xpertsystems/oil034-sample

OIL-034合成排放数据集（样本）是一个免费的、模式一致的企业排放数据集预览版，由XpertSystems.ai开发，专为石油和天然气行业的排放与可持续性应用设计。该数据集包含12个CSV文件，总计133,980行数据，覆盖110个设施、10个真实生产区域（包括二叠纪盆地、鹰福特、巴肯、马塞勒斯、海恩斯维尔、墨西哥湾沿岸、北海、西加拿大、中东和西非）和10种资产类型（如上游生产、压缩机站、天然气处理、管道终端、LNG终端、炼油厂、储罐农场、海上平台、CCUS设施和氢气单元），时间跨度为45天。数据集旨在支持多种机器学习任务，包括CO2/甲烷排放清单建模、超级排放源检测、火炬燃烧效率优化、碳捕集利用与封存（CCUS）性能建模、卫星羽流相关性分析、监管报告分析和碳强度分级。数据集基于多个行业标准进行校准，如EPA温室气体报告计划（40 CFR Part 98 Subpart W）、EPA AP-42排放因子、IPCC AR5 GWP-100甲烷转换、Pasquill-Gifford大气扩散模型、EPA 40 CFR 60 Subpart Ja火炬燃烧效率标准等。数据集中包含详细的物理耦合特征和标签，如排放风险评分、超级排放源标志（>100 kg/hr）、监管超标标志和碳强度等级（A/B/C/D）。此外，数据集还提供了多表关系机器学习的能力，通过facility_id和timestamp可连接12个表格。样本版本具有确定性生成特性，使用种子参数确保可重复性，但请注意样本存在一些局限性，如碳强度等级偏斜、甲烷均值较高等，这些在README中已详细说明。OIL-034 — Synthetic Emissions Dataset (Sample) is a free, schema-identical preview of XpertSystems.ais enterprise emissions dataset for CO2/methane emission inventory ML, super-emitter detection, flare combustion efficiency optimization, CCUS performance modeling, satellite plume correlation, regulatory reporting analytics, and carbon intensity grading. The sample covers 110 facilities across 10 real production regions (Permian Basin, Eagle Ford, Bakken, Marcellus, Haynesville, Gulf Coast, North Sea, Western Canada, Middle East, West Africa) and 10 asset types (upstream production / compressor station / gas processing / pipeline terminal / LNG terminal / refinery / tank farm / offshore platform / CCUS facility / hydrogen unit) over 45 days with 133,980 rows across 12 tables. The dataset is calibrated to named industry standards including EPA Greenhouse Gas Reporting Program, EPA AP-42 Emission Factors, IPCC AR5 GWP-100 methane conversion, Pasquill-Gifford atmospheric dispersion, and EPA 40 CFR 60 Subpart Ja flare combustion efficiency. It includes feature-coupled ML labels such as emissions risk score, super-emitter flag (>100 kg/hr), regulatory exceedance flag, and carbon intensity grade (A/B/C/D). The sample is deterministic via a seed parameter for reproducibility and is intended for emissions ML research, with noted limitations on sample-scale representativeness.