| For China’s high-water-cut continental sandstone oilfield,the dominant seepage channel is formed in the reservoir due to long-term water washing,which aggravates the inefficient and invalid circulation of injected water,seriously affects the improvement of recovery factor,and becomes the core problem affecting the efficient and low-consumption production of the oilfield.How to judge and identify the dominant seepage channel is the key to control the low efficiency cycle,in which accurate measurement and calculation of split flow in each direction of production and injection well in each layer is the basis for accurate identification of the dominant seepage channel and development adjustment.On the basis of referring to the existing dynamic splitting method,a new random dynamic splitting method is established in this thesis,which can obtain a more reasonable amount of split flow between Wells.In this thesis,based on the principle of material balance in oil field,the equations of the relation between the injection quantity of water well,the split-flow quantity of each layer and the production quantity of oil well are established.In the face of the equations of an unknown quantity more than the known quantity problem,this paper puts forward the splitting shunt with dynamic and static data to calculate initial value,and using minimum variance control error of the calculated value and measured value based multi-objective functions,Finally,the optimal solution of multi-objective function is solved to obtain the actual split flow value.In this way,according to the measured data,the random constraint condition is added to solve the problem of obtaining approximate solution by multi-solution or no solution equation,and a new method of random dynamic splitting of single-layer directional dynamic index is proposed.Dynamic splitting of water absorption,liquid production and oil production of dynamic and static control units of injection Wells and production Wells in typical blocks of Daqing oilfield is carried out and verified with profile test data.Compared with the measured value,the average prediction error is 7.4%,and the error is less than 20%. |