Study On Lost Circulation Control Measures Based On Data Analysis

Bohai Oilfield is the largest offshore crude oil production base in China,with crude oil production reaching 30.132 million tons in 2021.With poor formation integrity in the Bohai Oilfield,there are problems such as high frequency of lost circulation,multiple types of loss,and difficulty in lost circulation control in drilling process.At present,the selection of lost circulation control methods in the field mainly relies on the experience of engineers,and there are problems such as poorly targeted measures,low success rate of once plugging,long non-productive time(NPT)and high operating costs,which havd become the main factors limiting the safe and fast drilling in the Bohai oilfield.In this thesis,by collecting,compiling,analyzing and summarizing the historical data of lost circulation in the Bohai Oilfield in recent 10 years,on the basis of the existing classification methods for lost circulation control,combined with the actual engineering practice,the loss in Bohai Oilfield are divided into 3 major categories,namely pan-pore type,fracture type and fault type,according to the different types of loss channels.Based on the characteristics of lost circulation,loss rate and performance after emergency treatment,the well loss was further subdivided into 12 categories,and the basis for judging each type of loss is proposed,the basic characteristics of each type of loss are summarized,and a new method was established to comprehensively evaluate the effectiveness of lost circulation control by combining the success rate of plugging and the recovery rate of flow rate.Finally,the corresponding lost circulation control methods were recommended for 12 types of well loss,based on the difficulty of plugging construction,operation cost and plugging effect,and the recommended criteria for lost circulation control methods based on the comprehensive classification of loss were established.Combining the plugging analysis and the existing loss prevention technology in Bohai Oilfield,the loss prevention measures are proposed for different types of well loss.Adopting lost circulation control measures based on the comprehensive classification of loss type can improve the targeting of lost circulation control,increase the success rate of once plugging,and achieve the purpose of shortening non-production time,reducing operation costs and safe and fast drilling.To ensure the reliability of the recommended plugging method based on the comprehensive classification of lost circulation and to consider the influence of various parameters on loss and the plugging process,machine learning algorithms are introduced to build models to assist the plugging decision.Based on the three algorithms of Random Forest(RF),Gradient Boosted Decision Tree(GBDT)algorithm and XGBoost algorithm to establish the plugging effect evaluation model,19 characteristic parameters related to loss and plugging are selected as input parameters and plugging effect is the label,283 well losses of 167 wells in recent 10 years in Bohai oil field are used as data sources,876 sets of data are selected as sample sets,and the built-in modules and libraries of Python are used to to complete the modeling.The performance of the models was evaluated using confusion matrix,ROC curves and AUC values,and the results showed that all three models have good accuracy and stability,among which the XGBoost algorithm model has the best comprehensive performance.By using the plugging effect evaluation model and the recommended plugging methods based on comprehensive classification,can help field engineers to make reasonable evaluation of the plugging effect of different plugging methods in advance and effectively assist field engineers to choose the best lost circulation control method.