The article presents an innovative approach to oil and gas field infrastructure planning by integrating neural networks with geospatial analysis. The core solution relies on applying the pre-trained Segment Anything Model (SAM), which was specifically fine-tuned for high-precision segmentation of satellite imagery. The paper describes a methodology for fine-tuning of SAM using a specialized dataset of satellite images obtained via the Google Earth Engine platform and annotated with vector data from OpenStreetMap. The primary objective of this fine-tuning is to enhance the model’s capability to accurately identify infrastructure elements and natural landscape features, such as forests, water bodies, roads, and buildings. The key stage of the approach involves generating integrated desirability maps, which reflect the cumulative impact of various limiting factors, including cultural heritage sites, protected natural areas, sanitary protection zones, etc. A system of weighting coefficients was developed to quantify the significance of each factor in constructing these desirability maps. This approach enables automated identification and evaluation of limiting factors, significantly reducing time requirements and increasing decision-making accuracy. The methodology was successfully tested on the Mishkinskoe oil field site, demonstrating the automated placement of well pads and routing of linear communications, taking into account pipeline capacities and cost indicators of designated areas. The results underscore the high effectiveness of the proposed approach in optimizing the infrastructure planning of oil and gas fields.
References
1. Kirillov A., Mintun E., Ravi N. et al., Segment anything, Proceedings of 2023 IEEE/CVF International Conference on Computer Vision (ICCV), Paris, France,
01-06 October 2023, DOI: https://doi.org/10.1109/ICCV51070.2023.00371
2. Dosovitskiy A., Beyer L., Kolesnikov A. et al., An image is worth 16x16 words: Transformers for image recognition at scale, arXiv preprint, 2021,
DOI: http://doi.org/10.48550/arXiv.2010.11929
3. Google Earth Engine. Earth Engine API Documentation. Google Developers, URL: https://developers.google.com/earth-engine/apidocs
4. OpenStreetMap. Open Database of Geospatial Data, URL: https://www.openstreetmap.org/
5. Boeing G., OSMnx: New Methods for acquiring, constructing, analyzing, and visualizing complex street networks, Computers, Environment and Urban Systems, 2017, V. 65, pp. 126-139, DOI: https://doi.org/10.1016/j.compenvurbsys.2017.05.004
6. Sapozhnikov Ya.E., Mironova A.V., Optimization of field development and surface field construction system and the use of artificial intelligence methods (In Russ.), Neft’.Gaz.Novatsii, 2024, no. 4, pp. 66–70.
