A Novel GAN Architecture Reconstructed Using Bi-LSTM and Style Transfer for PV Temporal Dynamics Simulation

Abstract

The stochastic production simulation of photovoltaic (PV) power is crucial for the analysis of power balance in power planning, annual or monthly operational planning, and long-term transactions in the electricity market, especially in power systems with a high share of PVs. To model the uncertainty and temporal characteristics inherent in PV power, this letter introduces the style transfer and innovatively establishes bi-directional long short-term memory generative adversarial networks (GAN). Simulation results confirm the advantages of the proposed GAN over traditional convolutional neural network-based GANs in simulating the diversity and temporal characteristics of PV power.