Multiple Linear Regression Method To Study The Main Controlling Factors And Prediction Model Of Shunbei Carbonate Borehole Instability

The Shunbei Oil and Gas Field is a deep fault-karst oil and gas reservoir with typical fractured deep carbonate reservoir characteristics.During drilling and completion,wellbore collapse and drilling fluid loss and other wellbore instability problems are prone to occur,which seriously hinders the exploration and development process.,increasing drilling costs.Therefore,correct understanding and prediction of borehole instability is the key to efficient development of Shunbei oil and gas field.It is very difficult to take cores in carbonate reservoirs on site,and the cores taken are often cores from stable formations,resulting in insufficient representation of rock mechanical parameters in fractured reservoirs.Therefore,traditional laboratory experimental methods and numerical modeling methods cannot be used for wells in this area.In view of the above problems,based on the field data and other scholars’ research,the influencing factors are divided into geological,engineering and fluid factors,and the loss rate and well diameter enlargement rate are selected as the target prediction function,and the data of the influencing factors and the target function are selected.preprocessing.The linear correlation analysis method is used to eliminate irrelevant variables and collinear variables,the multiple linear regression method in the data method is used to initially establish a mathematical equation,the least square method is used to solve the equation coefficients,and the stepwise regression method based on t test is used to screen the main controlling factors.And use the python language to achieve the above process.The research results show that the main controlling factors with the leakage rate as the objective function are vertical depth,azimuth angle,geodetic coordinate X,well flushing fluid displacement,drilling time,WOB,initial cutting,φ3 and solid phase content.The azimuth angle has the greatest influence on the leakage rate;the main controlling factors taking the well diameter enlargement rate as the objective function are the well inclination angle,the geodetic coordinate X,the geodetic coordinate Y,the well flushing fluid displacement,the weight on bit,φ600 andφ100 in total.The coordinate Y has the greatest influence on the diameter expansion rate.Finally,the optimal prediction model is obtained by regression of the main control factors.In order to verify the accuracy of the prediction model,the field test set data was used to test the prediction model.The results showed that the prediction accuracy of the leakage rate was 91.09%,and the prediction accuracy of the well diameter expansion rate was 90.65%.In order to test whether the prediction model established by the multiple linear regression method is the optimal method,the neural network prediction model and the support vector machine prediction model are established for comparison and verification.The results show that the prediction model established by the multiple linear regression method has the highest accuracy.