Research on virtualization-based planning optimization method for distribution networks integrating multi-source power signal feature

To enhance medium-to-long-term planning capabilities of distribution networks under uncertain conditions, this study constructs a virtualization optimization method driven by multi-source power signals. Based on feature extraction and temporal alignment of heterogeneous data, a feature fusion model utilizing mutual information weighting is established to generate a unified embedding representation. A virtual resource node set is constructed through resource attribute mapping (), integrating multidimensional metrics such as capacity, volatility, and reliability to form a dispatchable high-dimensional modeling system. Building upon this, a multi-objective optimization framework is introduced, designing an objective function encompassing system investment costs, power losses, uncertainty mitigation, and supply reliability, while establishing a feasible constraint domain. A hybrid optimization process combining NSGA-II and CVaR evaluation mechanisms is developed to achieve rapid convergence of the Pareto solution set and quality control of the solution space. Simulation analysis conducted on the IEEE 33-node system demonstrates that the proposed method exhibits significant advantages in system loss control, regulation cost optimization, and robustness enhancement, validating the effectiveness of multi-source feature fusion and virtualization strategies in distribution network planning scenarios.