Research On Multi-Constraint Layed Injection-Production Optimization Model And Algorithm For Waterflooding Reservoirs

With the continuous development of Chinese oil reservoirs,water flooding oil reservoirs has entered the late stage of high water cut development,the contradiction between layers is prominent,the water cut keeps rising,which leads to the continuous decline of oil production.Due to the large difference in physical properties of small reservoirs within the interval,general water injection can no longer satisfy the complex geological environment and seepage process of the reservoir,and it is difficult to ensure the effectiveness of water injection.Therefore,layered water injection has become a hot research direction for improving oil recovery in the late period of high water cut.Efficient layered injection-production scheme can maintain reasonable reservoir pressure,improve production efficiency and promote stable production and water control in old oilfields,which has practical guiding significance for improving recovery effect and economic benefit in the late stage of water drive high water cut oilfields.This thesis conducts research on the problem of layered injection optimization.First of all,oil production rate is taken as the research goal,the single well layered data is screened and processed,data sets are divided according to the characteristics of sedimentary microfacies,and important factors of oil production rate are determined through correlation analysis and attribute reduction.Considering that the oil production rate is sequential,an improved temporal convolutional network based on attention mechanism and autoregressive component was proposed to predict oil production rate.The effectiveness of the model was verified by comparison experiments with stepwise improved model,traditional temporal convolutional network,long-short time memory network,gated recurrent unit,etc.Secondly,based on the principle of material balance,the expression of the relationship between oil production rate and injection-production ratio was established.The oil production rate,water cut and pressure recovery rate were selected as the constraint conditions,and the stratified injection-production optimization model based on the theory of net present value and the law of water drive decline was established.An improved Harris Hawk optimization algorithm based on Piecewise chaotic mapping,golden sine strategy and Cauchy variation strategy is proposed to solve the optimal oil production rate,water cut and pressure recovery rate which meet the maximum net present value.The effectiveness of the improved algorithm is verified by comparing with a variety of intelligent optimization algorithms.Thirdly,according to the optimization variables,geological reserves,injection-production ratio and other related parameters,the interval injection volume was calculated,and the interval injection volume of connected well was calculated with the oil well as the center.Then,the interval injection volume of connected well was accumulated with the water well as the center,so as to obtain the layered injection-production optimization scheme.Finally,based on the research of model and algorithm in this thesis,the optimization system design of layered injection and production in water flooding reservoir is described from the perspectives of system architecture,design mode and design principle,which provides a design idea for the digital and intelligent development of layered injection and production in oilfields.