基于深度强化学习的多端背靠背柔性直流系统直流电压控制
DC Voltage Control of Back-to-back Multi-terminal VSC-HVDC System Based on Deep Reinforcement Learning
-
摘要: 为了提高互联配电网多端背靠背柔性直流系统的直流电压控制精度,增强抗干扰能力,提出一种基于深度强化学习的直流电压控制方法,将深度学习神经网络与确定策略梯度融合,实现连续动作搜索,自适应调整电压控制策略。首先,建立多端背靠背柔性直流系统数学模型,分析直流电压控制的非线性和不确定性特征;然后,给出了基于深度强化学习的直流电压控制算法框架,设计了动作与状态空间、奖励函数、神经网络和学习流程;最后,通过仿真分析发现,相比传统比例-积分(PI)控制方法,所提方法具有更好的动静态性能,有效提高了直流电压的控制精度,减小了扰动下直流电压波动和功率超调,缩短了直流电压和功率的恢复稳定时间。Abstract: To improve the accuracy of DC voltage control and anti-disturbance ability of back-to-back multi-terminal voltage source converter based high voltage direct current（VSC-HVDC） in interconnected distribution networks, a DC voltage control method based on deep reinforcement learning is proposed. The deep learning neural network is integrated with the deterministic policy gradient to realize continuous action search and adaptive adjustment of voltage control strategy. Firstly, a mathematical model of back-to-back multi-terminal VSC-HVDC system is established, and the nonlinear and uncertain characteristics of DC voltage control are analyzed. Then, the framework of DC voltage control algorithm based on deep reinforcement learning is proposed. The action and state space, the reward function, the neural network and the learning process are designed. Finally, the simulation results show that the proposed method has better dynamic and static performances than the conventional proportion-integral（PI）control method. It can significantly improve the control accuracy of DC voltage, reduce the DC voltage fluctuation and power overshoot under disturbance, and shorten the recovery time of DC voltage and power.
下载:
