Research On Adaptive Control Method For Production Parameters Of Pumping Wells Based On Deep Learning

Some low-yield oil wells in my country have low formation energy and low productivity,and the operating efficiency of the oil production system decreases,resulting in a huge waste of electric energy.At present,the effective method to solve the above problems is to change the production parameters by adjusting the frequency of the pumping unit,so as to realize the self-adaptive matching between the pumping capacity of the pumping unit and the liquid supply of the formation.However,most of the existing adaptive control methods of pumping unit production parameters rely on single-factor control,and the accuracy and real-time performance need to be improved.With the comprehensive advancement of production informatization,the oilfield big data resources have been initially formed,and the remote control of pumping wells and real-time data acquisition have been realized.The data indicators collected in real-time provide richer,more comprehensive,and more timely data support for the parameter control of the pumping unit.If these data can be fully and effectively used,the effect of the parameter control of the pumping unit will be further improved.Therefore,this paper proposes an adaptive control method for pumping well production parameters based on deep learning.Combined with literature research and theoretical knowledge of oil production engineering,three data indicators of pumping unit ground dynamometer,dynamic fluid level,and fluid production speed are selected as the basis for self-adaptive control of pumping unit production parameters.Aiming at the problems that the traditional analytical calculation model of pump fullness is too simplified,the initial parameters are difficult to obtain,and the fluid distribution in the wellbore is complex,according to the deep learning theory,and intelligent identification method of pump fullness based on convolutional neural network is established.According to the change characteristics of various time series data of pumping unit,the time window reflecting short-term trend and the long-term trend is determined,and the extraction method of characteristic indicators such as average level,trend,and volatility in the time window is studied.After determining the decision function based on the single-factor decision-making influence mechanism,the calculation method of the single-factor decision factor for the control of the pumping unit production parameters for every single characteristic index is established.Then,according to the comprehensive influence of various factors on the control decision,a calculation method for the comprehensive decision factor for the control of the production parameters of the pumping unit is established.At present,based on the newly developed self-adaptive control method of pumping well production parameters,the program has been compiled and deployed in an oilfield management area,and 10 wells of mine have been applied.The changes of the production data indicators of the pumping wells before and after the regulation show that the adaptive regulation method of the production parameters of the pumping wells based on deep learning makes the production of the pumping wells more suitable for the formation fluid supply,and the accuracy and real-time performance of the regulation are significantly improved.,the electricity bill has dropped significantly,the maintenance-free period of oil wells has been extended,and the production efficiency has been significantly improved.