【摘要】 随着微电网分布式能源渗透率的不断提高，基于离散动作空间的微电网调度策略存在精度较低、优化效果和稳定性差等问题。基于此，提出一种基于深度确定性策略梯度（deep deterministic policy gradient,DDPG）算法的连续动作优化调度求解方法。通过建立状态空间、动作空间和奖励函数，将调度问题转化为强化学习问题，并采用DDPG算法进行求解。实验结果表明，相比传统的深度Q网络（deep Q-networks,DQN）算法，DDPG算法在奖励函数曲线的收敛性和精确性方面表现更好，调度策略也更优；优化目标上，DDPG算法在微电网综合运营成本和污染物排放量方面分别降低了约5%和34%。更多还原

【Abstract】 With the increasing penetration rate of distributed energy in micro-grid, the micro-grid scheduling strategy based on discrete action space has some problems, such as low precision, poor optimization effect and poor stability. Based on this, a continuous action optimization scheduling solution method based on deep deterministic policy gradient(DDPG) algorithm is proposed. By establishing state space, action space, and reward function, the scheduling problem is transformed into a reinforcement learning problem, and solved by DDPG algorithm. The experimental results show that compared with the traditional deep Q-network(DQN)algorithm, the DDPG algorithm performs better in the convergence and accuracy of the reward function curve,and the scheduling strategy is also better. In terms of optimization objectives, the DDPG algorithm reduces the comprehensive operating cost and pollutant emissions of micro-grids by about 5% and 34%, respectively.更多还原