Research On Big Data Aided Analysis And Prediction Method Of Oil-water Performance In Water Drive Reservoir

Water injection development is the most commonly used in all reservoir development methods.In the process of water injection development,the water flow will wash the mud between the reservoir voids,change the original pore structure,increase the permeability.At the same time,there will be local accumulation of argillaceous matter,which will reduce the permeability.Therefore,this change should be considered when predicting oil-water performance.The traditional production performance prediction model relies on experience of researchers,and the prediction error is large.The artificial intelligence neural network for oil-water dynamic prediction is difficult to consider the changes of reservoir conditions,and the prediction results are not accurate enough.Therefore,it is very important to establish an oil-water dynamic prediction neural network considering reservoir changes.Combined with graph convolution neural network and sequential neural network,and considering the law of seepage mechanics in reservoir,this thesis establishes an oil-water dynamic prediction method considering the change of reservoir conditions.It mainly includes three parts:(1)Based on the principle of graph convolution neural network,combined with the law of seepage mechanics and the initial reservoir conditions of the reservoir,the inter well connectivity model with adaptive graph correction is established.The model is used to study the inter-well connectivity of five reservoirs.Compared with professional software and ANN networks,the results show that the model can accurately represent the relative size of inter-well connectivity.(2)Combining the above the inter-well connectivity model with adaptive graph modified with time series neural network,a production performance prediction model based on graph correction is established,and the daily oil production of single well and daily water cut of single well of five reservoirs and the liquid production of block B of Bohai A reservoir are predicted.Compared with LSTM,GRU and T-GCN networks,the results show that the adaptive graph correction production performance prediction model considering the changes of reservoir is the most accurate.(3)Deeply study the model and analyze the factors affecting the accuracy of the model.The results show that when there are enough production performance data,the presence or absence of reservoir geological parameters has little impact on the accuracy of the model.When there are few production performance data,the reservoir geological parameters will greatly improve the accuracy of the model.The order of input data has no effect on the accuracy of the model.Predicting the characteristics of an appropriate number of production wells can improve the accuracy of the model to a certain extent.