Non-intrusive modeling for integrated energy system based on two-stage GAN

Abstract

Generally, an accurate model can describe the operating states of a system more effectively and provide a more reliable theoretical basis for the system optimization and control. Different from the traditional intrusive modeling, a non-intrusive modeling method based on two-stage generative adversarial network (TS-GAN) is proposed for integrated energy system (IES). By using this method, non-intrusive modeling for the IES including photovoltaic, wind power, energy storage, and energy coupling equipment can be carried out. First, the characteristics of IES are analyzed and extracted based on the meteorological data, energy output, and energy price, and then the characteristic database is established. Meanwhile, the loads are classified as uncontrollable loads and schedulable loads based on frequency domain decomposition to facilitate energy management. Furthermore, TS-GAN algorithm based on the Stackelberg game is designed. In the TS-GAN, the first-stage GAN is used to generate the operating data of each equipment identified by non-invasive monitoring, and the second-stage GAN distinguishes the accumulated data generated by first-stage GAN and further modifies the generator models of the first-stage GAN. Finally, the effectiveness and accuracy of the proposed method are verified by the simulation of an energy region.

Keywords

• non-intrusive monitoring
• system modeling
• generative adversarial networks
• integrated energy system
• stackelberg game