle plugging agent GZD and plant fiber GDJ were used as plugging materials,and four kinds of crack-proof plugging formulas with different formulas with different crack widths of 1.0×0.5mm,1.5×1.0mm,2.0×1.5mm and 3.0×3.0mm were obtained,through the laboratory experiments.Taking the data generated in the experiment as the training set and judging whether it is successful,the neural network model of the plugging formula prediction is established.The model is used to predict the plugging formula for other crack widths.Taking the data generated in the experiment as the training set and judging whether it is successful,the neural network model of the plugging formula prediction is established which is used to predict the plugging formula for other crack width.