Application Research On High-precision Modeling And Optimization Of Oil Field Production Process

Oil field is the mainstay of energy production,but also is the major object of energy consumption.Mechanical production is the main energy consumption and its efficiency is generally less than 30%.If each equipment can save a little energy that may achieve an amazing economic benefits.How to improve technology and management level of the pumping production are the key issues,which is concerned by oil field and needs to be solved urgently.With the development of digital oil field,more and more testing devices have been installed in well,which recorded a large number of working parameters,yield and energy consumption data in detail,which means we can use the data mining techniques to find potential law of the oil production and with mathematical model description.Then we achieved optimal operation parameters of the system by the intelligent optimization technology are achieved to keep system in the best running state to save energy and enhance the efficiency.Therefore,in this thesis we use production system as the research subject to study the key scientific issues in which data mining technology is used to construct model for pumping system and intelligent optimization technology is used for optimal production parameters.Through the theoretical research,simulation experiment,software development,and oil field machine,we can promoted in modeling autonomous,intelligent optimal and decision-making autonomous.The research content is described as follows:(1)In this thesis,we proposed dynamic evolution modeling of oil field machine process,which based on the Unscented Particle Filter neural network(UPFNN).Establishing accurate machine process model is the premise to realize production optimization.Due to the machine system influenced by machinery,formation and artificial uncertain factors,it is difficult to accurately grasp the relationship between parameters,environment variables and the performance of the system.Therefore this thesis put forward to use the Unscented Particle Filter update neural network weights and threshold,at the same time,we established Unscented Particle Filter neural network dynamic evolution model to match subspace of machine production system.This method uses the Unscented Kalman Filter to estimate the particle density,and updates the particle to improve the accuracy of particle filter,and to improve theaccuracy of model by neural network.(2)On the basis of the accurate machine process model,we raised the preference driven multi-objective optimization of mechanical oil extraction parameters.Oil field usually need to accord the global reservoir distribution scientific and rational design the production of each area to realize the scientific and national exploitation of oil reservoir.Therefore,the oil machine system cannot used the maximum oil production and minimum energy consumption as optimized direction.The production capacity is close to a given value and minimum energy consumption.What's more,the optimization of oil field production system is obtaining the optimal value of the objective function under various constraints,which is a complex nonlinear optimization problem,it is difficult to get the best answers by the traditional methods.With elite strategy non-dominated sorting genetic algorithm,we calculates congestion level to avoid individual parameter sharing problem in order to save the parent outstanding individuals by elitist preservation can achieve multi-objective parallel optimization.This is more advantageous than the traditional optimization method to deal with the problem of complex,high dimensional and difficult to resolve in industrial process.Therefore,first of all,this thesis combine machine process model by Unscented Particle Filter neural network with the preference function of the production system to constructed multiple objective optimal function,then we use the non-dominated sorting genetic algorithm to solve Pareto of the production parameters.In the last not the least,we use ordered weighted to obtain score each of Pareto to get the best solution.(3)Develop support system of the digital oil field pumping fleet scheduling optimization.In this thesis,we use the method which includes C# and MATLAB to develop a support system of digital oil field pumping oil cluster scheduling optimization that make theoretical research to guide the actual production.Then we implanted the support system into oil field production management platform to achieve the object that the machine mining process autonomous modeling.And the parameters can realize the intelligent optimization and self-decision.