Productivity Prediction Based On Big Data And Machine Learning

The reserves and production of water drive old oilfields play an important role in SINOPEC.After entering the ultra-high water cut period,the old oilfields still have a lot of remaining oil potential and are still the main force of oilfield development.Using big data technology and machine learning algorithm to establish the development index prediction method to adapt to the situation of ultra-high water cut and low oil price,to form a highprecision and efficient oil well productivity prediction technology,which will have great market demand.The development index prediction and oil well productivity prediction methods have wide adaptability for water drive sandstone reservoir in high water cut stage,and have the advantages of easy operation,good effect,less investment and high benefit.Through popularization and application,it has important significance for the benefit development of water drive reservoir in low oil price.In this paper,we use big data technology and machine learning algorithm to predict oil well productivity.First of all,we study and compare the traditional productivity analysis methods,excavate the relevant variables and parameters needed for productivity prediction,and summarize the previous prediction methods and percolation laws to obtain the dynamic change law of productivity,and combine it with the production decline law to carry out big data processing Learning filter.In this paper,the productivity forecasting method based on time series is further studied.First,based on the single factor productivity prediction of time series and oil well liquid production series,the preliminary training model is obtained,and compared with the actual data,the error analysis is carried out,and the feedback model is modified to obtain the single factor productivity prediction model.The second step is to carry out the multi factor productivity prediction of time series according to the factor screening results.By changing the influencing parameters,different well patterns have been set,and the productivity prediction models under different well patterns have been obtained.In this paper,aiming at the problem of production productivity prediction method in ultrahigh water cut period,a multi factor time series production productivity prediction model based on big data and machine learning algorithm is proposed.Compared with traditional production productivity prediction methods,the accuracy and speed of prediction are greatly improved,and real-time prediction and feedback prediction are realized.It enriches the theory and method of productivity prediction,and provides a theoretical basis for development design and productivity evaluation.