Reservoir Facies Modeling Based on Generative Adversarial Network

作者：Shao-zhong Lin, Senlin Yin, Yaowei Zhang, Juanxia Liu, Chaohai Tao · 发表于：2024 International Conference on New Trends in Computational Intelligence (NTCI) · 年份：2024 · DOI：10.1109/NTCI64025.2024.10776431

Three-dimensional geological modeling of reservoirs is of great significance for developing oil and gas resources, groundwater resources, and carbon dioxide geological storage. Geological facies models are the basis for accurately predicting underground oil reservoirs, geological carbon dioxide storage potential, and groundwater resources. Traditional geostatistical modeling methods can be consistent with geological models to some extent, but there are obvious shortcomings when the characteristics of geological models become complex. Therefore, this paper takes the braided river deltaic diversion channel and estuarine dam phase of the Hanjiang Formation in the Epping Depression of the Pearl River Estuary Basin as the research objective. Based on the geological background of the workings, 3D seismic data and seven logging wells, a conditionally bounded phase simulation based on an improved GAN model is proposed. The research results show that the phase model generated by modeling well facies data matches the input well facies data. The generated phase model has good diversity, and the trained generator can generate high-quality phase models with 100%. accuracy in reproducing well facies.