Multi-microgrid optimization operation strategy considering nonlinear conditions and renewable energy uncertainty: A data-driven method

Achieving stable operation of multi-microgrid (MMG) systems with complete privacy protection is a challenging problem, which is rendered more difficult when considering both the nonlinear conditions of the equipment and the uncertainty of renewable energy. To address these issues, this study proposed a novel data-driven method. First, the neural network was used to describe the relationship between the partial load and the efficiency and carbon emission factors of the equipment. Second, the least squares-generative adversarial networks (LS-GAN) and Monte Carlo were used to calculate the output interval of wind power and photovoltaics. Consequently, a method for calculating the normal distribution model based on historical information was proposed. Finally, a multi-agent soft actor-critic (MA-SAC) was proposed to realize peer-to-peer energy trading and privacy protection for the MMG system. The results of a case study show that the proposed method is sufficiently flexible to solve the problem of MMG optimization operation considering the uncertainty of renewable energy under nonlinear and nonconvex constraints. In addition, MA-SAC exhibits stronger optimization capabilities than the current mainstream multi-agent deep reinforcement learning algorithms.