A Method For Prediction,Diagnosis And Control Of Reservoir Damage

The choice of reservoir damage control scheme is mainly based on empirical decision,and its effect is contingency.In this paper,from the point of view of oil and gas data,the relationship between the control factors of reservoir damage and the production of oil and gas wells is analyzed by data mining technology.Thus,the production forecast,main factor diagnosis and the quantitative optimization under the condition of unmanned intervention are realized.It is called"cocoon stripping method for control parameters of reservoir or production damage",CSM for short.The parameters of reservoir damage control are defined as the parameters of internal and external reservoir damage sources,which are characterized by partial incompleteness,outliers,magnitude and dimension difference,various characteristics and high latitude.Therefore,the data processing methods such as missing value and abnormal value processing,data normalization,clustering and dimensionality reduction are selected.Then,it introduces how to construct CSM models based on multiple regression,artificial neural network and support vector machine.Finally,taking the N oilfield fracturing data of N as an example,17 models of Ols,BP,Elman,Rbf and SVM between oil production increment and its control parameters are established.The BP network model is the best,the sum of squared residuals is 81.83,and the average accuracy of prediction is 96.86%;Moreover,return rate is the main controlling factor that damages the fracturing effect of N oilfields;Optimize the fracturing parameters of Well#31,and found that when the returning rate is 50%,the fracturing fluid is 260 m~3,the prepositioning fluid is 103 m~3,and the sand volume is 37 m~3,the output can be increased from 1.7 t/d to 10 t/d,indicating the feasibility of CSM.The CSM realizes the quantitative decision of the reservoir damage control scheme,which provides a new way for studying reservoir damage.