Residual oil prediction according to seismic attributes and oil productivity index based on PCA and BPNN
Xiaohuan Yan, Jianxiong Dong, Zhiwen Niu, Dongzhou Liu, Aiquan Chen, Aliyeva Gunay, Jinglin Cui, Yajie Wang, Jiangong Chen · 2024
In mature oilfield with high water cut, dynamic modeling is usually used to predict residual oil, which takes more time and still retains uncertainty in areas without wells. For mature oilfield that have accomplished seismic acquisition in the later development stage, seismic attribute technology could be an efficient and effective method for residual oil prediction. However, in practical application, the prediction results based on seismic attributes cannot match well with production data. This paper proposes a new residual oil prediction workflow based on Principal Component Analysis (PCA) and Back Propagation Neural Network (BPNN) algorithm for residual oil prediction by using multiple seismic attributes and the oil productivity index. In this workflow, the multi-attributes have been optimized using the PCA algorithm, while the BPNN algorithm has been used to obtain a new residual oil prediction attribute with supervision from the productivity index. The new attribute has been proven to have great consistency with the high oil productivity index zone in mature oilfield A in Central Asia. By application, the ratio of high-production wells in this oilfield has successfully increased.