Thread Content
Recently, the 7th Conference on Digital and Intelligent Innovation Development in the Oil and Petrochemical Industries, hosted by the China Petroleum Enterprise Association, was held in Hangzhou, Zhejiang Province. Experts attending the meeting noted that artificial intelligence (AI) technology is accelerating its integration into the core operations of the oil and petrochemical industries, driving changes across the entire value chain from exploration and development to production management. Despite challenges such as the high complexity of industrial environments and stringent security requirements, AI is becoming an important force driving improvement in efficiency and quality as well as transformation and upgrading within these industries, through the approach of human-machine collaboration and collaborative innovation among industry, academia, research institutions, and end-users. A rational approach should be taken to implement AI in industrial settings; the unique characteristics of these environments require that the application of AI technologies in the petroleum and petrochemical industries proceed in a steady manner. Gu Weixi, director of the Intelligence Research Institute at the China Industrial Internet Research Institute, pointed out that the low tolerance for errors and high security requirements in industrial environments have led to a slower adoption of AI in the manufacturing sector compared to the consumer internet sector. Currently, the industry is facing a severe mismatch between supply and demand: companies with expertise in AI are not familiar with industrial processes, while manufacturing firms lack the AI technical capabilities, which hinders the deeper implementation of automation. To this end, the implementation of AI must adhere to the path of \"human-machine collaboration\". To overcome this challenge, the China Industrial Internet Research Institute took the lead in establishing a \"Platform for Matching Supply and Demand in New-Type Industrialization driven by Artificial Intelligence.\" To date, this platform has facilitated over 300 face-to-face meetings, and a number of successful cases have been created in fields such as automotive design, drug development, and petrochemicals. It helps to reduce the barriers to technology adoption and provides systematic support for the intelligent transformation of industry. Gu Weixi emphasized that AI is undergoing a historic transformation from specialized systems to general-purpose ones, exhibiting three key characteristics of \"bigness, versatility, and integration,\" signaling the full arrival of the era of large models. “The petroleum and petrochemical industry should approach the use of AI with caution: it should actively leverage AI’s potential to address knowledge asymmetries, while avoiding blind pursuit of new technologies. ”Yang Yudong, Secretary-General of the China Digital Governance Research Association, Senior Researcher at the Internet Development Institute of Beijing Normal University, and Guest Researcher at the Capital Science and Technology Development Institute, pointed out that AI technology has surpassed the 50% diffusion threshold and entered a critical stage of transition from growth phase to maturity phase. However, large models remain probability-based tools – they are stronger at detecting correlations than at performing causal reasoning. Regarding network security, Yang Yudong argued that the key to adding value to data lies in the capabilities of the tools. It is necessary to emphasize both management and technology, and to use scientific methods to enhance network resilience; at the same time, enterprises must define clear directions for security technologies and shift away from traditional security approaches. “The implementation of AI in the energy sector must follow the approach of ‘management first, then operation’. ”Yang Yudong proposed the \"four considerations\" principle: consider the era, and grasp the technical characteristics of the first and second halves of the Industrial Revolution ； Observe trends and use S-curves to assess technical maturity ； Consider the context, and focus on addressing the issue of information and knowledge asymmetry ； Look at the products to distinguish between enhanced and native technology paths. AI is shifting from a supplementary tool to a core element in rational AI implementation strategies; in fields such as oil and gas field development, AI technology is gradually evolving from a supplementary tool to a key supporting force. Zhao Hui, member of the Party Committee Standing Committee and vice president of Yangtze University, pointed out that intelligence is the essential path to promoting the precise and efficient development of oil and gas fields. Research shows that intelligence can increase recoverable reserves by 5% and reduce development costs by 10%~30%, playing a key role in improving the quality and efficiency of the industry. However, over 90% of the current reservoir numerical simulation software relies on imports, so it is imperative to develop an independent and controllable technical system. Zhao Hui explained that his team developed a graph structural surrogate model, PINN-GCEM, which integrates physical information. In applications in oil fields such as Bohai SZ36-1, its fitting accuracy exceeded 92%, and the computation speed increased by 1 to 2 orders of magnitude, enabling rapid simulation and parameter inversion on a second-by-second basis. In addition, the \"generalized connection element method\" independently developed by the team has been integrated into CNOOC’s first numerical simulation platform for offshore water-flooded sandstone oil reservoirs – \"Haichao Osim\", resulting in a computing efficiency that is more than 20 times higher than that of traditional methods. Zhao Hui said that the team has also developed a series of intelligent decision-making methods for optimizing injection and production parameters, designing well patterns, and controlling fracture networks, thereby facilitating the transition in oil and gas development from being experience-driven to one driven by both data and underlying principles. Chen Zhangxing, a foreign member of the Chinese Academy of Engineering, a member of the *** Academy of Engineering/Royal Academy of Sciences, and a member of the American ** Academy of Engineering, introduced the \"Gaia Large Model\" developed by his team. This model enables efficient solution finding and generalized application in various applications such as well logging interpretation, 3D modeling, and parameter optimization, thereby facilitating the creation of an integrated workflow that spans from intelligent interpretation to production management. State-owned enterprises are taking the lead in promoting the deep integration of AI with various industries, serving as the driving force behind industry transformation; many energy-related state-owned enterprises are actively advancing the application of AI, large-scale models, and the industrial internet. Shan Tongwen, the leader of the special project team in the Department of Science, Technology and Digital Transformation at China National Offshore Oil Corporation, explained that CNOOC is thoroughly implementing the \"Artificial Intelligence +\" initiative to foster comprehensive integration of artificial intelligence with the ocean energy industry. Among them, the Shenhai-1 gas field operates with intelligent remote control, enabling its annual production capacity to reach 4.5 billion cubic meters; it thus became the world’s first ultra-deep-water platform capable of continuous operation during typhoon seasons. Data is a core element of corporate digital transformation. According to Shan Tongwen, CNOOC has built the country’s first smart refinery through the use of \"5G + industrial Internet\", achieving a 100% quality compliance rate, zero major safety incidents, and a reduction in carbon emissions of over 200,000 tons per year. At the same time, an integrated digital platform for mid- and downstream sales was developed, resulting in total transaction amounts exceeding 2 trillion yuan and serving over 11 million customers. Wang Zizong, Chief Engineering Expert of Sinopec Group Co., Ltd., said that Sinopec has established a three-level model system of \"foundation–industry–specialty\". Its independently developed \"Great Wall Large Model\" comes in two versions with 700 billion parameters and 700 billion parameters respectively, and has passed **leading-level evaluations ； At the same time, a high-quality multimodal dataset covering text, images, and videos was created, with a volume of over 650TB. A computing power system of 781Pflops was established through a combination of leasing and in-house construction, as well as cloud-edge collaboration, with a domestic production rate of over 80%. Regarding the implementation of these solutions in practical scenarios, Wang Zizong explained that Sinopec has used AI technology to increase the efficiency of earthquake data processing by 10 times, reduced the development time for polyimide gas separation membranes by 80%, and improved the efficiency of developing lubricant products by 30% to 50%. To date, 25 use cases have been launched, and 313 business systems have integrated AI capabilities, thus initially fostering an innovative environment in which everyone makes use of AI. Hu Bingjun, General Manager of the Digital and Information Management Department at China National Petroleum Corporation, explained. The Kunlun large model developed by China National Petroleum Corporation adopts a \"1+4+N\" four-layer architecture; 60 large models have been trained using this framework. The parameters of the language-based large models have increased from 70 billion to 300 billion, while those of the vision and multi-modal large models have also seen a significant rise. **The pipeline network achieves precise monitoring and intelligent control of its operation by establishing an intelligent system that covers scheduling and command, equipment management, safety monitoring, and business decision-making ； **Energy relies on the industrial Internet to build a \"one network, one database, three platforms\" architecture, enabling data integration, model-driven operations, and three-level collaboration.