As the underlying industrial software, optimization solver is widely used in energy and power, industrial manufacturing, transportation logistics and other areas requiring complex calculation and planning, with high technical barriers and difficult research and development, and has been dominated by European and American enterprises for decades. Among them, electric power involves the national economy and people’s livelihood, and the optimization solution of power grid dispatching operation has the characteristics of multiple variables, complex constraints, and great difficulty in solving, so the development and breakthrough of domestic autonomous controllable solvers are urgently needed.

Therefore, this competition specially set up a domestic solver competition, the content of the power system unit combination and economic scheduling problems as the background, the decision variables for unit start-up and shutdown and unit output, constraints include unit operation constraints, system balance constraints, grid security constraints, etc., to minimize unit start-up and shutdown costs and operating costs.

This competition brought together all the domestic commercial optimization solver development teams and some power dispatching research institutes. After two rounds of test competition and one formal competition, Damo Yuan solver finally won the first prize. In addition, Tsinghua University and Qingneng Interconnection United, which won the second prize, are also provided with some technical support by Damo Institute self-developed solver.

Damowon solver has excellent performance in both precision index and speed index, which is due to the comprehensive use of innovative technologies such as heuristic neighborhood search algorithm based on relaxation and degradation movement and linear constraint propagation enhanced by maximum clik, and successfully captures and utilizes the special structure of power. Last year, the Dharma Institute self-developed solver team has also won the NeurIPS virtual power plant competition, GECCO&IEEE Energy Scheduling and other international power energy competitions.

IC697ACC720

Under the leadership of internationally renowned mathematician Yin Wutao, the Decision Intelligence Laboratory of Damo Institute has been committed to solver research and development since 2019, starting from zero, going through four years and iterating 26 versions, and finally officially launched version V1.0 in October 2023, becoming a self-developed solver with complete functions and leading performance.

At the same time, Damo Institute also continues to break through the original framework of the solver, combined with the realistic needs of China’s industrial scene, and has launched a self-developed modeling language, parameterizer, optimization platform, in August this year to create a large decision reasoning model, online with mathematical optimization technology ability of AI engineer MindOpt Copilot. Users do not need to know complex math or programming knowledge, directly use natural language to ask optimization questions, MindOpt Copilot can automatically mathematical modeling, the problem into linear programming and mixed integer linear programming optimization model, and code and call the software, calculate the best answer.

Yin Wutao, head of the Decision Intelligence Lab at Dharma Institute, said: “Since we started four years ago, Dharma Institute Solver has been targeting the next generation of technology, rather than just catching up with international leaders.”

At present, the self-developed solver of Damo Institute has been widely used in cloud computing, supply chain, e-commerce, finance and other fields. In the field of power energy, Damo Institute has cooperated with the State Grid to reach the international leading level in the performance of many power dispatching problems; Cooperated with the Power Dispatching Control Center of China Southern Power Grid to release the “Power dispatching intelligent Decision Platform”, which helped China Southern Power Grid to realize the dispatching from 15 minutes to seconds, the accuracy rate reached the level of experienced dispatchers, and supported the fourth and fifth power dispatching AI application competitions of China Southern Power Grid.