A Generative Geological Modeling Method for Intelligent Drilling Algorithm Training Process

Summary As global oil and gas exploration expands into increasingly complex areas, traditional directional drilling methods face significant challenges due to their limited real-time data processing and decision-making capabilities. The next generation of data-driven intelligent drilling steering technology has attracted much attention due to its autonomous decision-making and optimization capabilities. Its effectiveness and stability depend on interactive training or simulation of high-precision geological models. However, the traditional modeling method based on interpolation is limited by geological data, and the model has low resolution and lacks details, which cannot meet the needs of intelligent drilling algorithms for multiscene and high-frequency data stream interactive training. Aiming at the above problems, this paper proposes a generative geological modeling method based on spatiotemporal data fusion, which is inspired by video frame interpolation technology, to support the training process of intelligent drilling algorithms. The core of our method is a multistage image generation network with an encoder-decoder architecture, which can accurately and quickly interpolate geological section images. This method combines the interactive geological modeling mechanism to dynamically generate the geological section image corresponding to the real-time position of the bit during the simulation process. The resulting heterogeneous geological model provides an accurate drilling environment for the intelligent drilling algorithm. The experimental results of geological section generation show that the proposed model is superior to the existing models in peak signal-to-noise ratio (PSNR), structural similarity index (SSIM), and interpolation error (IE) evaluation indicators, which have higher image quality and structural similarity. At the same time, the proposed model has better performance in numerical accuracy indicators because it can more accurately fit geological features and spatial structure information and achieves a generation speed of 30 frames/second (FPS). The interaction experiment with the intelligent drilling algorithm further confirms that the proposed geological modeling method provides necessary data support for the training process of the intelligent drilling algorithm and lays a foundation for its optimization.