xpertsystems/oil018-sample

OIL-018合成多相流数据集（样本）是一个免费预览版，基于XpertSystems.ai的企业多相流数据集，专为石油和天然气上游生产中的多相流应用设计。该数据集用于流型分类、段塞预测、PVT属性估计、分离器优化和流动保障机器学习研究。样本覆盖了250口井，分布在12个全球盆地、7种地层类型和5种举升类型，模拟了60天，时间分辨率为240分钟，包含110,150行数据，分布在12个表格中。数据内容涵盖井的主数据、多相流时间序列、压力温度剖面、流型分类（基于Beggs & Brill方法）、段塞事件、分离器性能（基于API 12J标准）、PVT属性（基于Vasquez-Beggs相关式）、水合物/蜡风险、人工举升行为、流动保障异常和生产标签。数据集经过行业标准校准，包括Beggs & Brill、Hagedorn & Brown、API 12J等，并验证了物理一致性（如BHP始终大于WHP）。潜在应用包括流型多类分类、段塞检测、液体加载预测、PVT属性回归、分离器效率优化、水合物/蜡风险评分、异常分类、生产稳定性分类等。但样本存在一些限制，如流型以段塞和搅动为主、液体加载标志触发率高、生产稳定性等级不平衡等，适用于研究而非实时操作决策。OIL-018 — Synthetic Multi-Phase Flow Dataset (Sample) is a free, schema-identical preview of XpertSystems.ais enterprise multiphase flow dataset for flow regime classification, slugging prediction, PVT property estimation, separator optimization, and flow assurance ML. The sample covers 250 wells across 12 global basins, 7 formation types, 5 lift types, simulated over 60 days at 240-minute resolution, with 110,150 rows linked across 12 tables. Data includes well spine, multiphase flow timeseries, pressure-temperature profiles, flow regime classification (per Beggs & Brill 1973), slugging events, separator performance (per API 12J), PVT properties (per Vasquez-Beggs 1980), hydrate/wax risks, artificial lift behavior, flow assurance anomalies, and production labels. The dataset is calibrated to industry standards such as Beggs & Brill, Hagedorn & Brown, API 12J, etc., with physical consistency validation (e.g., BHP > WHP). Suggested use cases include flow regime multiclass classification, slugging detection, liquid loading prediction, PVT property regression, separator efficiency optimization, hydrate/wax risk scoring, anomaly classification, production stability classification, and multi-table relational ML. However, the sample has limitations like slug/churn-dominated flow regimes, high liquid loading flag rates, imbalanced stability grades, and is intended for research, not live operational decisions.