简化前馈网络用于负荷模型参数辨识的研究

RESEARCH ON PARAMETER IDENTIFICATION OF LOAD MODELS BY USING LINEAR BP NETWORKS

负荷的精确建模在电力系统分析中的重要性已被电力工作者所认识到。文中在对自适应的三层前馈网络(ABP)进行修改、简化后,得到简化(或线性)前馈网络(LBP)。并结合现场试验和仿真数据,对用线性前馈网络和辅助变量法对负荷建模的精度进行了比较。把线性前馈网络作为一种算法,对负荷动态模型即差分方程的参数进行了辨识。同时,应用 LBP 对负荷静态模型即幂函数和多项式模型的参数进行了辨识,还提出基于人工神经网络的负荷动静模型同时辨识法。

The importance of accurate load models has been known for many years in power system analysis. A simplified linear BP network (LBP) is put forward after a modification of typical three layers BP network. As a new method, the LBP networks have been applied to the parameter identification of static and dynamic models of loads, such as differential equation model; exponential model and polynomial function model. And this method is applied to field tests data and is verified. Parameters of static and dynamic models can be identified by LBP at the same time. The feasibility and validity of this method has been proved.