Data-driven constrained multi-objective dual-layer autonomous intelligent optimization and applications in integrated coal mine energy systems

Exact algorithms, known for high computational efficiency and stability, are among the mainstream methods for solving practical optimization problems, but their performance deteriorates significantly when tackling complex problems. Intelligent optimization algorithms with self-organizing, adaptability and self-learning characteristics are better suited for complex problems but have longer runtimes, which hinders rapid response. Given this, this paper proposes a data-driven constrained multi-objective dual-layer autonomous intelligent optimization method that utilizes the complementary advantages of exact and intelligent algorithms to solve constrained multi-objective optimization problems with different response time requirements. In the first layer, an evaluation rule based on preference information enables a unified comparison of the performance of exact and intelligent algorithms, thereby identifying algorithm categories suitable for solving problems based on their requirements. In the second layer, the most suitable algorithms are generated based on problem characteristics and response time requirements. The proposed method is applied to six benchmark test problems and the operation optimization problems of integrated coal mine energy systems, and the experimental results verify its effectiveness and feasibility.