Abstract: As the energy terminal of the Energy Internet, the model of we-energy is the basis for energy Internet optimization and dispatch. To solve the modeling problem of We-energy, this paper proposed a non-intrusive modeling method based on generative adversarial networks（GAN）. First, considering the energy generation characteristics of wind turbines, photovoltaics, energy storages and coupling equipments, the non-intrusive monitoring for We-energy was realized by joint sliding and gradient separation. Furthermore, combined with the generation capacity and discrimination capacity of generation adversarial network, the improved GAN was used to deal with the time-asynchronous problem of electricity-gas-thermal data, and to classify and identify energy equipments. And a dispatching model for We-energy was established according to equipment classification. Finally, the effectiveness and accuracy of the proposed method were verified by the simulation of an energy region in the north.