CNPC Builds a Cognitive Computing Platform to Promote AI Application in the Oil and Gas Industry

CNPC worked with Huawei to build a cognitive computing platform — E8. It uses AI technologies like knowledge graphs, natural language processing, and machine learning to build, calculate, and apply knowledge systems. This provides intelligent analysis methods for oil and gas E&P research as well as production management. The platform helps expand access to reserves while reducing costs. With the platform, decision makers can use findings from mass data to make quicker, better informed decisions.

The scientific discipline of AI officially began in 1956. It was first proposed by John McCarthy, Marvin Minsky, and a select few other scientists. This year, ChatGPT, an AI-powered language model, made its debut. AI technology, known as the engine of the fourth industrial revolution, has changed life as we know it. This is no different for the oil and gas industry.

AI Helps Break Through Bottlenecks in Oil E&P

Exploration is the first key step of oil and gas exploitation. The object of oil and gas E&P in China is morphing from conventional to unconventional oil and gas fields, and from shallow to deep. However, as the exploration process continues, high-quality resources are depleting. Many old oilfields have entered their middle and later stages with an ultra-high water cut. With remaining reservoirs becoming sparse, production is in serious decline. Therefore, updating and iterating E&P technology is the key to increasing the quality and efficiency of oil and gas production and removing development bottlenecks.

To this end, CNPC worked with Huawei to build a cognitive computing platform — E8. It uses AI technologies like knowledge graphs, natural language processing, and machine learning to build, calculate, and apply knowledge systems. This provides intelligent analysis methods for oil and gas E&P research as well as production management. The platform helps expand access to reserves while reducing costs. With the platform, decision makers can use findings from mass data to make quicker, better informed decisions.

CNPC's cognitive computing platform is a general, open, and scalable AI computing platform. It is designed based on four key factors: data, algorithms, computing power, and scenarios. It provides a one-stop AI development environment from data processing, machine learning, model release, all the way to inference application. The platform includes five modules: data processing and feature analysis, natural language processing and knowledge graph construction, 120 intelligent algorithms and visual machine learning pipelines, and the open intelligent service supermarket. It provides users with 35 common intelligent services and high-performance computing power, involving intelligent Q&A, knowledge search, knowledge recommendation, and text generation.

CNPC's cognitive computing platform solidifies and inherits E&P knowledge and expert experience, which can then be shared. Moreover, it provides an intelligent engine for oil and gas service innovation.

CNPC's Cognitive Computing Platform: AI Helps Explore Oil and Gas

Oil and gas are both buried thousands of meters underground, where the landscape is complex and obtaining geological data is costly. Take well logging as an example. Across thousands of meters underground, there may only be a few meters worth of recoverable oil formations. Geologists need to accurately judge the characteristics of underground structures and reservoirs. One way they can determine this is to use HPC technology to comprehensively calculate massive seismic data and then rely on analysis by experts. Otherwise, they just have to take a stab in the dark.

Thousands of wells are drilled each year in an oilfield. However, it usually takes one to two days for an expert to identify a new spot worth drilling. This is a huge workload for employees. To take this burden off their shoulders, they can use the cognitive computing platform.

By doing so, Changqing Oilfield can quickly construct a specialized model for oil logging. It activates and integrates sealed logging data and scattered stratigraphic data. The model helps intelligently identify oil, gas, and water layers, shortening the average identification time by 70%. Plus, the identification accuracy is comparable to logging interpretation experts.

Quality Practices Are Vital

During large-scale application, CNPC's cognitive computing platform has outstanding performance in seismic data processing, seismic data interpretation, well logging oil and gas formation identification, pumping well condition diagnosis, and so on.

At Daqing Oilfield, the machine learning method is used to predict the crude oil yield and water cut for old, new, and measured wells. The prediction model working at its peak is 90.74% accurate, and prediction efficiency is 10 times higher than that of the traditional method. At present, it is commonly used by the Daqing Oilfield Exploration and Development Research Institute as well as oil production plants for development deployment and dynamic analysis. At Dagang Oilfield, based on the quantitative diagnosis of oil well working conditions and remote real-time online management, oilfield management systems can go from post-diagnosis to pre-warning. The diagnosis accuracy of abnormal working conditions can reach over 90%, and the operation maintenance cost can be reduced by 20%.

To date, CNPC has used AI to form best practices and advance exploration in 22 E&P scenarios. Huawei will spare no effort to use innovative technologies to promote E&P transformation. In terms of E&P technologies, we will strive to help CNPC go from a follower to a leader in the space.