Study on High-dimensional Multi-objective Optimization Methods for a Power Integrated Energy System Digital Twin

The construction of a digital twin for integrated power and energy systems faces the core challenge of highdimensional space multi-objective optimization. This study proposes an optimization framework based on cross-dimensional modeling and an improved Non-dominated Sorting Genetic Algorithm III (NSGA-III), aiming to address the deficiencies of traditional methods in model accuracy, computational efficiency, and decision-making dimensions. By employing hierarchical modeling techniques to integrate the dynamic characteristics of electric, thermal, and gas energy flows, a multi-physics coupling model of the digital twin is constructed. The proposed approach optimizes the distribution of solution sets in high-dimensional objective spaces by combining an improved reference point strategy with adaptive dimensionality reduction technology. Experimental results show that, in microgrid scheduling scenarios, the proposed method improves the hypervolume indicator (HV) by $18.7 \%$ and increases computational efficiency by 32% compared to traditional NSGA-II and MOEA/D algorithms. The research findings provide theoretical support for real-time optimization and reliability analysis of integrated energy systems.