【摘要】 负荷模型的准确性对于电力系统仿真分析至关重要。针对目前负荷模型无功方面的研究较为薄弱的问题,提出一种功率耦合型的人工神经元网络负荷建模方法,首先利用传统方法建模,构造样本空间,即以恒阻抗模型和一阶电动机模型构成的综合负荷作为负荷模型,其时域仿真潮流解作为训练样本,然后以BP算法训练神经网络,用训练好的网络来仿真负荷功率。结果显示,当网络输出为负荷有功功率和负荷无功功率共同构成的二维向量时,可使负荷模型无功方面的精度有较大的提高。更多还原

【Abstract】 The accuracy of load modeling is critical to power system simulation. Due to the present situation that research on reactive load modeling is relatively weak, this paper proposed a load modeling method based on neural network with real and reactive power coupled. Firstly sampling space was created by traditional way of load modeling, the integration of constant impedance and inductive motor was taken as load model, and the time-domain power flow solutions was taken as training samples, then the neural network was trained in BP algorithm, and the trained network was used to simulate load power responses. The results show that the reactive accuracy of the load model can be greatly improved when the network output includes real and reactive load power.更多还原