Zhai, Wenbao (China University of Petroleum) | Li, Jun (China University of Petroleum) | Xi, Yan (China University of Petroleum) | Liu, Gonghui (China University of Petroleum) | Yang, Hongwei (China University of Petroleum) | Jiang, Hailong (China University of Petroleum) | Zhou, Yingcao (CNPC Engineering Technology R&D Company Limited)
Shale reservoir heterogeneity is more and more focused during shale gas development, especially deep shale gas reservoir buried in the depth of over 3,500 m. However, the evaluation methods of heterogeneity are not always available and poor applicability. In this study, a Principle Component Analysis (PCA)-Artificial Neural Network (ANN) model was presented. The evaluation steps of the model were also given. The validation of the model was confirmed by using a deep shale gas well located in Weiyuan area of Sichuan Basin, China. The results of the validation show that the model presented in this study can be in good agreement with the assessed values of heterogeneity obtained from microseimic events. The developed model's effectiveness was tested by comparing the results acquired from ANN without PCA, where the PCA reduces the dimension of input parameters to improve results of PCA-ANN over 80%. Therefore, the PCA-ANN model can help the engineers evaluate the deep shale reservoir heterogeneity, which provides a tool to give preliminary recommendations of the likelihood of improving the effectiveness of hydraulic fracturing. Implementation of the proposed model can serve as a cost-effective and reliable alternative for the deep shale reservoir.