Tsinghua University won the fifth power dispatching AI application competition with perfect scores and the competition ended successfully!

The power market clearing speed has set a new record. On December 27, China Southern Power Grid’s fifth power dispatching AI application competition concluded successfully in Guangzhou. The Tsinghua University team won the championship with a perfect score. Relying on "cloud + AI" technology, the overall clearing process was controlled within 600 seconds, improving efficiency by nearly 50%.

When supply and demand are balanced, it is called clearing. With the further advancement of power market transactions, new entities such as photovoltaics, energy storage and virtual power plants have joined in. How to quickly clear electricity as a commodity has attracted much attention in the industry.

In July 2022, the southern regional power market led by China Southern Power Grid was officially launched, becoming the first regional market in the country to be put into trial operation. Currently, 160,000 trading entities conduct power spot transactions on the platform, achieving free and real-time transactions. However, due to the large scale of clearing, complex constraints, and numerous variables, the power grid faces many new challenges in terms of clearing time and operating costs with the acceleration of the marketization process and technological advancement.

On November 14, 2023, China Southern Power Grid held the 5th Power Dispatch AI Application Competition. The theme was rapid clearing of the power spot market based on artificial intelligence. This competition attracted 30 teams from the power industry across the country, including scientific research institutions, universities and Internet ecological enterprises.

Liu Yingshang, deputy chief engineer of China Southern Power Grid Corporation and secretary of the Party Committee of China Southern Power Grid General Investigation Committee, said that in recent years, general large models of artificial intelligence have continued to emerge, forming a pattern in which a hundred flowers bloom. The innovative inspiration displayed by the contestants in the competition is gradually promoting the development of the power "dispatch AI" ecosystem. China Southern Power Grid will continue to uphold technological innovation thinking, give full play to the leading role of artificial intelligence, lead industrial transformation and upgrading, give full play to the role of the national team, adhere to innovation leadership, and further promote digital transformation and digital grid construction.

According to reports, the competition environment of this competition is provided by the scheduling cloud master node and the same city node. Among the six teams that entered the finals, four teams were developed using master nodes, accounting for 66%.

Nanwang Scheduling Cloud was put into production and operation in August 2019. It is based on the Alibaba Cloud Feitian operating system. As the first large-scale industrial cloud built in the national power industry at that time, it was able to expand computing resources in seconds. In the past four years, China Southern Network Dispatching Cloud has been supporting the stable and efficient operation of more than 200 business systems of China Southern Network, and has become a solid foundation for AI innovation.

Wang Zhengcheng, a member of the Tsinghua University team, said: "The overall user experience of the Nanwang Scheduling Cloud Platform is relatively smooth. There is a big difference between development in the cloud environment and scientific research in the laboratory environment, because it is difficult for laboratory servers to be elastic. Expansion and contraction: Although the computing and storage resources currently allocated to each participating team in the South China Open are limited, if large-scale training is to be done, you can quickly apply for expansion."

In order to simulate the actual clearing problem, China Southern Power Grid provides processed power grid data of thousands of generators for 30 days. Another participant, Zeng Hongtai, believes that solving the clearing problem is actually a mixed integer linear programming problem, and the key lies in using AI to determine integer variables.

On the MindOpt Studio AI platform, researchers from Alibaba DAMO Academy have developed an auxiliary programming capability. This capability can help the Tsinghua University team save a lot of time, no longer needing to spend time reading and importing data. Instead, they can focus their efforts on algorithmic innovation.

In the end, the Tsinghua University team won the championship with two perfect scores: "time average score" and "cost average score". They successfully controlled the overall clearing time to about 600 seconds and increased the clearing efficiency by nearly 50%. This fully proves the feasibility of AI to accelerate market clearing.

At present, the national power system is transforming and developing towards a new power system. As new energy photovoltaic and wind power installations continue to increase, the future power system will become rapidly volatile. Traditional dispatching methods are difficult to adapt to the operating characteristics of new power systems in the future. Faced with this rapidly changing system, technology in the power field is moving towards the integration and development of "cloud AI".

Yin Wotao, head of the Decision Intelligence Laboratory of Alibaba Damo Academy, said that the combination of AI and mathematical solving technology is driving changes in the field of new energy power dispatching. The application of dual decision engines has greatly improved the operating efficiency and reliability of the power system, and provided new technical support for intelligent green energy management. In the future, the cloud AI model will continue to promote the power industry to develop in a green, sustainable and intelligent direction.

The China Southern Power Grid Dispatching AI Application Competition is currently the most influential and most applicable AI competition in the power industry. It is hosted by the China Southern Power Grid Power Dispatching Control Center and the China Electrical Engineering Society's Power System Automation Committee.

China Southern Power Grid has always attached great importance to the application of cloud computing and AI technology in the power industry. The past four competitions have focused on AI and revolved around topics such as load forecasting and scheduling optimization decision-making. With the innovative results of previous competitions, China Southern Power Grid has realized AI load forecasting for single-track operation of the entire network and cultivated a flexible and open new energy power forecasting value ecosystem.

The above is the detailed content of Tsinghua University won the fifth power dispatching AI application competition with perfect scores and the competition ended successfully!. For more information, please follow other related articles on the PHP Chinese website!