Optimal scheduling of a wind energy dominated distribution network via a deep reinforcement learning approach

With the development of clean energy systems, large-scale renewable energy is being connected to the traditional distribution network, which also brings new challenges to the reliable and economic scheduling of the power grid. To address these challenges, this paper proposes an intelligent scheduling strategy for a wind energy dominated distribution network, which aims to reduce the fluctuation caused by the wind energy. First, the energy scheduling model and objective function of the distribution network system are established and the constraints of various types of components are considered. Then, deep reinforcement learning is introduced to realize real-time decision in distribution network to solve the problem of fluctuation caused by the uncertain wind power output. The energy scheduling method is developed into a Markov decision process based on deep deterministic policy gradient (DDPG) algorithm. Finally, the simulation is verified on the IEEE14 node system. The results verify that the proposed approach can effectively reduce power fluctuations in the distribution network. The superiority of the adopted DDPG algorithm is demonstrated by comparing with the deep Q network algorithm.