Research On Application And Optimization Of Machine Learning Algorithm In Oil Drilling Field

Intelligentization is the development trend of the future society.For traditional industries,how to actively learn new ideas,new theories and new technologies to win the opportunities at the dawn of the intelligent wave is crucial to the development of themselves.Oil drilling is a typical traditional industrial discipline.The prediction and optimization of drilling speed has always been one of the important topics in the industry.Increasing the drilling speed can effectively improve drilling efficiency,shorten drilling period and reduce drilling cost.In this paper,the machine learning algorithm is used to study the prediction and optimization of drilling speed.The main work is as follows:Firstly,the basic theories of the commonly used outlier detection algorithms are studied.These commonly used outlier detection methods are used to remove the outliers of the drilling data,then drilling speed prediction model is trained using the data after removing the outliers.And the performance of an outlier detection method is evaluated and analyzed by the performance of the model.Then,based on the above work,three better methods are integrated,and a new outlier detection algorithm is proposed.The verification shows that,after removing outliers using the proposed method,better model is trained with an unchanged detected outlier ratio.Secondly,based on the integrated outlier detection method,combined with the actual situation of drilling data,a modular drilling data preprocessing scheme is designed and implemented,which is mainly divided into four modules:data selection module,data cleaning module,outliers processing module,data segmentation and normalization module.The data selection module and the data cleaning module perform basic data deletion and modification operations.Then the outlier processing module detects and deletes outliers.Finally,the original data is divided into the training set,the verification set,and the test set in data segmentation and normalization module.These four modules are functionally independent of each other and form the basic drilling data pre-processing process together.Finally,based on the actual situation,using the layered architecture,a basic scheme of drilling speed prediction and optimization is proposed,which mainly includes the drilling data preprocessing layer,the drilling speed prediction training and learning layer,the drilling speed prediction result display and analysis layer,and drilling speed optimization layer.In the drilling speed prediction training and learning layer,according to the performances of models such as LightGBM,fully connected neural network and GRU network model,the influence of each input parameter on the drilling speed and the influence of time series information on the prediction result of drilling speed are discussed.In the drilling speed optimization layer,the trained fully connected neural network model using the first-order data is selected to simulate the drilling speed optimization.The results show that the proposed scheme is feasible and lays a good foundation for the future practical system application.