Relying on Sinopec’s Changheng AI large model, Jianghan Oilfield has developed a full-process “design–construction–evaluation” closed-loop fracturing optimization decision system, boosting efficiency, quality, output, and cost reduction in fracturing operations.

AI gives shale gas fracturing a smart brain

Recently, the Jiaye 18-12HF infill adjustment well in the Jiangdong Block of the Fuling Shale Gas Field achieved a high industrial gas flow of 143,000 cubic meters per day. This marks a major breakthrough in enhancing new-well productivity in the mature area of Fuling. “This achievement could not have been realized without the strong support of our fracturing optimization decision system,” explained Zhu Baiyu, an AI specialist at Jianghan Oilfield Engineering Institute, on November 28.

To further improve shale gas development efficiency, Jianghan Oilfield leveraged its experience from more than 10,000 fracturing stages deployed in Fuling and targeted efficiency, quality, output enhancement, and cost reduction. Using Sinopec’s Changheng AI model, it built a full-process closed-loop fracturing optimization system that supports data-driven iteration of fracturing technology. Three core technologies emerged: intelligent parameter optimization, real-time fracture-network diagnostics, and automated quantitative evaluation—transforming fracturing decision-making from experience-based to model-driven.

This system was selected as an outstanding case at this year’s China Oil & Gas AI Technology Conference, becoming a benchmark example for intelligent fracturing upgrades in the domestic industry. It has also been included in Sinopec’s catalog of typical AI application scenarios.

Intelligent optimization of fracturing design enables optimal parameter output

In the design for the Jiaoye 9-Z1HF well, engineers used reverse engineering to define the optimization parameter range and carried out precise optimization through the system. As a result, test production increased significantly, with daily gas output reaching 180,000 cubic meters, marking a breakthrough in the central gas layer.

A good plan is the foundation of good execution. The system introduces a “mechanism–data–economic” multi-driven pre-fracturing parameter optimization method, establishing links between geological-engineering data and production. Through data augmentation, data correlation increased from 47% to 85%, and production forecasting accuracy exceeded 85%, laying the groundwork for precise parameter selection.

“In the past, fracturing designs focused mainly on maximizing well production,” said Wu Yihao, deputy director of the Reservoir Stimulation Department. By incorporating cost, gas price, and other economic indicators to target the best input–output ratio, the system upgrades the parameter optimization model and shifts from traditional forward simulation to data-economic-driven optimization. This ensures the best economic benefit from fracturing parameters. Several wells this year have exceeded 300,000 cubic meters per day during testing.

Additionally, the system can automatically output four key design elements—stage and cluster layout, scale optimization, pumping schedule, and cost estimation—improving design efficiency by 30% and accuracy by 50%. It also streamlines report preparation and reduces repetitive work.

Real-time fracture diagnostics enable rapid and accurate decision-making

“Remote experts can now monitor the existing fracture network while guiding new stages in real time. When anomalies arise, we can quickly adjust parameters to optimize the fracture morphology and achieve better stimulation results,” described Zhang Fan, a reservoir stimulation expert, on October 11 while observing operations on well Jiaoye 165-5HF from the integrated command center.

Sand plugging is the biggest risk during fracturing. “If the flow path is blocked, fracturing fluid cannot enter the formation—like a clogged water pipe,” Zhu Baiyu explained. This forces the operation to stop for remedial work, costing both time and money.

The system incorporates pressure prediction and sand-plugging early-warning technology, supported by a large database. AI learned from 650 stages that experienced sand plugging and built a predictive model capable of forecasting pressure 1 minute in advance with 92% accuracy. The early-warning function keeps sand-plugging incidence below 0.5% and improves on-site command efficiency by 50%.

The system also integrates real-time fracture-network fitting technology. Learning from 1.5 million fracture propagation parameters, it enables fracture propagation analysis during pumping rather than waiting 3 hours after each stage. Accuracy reaches 85%. “We can predict the next fracture path and adjust in real time, increasing fracture complexity and enhancing stimulation,” Zhu said.

Rapid post-fracturing evaluation improves construction quality

“Previously, real-time evaluation tools were limited. We could only judge fracturing performance during flowback testing after operations were completed,” explained Wei Qi, deputy director of the Reservoir Stimulation Department. “Parameter optimization relied heavily on experience, and dynamic monitoring like microseismic or tracers required long timeframes and high costs.”

Now, the system’s big-data-based construction quality assessment framework provides automated quantitative evaluation after each stage. Using four categories and twelve criteria, the system upgrades from single-dimensional fracture analysis to four-dimensional (parameters, curves, pressure, fractures), enabling automated post-stage evaluation within one hour—shifting assessment from post-operation to real-time guidance.

“With real-time evaluation—evaluate one stage, track one stage, optimize one stage—the overall fracturing effectiveness naturally trends upward,” Wei said. Previously, systematic evaluation took place long after testing, mainly based on models and curve shapes, lacking clarity and immediacy.

Additionally, the system establishes routine quality-control mechanisms for construction evaluation and downhole material testing (fluids, proppant, plugs). This allows automated, non-manual scoring and significantly enhances integrated fracturing operation management.