CN-117127961-B - Mechanical drilling speed prediction method based on micro-well deviation characteristics

CN117127961BCN 117127961 BCN117127961 BCN 117127961BCN-117127961-B

Abstract

The invention discloses a mechanical drilling speed prediction method based on micro-well deviation characteristics, which comprises the steps of obtaining logging while drilling and logging data of an actual drilling well as a first data set, preprocessing the first data set, extracting optimal characteristics and micro-well deviation characteristics as input characteristics, preprocessing the first data set, performing data cleaning and correlation analysis, utilizing the input characteristics to complete training of a mechanical drilling speed prediction model, and utilizing the trained mechanical drilling speed prediction model to predict the mechanical drilling speed of a target well. The introduced micro-well deviation feature can obviously improve the ROP prediction precision of the model. By introducing the micro-well deviation feature into the MLP neural network, the model weight is optimized, and the accuracy of predicting ROP is improved. By introducing the micro-well deviation feature into the SVR model and selecting the optimal super-parameters, the prediction effect is remarkably improved. The method provides a new thought for improving the prediction and control of drilling parameters by utilizing machine learning, and can be popularized and applied to the optimization of drilling speed and efficiency.

Inventors

LIU WEIJI
FENG JIAHAO
LI KE
ZHU XIAOHUA

Assignees

西南石油大学

Dates

Publication Date: 20260508
Application Date: 20230831

Claims (5)

1. The mechanical drilling speed prediction method based on the micro-well deviation characteristics is characterized by comprising the following steps of: Acquiring logging while drilling and logging data of an actual drilling well bore as a first data set; Preprocessing the first data set, extracting optimal characteristics and micro-well deviation characteristics as input characteristics, wherein the preprocessing comprises data cleaning and correlation analysis; training a mechanical drilling speed prediction model by utilizing the input characteristics; predicting the mechanical drilling speed of the target well by using the trained mechanical drilling speed prediction model; the micro-well inclination characteristic represents the change trend of a certain section of well curvature above the drill bit; The extraction method of the micro-well deviation feature comprises the following steps: And calculating the curvature change rate of the wellbores at the deep positions of the wells, and superposing the curvature change rate of the wellbores within a specified distance from the drill bit to the upper part of the drill bit to form the micro-well deviation characteristics of the distance range.
2. The method for predicting the mechanical drilling rate based on the micro-well deviation characteristics according to claim 1, wherein the pretreatment method comprises the following steps: deleting the row of the missing value in the first data set, and processing through a smoothing filter; Carrying out pearson correlation coefficient quantization analysis on the first data set after the smoothing filter treatment, and eliminating the characteristic of low absolute value of the pearson correlation coefficient; And training and evaluating the rest features through a multi-layer perceptron model by using an exhaustion method, and selecting the optimal features according to the evaluation result.
3. The method for predicting rate of penetration based on the characteristic of micro-well deviation of claim 2, wherein the optimal characteristics are weight on bit, torque, rotary table rotational speed, inlet temperature, inlet conductivity, outlet conductivity, mud density, DC index, and borehole curvature.
4. The method for predicting the rate of penetration based on the characteristic of micro-well deviation according to claim 1, wherein the model for predicting the rate of penetration comprises a multi-layer perceptron with initial stage weight parameters and bias parameters optimized by genetic algorithm.
5. The method for predicting the mechanical drilling speed based on the micro-well deviation features of claim 1, wherein the mechanical drilling speed prediction model comprises a support vector regression model with optimal super-parameter combination optimized through a network search algorithm.

Description

Mechanical drilling speed prediction method based on micro-well deviation characteristics Technical Field The invention relates to the technical field of oil and gas reservoir development, in particular to a mechanical rotating speed prediction method based on micro-well deviation characteristics. Background With the continuous deep exploration and development of oil and gas, deep and complex stratum is more and more encountered, and many new challenges are presented to the drilling technology. Especially for the drilling process of ultra-deep well, the drilling depth and the complexity of stratum can greatly improve the drilling difficulty, so that the exploitation cost is high. In order to ensure high yields and high efficiency of oil and gas well development, remote monitoring of the drilling process and optimization of drilling parameters are indispensable. The rate of penetration (rate of penetration, ROP), which is one of the key parameters in assessing the efficiency of the drilling process, refers to the rate at which the drill bit breaks rock or sediment, typically expressed in meters per hour or feet per hour, which reflects the drilling efficiency and formation characteristics, and increasing ROP can save drilling time and costs while also reducing wear and damage to the drilling tool. Thus, improved prediction and control of ROP is critical to optimizing drilling. It is difficult for conventional empirical models to accurately describe the nonlinear relationship between the various complex factors that affect ROP. And the development of machine learning technology provides a new idea for predicting ROP by using a data training model. The machine learning model enables faster and more accurate nonlinear prediction than the traditional model. However, the existing method for predicting the ROP by utilizing machine learning still has room for improvement in prediction accuracy, and the influence of micro inclination (micro well inclination) of the well wall above the drill bit on the ROP is not considered yet. Vibration is generated when the drill bit encounters a micro-well deviation, and ROP is affected. Disclosure of Invention In order to solve the technical problems, the invention provides a mechanical drilling rate prediction method for introducing new features reflecting micro well deviation information above a drill bit into a machine learning model so as to improve the prediction accuracy of ROP, which specifically comprises the following technical scheme: the mechanical drilling speed prediction method based on the micro-well deviation characteristics comprises the following steps: Acquiring logging while drilling and logging data of an actual drilling well bore as a first data set; Preprocessing the first data set, extracting optimal characteristics and micro-well deviation characteristics as input characteristics, wherein the preprocessing comprises data cleaning and correlation analysis; training a mechanical drilling speed prediction model by utilizing the input characteristics; and predicting the mechanical drilling speed of the target well by using the trained mechanical drilling speed prediction model. In some preferred embodiments, the method of preprocessing comprises: deleting the row of the missing value in the first data set, and processing through a smoothing filter; Carrying out pearson correlation coefficient quantization analysis on the first data set after the smoothing filter treatment, and eliminating the characteristic of low absolute value of the pearson correlation coefficient; And training and evaluating the rest features through a multi-layer perceptron model by using an exhaustion method, and selecting the optimal features according to the evaluation result. In some preferred embodiments, the optimal characteristics are weight on bit, torque, rotary table rate of penetration, inlet temperature, inlet conductivity, outlet conductivity, mud density, DC index, and borehole curvature. In some preferred embodiments, the micro-well deviation feature characterizes the trend of change in curvature of a section of the wellbore above the drill bit. In some preferred embodiments, the method for extracting the micro-well deviation feature comprises the following steps: And calculating the curvature change rate of the wellbores at the deep positions of the wells, and superposing the curvature change rate of the wellbores within a specified distance from the drill bit to the upper part of the drill bit to form the micro-well deviation characteristics of the distance range. In some preferred embodiments, the rate of penetration prediction model includes a multi-layer perceptron that optimizes initial stage weight parameters and bias parameters by genetic algorithms. In some preferred embodiments, the rate of penetration prediction model comprises a support vector regression model that optimizes an optimal combination of super parameters via a web search algorithm. Advantageous ef