Research And Application Of Prediction Model Of Oil Well Wax Deposition Based On Machine Learning

Oil is one of the most important fossil energy in the world,which plays a very important role in the operation of modern industrialized society.However,wax is one of the most important components of crude oil.In the process of oilfield exploitation,it is often accompanied by the phenomenon of oil well waxing.Wax deposition in oil well is a phenomenon that has a negative impact on the normal production of oil well during the development and exploitation of oil field.This phenomenon will cause the blockage of oil flow channel and reduce the oil production in the process of oil well exploitation.Therefore,it is very important to make an intelligent early warning for the wax deposition status of oil wells and complete the repair of oil well equipment in advance.It has a critical value to improve the productivity efficiency,reduce the maintenance costs and intelligent management of oil fields.The main contents of this paper are as follows:(1)A-ENN mixed sampling method is proposed to solve the problem of serious sample imbalance caused by the large number difference between wax deposit samples and non wax deposit samples in oil well wax deposit data set.ADASYN algorithm will give too much weight to individual minority samples to reduce the decision space and the possible over-learning problem.ENN algorithm may ignore a large number of effective sample information.The combination of two over-sampling and under-sampling methods solves the above problems.Finally,the experimental results verify the effectiveness of the algorithm.(2)Aiming at the insufficient difference of base classifiers in Stacking ensemble algorithm,a weighted Stacking algorithm is proposed.In this method,different weights are given according to the different classification performance of the base classifier and improve the performance of the algorithm as a whole.In addition,according to the proposed algorithm,a weighted Stacking oil well wax deposition prediction model based on A-ENN is constructed.The model predicts the wax deposition of two oil wells and compares it with the real wax deposition time.The experimental results show that the model achieves more accurate prediction results,and has the feasibility of applying to the actual oilfield production.(3)Based on the prediction model of oil well wax deposition,a prediction system of oil well wax deposition is designed and developed.The system is mainly divided into three functional modules,the functional modules of oil well management,oil well wax prediction and work map attribute change.The system successfully transformed the prediction model of oil well wax deposition into a computer-aided tool,which can more effectively assist the oil field and workers to complete the wax removal work in time and ensure the efficient production of the oil field.