Ulmki, China, June 23, 2025 /prnewswire/ – June 20th, the state's Xinjiang Information & Telecommunication Company successfully completed the deployment and optimization of the knowledge services infrastructure, achieving unified management of professional and internal/external knowledge resources.
Currently, the application of artificial intelligence continues to deepen, but the knowledge service systems and processing modes remain challenged and struggles to meet the requirements of various disciplines. Key issues include the limitations of large-scale models, fragmentation of knowledge within the power industry, and low application efficiency. As a core infrastructure to support the construction of new power systems, the platform aims to establish a comprehensive knowledge hub across the business chain, integrate industry knowledge resources, and overcome the bottlenecks of technical applications. By leveraging RAG (searched generation) technology, a mechanism combining “large-scale model context learning” with “high-quality external knowledge input.” Accurate searching of fragments of knowledge and fusion of multi-source data has enhanced the recall accuracy to 94.7%, significantly outperforming traditional vector search methods. This ensures the permissions and accuracy of the output content.
The Knowledge Services Infrastructure Platform has three core features: Integrated internal and external knowledge base search, intelligent extraction of multimodal information, advanced question answering and data analysis capabilities. It adopts a visual configuration system, provides flexible RAG policy support, enabling intelligent matching across synchronous search and multi-source knowledge bases. To date, the company has collected 19 requirements for the construction of the RAG knowledge base and over 19,000 knowledge documents. Of these, 710 documents relating to intelligent judgment scenarios for power economic relations test the slicing of knowledge and knowledge space construction. The scenario application is successfully integrated through the API interface, with the slice data recall accuracy increased by more than 40% compared to the original method, significantly improving the accuracy of question answers. This transformation significantly improves work efficiency and quality, and changes dramatically from “experience-driven” to “knowledge-driven” across the professional domain.
Going forward, State Grid Xinjiang Information & Telecommunication Company will continue to improve its Knowledge Services infrastructure platform and promote its iterative capabilities. Knowledge internal storage management and regular operational support involves sustained efforts to ensure that infrastructure services remain applicable and user-friendly, while maintaining the authority and timeliness of knowledge content. This initiative effectively promotes the deep integration of artificial intelligence technology and power grid structures.
