News Information

In the context of the global energy structure transformation, the oil and gas exploration and development industry is currently facing unprecedented challenges and opportunities. With the advancement of technology and the development of society, the introduction and application of artificial intelligence (AI) technology have brought revolutionary changes to the oil and gas industry. The integration of AI technology not only optimizes traditional operation processes but also opens up new horizons for exploration and development. In the future, through top-level design, data integration, technological innovation, and the establishment of a cooperative ecosystem, it will also bring more efficient, safer, and more sustainable development to the oil and gas industry. This edition delves into the current application status of AI technology in oil and gas exploration and development and predicts its future development trends, aiming to provide references and inspiration for industry professionals. Please stay tuned. 

□ Sun Xudong, a digital technology expert from Sinopec Petroleum Engineering Technology Research Institute 

Against the backdrop where all countries are competing for general artificial intelligence (AGI) technology, the development strategy of China's artificial intelligence industry has been included in the government work report for 2024: "We should vigorously promote the construction of modern industrial systems, accelerate the development of new productive forces. At the same time, deepen the research and application of big data and artificial intelligence, launch the 'Artificial Intelligence+' initiative, and build a digital industrial cluster with international competitiveness." 

At present, major oil companies both at home and abroad have listed artificial intelligence as an important strategic development. China Petrochemical has responded to the new situation and requirements, promoting the digital transformation of oil and gas, planning the overall design of artificial intelligence, enabling it to deeply empower business development and form new productive forces. 

What is the application model and organizational ecosystem of AI technology in the oil and gas sector? 

In the oil and gas sector, artificial intelligence technology has a wide range of application possibilities. On one hand, AI, as a new technology, is directly applied to various sub-sectors of oil and gas exploration and development. On the other hand, through integration with automated robots, industrial internet, digital twins, and intelligent facilities, AI forms systematic solutions that can be applicable to different scenarios. 

In the field of oil and gas exploration, AI can conduct comprehensive analysis of massive geological, geophysical and geochemical data for resource evaluation; identify geological structures and reservoir properties in seismic interpretation; and reconstruct curve data, identify reservoirs and conduct oil, water and gas interpretations in well logging analysis. In oil production, AI can perform real-time processing and analysis of IoT data, identify working conditions, predict production capacity and regulate control schemes; in reservoir development, it replaces traditional reservoir modeling and digital simulation, establishes more accurate reservoir models and fluid flow models, and constructs more precise oil reservoir development plans. In drilling engineering, for the "measurement, transmission and control" closed-loop drilling decision-making, AI, based on geological data and real-time sensor data at the well site, applies new methods such as trajectory optimization, drill bit selection, drilling parameter optimization, and risk analysis and warning to achieve intelligent drilling parameter control and risk prediction. In management and decision-making work, AI can deeply learn and provide decision-making analysis methods and tools for management personnel in complex situations. 

In the fields of automation and robotics, AI combines computer vision (CV) and natural language processing (NLP) technologies to create the "embodied intelligence" of automated equipment. Deep learning-driven robots and drones can perform tasks in dangerous or inaccessible environments, such as underwater pipeline inspections or on-site monitoring. In the industrial internet domain, AI conducts real-time data processing and analysis based on cloud computing power, forming data-driven production operation automation, which is widely applied in earthquake construction, drilling engineering, and oil and gas extraction engineering. In the digital twin technology field, AI utilizes the Internet of Things, big data analysis, and virtual reality technology, through deep integration with geological and reservoir mechanism proxy models, to simulate the performance of geological objects and oil and gas reservoirs in real time. In the intelligent production equipment field, AI combines the Internet of Things and edge computing to form a closed loop of data, people, instructions, and control, promoting the trend of unmanned equipment and facilities, and is widely applied in drilling platforms and oil production platforms. 

What stages has the application of AI technology in the oil and gas industry gone through? 

The field of artificial intelligence officially came onto the historical stage at the Portsmouth Conference in 1956. Over the past few decades of development, it has gradually formed three major schools of thought: the symbolic school, the connectionist school, and the behaviorist school. 

The application of artificial intelligence technology in the oil and gas industry is gradually deepening into specific fields of exploration and development along the technical path of "machine learning - deep learning - large language models / industry-wide large models". 

The period of machine learning application 

Machine learning is an important technical branch within the connectionist school. It involves training based on historical data to dynamically form rules and then applying these rules (models) to solve problems. Technically, it is a simulation of the learning and training mechanisms of the brain. 

The application of machine learning technology in oil and gas exploration and development can be traced back to the 1990s. Machine learning methods such as pattern recognition, genetic algorithms, and BP neural networks (neural network models) have begun to be applied in areas such as processing and interpretation of drilling, logging and testing information, geological structure interpretation, reservoir attribute identification, optimization of drilling while drilling parameters and risk warning, analysis of production decline curves, and optimization of production processes. 

After 2000, Sinopec witnessed the emergence of a number of specialized applications based on data mining, such as test layer selection using genetic algorithms and reservoir "sweet spot" prediction using BP neural networks. Nowadays, machine learning methods have become an important approach in the research and decision-making of oil and gas exploration and development. They not only exist in the management and decision-making processes of various business procedures, but are also applied in almost all seismic processing and interpretation, geological modeling, reservoir numerical modeling, and petroleum engineering professional software. 

2. Period of Deep Learning Application 

Deep learning emerged around 2008 and was formed by the deepening development of a significant branch of machine learning - neural network technology. Through the neural network method that integrates multiple layers of neurons (hence the term "deep" was coined), deep learning has achieved remarkable success in image recognition and natural language processing. 

After 2015, traditional machine learning was on the rise, and artificial intelligence technologies represented by deep learning began to be widely applied in the oil and gas industry. Compared with traditional machine learning, deep learning does not require experts to participate in the construction of "feature parameters" in the method, enabling the full utilization of the advantages of oil and gas big data, especially in fields such as seismic, logging, reservoir development, and oil production, and bringing about unprecedented innovative forces. 

Over the past few years, deep learning has been widely applied in the oil and gas sector to solve problems with big data characteristics such as seismic processing and interpretation, reservoir identification, reservoir model simulation, real-time data analysis and optimization for oil and gas production. Key application areas include seismic data imaging analysis and automatic extraction of structures and properties using convolutional neural networks, automatic extraction of reservoir parameters using deep learning (for turbidite rock identification in Shengli Oilfield), and drilling trajectory design and drilling efficiency improvement, production optimization and production output prediction, as well as design of oil production injection and fracturing schemes using reinforcement learning. 

In the past two years, deep learning has achieved significant technological breakthroughs in the two fields of reservoir geological modeling and oil and gas reservoir numerical simulation by using surrogate models to replace traditional attribute algorithms and fluid mechanics equations. 

3. The Era of Large Language Models 

The large models currently being widely discussed generally refer to large language models (LLMs). Large language models represent the continuous development of deep learning in the field of text processing and are deep learning models with parameters measured in "billion (B)" as the basic unit. 

In 2023, large language models such as ChatGPT developed by OpenAI flourished. Their powerful functions are gradually shaping an intelligent form of "understanding the world" - general artificial intelligence. They not only generate natural language texts but also deeply understand the meaning of the texts and handle various natural language tasks, such as text summarization, question answering, translation, etc. Based on this capability, the focus of applications of large language models is on knowledge mining and retrieval, intelligent question answering, report generation and data analysis interpretation, interdisciplinary collaboration and communication, as well as intelligent analysis and decision-making. 

Since the second half of 2023, the application of large language models in the oil and gas sector has entered a stage of rapid development. It is based on a mature large language model and expands its industry capabilities. It can be applied to major decision-making scenarios such as exploration deployment, well location verification, development plan verification, production command, and emergency command, providing quick and accurate decision-making basis and intelligent problem analysis and strategy recommendations. 

The "Sheng Xiaoli" large-scale model of Sinopec Shengli Oilfield (released in December 2023 as the second version) is supported by multiple large language model backends and possesses over 20 functions such as querying oil and gas expertise, image file querying, production information querying, work progress querying, production anomaly analysis, and document-assisted writing. It has been proven through application that it can significantly reduce the cumbersome tasks of researchers and managers in searching for data. 

The China Petroleum Pipeline Bureau Design Institute, in collaboration with Baidu Company (Wenxin Large Model), has launched China's first artificial intelligence model for the oil and gas storage and transportation field - WisGPT (the first version was released in February 2024). Based on the industry knowledge, data, and application scenarios of oil and gas storage and transportation, this model can provide decision support for all aspects of oil and gas storage and transportation. 

China Petroleum Kunlun Digital Intelligence has collaborated with Alibaba Cloud's Tongyi Qianwan large model to promote the application of the large model in the oil and gas industry. This collaboration aims to provide intelligent decision-making support for the entire process of oil and gas exploration, development, production, and management. 

How will the future large-scale models of the oil and gas industry develop? 

At the 2024 GTC conference, AI scientist Li Feifei pointed out that future models for understanding the world will not be "one-dimensional" large language models, but rather a "four-dimensional structure" model based on "space + time". In the oil and gas industry, due to the highly specialized and segmented nature of the business domain and business processes, mature theoretical methods and research procedures have been formed over a long period of time, as well as research methods that conduct simulation analysis based on mathematical models and mechanism models. This field's high requirements for risk control, accuracy, and precision are beyond the scope that a simple large language model can support. 

At present, the theoretical design of industrial large models is still in the exploratory stage, but the process of applying large model technology to industrial scenarios is beginning. Industrial large models use the general large language model as the "brain", and through technologies such as industrial knowledge injection, mechanism integration, model integration, and closed-loop control, they form solutions tailored to specific application scenarios. For example, "Liangyang Industrial Large Model" is built on the base of Suning Big Model from iFLYTEK, featuring five core capabilities including industrial text generation, knowledge question answering, understanding computing, code generation, and industrial multimodality, and can achieve a full-process closed-loop for problem-solving in specific scenarios. "Pan Gu Large Model" is developed by Huawei based on different industry and scenario requirements, using various modalities of industry data and knowledge accumulation, and based on basic large model technology, it builds analytical capabilities for different industry domains, helping enterprises make efficient decisions in areas such as reservoir simulation, resource assessment, and drilling optimization. 

The development path of large-scale models in the oil and gas industry is to leverage the interactive and intelligent decision-making advantages of large language models, and deeply integrate oil and gas data and industry knowledge to construct an analysis and prediction model that integrates multiple intelligent entities and has multi-modal information processing capabilities. This enables precise simulation and production optimization, as well as high-precision decision-making capabilities. 

In practical applications, industrial large-scale models still encounter numerous technical and application challenges, including technological maturity, data availability, economic feasibility, and industry acceptance, among others. Therefore, a cautious attitude should be adopted towards the actual application effects and impacts of AI technology. The actual application and implementation of industrial large-scale models may still require more research and practice. 

It can be imagined that in the future, the construction of the large-scale oil and gas models might extend from a business perspective, such as geophysical large-scale models centered on seismic attributes, geological large-scale models centered on geological structures, reservoir large-scale models centered on oil reservoir development, and geological engineering large-scale models centered on wellbores and strata. 

What key issues need to be addressed in the application of AI technology in the oil and gas sector? 

To promote the development of artificial intelligence in the oil and gas sector, a top-level design is required. 

The development of "artificial intelligence +" in the oil and gas sector requires a long-term and feasible overall planning: starting from digital transformation, based on data issues, achieving technological innovation through ecological cooperation, and realizing talent collaboration through project-driven approaches. 

Digital transformation aims to enhance the efficiency of the entire value chain and regards artificial intelligence as the core of the enterprise's digital strategy. It uses artificial intelligence infrastructure and technological innovation as its content to build intelligent oil fields and intelligent factories. Data-driven decision-making requires the investment in building a big data platform. By collecting, analyzing, and interpreting massive geological, engineering, production, and environmental monitoring data, it supports the construction of artificial intelligence models. Technological cooperation and innovation should involve collaborating with technology companies (such as IBM, Microsoft, Google, etc.) to develop AI solutions and establish internal R&D teams for technological innovation and breakthroughs. Project-driven talent collaboration should cultivate the cross-border capabilities and collaborative skills of petroleum engineers and data scientists, and promote project implementation through application scenarios. 

2. The integration of data and the establishment of a data ecosystem have become crucial. 

The massive amount of data generated by oil exploration needs to be unified and standardized to ensure that data from different regions can be effectively integrated. At the same time, with the digitization and sharing of data, data privacy and security have become increasingly prominent issues. How to address problems such as data ownership authentication, lifecycle tracking, and data security, in order to protect data property rights and business secrets, is one of the key challenges for ensuring the sustainable development of the industry. 

The International Large-scale Open Data Environment Standard (OSDU) is a major reform currently underway in the oil and gas industry. It is a data standard and software platform jointly promoted by many influential oil companies and oil and gas service companies. Its aim is to optimize data management and analysis processes by providing a standardized and interoperable data platform. OSDU can help companies efficiently manage massive amounts of geological, seismic, reservoir, drilling, etc. data, significantly improving data processing efficiency and accelerating the exploration and development decision-making process. Moreover, OSDU also promotes the integration of data sources and applications from different data providers, software developers, and oil companies, facilitating more comprehensive and accurate underground resource assessment. To a certain extent, it is prepared for big data analysis and artificial intelligence applications. 

Currently, Sinopec is promoting data unification and standardization through the construction of EPDC (data center), and achieving data integration collection, storage and sharing in exploration and development through the integrated technology of data lake and data warehouse. This has enhanced the accessibility and availability of data, and also helps to accelerate the development of innovative big data analysis thinking patterns in the future. 

3. The implementation of industrial large-scale models faces multiple challenges. 

At present, there are still many challenges in the innovation of industrial large-scale models in the oil and gas sector. 

Establishing AI technology applications that are oriented towards commercial value and engineering scenario requirements is an important starting point. Although industrial integration based on large language models has been widely carried out in China, there is no public development strategy or products for large language model applications in the international field. 

From the perspective of extensive industrial applications, currently in China, the application market is still dominated by general large language models. There is not enough penetration into specialized fields, and there has not yet been the formation of a standardized and systematic industrial application model for large models. Especially for industrial large models trained from the bottom up, there are certain technical barriers. 

As the application of large models in the industrial sector has become an industry consensus, large models have shown a development trend of basing on basic large models as the technical foundation and taking industrial applications as the entry point. The concept and implementation cases of industrial large models are also constantly emerging. In the future, large models will also embark on more practical exploration in areas such as data governance, data security, and business models. In the next 5 to 10 years, industrial large models may become an important paradigm for the intelligence of the oil and gas industry, promoting industrial upgrading.

A Lesson from Another Country 

Over the past few years, the organizational model for the application of artificial intelligence in the oil and gas industry has been extensive and highly efficient, featuring strong alliances. Oil companies have engaged in in-depth cooperation with information technology companies and artificial intelligence companies in an ecosystem-building manner, jointly developing solutions for specific segments of the oil and gas industry. 

Shell utilizes Microsoft cloud services and artificial intelligence technology to optimize exploration and oil extraction processes, accelerating its digital transformation. Both Shell and Chevron have collaborated with C3.ai to develop artificial intelligence-based energy management applications, aiming to enhance energy efficiency and reduce carbon emissions. 

The Norwegian National Oil Company has carried out in-depth cooperation with Kongsberg Maritime in underwater robot technology. It has collaborated with the Norwegian Ak Group to develop the world's first unmanned offshore oil platform, and has also cooperated with Microsoft's cloud AI platform to develop digital transformation applications. 

Saudi Aramco has collaborated with Siemens Energy to develop an intelligent energy management plan to achieve the goal of green energy; it has also cooperated with Google to apply machine learning to enhance exploration, production and operation efficiency, such as precise prediction of underground reservoirs. 

Total has collaborated with Google to leverage AI and data analysis technologies to enhance the efficiency of oil exploration and production. 

The Brazilian National Petroleum Corporation has collaborated with IBM to develop a machine learning platform, which is used to optimize the drilling process, reduce non-productive time, and increase the production of oil and gas. 

ConocoPhillips has utilized Microsoft's artificial intelligence and Internet of Things capabilities to achieve efficient production management, equipment maintenance, and asset optimization.