基于DDPG算法的微网负载端接口变换器自抗扰控制

直流微电网是新能源综合利用的重要形式，但其中的分布式接口往往存在着强随机性扰动，这给直流变换器的稳压控制带来了诸多问题。为了尽可能地抑制控制器参数固定时这种不确定性特征引起的不利影响，提出了一种利用深度确定性策略梯度(deep deterministic policy gradient, DDPG)算法整定线性自抗扰控制器参数的方法。依靠引入了智能算法的自抗扰微电网控制系统，实现了控制器参数的自适应调整，从而实现了微电网接口变换器的稳定运行。通过仿真对比了各类典型工况下，DDPG-LADRC与传统线性自抗扰控制器(linear active disturbance rejection control, LADRC)、双闭环比例-积分控制器的性能差异，验证了所提控制策略的有效性。而参数摄动下的鲁棒性分析结果结合多项指标下的系统整体性分析，充分体现了控制器参数的智能化调整所带来的多工况自适应性增益的优越性，具备较强的工程价值。

Active disturbance rejection control of a microgrid load-side interface converter based on a DDPG algorithm

The DC microgrid is an important form of comprehensive utilization of new energy, but the distributed interface often has strong random disturbance, which brings many problems to the voltage stabilization control of a DC converter. To suppress the adverse effects caused by this uncertainty feature when the controller parameters are fixed as much as possible, this paper proposes a method for setting linear ADRC controller parameters using a deep deterministic policy gradient algorithm (DDPG), and relies on the automatic rejection microgrid control system introduced by introducing intelligent algorithms to realize the adaptive adjustment of controller parameters, so as to realize the stable operation of the microgrid interface converter. Simulation curves are used to compare the performance differences between DDPG-LADRC and traditional LADRC and the double-loop PI controller proposed in this paper in various typical working conditions. The effectiveness of the control strategy proposed in this paper is verified. The robustness analysis results under parameter perturbation combined with the overall analysis of the system under multiple indicators fully reflect the superiority of multi-condition adaptive gain brought by the intelligent adjustment of controller parameters, and have strong engineering value.