Data-Driven Dynamical Control for Bottom-up Energy Internet System

With the increasing concern on climate change and global warming, reduction of carbon emission becomes an important topic in many aspects human society. The development energy Internet (EI) makes it possible to achieve better utilization distributed renewable sources with power sharing functionality introduced by routers (ERs). In this paper, a bottom-up EI architecture is designed, novel data-driven dynamical control strategy proposed. Intelligent controllers augmented deep reinforcement learning (DRL) techniques are adopted for operation each microgrid independently bottom layer. Moreover, concept curriculum (CL) integrated into DRL improve sample efficiency accelerate training process. Based exchange plan determined layer, considering stochastic nature electricity price future market, optimal dispatching scheme upper layer decided via model predictive control. simulation has shown that, under architecture, compared conventional methods such as proportional integral flow, proposed method reduces overall generation cost 7.1% 37%, respectively. Meanwhile, CL-based can significantly speed up convergence during DRL. Last but not least, our increases profit trading between ERs main grid.