Research On Early Warning Method Of Casing Damage In Fault Block Oilfield Based On Big Data

Casing damage is affected by geological conditions and engineering factors.Due to the different geological conditions of each oilfield,the types and mechanism of casing damage are different,which makes physical modeling more difficult.The damage to the casing is mainly caused by the change of the external force on the casing and the long-term action on the casing.The dynamic and static parameters of geology,engineering and production are the reasons for the change of external force.In addition,corrosion is also an important factor in casing damage.After decades of digital construction in domestic oilfields,each oilfield has accumulated a large amount of casing damage data,which contains various casing damage laws.The data-driven intelligent casing damage prediction method can analyze casing damage trends in real time according to changes in production parameters,and take real-time adjustment measures to prevent casing damage based on the analysis results,and improve the production rate and life of oil wells.Based on the data of production wells in the fault block oilfield of Shengli Oilfield,this thesis collects various geological,engineering,and production parameters related to the mechanism of casing damage,including geological parameters such as formation pressure,formation temperature,formation dip,engineering parameters such as cementing quality,casing outer diameter,casing wall thickness,perforation density,production parameters such as cumulative fluid production and cumulative injection volume.Through data cleaning and standardized processing of the collected data,a casing damage data set is established.After correlation analysis,the main factors of casing damage in a fault block oilfield for a given data set are determined,including cementing quality,ion number of produced fluid,injection production balance difference of formation and number of drilling faults.The intelligent prediction model of casing damage is established by using single machine learning classification algorithms such as logistic regression,decision tree and k-nearest neighbor and integrated machine learning classification algorithms such as random forest and Ada Boost.After cross-validation and comparison,the prediction accuracy of the random forest casing loss prediction model reaches 99%,and better prediction results can be achieved by using this model.For the production wells that casing damage may occur,the random forest,Adaboost,and GBRT regression algorithms in the machine learning algorithm are used to establish a casing damage depth intelligent prediction model.After verification,the determination coefficient of GBRT casing damage depth prediction model is 91.2%,and the prediction accuracy is high.The data-driven method can realize fast modeling according to the data of a given block,and can consider the influence of time factor on casing damage.It can be used to predict casing damage trend under different parameters and determine dynamic parameter limits.It is of great significance to the construction of intelligent oil field,reduce the cost of oil field exploitation and improve the benefit of oil and gas field development.