Rio de Janeiro, a forum stirred up waves in the energy industry. The Forum on Empowering New Power Systems with Artificial Intelligence Technology and the 6th Technical Seminar of the China Brazil Power Innovation Consortium were held here, with nearly 200 representatives from all sides. The brightest focus of the event was the 6541 Master Plan, a set of AI strategic frameworks centered on the Guangming Power Model, which was first systematically disclosed by State Grid Corporation of China.

6541, open it and take a look

6541 is not a mysterious code, but the complete framework of State Grid's artificial intelligence strategy, which includes 6 major business areas, 5 types of technical capabilities, 4 key elements, and 1 set of guarantee system. Six major fields cover planning and construction, power grid operation, equipment management, operation control, customer service, and business management, involving more than 600 specific scenarios, from home power outage repair to inter provincial ultra-high voltage transmission dispatch, all of which involve AI participation. The five types of abilities include perceptual intelligence, cognitive intelligence, decision-making intelligence, embodied intelligence, and scientific intelligence, allowing AI to not only see problems, but also understand them, decide how to solve them, and even make robots work on site. The four key elements are the unified technical foundation composed of models, samples, platforms, and computing power, and a set of guarantee systems covering organization, operation, standards, and ecology.

Real landing skills

The implementation carrier of the framework is the first multi-modal power industry model in China, the Guangming Power Model, which was released in December 2024 and has a parameter scale of billions. It integrates various forms of data analysis such as text, images, and videos. It has an average improvement of 20% in core abilities such as power knowledge memory and understanding, multimodal fusion analysis, and business logic reasoning compared to the base model, and an average improvement of 15% in professional abilities compared to mainstream large models. State Grid Corporation of China has fully embedded AI into its top-level strategy, with over 930 intelligent agents implemented and more than 270 core business processes embedded. The cumulative number of model calls has exceeded 1 billion times. In the field of distribution network repair, intelligent agents have been promoted to 27 provinces and cities across the country, with a 35% increase in repair efficiency. Keywords: Power Model, China Brazil Power Cooperation

China Brazil New Paradigm

The real highlight of this forum is the artificial intelligence cooperation initiative in the power industry released by the China Brazil Power Innovation Alliance, which clearly proposes to build an AI demand and case resource library guided by the actual needs of the industry, focusing on scenarios such as prediction and scheduling, intelligent inspection and asset management, digital twin, security and resilience, and promoting joint research and development and pilot verification. The number of members in the consortium has increased from the initial 15 to 38. State Grid has established an independent and controllable two-level intelligent computing center, and built a high-quality sample library of millions of images and millions of texts. As the largest electricity market in Latin America, Brazil has seen rapid growth in new energy installed capacity, and there is an urgent need for AI enabled power system management. The essence of this cooperation is the precise docking of China's power AI capabilities with the Brazilian market, from demand base construction to joint research and development to pilot verification, forming a complete closed loop from research and development to implementation, attempting to establish a replicable international cooperation paradigm, marking that energy AI is moving from domestic to international, from technology verification to large-scale output.Editor/Min Jing