高比例光伏微网无功均分控制中的Q学习方法

Q-learning Method in Reactive Power Sharing Control of High-proportion Photovoltaic Microgrid

针对采用传统下垂控制的分布式电源无功功率分配不均,提出了一种高比例光伏微网无功均分控制中的Q学习方法。该策略融合人工智能算法的随机搜索机制以及Q学习算法的迭代机制。首先,针对采用下垂控制的分布式光伏,以微网总无功偏差量作为奖励函数的依据,构建电压幅值和无功功率之间的反馈。其次,根据最大奖励Q值对应动作控制分布式光伏输出电压幅值。最后,协调控制分布式光伏无功输出,实现含高比例光伏微网的全局最优控制。仿真结果证明了该策略可以有效提高无功均分控制效果,减少系统无功环流,提高电压质量。

In view of the uneven distribution of reactive power under the traditional droop control of distributed genera⁃tion(DG),a Q-learning method in the reactive power sharing control of high-proportion PV microgrid is proposed.This strategy combines the random search mechanism of the artificial intelligence algorithm and the iterative mechanism of the Q-learning method.First,for the distributed PV under droop control,the feedback between voltage amplitude and reactive power is constructed based on the total reactive power deviation of microgrid.Second,the distributed PV out⁃put voltage amplitude is controlled according to the corresponding action of the maximum reward Q value.Finally,the distributed PV reactive power output is coordinated and controlled to achieve the global optimal control of high-propor⁃tion PV microgrid.Simulation results show that the proposed strategy can effectively improve the control effect of reac⁃tive power sharing,reduce the system’s reactive power circulation,and improve the voltage quality.