With the gradual increase in the penetration rate of electric vehicles (EV)

the impact of large-scale EVs integrating into the power grid on the distribution network has become increasingly evident. Meanwhile

due to the ever-intensifying uncertainty of EV loads

the safe and stable operation of the distribution network is now facing a severe threat. To address this issue

this paper proposes an active distribution network's active-reactive power collaborative optimization method that considers the uncertainty of EV loads and reactive power compensation. Initially

the Wasserstein distance and gradient penalty are employed to improve the generative adversarial network (GAN)

reducing the possibilities of gradient vanishing and non-convergence of the model. By introducing conditional variables and the self-attention mechanism

typical charging scenarios of electric vehicles are generated through the improved GAN to depict the uncertainty of EV charging loads. Then

electric vehicles are modeled to explore their reactive power compensation capabilities

and the demand response mechanism is adopted to guide EV users. Dynamic weight coefficients are used to construct a collaborative optimization model for the active and reactive power of the distribution network. Finally

the niche technique is combined with the improved particle swarm optimization algorithm to solve this model. The case analysis demonstrates that the proposed method reduces the distribution network's power loss by 18.14% and raises the overall voltage to above 0.95 pu

effectively reducing the occurrence of voltage violations while improving the system's economy.