Research On Measurement Method Of Oil Well Liquid Production Based On Multi-source Information-fusion

The prediction of oil well production is of great significance for maintaining normal production order and avoiding confusion in oilfield production management.The traditional liquid production measurement methods include separator glass tube oil measurement method,tipping bucket weighing crude oil measurement method,mass flow measurement method and so on.These methods have some shortcomings,such as high cost of measurement equipment,limited conditions and so on.Therefore,this paper adopts the data-based oil well liquid production measurement scheme,analyzes some existing problems and establishes corresponding solutions.(1)For the traditional use of single information source modeling results are usually difficult to achieve the desired accuracy,and the measurement results are one-sided and uncertain.A multi-source information fusion modeling method is proposed.The least squares support vector machine(LS-SVR),Wavenet liquid production time series and indicator diagram model are used as sub information sources.These sub information sources consider the relationship between oil well liquid production and each variable,the time series of oil well liquid production,and the relationship between oil well liquid production and mechanism process,To achieve the purpose of making more use of modeling data information.Simulation results show that the multi-source information fusion method can improve the accuracy and generalization of liquid production measurement better than the single information source method.(2)For the traditional multi-source information fusion method,the influence of the increase of the output error of the sub information source on the fusion result is less considered,which leads to the poor accuracy of the output result.Therefore,a multi-source information fusion method based on deviation degree is proposed.Firstly,according to the modeling principle of different information sources,the deviation degree of corresponding sub information sources is calculated,and then the deviation degree is normalized to [0,1] Finally,by establishing the relationship between the deviation degree and the fusion result,the final multi information fusion result is calculated.Simulation results show that the multi-source information fusion method based on deviation degree can effectively reduce the impact of the increase of output error of small information sources on the fusion results and improve the accuracy of prediction results.(3)Aiming at the problem that the static model can not meet the dynamic production of oil well,a multi-source information fusion dynamic updating method is proposed.By adding data analysis and evaluation module,the output of LS-SVR and Wavenet information sources is judged,and the effectiveness of the model is judged according to the results.The principle of producing similar output by using similar input and the measured liquid production value is input to the training regularly The LS-SVR and Wavenet information sources are updated respectively in the way of set to achieve the goal of online updating of sub information sources and enhance the dynamic updating ability of the system.The simulation results show that the dynamic model has higher prediction accuracy than the static model.