At the recently held "2025 Digital Transformation Empowering Smart Power Plant Construction Exchange Conference", customer representatives from companies such as Guodian Power Development Co., Ltd. and Jiangsu Guoxin Yangzhou Power Generation Co., Ltd. shared benchmark cases of digital transformation jointly created with Luculent. Luculent experts presented the innovative applications of the SuShine YaoGuang large - scale model in the industrial field, providing a replicable transformation path for the power industry from multiple dimensions of technological empowerment and practical implementation.
AI - Driven Digital Transformation in Thermal Power Safety
In the digital cockpit of Guodian Power, AI is driving a smart transformation in safety production from "human - based governance" to "data - based governance".
Fu Yu, the director of the Production Technology Department of Guodian Power, introduced in his report that the digital control platform for thermal power safety production of Guodian Power, as an important part of the company's transformation plan, has constructed a "4 - center" professional digital control model for safety production. It takes the "Company Production Control Center" as the command center, relies on the "Operation and Maintenance Center" (power plants), and is supported by the "Regional Material Center" and the "Regional Maintenance Center". This model not only meets long - term planning and assists in scientific decision - making but also achieves management closure, strengthens information concentration, revitalizes digital assets, and responds to policy requirements. "This is not just a simple technological upgrade but a genetic recombination of management thinking. Guodian Power is pressing the 'accelerator' and sprinting full - speed towards a new era of safe, efficient, and green smart energy."
Fu Yu introduced that the platform has been applied in 12 thermal power units and 37 generating units. More than 50,000 "AI + mechanism" intelligent early - warning models have built a three - dimensional protection network of "operating parameters - equipment failures - economic optimization - working condition optimization", making early - warning more accurate and efficient, and enabling safety production to shift from "fire - fighting after the event" to "pre - event prevention". At the same time, the platform precipitates and transforms expert knowledge and experience into a "Four - Library Encyclopedia". The anomaly library records the "medical records" of equipment, the risk library draws a "map" of potential hazards, the hazard library establishes "archives" of problems, and the knowledge library inherits the "mental methods" of experts. Relying on this platform, Guodian Power has also organized more than 10 proposition model competitions among its thermal power units, further enhancing employees' digital iteration capabilities through "learning through competition".
Full - chain Domestic Upgrade of Intelligent Monitoring
At the application report meeting of the smart power plant project of Jiangsu Guoxin Yangzhou Power Generation Co., Ltd., Yang Peng, an expert from the Power Generation Department, introduced a "milestone" breakthrough recently achieved by the company - the successful completion of a comprehensive domestic upgrade of the intelligent monitoring 3.0 system.
"The 3.0 version has added 68 AI models and 20 non - outage and non - derating early - warning models, and both the early - warning accuracy rate and timeliness rate have increased to a high level of 98%. It is like an 'unmanned driving system', escorting the safety production of the power plant," Yang Peng said. All these changes are due to the new domestic system kernel - using the Galaxy Kirin V10 operating system, adding a data backup server, upgrading the time - series database to Trend DB V5.0, upgrading the digital - intelligent platform to LiEMS8.0, with rich built - in AI algorithms, greatly improving performance and safety, and achieving a "double leap" in information security and self - controllability.
Yang Peng also said that the 3.0 version of the system has also been comprehensively optimized and upgraded in terms of user operation experience. The early - warning query function can now accurately count according to the content and conduct a detailed historical traceability of each early - warning, making the display of early - warning information clearer and more intuitive. The early - warning traceability method has changed from text description to a snapshot of the early - warning trigger link on the canvas. Users can replay the whole process of early - warning triggering like watching a high - definition movie, making early - warning analysis more efficient and powerful. Jiangsu Guoxin Yangzhou Power Generation Co., Ltd. will continue to conduct in - depth research in areas such as unit operation optimization, equipment energy consumption degradation analysis, and energy conservation and carbon reduction, continuously tap the potential of the intelligent monitoring system, and use self - controllable innovative technologies to promote the construction of a new energy system to take new steps.
Luculent SuShine YaoGuang Large - Scale Model: Unleashing New Productivity of Industrial AI
"The emergence of DeepSeek has promoted the implementation of industrial large - scale models. However, general large - scale models are prone to 'hallucinations' and have limitations and challenges in identifying professional terms and handling complex business logic. This requires professional industry - specific large - scale model platforms. SuShine YaoGuang is fully integrated with industry - specific big data and enterprise - internal knowledge bases, forging a dedicated large - scale model for industrial enterprises that 'understands industry jargon'."
Bian Zhigang, the vice - dean of the SuShine Industrial Internet Research Institute, introduced in his keynote report that the YaoGuang large - scale model provides a complete capability framework around knowledge reasoning, multi - modal recognition, and process optimization decision - making. Especially for scenarios such as equipment failure early - warning, intelligent monitoring, and energy - saving optimization, it uses the large - scale model to dispatch small models to achieve collaborative operation among different intelligent agents. Taking the failure early - warning scenario as an example, each intelligent agent accurately perceives real - time signals like an "industrial nerve ending", calls the calculation results of small models, completes local calculation and feature extraction, and finally converges the results to the center for global deduction. The large - scale model, as the "industrial brain", conducts comprehensive diagnosis in combination with historical data and process knowledge bases and generates accurate operation and maintenance plans. This architecture not only solves the problem of connecting complex business logic but also significantly improves the credibility of model decisions.
Bian Zhigang pointed out that the YaoGuang large - scale model provides one - stop industrial large - scale model development services, supports the management of general large - scale models such as DeepSeek, and has a built - in mature large - scale model toolchain and a unified model scheduling and management platform, making model fine - tuning and prompt - word engineering more convenient and standardized. The product is mature and ready - to - use. It has precipitated more than 30 mature scenario applications around five major types of scenarios, including production operation, intelligent equipment operation and maintenance, safety production, energy - efficiency optimization, and business management, enabling enterprise large - scale models to be quickly implemented and achieve results.
The conference also invited representatives from leading enterprises in the power industry such as China Energy Group, China Yangtze Power, CNOOC Gas & Power Group, China Huadian, and Three Gorges New Energy, as well as experts and scholars from the Electric Power Planning and Design Institute and Tsinghua University to gather for a dialogue, conducting in - depth exchanges around energy digital transformation, technological innovation, and scenario practice.
This conference is not only a concentrated display of the industry's digital intelligence achievements but also the starting point for ecological collaborative innovation. Luculent sincerely thanks the Thermal Power Committee of the Chinese Society for Electrical Engineering for building this high - end dialogue platform and thanks customers and partners for their choice and trust. In the future, the company will continue to deepen the integration and innovation of technologies such as AI large - scale models and cloud - edge collaboration with industrial scenarios, unite the forces of industry, academia, research, and application, and promote the industry to move towards a new future of high - quality development.
