en.Wedoany.com Reported - The 2026 World Artificial Intelligence Conference and High-Level Meeting on Global AI Governance was held in Shanghai from July 17 to 20. During the conference, the State-owned Assets Supervision and Administration Commission of the State Council hosted the "AI Empowering New Quality, Full-Scale Renewal" Enterprise AI High-Quality Development Forum on July 17, releasing the outstanding achievements of the "Hundred-Domain Intelligent Data" Central Enterprise High-Quality Industry Datasets. The "Multimodal Dataset for Intelligent Management of Photovoltaic Engineering Progress" from China Energy Engineering Group Yunnan Electric Power Design Institute was successfully selected.
This dataset is a representative data asset in the field of digital intelligence for new energy engineering. It stood out among numerous submissions from central enterprises and became one of the first batch of high-quality datasets released under the "Hundred-Domain Intelligent Data" initiative, marking progress by China Energy Engineering Group in the valorization and scenario-based application of engineering data elements.
The "Multimodal Dataset for Intelligent Management of Photovoltaic Engineering Progress" addresses pain points in photovoltaic engineering progress management, such as scattered data, delayed information, and difficulties in verifying engineering quantities. It constructs a multimodal data system encompassing drone aerial imagery, on-site videos, and construction documents, covering key construction phases including drilling, pouring, support installation, and module assembly, providing standardized data support for automatic calculation of engineering quantities and dynamic progress control.
In terms of data scale and quality, the dataset totals over 5TB, containing more than 50,000 aerial images and over 10,000 annotated images, with a daily increment of 10GB. The comprehensive data quality assessment score is 95.2 points, and the scene recognition accuracy exceeds 95%. It features large volume, high quality, and sustainable iteration, providing a solid data foundation for training intelligent recognition models for photovoltaic engineering progress.
In terms of application effectiveness, the full-chain digital intelligence management platform built on this dataset can automatically identify construction elements, calculate engineering quantities, and display progress. Currently, the dataset has been applied in over 10 domestic and international photovoltaic EPC projects, including the CGN Northern Laos Clean Energy Base and the Inner Mongolia Alxa High-Voltage Outbound New Energy Base, both under the general contracting of Yunnan Electric Power Design Institute. The average project duration has been shortened by 8% to 12%, saving 15 to 20 days, with comprehensive cost savings of 1.5 to 2 million yuan. Progress data update efficiency has increased by over 30%, effectively reducing manual inspection workload and on-site safety risks, driving project management from manual statistics to real-time perception and from experience-based judgment to data-driven decision-making.
This dataset is based on the research findings of the scientific project "Research on Specific Feature Extraction Methods from UAV Digital Orthophoto Images Based on Spatial Machine Learning" by the Yunnan Electric Power Design Institute project team. The results deeply integrate UAV aerial surveying, spatial machine learning, and photovoltaic engineering progress management, forming a replicable and scalable path for engineering dataset development. It achieves the transformation from research methods to engineering practice, providing a more solid data foundation and technical support for "AI + Construction" and the digital intelligent management of new energy engineering.
