Method for predicting and optimizing penetration rate in oil and gas drilling based on cart algorithm

A method for predicting and optimizing a penetration rate in oil and gas drilling based on a CART algorithm, comprising the following steps:

• Step 1:
• selecting a block and sorting all available drilling data, that is, whole-meter well-logging data and well-measuring data, into a unified Excel table or TXT text as original drilling data;
• Step 2:
• performing data preprocessing by taking a well depth, bit type, bit size, and deformation type as the basis for dividing different spuds, wherein drilling parameters that affect the penetration rate include drilling engineering parameters and original formation parameters, the drilling engineering parameters include a weight on bit, rotation speed, torque, drilling fluid density, displacement and riser pressure, and the original deformation parameters include a sonic time difference and natural gamma;
• with 8 drilling parameters as different characteristic attributes, obtaining an initial data set D1 by taking drilling data contained in each characteristic attribute as an input variable X and the penetration rate as an output variable Y;
• Step 3:
• performing data correlation analysis:
• analyzing the correlation between the input variables in the initial data set D1 and the penetration rate by using a correlation coefficient method, calculating correlation coefficients between the 8 input variables and the penetration rate in sequence, and sorting the input variables according to the values of the correlation coefficients to obtain a training data set D2 of different spuds;
• Step 4:
• establishing a regression tree model between the input variables in the training data set D2 of different spuds and the penetration rate by using the CART algorithm, training the regression tree model by randomly dividing 80% of data in the training data set D2 as a training set, and testing the trained regression tree model by using the remaining 20% of data as a test set, wherein it is considered that the model is available if a test score reaches 80 or above;
• Step 5:
• predicting a penetration rate, and analyzing, by using the regression tree model with a binary tree structure obtained by the CART algorithm, information of each leaf node of the generated binary tree, wherein an average value of the leaf nodes is used as a predicted value of the penetration rate;
• Step 6:
• recommending drilling parameters:
• characterizing the intensities of impacts of the input variables on the penetration rate from top to bottom by using the regression tree model with the binary tree structure obtained by the CART algorithm, wherein the topmost end represents the strongest impact, and the bottommost end represents the weakest impact;
• providing an optimal recommended value of the input variable for nodes in each layer;
• traversing a node division result of each layer from top to bottom to obtain different recommended values of drilling parameters;
• and
• Step 7:
• substituting the recommended values of drilling parameters into the regression tree model based on the CART algorithm to obtain an optimized value of the penetration rate;
• taking a difference between the optimized value of the penetration rate and a predicted value of the penetration rate being less than 10% as an optimal judgment condition;
• recommending to use the drilling parameters obtained in Step 6 if the optimal judgment condition is reached;
• if not, returning to the Step 3 and performing correlation analysis again.