中国石油网消息：9月4日获悉，由勘探开发研究院和东方物探联合攻关建立的油气地震勘探大模型高质量数据集，在长庆、塔里木等国内外油气田20多个勘探工区的地震勘探实践中取得较好的应用效果，地震解释效率提高9至12倍，有望推动传统地震解释工作模式变革。

China Oil News: On September 4th, it was learned that the high-quality data set of the large-scale oil and gas seismic exploration model jointly developed by the Exploration and Development Research Institute and Eastern Geophysical Company has achieved good application results in seismic exploration practices of more than 20 exploration areas in domestic and foreign oil fields. The seismic interpretation efficiency has increased by 9 to 12 times, and it is expected to promote the transformation of the traditional seismic interpretation work mode.

油气勘探领域积累了海量的叠前叠后地震数据、测井曲线及解释成果数据、地质综合数据等多源异构数据，如何有效训练庞大且复杂的多学科数据，一直是制约模型训练效率及预测准确率的关键难题。

In the field of oil and gas exploration, there are a vast amount of multi-source heterogeneous data such as pre-stack and post-stack seismic data, logging curves and interpretation results, as well as geological comprehensive data. How to effectively train such a large and complex multi-disciplinary data has always been a key challenge that restricts the efficiency of model training and the accuracy of predictions.

为此，勘探开发研究院和东方物探创新提出油气地震勘探大模型高质量数据集构建方法，并应用该高质量数据集在地震大模型构建技术研究等方面取得了显著成效。

Therefore, the Exploration and Development Research Institute and Eastern Geophysical Company innovatively proposed a method for constructing high-quality datasets for oil and gas seismic exploration in the large-scale seismic model, and applied this high-quality dataset to achieve remarkable results in the research of seismic model construction technology.

据悉，该数据集聚焦地震勘探业务，收集包含松辽、鄂尔多斯、塔里木、四川、准噶尔、柴达木、渤海湾等国内主要含油气盆地高分辨率地震原始和成果数据，业务覆盖碳酸盐岩缝洞体、碳酸盐岩生物礁滩、火山岩、页岩、砂岩岩性体等主要油气勘探目标，数据集总量达到220TB。

It is reported that this data collection focuses on seismic exploration business, and collects high-resolution seismic raw data and results from major domestic oil and gas basins such as Songliao, Ordos, Tarim, Sichuan, Qargalai, Qaidam, and Bohai Bay. The business covers major oil and gas exploration targets including carbonate rock fracture and cave bodies, carbonate rock biological reefs and beaches, volcanic rocks, shale, and sandstone lithological bodies. The total volume of the data set amounts to 220TB.

该数据集支撑地震解释大模型预训练、智能二维地震波场正演应用场景建设。目前，数据集应用于9个地震处理、解释业务应用场景建设，包括基于智能化科学计算的正反演、碳酸盐岩走滑断裂识别、碳酸盐岩缝洞储集体智能预测、河道识别、致密砂岩岩性智能预测、断层智能解释等，均取得良好的应用效果。

This dataset provides the foundation for the pre-training of large-scale seismic interpretation models and the construction of intelligent two-dimensional seismic wavefield forward modeling applications. Currently, the dataset is applied in the construction of 9 seismic processing and interpretation business scenarios, including inversion and forward modeling based on intelligent scientific computing, identification of carbonate rock strike-slip faults, intelligent prediction of carbonate rock fracture and cave reservoirs, river channel identification, intelligent prediction of tight sandstone lithology, and intelligent interpretation of faults, etc., and has achieved excellent application results.