Prediction And Application Of Oilfield Production Based On Time Series Analysis Method

The prediction of oil field production is of vital importance to guiding the production of oil field.At present,the multi-feature modeling method is widely used in the prediction of oilfield production in different regions,although the prediction accuracy of the multi-feature modeling method is accurate.But because around the actual condition is different,factors which influence the production of oilfield is complicated,oil field area is large,in the prediction of oil production,extraction of feature modeling methods,more practical more bad and don't have universality.Oil production of oil and gas in the Internet of things project(A11),manages several large domestic production of oil field,the universality of how to establish effective prediction model,how to improve the prediction accuracy,is a focus in the study of this article.On the basis of summarizing the previous research results,this thesis deeply discusses the application of time series analysis method in oilfield production forecast by combining theory with practical method.(1)in order to establish a valid predictive model for A11,this paper proposes a time series method to analyze and model.Time series analysis method does not need to extract the external features,only care about the internal characteristics of the yield curve itself,considering again,complexity of the components of oil field production data and the limitation of the number of sample data,thus can be used in a time series analysis method to analyze the oilfield production data modeling.Firstly,the theoretical basis and modeling process of classical time series model AR,MA,ARMA and ARIMA are studied.Then depends on the actual oil field production data,abstract oil field production data,for a single moment instantaneous value,using data from a 365 days a year are modeled as the training sample,after 20 days of data as the checking sample,to the fine degree of assessment model,according to the characteristics of the oil field data,the final evaluation through the experiment the ARIMA model has good applicability for oilfield production prediction,prediction accuracy up to 98.16%.(2)in order to improve the accuracy of the model,the time series combination prediction model of ARIMA-BP neural network was proposed.In the actual production environment time series,more or less both contain linear and non-linear parts,and the oilfield daily output is the typical representative.The ARIMA model can fit and predict the linear time series well,and the BP neural network model is better for the prediction of nonlinear time series.ARIMA-BP combined forecasting model is by using ARIMA model to simulate the time series of linear fitting,to extract the residual time series,then use the BP neural network modeling,linear composition and nonlinear components modeling using ARIMA model and BP neural network model respectively,forecast forecast sum,respectively.through the experiment show that the BP model prediction accuracy is 98.86%,the composite model prediction accuracy is 99.28%,the composite model than a single model to predict the effect is more ideal.