一种超导磁储能装置控制方法

A control method for superconducting magnetic energy storage device

电力系统发生故障时需要快速平抑振荡以保障系统稳定性.超导磁储能装置(Superconducting Magnetic Energy Storage Device,SMES)能够迅速跟踪功率波动,进行4象限功率补偿,有效提高系统暂态稳定性.针对电压源型变流器的超导磁储能装置,提出了一种基于执行依赖启发式动态规划(Action Dependent Heuristic Dynamic Programming,ADHDP)智能算法的超导储能外环控制方法.该方法通过强化学习,自适应调整结构参数,实现系统的最优控制.在MATLAB/Simulink环境下建立了单机无穷大系统和2机系统仿真模型,分别对传统的PI控制、固定学习率ADHDP控制器和自适应学习率ADHDP控制器的控制效果进行对比分析.仿真结果表明,自适应学习率ADHDP具有较明显的优势;另外,在系统连续故障情况下,ADHDP表现出了较好的学习功能,能够获得比上次更好的控制效果.

when a fault occurs in the power system, the power oscillation must be damped as soon as possi- ble to ensure the stability of the system. Superconducting magnetic energy storage （SMES） device can track the power fluctuation rapidly and accomplish power compensation in all four quadrants. It is an effec- tive device to enhance the stabilization of the electric system. Based on the voltage source converter （VSC）, an intelligent algorithm called as action dependent heuristic dynamic programming （ADHDP） is proposed for the outer-loop control of the SMES. Through the reinforcement learning, the designed con- troller can adjust its parameters adaptively and obtain the best performance as well. Finally, simulations are carried out in the MATLAB/Simulink. The comparison between the conventional proportional integral （PI） method, the fixed learning rates ADHDP method and the adaptively learning rates ADHDP method is conducted in the single-machine infinite bus system and two machine system. The results show that the a- daptively learning rates ADHDP controller has obvious advantages. Moreover, the ADHDP controller shows a good learning ability by obtaining a better control performance under the condition of continuous faults.