基于深度学习的恒功率负荷直流微电网稳定性分析

Deep Learning Based Stability Analysis of DC Microgrid with Constant Power Loads

针对恒功率负荷(CPL)因负阻抗特性导致直流微电网母线电压失稳的问题,从限制CPL取值范围出发,引入深度学习算法分析系统的稳定性。首先,通过采用直流微电网机理模型仿真,建立初始数据库;然后,对数据库进行预处理,在此基础上,利用最大信息系数进行相关性分析,得到系统参数与CPL临界值之间的相关性强弱信息,并实现输入特征降维;接着,采用自定义损失函数与量化预测性能等方法设计适用于CPL临界值预测的深度学习策略。最后,针对不同场景的直流微电网进行仿真测试,并通过硬件在环实验验证所提深度学习策略的有效性和优越性。

Aiming at the bus voltage instability problems of DC microgrids caused by constant power loads(CPLs)due to negative impedance characteristics,starting from limiting the value range of CPL,a deep learning algorithm is introduced to analyze the system stability.Firstly,the mechanism model of the DC microgrid is adopted to conduct the simulations,thus setting the initial database.Secondly,the database is preprocessed,and on this basis,the maximal information coefficient is used for correlation analysis to obtain the relationship strength between the system parameters and CPL critical values,as well as to achieve the dimension reduction of input features.Thirdly,a deep learning strategy suitable for CPL critical value prediction is designed by using a custom loss function and quantified prediction performance methods.Finally,the simulation tests are conducted on the DC microgrid in different scenarios,and the effectiveness and superiority of the proposed deep learning strategy are validated through the hardware-in-the-loop experiments.