A novel clustering approach combining Kmeans++ and PAM is introduced to segment the daily load curve chronologically for the dynamic reconfiguration of distribution networks incorporating time-varying wind solar power and loads. Multi-objective dynamic reconfiguration model of distribution network based on the optimal objectives of comprehensive cost, voltage offset and load balance. To enhance the computational efficiency of the model, an improved Nutcracker optimization algorithm (INOA) was proposed, which used Tent mapping + quasi-reflection learning to provide high-quality initial population. Dynamic fitness-distance balance selection method and tangential flight strategy are introduced to enhance the global search capability. The Cauchy-Gaussian variation perturbation is incorporated to augment the algorithm"s capability to escape from local optima. Using the IEEE 33-node system as a basis, the outcomes indicate that the suggested approach effectively achieves optimal load distribution and efficiently addresses the restructured model.