Research On Auto History Matching Techniques Based On Production-injection Profile

Auto history matching is the technique to obtain the real geology model by minimizing the difference between reservoir production history data and simulation dynamics data.This process is accomplished by computer means using optimization theory to automatically correct the reservoir physical parameters to reflect the real geological conditions of the reservoir.However,it is difficult to solve this kind of large-scale complex problem based on the gradient method which is commonly used in the automatic history fitting at home and abroad especially for those problems of multi-layer model with strong interlayer contradiction.The fitting speed and quality are far from satisfaction.On the other hand,the current automatic history fitting is mostly limited to the inversion of static permeability field,and can not realize the dynamic tracking inversion process of permeability field.But currently most of oil fields at China have entered the later periods of production which is characterized of high water cut.Taking Daqing and Shengli Oilfield as an example,the average water content of the main blocks is more than 90%,and the oil production of these blocks accounts for 50% ~ 60% of the total oil field.The long-term water injection development of the reservoir makes the rock surface and foreign fluid contact continuously in which some physical and chemical effects would inevitably occur during this process and cause the reservoir parameters change.These changes would form a response in the production profile.Therefore,how to track these changes and provide a more accurate reservoir description is important and urgent for oilfield prediction and management in medium and high water cut period of the oil fields.In this paper,the mathematical algorithm is optimized from the perspective of speed and efficiency,and ensemble Kalman filter algorithm is chosen to accomplish our idea.This algorithm is a time series algorithm in which the process of data assimilation is also the process of real-time updating of the model.At the same time,due to the expansion of the amount of data,we need to improve the performance of the algorithm by updating its theroy.In this paper,we propose an improved EnKF scheme from the perspective of near well zone and remote well zone which is based on covariance localization method and sparse optimization method respectively.Covariance Localization is realized by calculating the critical radius length and introducing the localized correlation function in the horizontal and vertical directions in the reservoir model.The critical radius length considers the correlation radius of the priori geological model and the observation radius of the observed data.This method could filter the correlation noise generated by the observed data and reduce the pseudo-correlation in the calculation process of the covariance matrix,which at the same time reduces the possibility of filtering divergence and improves the inversion effect of the near-well area.Sparse optimization method uses as few atoms as possible to express the signal in a given learning dictionary so as to obtain a more concise signal representation,and make it easier for us to obtain the information contained in the signal.And this operation is based on establishing and building a learning dictionary firstly.This method could reduce the computational and storage cost,improve the interpretability of the learning model,and enhance the overall effect of the reservoir,especially for the remote well.The results show that the improved EnKF algorithm which is based on covariance localization and sparse optimization improves the inversion performance and can be applied to the automatic history fitting process to improve the efficiency of fitting.On the one hand,it can improve the accuracy of inversion permeability field.On the other hand,the algorithm has a stronger algorithm stability,especially in a certain range of change,it could capture the change of permeability and track its trends of permeability field so as to realize the tracking process of dynamic permeability field.