AI in Oil, Gas & Consumable Fuels

PetroChina's Research Institute of Petroleum Exploration & Development built a shale oil large language model on Qwen2.5-72B and Qwen3-32B using a 21-billion-token domain corpus, QLoRA fine-tuning with dynamic context parallelism and a reflective RAG loop, reaching 86.35 on objective and 75.54 on subjective shale oil tests versus 77.32 and 12.12 for GPT-4o. A follow-on agent built on the Model Context Protocol orchestrates 23 geoscience tools for seismic processing and interpretation and scored 94.377 on a four-part tool-calling benchmark, above ChatGPT-5.4 and Kimi k2.6.

One of the largest oil and gas companies in Vietnam deployed Nanoprecise's MachineDoctor wireless sensor system with 6-parameter vibration monitoring on rotating equipment, detecting a Stage 3 unbalance fault on Blower H-1741 and avoiding $130,000+ in production losses.

A Journal of Petroleum Technology case study describes Chevron's phased 2020 deployment of OPX Ai's IOCaaS across dozens of gas wells, compressors and a central facility in Alberta. SCADA, historian and ERP data feed AI models that flag exceptions for engineers. The operator reports a 5% LOE reduction in the first year and gradual hand-over of some set-point decisions to the system.

Shell and Huracán Technical Advisory ran an internship proof of concept that adapts Huracán's RET-RAG semantic search tool to Shell's TechXplorer Digest magazine archive of more than 50 volumes. The pilot covered the last few issues, pairing retrieval-augmented generation with locally hosted LLMs and inline citations.

Shell deployed AI-driven predictive maintenance across 10,000+ assets ingesting 20 billion rows weekly from 3 million sensors, running 10,000+ production ML models on Azure Databricks and C3 AI platform at global industrial facilities.

The CNPC Exploration and Development Research Institute and Xinjiang Oilfield developed a one-billion-parameter model for pumping wells, described as the first of its kind in the domestic oil and gas industry. It is applied at scale on over 3,800 wells and diagnoses abnormal conditions with over 90% average accuracy.

Shell deployed AI predictive maintenance across its global oil and gas operations, monitoring over 10,000 critical assets with 20 billion sensor readings per week, producing 15 million predictions daily.

Equinor said AI delivered USD 130 million in value and savings for it and its partners in 2025 and more than USD 330 million since 2020. Predictive maintenance on 700+ rotating machines with 24,000 sensors accounts for USD 120 million since 2020; AI well planning saved USD 12 million at Johan Sverdrup phase 3, and an AI tool raised seismic interpretation capacity tenfold.