基于深度确定性策略梯度与模糊PID的直流微电网VRB储能系统就地层功率控制

针对直流微电网全钒液流电池(vanadium redox flow battery, VRB)储能系统在实际运行时就地控制层中的功率控制器存在时滞、精度低及抗干扰能力差等问题，提出了一种基于深度确定性策略梯度与模糊PID的功率跟踪控制策略。首先，建立VRB的等效电路模型来描述功率传输特性，并设计了由模糊PID与深度确定性策略梯度(deep deterministic policy gradient, DDPG)算法组成的复合控制器。将模糊PID作为主控制器对功率环进行控制，DDPG作为辅助控制器来补偿功率跟踪误差。然后，设计了VRB储能系统就地层功率跟踪控制器，采用麻雀搜索算法(sparrow search algorithm, SSA)对PID参数和模糊规则进行优化，并通过阶跃信号对优化后的系统输出响应进行测试。同时将分配指令功率与储能单元给定功率偏差作为数据集在DDPG中进行训练，以提高主控制器的响应速度和抗干扰能力。最后，通过在3种不同场景的算例下进行仿真，验证了控制策略的有效性及稳定性。结果表明：所提控制策略在电池充放电时，能够快速地跟踪到功率指令值；实时跟踪时，跟踪功率值与调度指令值偏差小于±2%；受到扰动时，能准确修正功率偏差，满足实际要求。

Power control of the ground layer of a VRB energy storage system in a DC microgrid based on depth deterministic policy gradient and fuzzy PID

A power controller has problems of time delay, low control accuracy, and poor anti-interference ability in the actual operation of a vanadium redox flow battery (VRB) energy storage system in the ground layer of a DC microgrid. Thus a power cooperative control strategy based on the depth deterministic policy gradient and fuzzy PID is proposed. First, the equivalent circuit model of the VRB is established to describe the power transmission characteristics, and a composite controller composed of a fuzzy PID and a deep deterministic policy gradient (DDPG) algorithm is designed. In addition, the fuzzy PID is used as the main controller to control the power loop, and the DDPG is used as the auxiliary controller to compensate for the power tracking error. Then, the power tracking controller of the VRB energy storage system in the ground layer is designed, the PID parameters and fuzzy rules are optimized by the sparrow search algorithm (SSA) algorithm, and the output response of the optimized system is further tested by step signal. At the same time, the deviation between the command power of the distribution layer and the given power of the energy storage unit is used as a data set for training in the DDPG to improve the response speed anti-interference capability of the main controller. Finally, the effectiveness and stability of the control strategy are verified by simulation in three different scenarios. The results show that the proposed control strategy can quickly track the power command value during battery charging and discharging. The deviation between tracking power value and dispatching instruction value is less than ±2% during real-time tracking; the power deviation can be corrected accurately to meet the actual requirements when the power controller is disturbed. This is in line with the actual requirements.